Nanoparticles for in-vitro and in-vivo biosensing and imaging

UNIVERSITÀ DEGLI STUDI DI MILANO BICOCCA
Scuola di Dottorato di Scienze
Corso di Dottorato di Ricerca in Fisica e Astronomia
Laura Sironi
Matricola 041003
Nanoparticles for in-vitro and
in-vivo biosensing and imaging
Tutore
Prof. Giuseppe Chirico
Coordinatore
Prof. Claudio Destri
Ciclo XXIII 2008-2010

Contents
1 Metal nanoparticles 11
1.1 Nanoscaled materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2 Historical overview: gold through the centuries . . . . . . . . . . . . . . . 12
1.3 Physical and chemical properties . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Optical properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Mathematical origin of the plasmon resonances . . . . . . . . . . . . . . . 17
1.5.1 Mie theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.6 Mie-theory approximations . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6.1 Size dependence: small particles . . . . . . . . . . . . . . . . . . . 21
1.6.2 Size dependence: larger particles . . . . . . . . . . . . . . . . . . . 22
1.7 Intrinsic size effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.7.1 Specific phenomena influencing the surface plasmon absorption . . 25
1.8 Shape-dependent properties : the Gans theory . . . . . . . . . . . . . . . 26
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2 Principles 35
2.1 Basics of Fluorescence Optical Spectroscopy . . . . . . . . . . . . . . . . . 35
2.1.1 Fluorescence lifetimes and quantum yields . . . . . . . . . . . . . . 37
2.2 Radiative decay engineering . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3 Review of metallic surface effects on fluorescence . . . . . . . . . . . . . . 41
2.3.1 Theory for Metallic ParticlesFluorophore Interactions . . . . . . . 41
2.4 Two-photon excited fluorescence . . . . . . . . . . . . . . . . . . . . . . . 43
2.4.1 One-photon and two-photon excitation Point Spread Functions . . 46
2.4.2 OPE versus TPE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3 The techniques 54
3.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.1 General insight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2

CONTENTS 3
3.2 Dynamic Light scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3 Depolarized light scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4 Method of Cumulants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.5 MemExp program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6 MemExp:a MEM/NLS algorithm . . . . . . . . . . . . . . . . . . . . . . . 63
3.7 Fluorescence Fluctuation Spectroscopy (FFS) . . . . . . . . . . . . . . . . 66
3.8 Fluorescence Correlation Spectroscopy . . . . . . . . . . . . . . . . . . . . 67
3.8.1 The auto-correlation function G(τ): Brownian diffusion . . . . . . 68
3.8.2 The auto-correlation function G(τ): beyond diffusion . . . . . . . . 71
3.8.3 Pseudo cross-correlation function . . . . . . . . . . . . . . . . . . . 74
3.9 Photon Counting Histogram (PCH) . . . . . . . . . . . . . . . . . . . . . 76
3.9.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.9.2 Freely diffusing particles . . . . . . . . . . . . . . . . . . . . . . . . 77
3.9.3 Photon Moment Analysis (PMA) . . . . . . . . . . . . . . . . . . . 81
3.10 Optical pathway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.11 Setup calibration with a reference dye . . . . . . . . . . . . . . . . . . . . 82
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4 Nano-bio sensors for protein detection 90
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.2 p53 protein properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.2.1 P53 structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.2.2 Control of the p53 protein half-life . . . . . . . . . . . . . . . . . . 97
4.2.3 p53 induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.2.4 P53 often ushers in the apoptotic death program . . . . . . . . . . 99
4.3 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.3.1 Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.3.2 Experimental methods . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.4 Fluorescence Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.5 Photon Counting Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.6 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.6.1 Dynamic Light Scattering: cumulant analysis . . . . . . . . . . . . 107
4.7 Dynamic Light Scattering of p53 construct: MEM analysis . . . . . . . . 108
4.7.1 Absorption spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.7.2 Characterization of complexed FITC in solution . . . . . . . . . . 112
4.8 Detection of the model protein BSA . . . . . . . . . . . . . . . . . . . . . 112
4.8.1 Gold NP-FITC complexes in the absence of BSA . . . . . . . . . . 112
4.8.2 Protein interactions with the FITC gold NP complexes . . . . . . 114

4 CONTENTS
4.9 Fluorescence Spectroscopy: Burst analysis . . . . . . . . . . . . . . . . . . 114
4.9.1 Aggregates size estimate . . . . . . . . . . . . . . . . . . . . . . . . 114
4.9.2 Excited state lifetime in the presence of BSA . . . . . . . . . . . . 115
4.10 Basic gold nanocrystal protein sensor . . . . . . . . . . . . . . . . . . . . . 120
4.11 p53 protein detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.12 Fluorescence burst analysis in solution . . . . . . . . . . . . . . . . . . . . 121
4.13 Dependence of the FITC Lifetime on the p53 Concentration . . . . . . . . 123
4.14 In Vitro Selectivity of the p53 Assay . . . . . . . . . . . . . . . . . . . . . 125
4.15 In Vivo Test of the p53 Assay . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.16 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5 Anisotropic nanoparticles 132
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.2 Gold nanoparticles luminescence . . . . . . . . . . . . . . . . . . . . . . . 133
5.2.1 Two-photon luminescence (TPL) . . . . . . . . . . . . . . . . . . . 135
5.3 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.3.1 Nanoparticles synthesis . . . . . . . . . . . . . . . . . . . . . . . . 137
5.3.2 Cell culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
5.4.1 UV-Vis characterization . . . . . . . . . . . . . . . . . . . . . . . . 139
5.4.2 Transmission Electron Microscopy (TEM) characterization . . . . 141
5.4.3 Z-potential characterization . . . . . . . . . . . . . . . . . . . . . . 144
5.5 Dynamic Light Scattering characterization . . . . . . . . . . . . . . . . . . 144
5.5.1 Solutions of LSB (LSB micelles) . . . . . . . . . . . . . . . . . . . 144
5.5.2 Dynamic light scattering of the NPs . . . . . . . . . . . . . . . . . 146
5.6 FCS experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5.7 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.8 Concentration evalutation . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.9 TPL dependence on excitation power . . . . . . . . . . . . . . . . . . . . . 157
5.10 Emission Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.11 Excitation spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.12 Dependence of TPL on the Polarization of the exciting field light beam . 161
5.12.1 Citotoxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.13 Cellular uptake . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.14 Photothermal effects of gold NRs . . . . . . . . . . . . . . . . . . . . . . . 171
5.15 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.15.1 Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . 171

CONTENTS 5
5.15.2 Spectral characterization . . . . . . . . . . . . . . . . . . . . . . . 173
5.15.3 Fluorescence Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . 173
5.15.4 Dye-Lifetime measurement . . . . . . . . . . . . . . . . . . . . . . 173
5.16 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.17 Rhodamine-B characterization . . . . . . . . . . . . . . . . . . . . . . . . 174
5.17.1 Fluorescence emission measurement . . . . . . . . . . . . . . . . . 174
5.17.2 Lifetime measurement . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.18 Assay characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.19 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
6 In-vivo microscopy 182
6.1 Intravital microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
6.2 Problems in IVM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
6.3 The biological problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
6.3.1 NK-DC interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
6.3.2 Sample preparation methods . . . . . . . . . . . . . . . . . . . . . 191
6.4 Data acquisition and analysis . . . . . . . . . . . . . . . . . . . . . . . . . 192
6.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
6.5.1 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . 192
6.5.2 Lymph nodes topography . . . . . . . . . . . . . . . . . . . . . . . 200
6.5.3 Validation of the experimental setup performances: T and DC cells 201
6.5.4 NKs in steady state conditions . . . . . . . . . . . . . . . . . . . . 202
6.6 DC-NK cells dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.8 Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.9 Immune system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.10 PRRs and control of adaptive immunity . . . . . . . . . . . . . . . . . . . 218
6.10.1 TLRs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
6.11 Dendritic cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
6.12 Natural Killer cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
6.13 Lymph-nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
6.13.1 Lymph-node architecture . . . . . . . . . . . . . . . . . . . . . . . 224
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
6.14 Laser sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
6.14.1 Millennia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
6.14.2 Titanium-Sapphire (Ti-Sa) optical resonant cavity . . . . . . . . . 236
6.15 Microscopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
6.15.1 Nikon TE300 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

6 CONTENTS
6.15.2 Olympus BX51 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
6.16 The non-descanned detection system . . . . . . . . . . . . . . . . . . . . . 243
6.16.1 The ND-unit design . . . . . . . . . . . . . . . . . . . . . . . . . . 244
6.17 Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
6.17.1 Single Photon Avalanche Diode (SPAD) . . . . . . . . . . . . . . . 246
6.18 Dynamic Light Scattering:optical system . . . . . . . . . . . . . . . . . . . 248
6.18.1 Softwares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
6.19 TimeHarp and Symphotime softwares . . . . . . . . . . . . . . . . . . . . 252

Introduction
In the last two decades several groups have investigated the changes of chemical and
physical properties of materials with size in nanometric scales. These studies have highlighted
a number of possible applications for nanostructures, which are now employed,
for example, in biology and medicine for imaging, disease detection, diagnosis, sensing
and therapy.
Noble metal (especially gold and silver nanoparticles) are particularly versatile for these
applications due to the phenomenon known as surface plasmon resonance (SPR), an
in-phase oscillation of all the conduction band electrons that resonates with the light
wave electric field. The resonance frequency depends on the size, shape, orientation and
dielectric constant of the nanoparticle. This coupling of SPR with the electromagnetic
field leads to a great enhancement of all the nanoparticle radiative properties, such as
absorption and scattering. The extinction cross section of these nanoparticles is 10 5 -10 6
larger than that of organic dye and in contrast to common molecular chromophores they
are extremely photostable and, depending on the shape, they convert efficiently light
into heat.
The SPR, tunable in the visible (for spherical NPs) and near-infrared region (for anisotropic
Nps) of the electromagnetic spectrum, can also interact, for gold nanoparticles a few
nanometers in size, with the fluorescence emission of dyes and substantially modify their
brightness and excited-state lifetime. Depending on the fluorophores-NP distance and
the NPs anisotropy one can obtain fluorescence enhancement or quenching. In both
cases we expect that any change in the dielectric constant of the NP surface, induced,
for example, by a biorecognition process that occurs on the surface itself, can produce a
change in the emission properties of the fluorophores.
SPR effect becomes also particularly important when combined with two-photon excitation
(TPE), which consists in the simultaneous absorption of two photons, each carrying
about half the energy necessary to excite the molecule. This excitation technique has
been particularly useful in the context of this work because it affects greatly the luminescence
quantum yield of anisotropic NPs: due to non-linear phenomena such as the
TPE, the luminescence (TPL) intensity is enhanced (when coupled with an appropriate
7

8 Introduction
plasmon resonance) by many orders of magnitude with respect to the single photon excitation
of at surfaces, improving the usefulness of these nanoparticles for in-vivo imaging
in the NIR region of the electromagnetic spectrum.
TPE offers a series of unique features for biological investigation both in vitro and in
vivo. First, the two-photon absorption bands of the dyes commonly used in biological
studies are wider than their one-photon analogous allowing the simultaneous excitation
of multiple fluorophores with a single excitation wavelength. Second, the stimulating
light beam has a high penetration depth because of the long infra-red wavelengths used,
allowing experiments in turbid media. Third, excitation takes place only at the plane of
focus, due to the scaling of the probability of simultaneous photon absorption with the
square of the light intensity. As a consequence TPE avoids the simultaneous absorption
of photons outside the specimen drastically reducing both photo-toxicity and fluorophore
bleaching. These advantages make nowadays TPE a well established tool for scientific
biological and medical research and can be coupled to anisotropic gold nanoparticles.
According to these considerations we have developed our project on two lines related to
the use of gold nanoparticles for sensing and non linear imaging.
In the first part of this work we have reviewed the materials and methods used: in
Chapter 1, the most relevant features of the surface plasmon resonance and the properties
of spherical symmetric and asymmetric nanoparticles are presented. The interaction
between fluorophores and nanoparticles and the basics of fluorescence emission
and two-photon excitation employed in the reported measurements are discussed, while
the physical basis of the Dynamic Light Scattering and the Fluorescence Correlation
Spectroscopy (FCS), the main technique employed in this PhD thesis, are described in
Chapter 3. The instruments employed for the measurements are instead described in
Appendix A.
The aim of the first part of this research project (reported in Chapter 4) is to exploit
changes of the dye excited-state lifetime and brightness induced by its interaction with
the gold surface plasmons for detection of tiny amounts of protein in solution under
physiological conditions. The system we investigated is based on 10 and 5 nm diameter
gold NPs coupled (via a biotin-streptavidin linker) to the FITC dye and to a specific
protein antibody. The interaction of the fluorophore with the gold surface plasmon resonances,
mainly occuring through quenching, affects the excited state lifetime that is
measured by fluorescence burst analysis in standard solutions. The binding of protein
to the gold NPs through antigen-antibody recognition further modifies the dye excitedstate
lifetime. This change can therefore be used to measure the protein concentration.
In particular, we have tested the nanodevice measuring the change of the fluorophore
excited-state lifetime after the binding of the model protein bovine serum albumin (BSA);

Chapter 9
then we have applied the nanoassay in order to recognize the p53 protein, whose detection
in the body is highly valuable as marker for early cancer diagnosis and prognosis,
both in standard solutions and in total cell extracts. The selectivity of the construct
with respect other globular proteins has been also addressed. The data indicate that the
FITC excited-state lifetime is a very sensitive parameter in order to detect tiny amounts
of protein in solution with an estimated limit of detection of about 5 pM, mostly determined
by the statistical accuracy of the lifetime measurement.
In the second part of the project (discussed in Chapter 5) we focused on the exploitation
of anisotropic gold nanoparticles as probes in cellular imaging. We have then studied
their photoluminescence (TPL) properties under two photon excitation. We have focused
on gold nanorods that can easily be obtained by synthesis with the standard surfactant
CTAB (cetyl trimethulammonium bromide).
The synthesis of asymmetric branched gold nanoparticles, obtained using for the first
time in the seed growth method approach a zwitterionic surfactant, laurylsulphobetaine
(LSB), has been developed in collaboration with the University of Pavia. We have shown
that LSB concentration in the growth solution allows to control the dimension of the
NPs and the SPR position, that can be tuned in the 700-1100 nm Near Infrared range.
The samples have been analized with several structural techniques to obtain a complete
characterization: from the data obtained through the absorption spectra in the
UV-Visible region, TEM images of the solutions, Fluorescence Correlation Spectroscopy
(FCS) and Dynamic Light Scattering (DLS) experiments. From the analysis of the data
we reached information on the nanoparticles shapes, dimensions and aggregation.
Three different populations have been found: nanospheres with diameter lower than 20
nm, nanostars characterized by large trapezoidal branches, and asymmetric branched
nanoparticles with high aspect ratio.
For imaging applications, a spectroscopic characterization was performed by employing
two photon excitation (TPE): the dependence of the TPL intensity on the power, wavelength
and polarization of the incident light intensity was studied.
TPL was finally exploited to study the cellular uptake of the nanoparticles in different
cell lines (macrophages and HEK cells). Beyond their use as constrast agent in imaging
and in photothermal therapy, NPs can be employed as drug carriers and can stimulate
and/or suppress the immune responses. Nanoparticles can also be engineered to serve as
vaccine carriers and adjuvants (agents added to a vaccine to augment immune responses
toward antigens), which can act in several ways to increase both native and adaptative
immune responses that will generate an effective immunological memory.
In this scenario and in order to employ in the next future the peculiar emission properties
of anisotropic nanoparticles to visualize, with improved sensitivity, real time cellular

10 Introduction
and molecular processes and to evaluate their immune properties, an initial study of cell
dynamics in-vivo in standard condition (with cells labelled with traditional fluorescent
dyes) has been performed in parallel to the experiments described above (Chapter 6).
With this aim, we have setup Intravital TPE Microscopy (IVM) experiments, which
afford a view into the lives and fates of diverse immune cell populations in lymphoid
organs and peripheral tissues.
In particular, the interaction between two cell lines belonging to the immune system
(described briefly in Appendix B), dendritic (DC) and natural killer (NK), has been
studied. In this field, I have developed a novel analysis method that enable the characterization
of NK dynamic behavior, based on the supervising of a set of parameters
(mean and istantaneous velocity, confinement ratio, NK-DC distance, root-mean square
displacement). This approach has allowed to point out that immune NK cells properties
can be activated as a consequence of a direct interaction with DC cells.

Chapter 1
Metal nanoparticles
1.1 Nanoscaled materials
The research in nanoscaled matter began to grow exponentially when it became recognized
that the properties of materials change drastically as their sizes decrease from the
bulk to small clusters of atoms [1]. Suitable control of the properties of nanometer-scale
structures can lead to new science as well as new devices and technologies [2]. Two
principal factors cause the properties of nanoscaled materials to differ significantly from
their behavior in bulk. First, the increased relative surface area and, second, the size
dependent properties begin to dominate when matter is reduced to the nanoscale. Not
only these effects change the chemical reactivity and strength drastically, but also the
electrical, optical and thermal response of the matter.
The most striking phenomenon encountered in metal nanoparticles are electromagnetic
resonances due to the collective oscillation of the conduction electrons. These
so-called localized Surface Plasmon Resonances (SPR) induce a strong interaction with
light; the wavelenght at which this resonance occurs depends on the local environment
and on the shape, size and orientation of the particle. The fundamental aspects and the
potential technological applications of SPR have been widely investigated in the past
decade, during which metal nanoparticles have been used for catalysis, sub-wavelenght
optical devices, surface enhanced Raman spectroscopy, diagnostic application, biological
imaging, sensing and for cancer treatment. Moreover, the compatibility of NPs and
biological systems, proteins and oligonucleotides is well established. Small particles
of different sizes and shapes have been shown to be suitable markers, contrast agents
(for instance in optical coherence tomography) and therapeutic agents for biomedical
applications [3],[4],[5]. For all these issues, it is advantageous to be able to tune the
particle plasmon resonance to the near-infrared region between 650 and 900 nm, where
water and hemoglobin have their lowest absorption coefficient [6],[7],[8]. This tunability
11

12 Metal nanoparticles
is provided by engineering the shape (principally spheric and elliptic) and the dimension
of the particles.
The following sections describe briefly the significant properties and the origin of the
plasmon resonances in spherical and ellipsoidal metal nanoparticles.
Figure 1.1: Nanoparticles in comparison with other biological entities.
1.2 Historical overview: gold through the centuries
The fact that gold compounds could impart a red color to the objects was well known;
Egyptian manuscripts from the Greco-Roman era refer to it [9]. The Egyptians, Greeks
and Romans used many colored pigments for the decoration of their buildings, ceramics
and glass-ware. The use of gold and silver particles in glassblowing is evident in the
famous Lycurgus Cup (Fig.1.2). Gold NPs scatter green and transmit red light so the
cup appear to be red or green depending on the position of the light source (inside or
outside) [10]. Chemical analysis of the Lycurgus cup shows that it is similar to most
other Roman glass, but it contains very small amounts of gold (about 40 parts per
million) and silver (about 300 parts per million). In figure 1.2 (right), a TEM image, a
silver-gold alloy coming from a sample of the Lycurgus cup is clearly visible [11]. The
crystalline nature of the Ag/Au particles and their fine dispersion in the glass suggests
that this colloidal metal was precipitated out from solution by heat treatment.
In the Middle Ages, chemistry developed mostly through the protoscience of Alchemy.
Alchemical processes required the construction of scientific apparatus and the invention
of many laboratory techniques like heating, refluxing, extraction, sublimation and distillation
to treat metals with various chemical substances [12]. In the early thirteenth
century the improvement of distillation methods resulted in the discovery of the mineral
acids which greatly increased the power of the alchemist to dissolve substances and to

Chapter 1 13
carry out reactions in solution. The ’royal’ solvent for gold was found to be ”aqua regia”,
created by adding salt ammoniac (ammonium chloride) to aqua fortis (HNO 3 ) [13].
Figure 1.2: Left: The Lycurgus cup dates from the fourth century AD. In reflected light (daylight) the
glass appears to be green, but when light is transmitted from the inside of the vessel it is red [10]. Right:
TEM image of a Ag/Au alloy [11]
Paracelsus (16th century), used gold in the preparation of drugs to relieve suffering,
as a cure for ailments and particularly for heart disorders. Recipes for the preparation
of medicinal colloidal gold solutions were well known by the early seventeenth century
[13]. Closely to this use in medicine, colloids were employed to produce ruby glass;
the florentine priest Antonio Neri, in 1612, wrote the first treatise about ruby glasses.
A contemporary chemist, Johann Kunckel, published his ’Ars Vitraria Experimentalis’,
which has been a standard treatise on glass technology for many years. It concerns the
production of ruby glass by the use of the purple precipitate, referred to as the ’Purple
of Cassius’ because in 1685 a doctor called Andreas Cassius published a basic alchemical
work ”De Auro”, where he described a method for the precipitation of gold [9].
A more scientific approach to colloidal gold
The first scientific study of metal nanoparticles is dated back to the seminal work of
Michael Faraday around 1850. He was the first to recognise that the red colour of gold
colloid was due to the minute size of the Au particles and that one could turn the preparation
to blue by adding salt to the solution. He obtained gold colloids reducing AuCl − 4
by phosphorus, following a procedure already reported by Paracelsus in the 16th century.
Faraday’s discovery that metals could form colloids was a breakthrough, although the
importance of his observations was not fully realized at that time; today his studies are
generally considered to mark the foundations of modern colloid science [14] [15]. The

14 Metal nanoparticles
word ”colloid” was first introduced by a London colleague of Faraday, Thomas Graham
(1805-1869), who is referred to as ”the father of colloid chemistry”. Subsequently the
German physicist Gustav Mie (1869-1957) published in 1908 an important article on
light scattering in matter, describing the optical properties of small metal particles; he
calculated the absorbance of colloidal gold particles as a function of the particle size
using classical electromagnetic theory.
In 1925 Richard Adolf Zsigmondy (1865-1929) was awarded by the Nobel price [16]
for his demonstration of the heterogenous nature of colloid solutions and for the methods
he used, which have since became fundamental in modern colloid chemistry. Zsigmondy
prepared practically equally-sized red gold hydrosols by the reduction of gold chloride
with formaldehyde in a weakly alkaline solution. Finally, he showed that colloidal gold
particles have a negative electrical charge, which largely determined their stability. This
charge can be screened or exchanged by adding salts resulting in an immediate particle
aggregation whereby the system coagulates.
Since colloidal gold has a particle size falling below the resolution limit of the optical
microscope, the world of colloids has been opened up to a detailed study with the invention
of the electron microscope at the beginning of World War II [17]. An interesting
investigation of various preparations of colloidal gold using the electron microscope as
the main tool started in 1948 at Princeton University and the RCA Laboratories by John
Turkevich and colleagues, who studied the reduction of gold salt with sodium citrate extensively.
G. Frens used the citrate reduction method studied by Turkevich to control the
size of gold particles ranging from 16 nm to 150 nm simply by varying the concentration
of sodium citrate added to the solution during the nucleation of the particles [18]. The
real renaissance of gold nanoparticles, however, started when M. Brust and coworkers
reported a simple reductive method using the borohydride reduction of chloroauric acid
in the presence of alkane thiols [19]: functionalized groups on gold colloids surfaces allow
their incorporation in three dimensional networks and make them useful in a wide range
of applications in the fields of sensors and molecular electronics.
1.3 Physical and chemical properties
Gold and silver are known for being generally inert and, especially gold, for not being
attacked by O 2 to a significant extent; this makes AuNP and AgNP stable in ordinary
conditions [20]. Gold nanoparticles are resistant also to strong oxidizing or highly acid
environments, on the other hand ”aqua regia” or solutions containing CN − can immediately
dissolve them [21]. Both Au and Ag are reactive with sulfur; in particular in
the case of organic thiols, ligation to nanoparticles surface is particularly effective for
the contemporary presence of a σ type bound, in which sulfur is the electron density

Chapter 1 15
donor and the metal atom is the acceptor, and a π type bound, in which metal electrons
are partially delocalised in molecular orbitals formed between the filled d orbitals of the
metal and the empty d orbitals of sulfur[20]. Since solid to liquid transition begins at
interfaces, a well known feature of nanometric particles is the lower melting temperature
with respect to the bulk. For instance gold undergoes a decrease in melting temperature
of about 400 ◦ C going from 20 nm to 5 nm particles and about 50 ◦ C going from the
bulk to 20 nm particles.
When nanoparticles size becomes comparable to the de Broglie’s wavelength of an electron
at the Fermi energy (0.5 nm for gold and silver), quantum size effects on the optical
properties of metal nanoparticles induce fluorescence emission, due to transitions between
discrete electronic states [22].
Thermal conductivity is enhanced for small particles due to higher surface to volume
ratio, and phonons energy become higher for very small particles [21].
Finally, when a metal particle decreases in size to become a nanoparticle or a nanocrystal,
a larger proportion of atoms is found at the surface. This observation together with
the fact that catalytic chemical reactions occur at surfaces leads to much more pronunced
reactivity when a given mass of gold is used in nanoparticulate form compared to the
case in which microsocpic particles are used [21] [23].
1.4 Optical properties
Due to their metallic nature, AuNp and AgNp possess a huge number of easily polarizable
electrons, that means strong interactions of the matter with light and high non-linear
optical properties. [24],[25],[26],[27],[28]
As mentioned in section 1.1, the most striking phenomenon in metal NPs is the surface
plasmon resonance (SPR), namely the coherent displacement of the conduction band
electrons from their equilibrium positions around their positive ionic core, in the presence
of a resonant electromegnetic wave (Figure 1.3).
These plasmon resonances give rise to the Surface Plasmon Absorption (SPA) band,
which lays in the visible spectral window for gold and silver particles with size 2-100 nm
and confers the characteristic bright yellow and red colour to AgNp and AuNp colloidal
solutions respectively. Moreover, the large number of electrons involved in the SPA accounts
for the incredibly high extinction cross section of these nanoparticles, which is
10 5 -10 6 larger than that of organic chromophores [29]. Another general photonic peculiarity
of AuNp and AgNp is their solid state behaviour: in contrast to common molecular
chromophores they are extremely high photostable and they fast convert light into heat.
[8] [30]

16 Metal nanoparticles
Figure 1.3: Left: Schematic of plasmon oscillation for a sphere. From [8]. Right: Poynting vector
or energy flow (lines) around a subwavelength metallic colloid illuminated at the plasmon wavelength
(top) and at a wavelength longer than the plasmon wavelength (bottom). The vertical lines on the left
indicate the diameter of the cross sections for absorption. This diagram does not show the energy flow
due to scattered light. From [30]. A colloid is illuminated with light with a wavelength (λ) that matches
the Plasmon resonance at λ P (top) or with λ ≻ λ P (bottom). The lines show the energy flow near
the particle.
For λ = λ P , the colloid has an optical cross section much greater than its geometrical
cross section. This effect gives rise to the energy flow into the particle, which results in the enhanced
fields near the illuminated colloids. For λ ≻ λ P , the optical cross section can be similar to or smaller
than the geometrical cross section (bottom), which is typical of organic fluorophores and semiconductor
nanoparticles (quantum dot).
Other than size, the SPA is primarily affected by particles shape and physicalchemical
environment properties, including the reciprocal distance between particles.
Figure 1.4 shows the creation of a surface plasmon oscillation in a simple manner.
The electric field of an incoming light wave induces a polarization of the free conduction
electrons with respect to the much heavier ionic core of a spherical nanoparticle. A net
charge difference occurs at the nanoparticle boundaries (i.e. at the surface) which in
turn acts as a restoring force. In this manner a dipolar oscillation of the electrons is
created with period T.
Figure 1.4(b) shows the SPA of 22, 48 and 99 nm gold nanoparticles [31] prepared
in aqueous solution by the reduction of gold ions with sodium citrate [32] [33]. The
absorption spectra of the different particles are normalized at their SPA maximum for
comparison. The molar extinction coefficient is of the order of 1·10 9 M −1 cm −1 for 20 nm
nanoparticles and increases linearly with increasing volume of the particles [34]. Note
that these extinction coefficients are three to four orders of magnitude higher than those
for the very strongly absorbing organic dye molecules [31].

Chapter 1 17
Figure 1.4: Surface plasmon absorption of spherical nanoparticles and its size dependence. (a) A
scheme illustrating the excitation of the dipole surface plasmon oscillation.The electric field of an incoming
light wave induces a polarization of the (free) conduction electrons with respect to the much heavier
ionic core of a spherical gold nanoparticle.A net charge difference is only felt at the nanoparticle boundaries
(surface) which in turn acts as a restoring force. In this way a dipolar oscillation of the electrons
is created with period T. This is known as the surface plasmon absorption. (b) Optical absorption spectra
of 22, 48 and 99 nm spherical gold nanoparticles. The broad absorption band corresponds to the surface
plasmon resonance. From [35]
1.5 Mathematical origin of the plasmon resonances
The mathematical origin of the plasmon resonance was described by Mie in 1908 [36].
He solved the problem of light diffraction by a single sphere using classic electrodynamic
theory.
1.5.1 Mie theory
Solving the problem of absorption and scattering of light by a small particle means
solving Maxwell’s equations with boundary conditions. The general formulation of the
problem is shown in Figure 1.5.
Assuming harmonic time dependence of the light source, it is possible to rewrite
Maxwell’s equations into the Helmotz wave equation
∇ 2 E + k 2 E = 0 (1.1)
∇ 2 H + k 2 H = 0 (1.2)
where k is the wave number. Both the particle and the medium can be described

18 Metal nanoparticles
by the dielectric function ɛ and the magnetic permeability µ (generally, the relative
magnetic permeability of the materials under study is close to 1), that enter in the wave
number as k 2 = ω 2 ɛµ. At the boundary between the particle and the medium, ɛ and µ
are discontinuous. It follows that the normal components of the field are discontinuous
whereas tangential ones are continuous. For points x on the particle surface, we can
write (ˆn is the normal vector)
[E 2 (x) − E 1 (x)] × ˆn = 0 (1.3)
[H 2 (x) − H 1 (x)] × ˆn = 0 (1.4)
Figure 1.5: Sketch of the problem as it is treated in section . A particle with optical constants ɛ p and
µ p is embedded in a medium with optical constants ɛ m and µ m, and illuminated by a plane wave, which
generates an electric field E 1 and a magnetic field H 1 inside the particle. The particle radiates a scattered
field in all directions, which leads, together with the applied fields, to an electric field E 2 and a magnetic
field H 2 outside of the particle.
Only restricted to spherical particles, this problem is exactly solvable as shown in
1908 by Gustav Mie [36] (a complete derivation of Mie theory is given by Bohren and
Huffman [30]), and the scattering matrices can be derived. From these information about,
e.g., the direction and polarization dependence of the scattered light can be extracted.
An important parameter that can be also calculated is the cross section, a geometrical
quantity that relates the incident light to the scattered, absorbed or extincted power.
The absorption, scattering and extinction cross sections (σ abs , σ sca and σ ext respectively)
[37] [38],[39],[30] for an arbitrary spherical particle with dielectric function ɛ p are defined
as:
σ sca = P sca
I inc
σ abs = P abs
I inc
σ ext = P ext
I inc
(1.5)
Since the extincted power is the sum of the scattered and absorbed power, the absorption
cross section is simply

Chapter 1 19
where the scattering and extinction cross sections are:
σ sca = 2π
k 2
σ abs = σ ext − σ sca (1.6)
∞ ∑
n=1
(2n + 1)(|a n | 2 + |b n | 2 ) (1.7)
σ ext = 2π
k 2 Re(a n + b n ) (1.8)
where a n and b n are given by
a n = mψ n(mx)ψ ′ n(x) − ψ n (x)ψ ′ n(mx)
mψ n (mx)ξ ′ n(x) − ξ n (x)ψ ′ n(mx)
(1.9)
b n = ψ n(mx)ψ ′ n(x) − mψ n (x)ψ ′ n(mx)
ψ n (mx)ξ ′ n(x) − mξ n (x)ψ ′ n(mx)
(1.10)
in equations 1.9 and 1.10 ψ and ξ are the Ricatti-Bessel functions of order n [30],
x=kR is a size parameter (R is the radius of the particle) and m= √ ɛ p /ɛ m is the square
root of the ratio between the dielectric functions of the particle ɛ p and of the medium
ɛ m . The prime indicates a derivation to the parameter in parentheses. Equations 1.9
and 1.10 signifies that the electic and magnetic fields inside the sphere can be expressed
as a multipolar series of spherical armonics with different symmetry, identified by the
multipolar order n (n equal to 1 corresponds to dipolar sphere excitation, n equal to 2
correspond to quadrupolar oscillation and so on).
The x parameter determines if the sphere is in the quasistatic (dipolar: R

20 Metal nanoparticles
In eq. 1.11, ω p is the so-called plasma frequency and ɛ ∞ includes the contribution
of the bound electrons to the polarizability. The plasma frequency is given by
ω p = √ ne 2 /ɛ 0 m ∗ with n and m ∗ being the density and effective mass of the conduction
electrons, respectively.
However, the complex electronic structure of metals is not described very accurately by
this model, so most calculations involving metals use a dielectric function known from
measurements. The measurements by Johnson and Christy [41] are generally considered
to be most reliable and were obtained on metal films under high vacuum conditions
Figure 1.6 shows a comparison between the Drude model and the values measured
by Johnson and Christy [41]: whereas the low energy values are well described by the
Drude-Sommerfeld model, additional contributions are present at higher energies (E>2
eV). The reason for this discrepancy is related to the excitation of electrons from deeper
bands into the conduction band, the so-called interband excitations. An additional
reason for the derivation from the Drude-Sommerfeld behaviour is that the conduction
band is increasingly non-parabolic for higher energies.
Figure 1.6: (a) Real part of the dielectric function of gold; (b) imaginary part. Experimental data
(JC) from Johnson and Christy (1972) [41] compared to calculated values using the quasi-freeelectron or
Drude model with the parameters indicated in the graph. The shaded area around the experimental data
indicates the measurement uncertainty. The interband contribution (broken line) is calculated from the
difference of the experimental data to the Drude model.
The effect of the dielectric constants of the metal and the surrounding medium on
the polarizability is illustrated in the simulations in figure 1.7; the solid line shows the
calculated extinction spectrum of a spherical Au particle with a diameter of 50 nm in
water. The peak in the extinction is found at a wavelength of about 520 nm, i.e., green
light is absorbed, giving the particles their red color. The spectrum of a Ag sphere of the
same size in the same medium is also shown in Figure 1.7 (dashed line). The extinction
peak is found at a smaller wavelength (at 420 nm), a direct consequence of the different
dielectric constants for Ag than for Au.
Changing the embedding medium to glass instead of water (i.e., ɛ m from 1.45 to 1.33)
causes a peak red shift for Ag by about 25 nm (see dash-dotted line). The glass effectively

Chapter 1 21
screens the surface charges resulting in a smaller restoring force and thus to a lower
resonance frequency.
Figure 1.7: Calculated extinction spectra of Au spheres (solid line), Au rods (dotted line) and Ag spheres
in water (dashed line) or silica (dash-dotted line). The refractive index of silica is taken as 1.45, the
radius of the particles is 25 nm. The rods have aspect ratio of 2 with a length of 80 nm. Ref [42]
1.6 Mie-theory approximations
Although Mie theory gives an exact solution for any spherical particle, in some cases it
can be useful to obtain simpler formulas for the cross sections using some approximations,
as discussed below.
1.6.1 Size dependence: small particles
For nanoparticles much smaller than the wavelength of light only the dipole oscillation
contributes significantly to the extinction cross-section. In this regime, which is called
the Rayleigh limit, it is required for the size parameter x=kR that [43],[15],[37],[38],[30]
|m| x

22 Metal nanoparticles
be frequency independent, the latter is complex and is a function of the energy. The
resonance condition is fullfilled when ɛ 1 (ω)=-2ɛ m , if ɛ 2 is small or weakly dependent on
ω [37]. This resonance is independent of particle size.
The dipole approximation is also often described as the quasistatic approximation, which
assumes that the entire surface of the nanoparticle is experiencing a constant electric
field. The above equation has been used extensively to explain the absorption spectra of
small metallic nanoparticles in a qualitative as well as quantitative manner [37].
1.6.2 Size dependence: larger particles
For larger nanoparticles (greater than about 20 nm in the case of gold) where the dipole
approximation is no longer valid, the plasmon resonance depends explicitly on the particle
size as the size parameter x (eq. 1.7-1.10)is a function of the particle radius r. The larger
the particles become, the more important are the higher-order modes in eq. 1.7 as the
light can no longer polarize the nanoparticles homogeneously. These higher-order modes
peak at lower energies and therefore the plasmon band red shifts with increasing particle
size [37],[38]. At the same time, the plasmon bandwidth increases with increasing particle
size [37],[38]. This is illustrated experimentally from the spectra shown in Figure 1.8.
As the optical absorption spectra depend directly on the size of the nanoparticles, this
is regarded as an extrinsic size effects [37].
Figure 1.8: Absorption spectra for increasing radius.
1.7 Intrinsic size effect
The situation concerning the size dependence of the optical absorption spectrum is more
complex for smaller nanoparticles for which only the dipole term is important. As can

Chapter 1 23
easily be seen from equation 1.13, the extinction coefficient does not depend on the
particle dimension. However, a size dependence is observed experimentally [37],[38].
Figure 1.9 shows the absorption spectra of four different size gold nanoparticles where
λ max of the plasmon absorption redshifts with increasing particle diameter.
Figure 1.9: Size effects on the surface plasmon absorption of spherical gold nanoparticles. The UV-vis
absorption spectra of colloidal solutions of gold nanoparticles with diameters varying between 9 and 99
nm show that the absorption maximum red-shifts with increasing particle size (a), while the plasmon
bandwidth follows the behavior illustrated in (b). The bandwidth increases with decreasing nanoparticle
radius in the intrinsic size region and also with increasing radius in the extrinsic size region as predicted
by theory. In panel (c) the extinction coefficients of these gold nanoparticles at their respective plasmon
absorption maxima are plotted against their volume on a double logarithmic scale. The solid line is a
linear fit of the data: a linear dependence is observed, in agreement with the Mie theory. Ref [44]
Furthermore, it is seen that the plasmon absorption bandwidth decreases with increasing
particle size and then increases again with a minimum for the 20 nm nanoparticles
(Figure 1.9b). A modification to the Mie theory require that the dielectric function
is assumed to become size-dependent when the diameter of the NP is smaller than the
mean free path of the conduction electrons, [ɛ = ɛ(ω, r)] [45]. This means the absorption
cross section to be size-dependent within the dipole approximation and thus is regarded
as the intrinsic size effect.
Contrary to the extrinsic size effects on the extinction cross section, that are simply
deduced from electromagnetic theory, intrinsic size effects are only due to the dependence
of metal dielectric constant on the size [50 finaal]. In case of spherical AuNP and AgNP
the extrinsic size effects are perceptible for sized above 20 nm in diameter and become
important for a size ≥ 60 nm, while intrinsic ones appear in the SPA for a size smaller
than 20 nm.
In general, the optical properties of noble metals are determined by conduction elec-

24 Metal nanoparticles
trons and d-band electrons. Therefore the dielectric constant in the UV-Vis-NIR regime
is composed of two terms [37]:
ɛ ∞ (ω) = 1 + χ s (ω) + χ d (ω) (1.14)
where χ d (ω) is the contribute of d-bands electrons and χ s (ω) is that of s-band conduction
electrons and can be expressed by a simple Drude-Sommerfield model:
χ s (ω) = −
ω2 p ωP 2
ω 2 + Γ 2 + i
Γ ∞
∞ ω(ω 2 + Γ 2 ∞)
(1.15)
where Γ ∞ is the bulk metal damping frequency. According to the Matthiessen rule,
conduction electron damping frequency of bulk metals is the sum of three main independent
processes: electron-electron scattering (τ e−e ), electron-phonon scattering (τ e−p )
and electron-defects scattering (τ e−d ) [46]:
Γ ∞ = 1
τ e−e
+ 1
τ e−p
+ 1
τ e−d
(1.16)
For metal particles with nanometric size the usual scattering processes change and
new contributes appear, mainly due to surface effects. The dominant process consist
of electron scattering at particle surface, that is no more negligible when conduction
electrons mean free path (≈ 45 nm for Au and Ag) becomes comparable to particle size.
This scattering contributes overshadow other phenomena originated by changes in the
phonons spectra, quantum size effects, changes in the phonon-electrons coupling for the
high surface charge during plasmon oscillation and so on.
Usually surface scattering
produces the complete quenching of SPA in AuNP and AgNP smaller than about 2 nm
and 1,5 nm respectively [21]. The most common way to account for the surface effect
on the damping frequency Γ consists in expressing the Matthiessen formula as a size
equation [37], [44]:
Γ(r) = Γ ∞ + A v F
r
(1.17)
where v F is the velocity of the electrons at the Fermi energy and A is an empirical
adimensional parameter usually close to 1, used to account for some factors affecting the
width for the SPA in specific cases. When replacing Γ ∞ with Γ(r) in equation 1.15 two
terms can be isolated in the expression of the dielectric constant [37]:
[
ɛ(ω, r) = ɛ ∞ (ω) + ωP 2 1
(
ω 2 + Γ 2 ∞
] [
1
ω
2
−
ω 2 + Γ(r) 2 ) + i P
ω ( Γ(r)
ω 2 + Γ(r) 2 − Γ ]
∞
ω 2 + Γ 2 )
∞
(1.18)

Chapter 1 25
The first is equal to the bulk constant and only the second accounts for the size
effect on the dielectic response of nanometric object. For Au and Ag the ɛ ∞ (ω) is
experimentally known with high precision [41], hence correction of ɛ(ω, r) for the size is
straightforward.
This model gives the correct 1/r dependence [47][37] of the plasmon bandwidth as a
function of size for nanoparticles described by the dipole approximation in the intrinsic
size region (r ≺ 20 nm). The best advantages of this theory are that it provides a
very good description for the dependence of ɛ m on size and that the modification of the
dielectric constant is done in a straightforward manner.
1.7.1 Specific phenomena influencing the surface plasmon absorption
Beside surface electronic scattering, a certain number of specific not negligible factors
affect the SPA. When a single nanoparticle is polycrystalline, electron scattering at the
grain boundaries increases the dumping frequency and produces slightly larger SPA [48].
High temperatures produce a small broadening of the SPA because the electron-electron
scattering frequency is dependent on the Fermi-Dirac electron distribution as [44]:
1
τ e−e
∝ (E − E F ) 2 (1.19)
where E F is the Fermi energy and E is the electron energy. Another specific effect
influencing the SPA arises from the fact that metal nanoparticles change the overall
dielectric properties of a material when they are present in high concentration. In general,
the Mie theory is valid under strict conditions: the concentration of the nanoparticles
in a solvent or solid matrix must be very low [37] [38], the individual particles have to
be non-interacting and separated each other. Under these assumptions, the electric field
created around one particle by the excitation of a surface plasmon resonance is not felt
by the other surrounding particles. If the interparticle distances become smaller than
the particle dimension or if aggregation occurs, the plasmon resonance red shifts and
often a second absorption peak at a longer wavelength is observed [37].
The most intense and common specific mechanism influencing the surface plasmon
absorption is called Chemical Interface Damping (CID) [37] [49]. It produces a sensible
widening and red shift of the SPA when adsorbates are present on particle surface.
This holds both for chemisorption, as in case of thiols stabilized nanoparticles, and for
physisorption, as in the case of citrate or alchilamine stabilized nanoparticles. The SPA
broadening due to CID is explained considering that adsorbates offer new relaxation
pathways for both excited electrons and phonons in the metal.

26 Metal nanoparticles
1.8 Shape-dependent properties : the Gans theory
If the size and the environment effects are clearly important, shape effects on the optical
absorption spectrum of gold nanoparticles seem to be even more pronounced [50]
[51] [52]. The plasmon resonance absorption band splits into two bands as the particles
aspect ratio (i.e. the ratio between the long axis, lenght, and the minor axis, width)
increases; moreover, the energy separation between the resonance frequencies of the two
plasmon bands increases [38] [39] [30]. The high-energy absorption band at around 520
nm corresponds to the oscillation of the electrons perpendicular to the minor rod axis
and is referred to as the transverse plasmon absorption. This absorption band is relatively
insensitive to the nanorod aspect ratio [50] [51] [52] and coincides spectrally with
the surface plasmon oscillation of the spherically symmetric nanocrystals. The other
absorption band at lower energies is caused by the oscillation of the free electrons along
the major rod axis and is known as the longitudinal surface plasmon absorption.
Figure 1.10 (inset) shows the absorption spectra of two nanorod samples having aspect
ratios of 2.7 and 3.3; the longitudinal plasmon band maximum (open circles) red
shifts with increasing aspect ratio while the transverse absorption band maximum (open
squares) does not change [50] [51] [52] [53].
Figure 1.10: Size-dependent surface plasmon absorption of gold nanorods. Optical absorption spectra
of gold nanorods with mean aspect ratios of 2.7 and 3.3 are shown. The short-wavelength absorption band
is due to the oscillation of the electrons perpendicular to the major axis of the nanorod while the longwavelength
band is caused by the oscillation along the major axis. The absorption bands are referred to as
the transverse and longitudinal surface plasmon resonances respectively.The former is rather insensitive
towards the nanorod aspect ratio in contrast with the longitudinal surface plasmon band which red shifts
with increasing the aspect ratio. This is illustrated in the inset where the two band maxima (square for
the, transverse mode; circle for the longitudinal mode) are plotted vs. the nanorod aspect ratio R. [35]
The optical absorption spectrum of a collection of randomly oriented gold nanorods
with aspect ratio R can be predicted using an extension of the Mie theory. Within the

Chapter 1 27
dipole approximation according to the Gans [54] treatment, the extinction cross-section
σ ext for elongated ellipsoids is given by the following equation [38] :
σ ext =
2πNV ɛ3/2 m
3λ
∑
j
( 1 )ɛ
Pj
2 2
[ɛ 1 + 1−P (1.20)
j
P j
ɛ m ] 2 + ɛ 2
whereby the P j values are the depolarization factors that depend on the nanorod the
three axes A, B, and C as indicated in Figure 1.11. B and C correspond to the particle
diameter (d), while the A axis represents the particle length (L).
Figure 1.11: Schematic representation illustrating the optical response of rodlike nano particles to an
electric field E. Two oscillating modes can be possible: (a) the transverse oscillation along the B or C
axis and (b) the longitudinal oscillation along the A axis [55].
P j for nanorods along A, B, and C axes are respectively
P A = 1 − e2
e 2
[ 1 2e ln(1 + e ) − 1] (1.21)
1 − e
with
e =
P B = P C = 1 − P A
2
√
1 − ( d L )2 =
√
(1.22)
1 − 1 R 2 (1.23)
where R = L/d is the aspect ratio. Link and coworkers [52] [56] calculated the
absorption spectra of gold nanorods using Equation 1.20 for various aspect ratios starting
from the measured dielectric functions of gold [57]. When the medium dielectric constant
was chosen to be 4, the maximum of the longitudinal plasmon band red-shifts 150 nm
for the aspect ratio increasing from 2.6 to 3.6 (Figure 1.12).
Moreover, Figure 1.12 shows that the increase in the peak position of the longitudinal
plasmon band with increasing nanorod aspect ratio follows a linear trend. The data
shown here correspond to the absorption maxima of the longitudinal plasmon band as

28 Metal nanoparticles
Figure 1.12: Simulation of the surface plasmon absorption for gold nanorods of different aspect ratio
by Link and coworkers. [52].
calculated in Figure 1.12b. The dotted line represents a plot of an equation derived by
Link in order to predict the maximum of the longitudinal plasmon band λ max [52]
λ max = (33.34R − 46.31)ɛ m + 472.31 (1.24)
It follows from Equation 1.24 that the maximum of the longitudinal plasmon resonance
also linearly depends on the medium dielectric constant ɛ m . Hence, with increasing
the medium dielectric constant, there is also a red-shift of λ max for a fixed aspect ratio
(Figure 1.12 c with R=3.3).
The fact that gold nanorods exhibit a longitudinal SPR in the near-infrared has
drawn interest of the biophysical community in the use of this particles in in-vivo study
because of larger penetration depth of the laser light: in the region between 650-900 nm,
≥ 300 µm depending on tissue types.
Solid gold nanorods have several advantages over other contrast agents; their synthesis
with various aspect ratios and tunable absorption wavelength in the near infrared region
[58] [59] [8] is quite simple and well-established. The typical size of the nanorods (50-200
nm) is potentially useful in applications such as drug delivery and gene therapy. In addition,
the biosafety of gold is well known and it has been used in vivo since the 1950’s [60]
and recently the low cytotoxicity of gold nanoparticles in human cells has been studied
in detail by Wyatt and coworkers [61].
Moreover, gold nanorods can be used as photo-absorbers in NIR and realize photothermal
therapy. Thus, upon laser irradiation at the surface plasmon absorption band, the
nanoparticles absorb energy and then immediately transfer it into heat. If the nanoparticles
are incorporated or incubated with biomolecules, cells or tissues, heat causes the
sharp increase on the local temperature around the nanoparticles and thus damages to
the surrounding materials. This process is called photothermal destruction and can be

Chapter 1 29
used for disease or cancer therapy, providing a novel class of photo-absorber in medical
applications. In fact, compared to the strongly absorbing Rhodimine 6G (ɛ =
1.16·10 5 M −1 cm −1 at 530 nm [62]) dye molecules, gold nanoparticles absorb about 10 3
stronger (for nanospheres of 40 nm in diameter, ɛ = 2.74·10 9 M −1 cm −1 at 530 nm, [58]).

30 Metal nanoparticles

Bibliography
[1] Nanotoxicology: nanotechnology grows up. Science, 304:1732–1734, 2004.
[2] C.N.R. Rao and A.K. Cheetham. Science and technology of nanomaterials: current
status and future prospects. J. Mater. Chem., 11:2887–2894, 2001.
[3] P. Alivisatos. The use of nanocrystals in biological detection. Nat. Biotechnol.,
22:47, 2004.
[4] J.G. Fujimoto. Optical coherence tomography for ultrahigh resolution in vivo imaging.
Nat. Biotechnol., 21:1361, 2003.
[5] J. Chen; F. Saeki; B.J. Wiley; H. Chang; M.J. Cobb; Z.Y. Li; L. Au; H. Zhang; M.B.
Kimmey; X. Li and Y. Xia. Gold nanocages: bioconjugation and their potential use
as optical imaging contrast agents. Nano Lett., 5:473, 2005.
[6] R. Weissleder. A clearer vision for in vivo imaging. Nat. Biotechnol., 19:316, 2001.
[7] R. Jin; Y. Cao; C.A. Mirkin; K.L. Kelly; G.C. Schatz and J. G. Zheng. Photoinduced
conversion of silver nanospheres to nanoprisms. Science., 294:1901, 2001.
[8] Kelly KL; Coronado E; Zhao LL; Schatz GC. The optical properties of metal
nanoparticles: the influence of size, shape, and dielectric environment. J Phys
Chem B, 107:668–677, 2003.
[9] L.B. Hunt. The True Story of Purple of Cassius. Gold Bull., 9(4):134–139, 1976.
[10] F.E. Wagner; S. Haslbeck; L. Stievano; S. Calogero; Q.A. Pankhurst and K.P. Martinek.
Before striking gold in gold-ruby glass. Nature, 407:691–692, 2000.
[11] D.J. Barber and I.C. Freestone. Archaeometry, 32:33–45, 1990.
[12] H.M. Leicester. The Historical Background of Chemistry. Dover Publications, Inc.,
1971.
[13] G.B. Kauffman. The Role of Gold In Alchemy. Part II. Gold Bull., 18(2):69–78,
1985.
[14] M. Faraday. Experimental relations of gold (and other metals) to light. Philos.
Trans. R. Soc., pages 145–181, 1857.
[15] P. Mulvaney. Surface Plasmon Spectroscopy of Nanosized Metal Particles. Langmuir.,
12:788–800, 1996.

32 Bibliography
[16] R.A. Zsigmondy. Properties of Colloids. Nobel Lecture., 11, 1926.
[17] J. Turkevich. Colloidal Gold Part I: Historical and Preparative Aspects, Morphology
and Structure. Gold Bull., 18(3):86–91, 1985.
[18] G. Frens. Controlled Nucleation for the Regulation of the Particle Size in Monodisperse
Gold Suspensions. Nature Phys. Sci., 241:19–22, 1973.
[19] M. Brust; M. Walker; D. Bethell; D. J. Schiffrin and R. Whyman. Synthesis of Thiolderivatised
Gold Nanoparticles in a Two-phase Liquid-Liquid System. J. Chem.
Soc., pages 801–802, 1994.
[20] Greenwood N.N.; Earnshaw A. Chemistry of the Elements. Elsevier Science: Oxford,
1997.
[21] Daniel ; Astruc. Chemical Review, 104:293–346, 2004.
[22] Zheng J.; Nicovich P.R.; Dickson R.M. Annual Review of Physical Chemistry,
58:409–431, 2007.
[23] Kvitek L.; Prucek R. Journal of Material Science, 2005.
[24] Lakowicz JR. Radiative decay engineering: biophysical and biomedical applications.
Anal Biochem., 298:1–24, 2001.
[25] Lakowicz JR; Shen Y.; DAuria S.; Malicka J.; Fang J.; Gryczynski Z.; Gryczynski I.
Radiative decay engineering 2. Effects of silver island films on fluorescence intensity,
lifetimes, and resonance energy transfer. Anal Biochem., 301:261–277, 2002.
[26] Geddes C.D.; Aslan K.; Gryczynski I.; Malicka J.; Lakowicz JR. Radiative decay
engineering. In: Geddes CD, Lakowicz JR (eds) Topics in Fluorescent Spectroscopy
vol 8: Radiative Decay Engineering. Springer Science-Business Media, Inc, New
York, page 405448, 2005.
[27] Geddes C.D.; Aslan K.; Gryczynski I.; Malicka J.; Lakowicz JR. Noble-metal surfaces
for metal enhanced fluorescence. In: Geddes CD, Lakowicz JR (eds) Topics
in Fluorescent Spectroscopy vol 8: Radiative Decay Engineering. Springer Science-
Business Media, Inc, New York, page 365401, 2005.
[28] Lakowicz JR. Radiative decay engineering 5: metal-enhanced fluorescence and plasmon
emission. Anal Biochem, 337:171194, 2005.
[29] Prasad P.N. Nanophotonics. Wiley, 2004.

Chapter 1 33
[30] Bohren CF; Huffman DR. Absorption and Scattering of Light by Small Particles.
John Wiley and Sons, Inc., New York,, page 530, 1983.
[31] Link S. and El-Sayed M.A. J. phys. Chem. B, 103:4212, 1999.
[32] Turkevich J.; Stevenson P.C.; Hillier J. Discuss. Faraday Soc., 11:55, 1951.
[33] Turkevich J.; Garton G. and Stevenson P. C. J. Colloid Sci., Suppl., 1:26, 1954.
[34] Link S. and El-Sayed M.A. J. phys. Chem. B, 103:8410, 1999.
[35] Link S.; El-Sayed M.A. Shape and size dependence of radiative, non-radiative and
photothermal properties of gold nanocrystals. Int. Reviews in Physical Chemistry,
19 (3):409–453, 2000.
[36] Mie G. Articles on the optical characteristics of turbid tubes, especially colloidal
metal solutions. Annalen Der Physik., 25 (3):377–445, 1908.
[37] Kreibig U. and Vollmer M. Optical Properties of Metal Clusters. Berlin : Springer,
1995.
[38] Papavassiliou G.C. Prog. solid st. Chem., 12:185, 1980.
[39] Kerker M. The Scattering of Light and Other Electromagnetic Radiation (New York:
Academic Press), 1969.
[40] Ashcroft N.W. and Mermin N.D. Solid State Physics. Philadelphia, Pennsylvania:
Saunders College, 1976.
[41] P.B. Johnson and R.W. Christy. Optical-Constants Of Noble-Metals. Phys. Rev.
B, 6(12):4370–4379, 1972.
[42] Johanna Jacoba Penninkhof. Tunable plasmon resonances in anisotropic metal
nanostructures. PhD Thesis, 2006.
[43] Henglein A. J. phys. Chem., 97:8457, 1993.
[44] S. Link and M.A. El-Sayed. Simulation of the Optical Absorption Spectra of Gold
Nanorods as a Function of Their Aspect Ratio and the Effect of the Medium Dielectric
Constant. J. Phys. Chem. B, 103:3073–3077, 1999.
[45] U. Kreibig and C.V. Fragstein. The Limitation of Electron Mean Free Path in Small
Silver Particles. Z. Phys., 224:307323, 1969.
[46] Kittel C. Introduction to solid state physics. Wiley, 1996.

34 Bibliography
[47] Doremus R.H. J. chem. Phys., 40:2389, 1964.
[48] Tang Y; Ouyang M. Nature Materials, 6:757–759, 2007.
[49] Llnaert T.; Mulvaney P.; Henglein A. J.Phys.Chem., 97:679–682, 1993.
[50] Yu Y.; Chang S.; Lee C. and Wang C.R.C. J. phys. Chem. B, 101 (34):6661, 2005.
[51] Van der Zande B.M.I.; Bohmer M.R.; Fokkink L.G.J. and Schonenberger C. J. phys.
Chem. B, 101:852, 1997.
[52] Mohamed M.B.; Ismael K.Z.; Link S. and El-Sayed M.A. J. phys. Chem. B, 101:852,
1997.
[53] Link S.; Mohamed M.B. and El-Sayed M.A. J. phys. Chem. B, 103:3073, 1999.
[54] Gans R. Annln Phys., 47:270, 1915.
[55] Foss C.A.; Hornyak G.L.; Stockert J.A. and Martin C.R. J. Phys. Chem, 98:2963,
1994.
[56] Lisiecki I.; Billoudet F. and Pileni M.P. J. Phys. Chem, 100:4160, 1996.
[57] Lisiecki I. and Pileni M.P. J. Am. Chem. Soc., 115:3887, 1993.
[58] Link S.; El-Sayed M.A. J. Phys. Chem. B, 103:8410, 1999.
[59] Murphy C.J.; Sau T.K.; Gole A.M.; Orendorff C.J.; Gao J.; Gou L.; Hunyadi S.E.; Li
T. J. Phys. Chem. B, 109 (29):13857, 2005.
[60] Sherman A.I.; Ter-Pogossian M. Cancer, 6:1238, 1953.
[61] Connor E.E.; Mwamuka J.; Gole A.; Murphy C.J.; Wyatt M.D. Small, 1:325, 2005.
[62] Hao E.; Schatz G.C.; Hupp J.T. Fluorescence, 14:331, 2004.

Chapter 2
Principles
2.1 Basics of Fluorescence Optical Spectroscopy
Luminescence is the emission of light from any substances and occurs from electronically
excited states [1]. It is divided in two categories: fluorescence and phosphorescence,
depending on the excited state nature. Singlet excited states give rise to fluorescence;
triplet states to phosphorescence. In the singlet state S 1 the electron e − in the excited
orbital is paired (of opposite spin) to the second e − in the ground state, S0, therefore
the transition to S0 (S 1 → S 0 ) is allowed and occurs rapidly: 10 −8 s ; so the typical
fluorescence lifetime is 1-10 ns (for lifetime one means the average time a molecule
spends in the excited state before coming back to the ground state). Phosphorescence is
generated by transition from the triplet excited state T 1 in which the e − has the same
spin orientation of the e − in S 0 : than the transition T 1 → S 0 is forbidden and its rate
is slow compared with that of S 1 → S 0 : 10 −3 - 10 0 s so the phosphorescence lifetime is
from millisecond to second, significantly larger than that of fluorescence; this makes these
two emission processes simply separable from a spectroscopic point of view [1]. Usually
fluorescence comes from aromatic molecules: among the most used are the coumarins,
the fluoresceines and the rhodamines. The first observation of fluorescence goes back
to 1845 by Sir John Frederick William Herschel [2]. Fluorescence data are presented
as emission spectra, i.e. a plot of the intensity vs. wavelength. This characteristic
depends upon the chemical structure of the fluorophore and shows the structure due
to the individual vibrational energy levels of the ground and the excited state. The
absorption end emission of light can be described in terms of the Jablonsky diagram [3]
(Figure 2.1).
35

36 Principles
Figure 2.1: Jablonski diagram.
The ground state, the first and second exited singlet and the first excited triplet states
are indicated respectively as S 0 , S 1 , S 2 , and T 1 ; for each state there exist several vibronic
levels (0, 1, 2, and so on). At room temperature the energy gap between S 0 and S 1 is too
large for thermally population of S 1 , so a light source is necessary to induce fluorescence.
After the absorption of light (hν A ), that occurs in 10 −16 s (i.e. instantaneously), several
processes take place:
• Internal conversion (IC): a fluorophore is excited to a higher vibrational level of
either S 1 or S 2 ; after that the molecule relaxes rapidly to the lower vibrational level
of S 1 : this process takes place in 10 −12 s.
• Intersystem crossing (ICS): the fluorophore undergoes a spin conversion to T 1 ; the
transitions T 1 → S 0 are forbidden and have the low rates typical of phosphorescence.
However an alternative route to de-excitation of T 1 is possible. In times of
the order of few microseconds, the triplet state T 1 may convert back to an excited
vibrational substate of S 1 , from which the molecule decays in few picoseconds to the
ground vibrational state of S 1 . The decay from S 1 to an excited vibrational state
of S 0 occurs instead in the nanosecond time range and is followed by a picosecond
transition to the ground vibrational state of S 0 [1].
• Quenching: deexcitation is a consequence of interactions with solute molecules,
especially with O 2 .
• Energy transfer: dipole-dipole interactions between molecules can occur if their
distance is less than few nm, causing energy transfer between fluorophores.

Chapter 2 37
A direct consequence of the Jablonski diagram (figure 2.1) is that the emission spectrum
is the mirror image of the excitation one.
Other two features of fluorescence
emission spectra deserve consideration: the Stokes shift [4] and the Kasha’s rule [5].
Looking at the Jablonsky diagram in figure 2.1, it is clear that between absorption and
fluorescence emission there is a loss of energy due to the relaxation to the lower vibrational
level of S 1 by internal conversion; moreover the transition S 1 → S 0 involves higher
vibrational levels of S 0 that rapidly relax to the ground state resulting in a further loss of
energy; as a result the emission of light takes place at lower energy (higher wavelengths)
with respect to excitation. This phenomenon is called Stokes shift and was first observed
by Sir George Gabriel Stokes in 1852 [4]. Moreover, the emission spectra are independent
from excitation wavelength (Kasha’s rule) [5]; this is because after absorption the
fluorophores rapidly relax to the lowest vibrational state of S 1 from which any transition
to S 0 starts. The relaxation is probably the result of the overlapping among numerous
states of nearly the same energy.
2.1.1 Fluorescence lifetimes and quantum yields
The fluorescence lifetime τ and quantum yield φ are perhaps the most important characteristics
of a fluorophore. The fluorescence quantum yield is the ratio of the number of
photons emitted to the number absorbed: substances with large φ, such as rhodamines,
display also the bright emission.
The lifetime determines the time available for the
fluorophore excites state to interact with its environment, and hence the amount of information
that would be available from its emission. The meaning of τ and φ is best
represented by a simplified Jablonski diagram (Figure 2.2), where the attention is focused
on those processes responsible for the relaxation to the ground state. In particular, we
are interested in the radiative (Γ) and nonradiative decay rate to S 0 (k nr ). The processes
governed by Γ and k nr both depopulate the excited state. The fraction of fluorophores
which decay through emission, and hence the quantum yield, is given by:
Γ
Q =
(2.1)
Γ + k nr
Q is always < 1 due to the Stokes losses of energy and becomes close to the unity
when k nr

38 Principles
the time spent in S 1 : for a single experimental decay the 63% relax at t < τ and 37% at
t > τ.
Figure 2.2: A simplified Jablonski diagram.
An opportunity to control the radiative rates arises from the interactions of fluorophores
with nearby metallic surfaces or particles. Nearby metal surfaces can respond
to the fluorophore oscillating dipole and modify the rate of emission and the spatial distribution
of the radiated energy. The electric field felt by a fluorophore is affected by the
interactions of the incident light with the nearby metal surface and also by interaction
of the fluorophore oscillating dipole with the metal surface (excitation enhancement).
Additionally, the fluorophores oscillating dipole induces a field in the metal (emission
enhancement). These interactions can increase or decrease, depending on distance, the
field incident on the fluorophore, its radiative decay rate and the spatial distribution of
the radiated energy.
2.2 Radiative decay engineering
We refer to the interactions of fluorophores with novel metal particles as radiative decay
engineering (RDE) because the radiative decay rates of the fluorophores are modified by
placing them in close proximity to the metal [6] [7] [8].
Prior to describing the unusual effects of metal surfaces on fluorescence it is valuable to
describe what we mean by Radiative Decay Engineering (RDE) and in particular spectral
changes expected for increased radiative decay rates. Spectroscopists are accustomed to
perform experiments using 5 to 10 A ◦ -sized fluorophores in macroscopic solutions, typically
transparent to the emitted radiation. There may be modest changes in refractive
index, such as for a fluorophore in a membrane, but such changes have only a minor
effect on the fluorescence spectral properties. In such nearly homogeneous solution, the
fluorophores emit into free space and are observed in the far field. The spectral properties
are well described by Maxwells equations for an oscillating dipole radiating into free
space. However, the interactions of electromagnetic radiation with physical objects can

Chapter 2 39
be considerably more complex. Like a radiating antenna, a fluorophore is an oscillating
dipole, whose near-field is characterized also by high spatial frequencies. The plasmonic
resonances of nearby metal surfaces can couple to these components of the oscillating
dipole and modify the rate of emission and the spatial distribution of the radiated energy.
Therefore the electric field felt by a fluorophore is affected by interactions of the incident
light with the nearby metal surface and also by interaction of the fluorophore oscillating
dipole with the metal surface. In other words, the fluorophore-oscillating dipole induces
a field in the metal. These interactions can increase or decrease the field incident on the
fluorophore and the radiative decay rate. Depending upon the distance and the geometry,
metal surfaces or particles can result in quenching or enhancement of fluorescence
by factors of up to 1000 [9] [10] [11]. The effects of metallic surfaces on fluorophores are
due to at least three mechanisms (figure 2.3).
Figure 2.3: Effects of a metallic particle on transitions of a fluorophore. Metallic particles can cause
quenching (k nr), can concentrate the incident light field (E m), and can increase the radiative decay rate
(Γ m). Ref.[12]
One is energy transfer quenching to these metals with a d 3 dependence (where d
is the distance between the metal surface and the fluorophore) [10]. The quenching
can be understood by damping of the dipole oscillators by the nearby metal. A second
mechanism is an increase in emission intensity due to the extinction field amplification
related to the so called ’tip effect’ due to the surface roughness of the metal. However,
another more important effect of the interaction between the metal surfaces and particles
is possible. The nearby metal can increase the intrinsic radiative decay rate Γ of the
fluorophore [12].
Assume that the presence of a nearby metal (m) surface increases the radiative rate

40 Principles
by an additional new term Γ m (Fig. 2.4). In this case the quantum yield and lifetime of
the fluorophore near the metal surface are given by
Q m =
Γ + Γ m
Γ + Γ m + k nr
(2.3)
τ m = (Γ + Γ m + k nr ) −1 (2.4)
Figure 2.4: Modified Jablonski diagrams which include metalfluorophore interactions.
arrows represent increased rates of excitation and emission. Ref. [12]
The thicker
These equations result in unusual predictions for the fluorophore emission near a
metal surface. As the value of Γ m increases, the quantum yield increases while the lifetime
decreases. We may observe then a variety of favorable effects due to metal particles, such
as increased fluorescence intensities, increased photostability 1 , and increased distances
for FRET. We refer to these favorable effects as metal-enhanced fluorescence (MEF).
The most dramatic relative changes are found for fluorophores with the lowest quantum
yields. If Q = 1.0, then changing Γ m has no effect. If Q is low, such as 0.1 in Fig. 2.5,
the metal-induced rate Γ m increases the quantum yield. At sufficiently high values of
Γ m , the quantum yields of all fluorophores approach 1.0. Examination of Fig.2.5 reveals
that larger values of Γ m /Γ are required to change the lifetime or quantum yield of low
quantum yield fluorophores. This effect occurs because, for the same unquenched lifetime
τ 0 , lower quantum yields imply larger values of k nr . Larger values of Γ m are required to
compete with the larger values of k nr .
1 less time at the excited state and as consequence lower bleaching

Chapter 2 41
Figure 2.5: Effect of an increase in the metal-induced radiative rate on the lifetime and quantum yields
of fluorophores. To clarify this figure we note that for Q=0.5, Γ=5 10 7 /s and k nr=5 10 7 /s. For Q=0.1,
Γ=1·10 7 /s and k nr= 9·10 7 /s. Ref.[12]
2.3 Review of metallic surface effects on fluorescence
The possibility of altering the radiative decay rates, through the change of the excitation
field phase, was demonstrated by measurements of the decay times of europium (Eu 3+ )
complex positioned at various distances from a planar silver mirror [13] [14] [15] [16]. The
lifetime oscillates with distance but the decay remains a single exponential independent
of the distance (Fig. 2.6). This effect can be explained by changes in the phase of the
reflected field with distance and the effects of this reflected field on the fluorophore. A
decrease in lifetime is found when the reflected field is in-phase with the fluorophores
oscillating dipole. An increase in the lifetime is found if the reflected field is out-of-phase
with the oscillating dipole. As the distance increases, the amplitude of the oscillations
decreases. The effects of a plane mirror occur over distances comparable to the excitation
and emission wavelengths. At short distances below 20 nm the emission is quenched.
This effect is due to coupling of the dipole to oscillating surface plasmons.
2.3.1 Theory for Metallic ParticlesFluorophore Interactions
Several groups have considered the effects of metallic spheroids on the spectral properties
of nearby fluorophores [17] [18] [19] [20] [21]. A typical model is shown in Fig. 2.7,
for a prolate spheroid with an aspect ratio of a/b. The particle is assumed to be a
metallic ellipsoid with a fluorophore positioned near it. The fluorophore is located outside
the particle at a distance r from the center of the spheroid and a distance d from the
surface, and can be oriented parallel or perpendicular to the metallic surface. The
presence of a metallic particle can have dramatic effects on the radiative decay rate of a
nearby fluorophore. Figure 2.7 shows the radiative rates expected for a fluorophore at

42 Principles
various distances from the surface of a silver particle and for different orientations of the
fluorophore transition moment. The most remarkable effect is found for a fluorophore
perpendicular to the surface of a spheroid with a/b = 1.75. In this case the radiative
rate can be enhanced by a factor of 1000-fold or greater. The effect is much smaller
for a sphere (a/b = 1.0), and even less for a more elongated spheroid (a/b = 3.0) when
the optical transition is not in resonance with the particle. In this case the radiative
decay rate can be decreased by over 100-fold. If the fluorophore displays a high quantum
yield or a small value of k nr , this effect could result in 100-fold longer lifetimes. The
magnitude of these effects depends on the location of the fluorophore around the particle
and the orientation of its dipole moment relative to the metallic surface. The dominant
effect of the perpendicular orientation is thought to be due to an enhancement of the
local field along the long axis of the particle.
Figure 2.6: Lifetime of Eu 3+ ions in front of a Ag mirror as a function of separation between the Eu 3+
ions and the mirror. The solid curve is a theoretical fit.Ref.[12]
Figure 2.7: (A)Fluorophore near a metallic spheroid. (B)Effect of a metallic spheroid on the radiative
decay rate of a fluorophore. Ref.[12]

Chapter 2 43
2.4 Two-photon excited fluorescence
Two-photon excitation was suggested for the first time by Maria Goppert-Mayer in 1931
[22]. Using the perturbation theory, she solved the quantum mechanical equations for
light absorption and emission extending the quantum mechanical treatment to electronic
processes that involve two annihilation (or two creation) operators: the two-photon
absorption and emission processes. The transition probability of a two-photon electronic
process was derived using second order, time-dependent, perturbation theory. The Maria
Goppert-Mayer derivation proves that the probability of a two-photon absorption process
is quadratically related to the excitation light intensity, and that it involves an interaction
with a meta-stable intermediate state. This interaction occurs within the lifetime of this
virtual state which is described as a non-linear superposition of all the vibro-electronic
states. Both photons interact together to induce the transition from the ground state to
the excited state (figure 2.8).
Figure 2.8: Jablonski diagram for a two-photon absorption process, compared with one-photon absorption.
Since this event depends on the fact that the two photons both interact with the
molecule simultaneously (within 10 −16 s), resulting in a quadratic dependence on the
light intensity, rather than the linear dependence of conventional one-photon excitation,
two photon excitation is often termed non-linear. The intensity-squared dependence
guarantees the localized nature of two-photon excitation [23].
If the excited molecule is fluorescent, it can emit a single photon of fluorescence as if it
were excited by a single higher energy photon: the fluorescence emission is independent
of the excitation modality. The transition to an excited state is possible only if the
energy of the exciting photon, hc=λ 1 , is equal to the energy gap, ∆E, between the two
states involved in the transition:

44 Principles
∆E = hc
λ 1
(2.5)
Since in TPE two photons are needed to induce the transition, they can combine
and give rise to the absorption process only if they interact with the same molecule at
the same time. The timescale for this process, given by the Heisemberg principle, is
10 −15 -10 −16 seconds [24] for photon energies in the visible-IR region. It is not necessary
for the wavelengths of the two photons to be the same. They have just to satisfy the
following relation:
λ 1 ≈
( 1
λ A
+ 1
λ B
) −1
(2.6)
where λ 1 is the wavelength needed for OPE and λ A and λ B are the wavelengths of
the photons involved in TPE. Usually, for practical reasons, the choice is to have λ A =
λ B = 2λ 1 [25] (see Figure 2.8).
The application of multi-photon excitation to optical sectioning is based on the superlinear
dependence of the molecular excitation rate on the laser irradiance. The probability
for a molecule to absorb two or more photons depends on the probability, p n , of finding
two or more photons simultaneously in the volume occupied by the molecule. Let us
suppose to have a cubic volume of side l completely filled by a laser beam of mean irradiance
I and wavelength λ. If m is the mean number of photons in l 3 , the mean energy
in the cubic volume is given by:
〈E m 〉 = m hc
λ
(2.7)
since the cross sectional area is l 2 and the time required to a photon to pass through
l 3 is l/c, the relation between the average number of photons m and the mean irradiance
〈I〉 is:
〈I〉 = 〈E m〉
l 2 l/c = mhc2
λl 3 (2.8)
Since the interaction of the light with the molecule occurs on a volume of the order
of the molecular volume which is related to the molar volume V M and the Avogadro’s
number, N A , as V M /N A , we can write the relation between the laser irradiance and the
number of photons per molecular volume as:
or
I = mhc2 N A
λV M
(2.9)

Chapter 2 45
m = λV MI
hc 2 N A
(2.10)
This means that there exists a direct relation between I and m. If λ = 800 nm, I
≈ GW/cm 2 and V M ≈ 10 −3 m 3 / Mole we can estimate m≈ 10 −5 .
A Poissonian distribution can then be used to calculate the probability to have a number
n of photons per molecular volume at the same time, p m (n):
p m (n) = mn
n! e−m (2.11)
Due to the low value of m the exponential function can be expanded in a Taylor
series. For TPE, n = 2, and we compute:
p 2 = 1 2 m2 ∝ ΓI 2 (2.12)
where Γ is a constant. This equation shows clearly the quadratic dependence of
the excitation probability of a TPE process on the laser irradiance. The fluorescence
emission (under TPE) depends quadratically on the excitation probability and linearly
on the molecular (two-photon) cross section and is given by:
and thus
[ ] 2
I f (t) ≈ δ 2 I(t) 2 ≈ δ 2 P (t) 2 π (N.A.)2
(2.13)
hcλ
〈I f (t)〉 = 1 T
∫ T
0
[ ] 2
I f (t)dt = δ 2 P (t) 2 π (N.A.)2 1
hcλ T
∫ T
0
P (t) 2 dt (2.14)
where T is an arbitrary time interval for continuous wave (CW) excitation and T =
1/f p for a pulsed laser and P(t) is the time shape of the pulse. For an IR square pulsed
laser source with 〈P 〉 = τ p f p P (t) = τ p f p P peak we have:
〈I f (t)〉 = δ 2
< P > 2
τ 2 p f 2 p
] 2 [π (N.A.)2 1
hcλ T
∫ τp
0
< P > 2 [ ] 2
dt = δ 2 π (N.A.)2 (2.15)
τ p f p hcλ
while for a CW laser:
[ ] 2
〈I f (t)〉 = δ 2 < P > 2 π (N.A.)2
(2.16)
hcλ
A comparison between equations 4.1 and 4.2 shows that, in order to operate at an
excitation efficiency similar to that of a pulsed IR laser, a CW laser should operate at
a power larger by a factor (τ p f p ) −1/2 . Thus an output of 10 W for a CW illumination

46 Principles
would correspond to 30 mW for a pulsed excitation with f p ≈ 100 MHz and τ p ≈ 100
fs [26].
However this is not practically the case since the TPE cross-sections for fs
pulsed excitation and for CW or picosecond, ps, excitation are not the same 2 , and
practically CW TPE has been reported also on biological samples [27]. We can compute
the fluorescence rate by considering that the number of photon couples that a fluorophore
absorbs per laser pulse is [28]:
n a ∝ δ 2
< P > 2
τ p f 2 p
[ ] (N.A.)
2 2
(2.17)
2ηcλ
where η = h/2π. The fluorescence rate is then theoretically given by 4.3 times the
laser repetition rate. In order to obtain an optimal fluorescence rate and to keep at a
minimum the ground state saturation, a laser repetition f p ≈ 100 MHz (i.e. ≈ 10 ns
between two consecutive pulses) has to be used since the typical excited state life-time
is τ lf
∼ = 1 ns. The ground state saturation occurs for the condition, < If > τ lf
∼ = 1.
Moreover we must have τ p ≈ 10 −13 s or τ p

Chapter 2 47
P SF OP E (u, v) = |h(u, v)| 2 = h(u, v)h ∗ (u, v) (2.20)
Figure 2.9: Axial confinement of two-photon excitation PSF
One-photon excitation does not provide optical sectioning capabilities since the total
power released per cross-section of the laser beam along the optical axis is constant as
can be verified from 2.20:
∫
|h(u, v)| 2 dv = constant (2.21)
cross−section
The optical sectioning is introduced by spatial filtering the emission in front of the
detector. On the other hand, TPE is axially confined [30] since its intensity PSF is the
square of 2.20:
P SF T P E (u, v) = |P SF OP E (u, v)| 2 (2.22)
Therefore at fixed u, the integral over v assumes a half bell shape [31](see figure 2.9).
More precisely the excitation rate (proportional to PSF T P E ) falls off as the fourth
power of the distance from the focal plane [26].
The TPE excited volume can be calculated by approximating the PSF T P E as a threedimensional
Gaussian volume. This is not a volume with distinct walls but rather the
volume over which most of the TPE laser excitation is released. Integrating over all
space the three-dimensional Gaussian yields:
where ω xy and ω z are calculated as:
V G3D
T P E = π 3/2 ω 2 xyω z (2.23)

48 Principles
ω xy = 0.320λ √
2NA
ifNA ≤ 0.7
ω xy = 0.325λ √
2NA 0.91
ifNA ≻ 0.7
ω z = 0.532λ [
]
1
√
2 n − √ n 2 − NA 2
(2.24)
(2.25)
On the other hand, by integrating the Gaussian Lorentzian beam we obtain V G−L
T P E
= πω 4 0 /λ ∼ = 0.113µm 3 for a 1.2 NA lens at λ = 900nm [32].
2.4.2 OPE versus TPE
For optical sectioning, i.e. for microscopy applications, the most important consequence
of two-photon excitation, TPE, or non-linear excitation in general, is the fact that the
molecular excitation is limited to a sub-femtoliter region around the focal plane. While
in one-photon excitation, OPE, the excitation volume is not confined in the focal plane
(see Figure 2.10).
Figure 2.10: Excited volumes by two photon laser source of λ= 760nm (upper) and by one photon laser
source λ= 380nm (lower) in fluorescein solution.
Under OPE the optical sectioning is obtained by spatial filtering the emission in front
of the detector in the so called confocal detection scheme. The rejection of out of focus
plane is achieved by placing a screen with a pinhole in front of the detector. The focal
point of the objective lens forms an image onto the pinhole screen: the specimen plane
and the pinhole screen are conjugated planes (and hence the name confocal). The ability
of a confocal microscope to create sharp optical sections makes it possible to build 3D

Chapter 2 49
renditions of the specimen.
The major drawback of using a lamp source when creating a point image onto the specimen
is that only a minor fraction of the emitted photons are actually contributing to
the image. Thus, to reduce the noise on the image, each pixel must be illuminated for
a long time. In turn, this increases the length of time needed to create a point-by-point
image and the photodamage. The solution is to use a light source of very high intensity.
The modern choice is a laser light source, which has the additional benefit of being
available in a wide range of wavelengths [33]. Unfortunately, fluorescence emission has
not an infinite duration: molecules can undergo photobleaching. The phenomenon of
photobleaching occurs when a fluorophore permanently looses the ability to fluoresce
due to photon-induced chemical damage and covalent modification. Upon transition
from an excited singlet state to the excited triplet state, fluorophores may interact with
another molecule or with the light to produce irreversible chemical modifications. The
triplet state is relatively long-lived with respect to the singlet state, thus allowing excited
molecules more time to undergo chemical reactions with reactive components in the environment.
This is the reason why reducing S 1 -T transitions, inhibits photobleaching.
The average number of excitation and emission cycles that occur for a particular fluorophore
before photobleaching is dependent upon the molecular structure and the local
environment. The excitation power released per plane is independent of the focusing
plane position along the optical axis. Thus, the major drawback of OPE confocal microscopy
is the photobleaching of molecules in out of focus planes: when the focal plane
moves to their plane they are no longer fluorescent. Since TPE requires the molecule to
interact simultaneously with two photons (within 10 −16 s), a high flux of photons (10 24
photons cm −2 s −1 [26]) is needed at the focal plane. Only in the ‘90s, the development
of mode-locked laser sources, that provide high peak power femtosecond pulses with a
repetition rate of ∼ = 100MHz and the introduction of high efficiency detectors (cooled
photo-multiplier tubes, or single photon avalanche diodes), allowed laser scanning microscopy
to take full advantage of TPE.
Non-linear excitation, in fact, offers a series of unique features that make TPE very
suitable for many applications. First, the two-photon absorption bands of the dyes commonly
used in biological studies are wider than their one-photon analogous, allowing
the simultaneous excitation of multiple fluorophores with a single excitation wavelength
[23]. Second, the TPE excitation light beam has a high penetration depth in thick specimens
because IR radiation is scattered to a less extent than visible light by the tissues
[26]. The scattering is the deflection of the light rays away from their original direction.
Its angular distribution depends on the refractive index inhomogeneities, objects size
and beam wavelength λ. For small objects in homogeneous media the scattering angu-

50 Principles
lar distribution is described by Rayleigh’s law: it is isotropic and strongly λ-dependent
∝ λ −4 (for dipoles) ∝ λ −2 (for larger scatterer as in biological tissues). Third, excitation
takes place only at the focal plane, due to the scaling of the probability of simultaneous
photon absorption with the square of the light intensity. As a consequence TPE avoids
the simultaneous absorption of photons outside the specimen drastically reducing both
photo-toxicity and fluorophore bleaching [30]. Moreover the fact that excitation, and
therefore emission, take place only in a tiny well defined volume, makes unnecessary to
place a pin-hole aperture in front of the detectors for the purpose of rejecting the signal
arising from out of focus region. This latter fact simplifes the optical set-up and reflects
also in a remarkable reduction of the background noise [26].

Bibliography
[1] Lakowicz J.R. In principles of fuorescence spectroscopy. Kluwer Academic Plenum
Press, New York., 1999.
[2] Herschel Sir J.F.W. Uber den mechanisms des photolumineszenz von farbstoffphosphoren.
Phil. Trans. R. Soc. London, 135:143–145, 1845.
[3] Jablonski A. On a case of superficial colour presented by an homogeneous liquid
internally colourless. Z. Phys., 94:38–46., 1935.
[4] Stokes G.G. On the change of refrangibility of light. Phil. Trans.R. Soc. London,
142:463–562., 1852.
[5] Kasha M. Characterization of electronic transitions in complex molecules. Disc.
Faraday Soc., 9:14–19., 1950.
[6] Geddes C.D.; Aslan K.; Gryczynski I.; Malicka J.; Lakowicz JR. Radiative Decay
Engineering. Topics in Fluorescent Spectroscopy, 8:405448., 2005.
[7] Geddes C.D.; Aslan K.; Gryczynski I.; Malicka J.; Lakowicz JR. Noble-metal surfaces
for metal enhanced fluorescence. Topics in Fluorescent Spectroscopy, 8:365401.,
2005.
[8] Joseph R. Lakowicz. Plasmonics in Biology and Plasmon-Controlled Fluorescence.
Plasmonics, 1:533., 2006.
[9] Glass A.M.; Liao P.F.; Bergman J.G. and Olson D.H. Interaction of metal particles
with adsorbed dye molecules: Absorption and luminescence. Optics Letts.,
5(9):368370., 1980.
[10] Campion A.; Gallo A.R.; Harris C.B.; Robota H.J. and Whitmore P.M. Electronic
energy transfer to metal surfaces: A test of classical image dipole theory at short
distances. Chem. Phys. Letts., 73(3):447450., 1980.
[11] Sokolov K.; Chumanov G. and Cotton T.M. Enhancement of molecular fluorescence
near the surface of colloidal metal films. Anal. Chem., 70:38983905., 1998.
[12] Joseph R. Lakowicz. Radiative Decay Engineering: Biophysical and Biomedical
Applications. Analytical Biochemistry, 298:124., 2001.
[13] Drexhage K.H. Interaction of light with monomolecular dye lasers. In Progress in
Optics (Wolfe, E., Ed.), 70:161232., 1974.

52 Bibliography
[14] Amos R.M. and Barnes W.L. Modification of the spontaneous emission rate of Eu31
ions close to a thin metal mirror. Phys. Rev. B, 55(11):72497254., 1997.
[15] Barnes W.L. Fluorescence near interfaces: The role of photonic mode density. J.
Modern Optics, 45(4):661699., 1998.
[16] Amos R.M. and Barnes W.L. Modification of spontaneous emission lifetimes in the
presence of corrugated metallic surfaces. Phys. Rev. B, 59(11):77087714., 1999.
[17] Philpott M.R. Effect of surface plasmons on transitions in molecules. J. Chem.
Phys., 62(5):18121817., 1975.
[18] Chance R.R.; Prock A. and Silbey R. Molecular fluorescence and energy transfer
near interfaces. Adv. Chem. Phys., 37:165., 1978.
[19] Gersten J. and Nitzan A. Spectroscopic properties of molecules interacting with
small dielectric particles. J. Chem. Phys., 75(3):11391152., 1981.
[20] Weitz D.A.; Garoff S.; Gersten J.I. and Nitzan A. The enhancement of Raman
scattering, resonance Raman scattering, and fluorescence from molecules absorbed
on a rough silver surface. J. Chem. Phys., 78(9):53245338., 1983.
[21] Chew H. Transition rates of atoms near spherical surfaces. J. Chem. Phys.,
87(2):13551360., 1987.
[22] Goppert-Mayer M. Uber elementarakte mit zwei quantensprunngen. Ann. Phys.
(Leipzig), 9:1273–295., 1931.
[23] C. Xu and W.W.Webb. Journal of the Optical Society of America B, 13(03):481.,
1996.
[24] Louisell W.H. Quantum statistical properties of radiation. Wiley New York, 1973.
[25] Diaspro A. Methods in cellular imaging. Oxford University Press, New York., 2001.
[26] Diaspro A. Confocal and two-photon microscopy: foundations, applications and
advances. John Wiley and Sons Inc., New York., 2002.
[27] Booth M.J. and Hell S.W. Continuous wave excitation two-photon fluorescence
microscopy exemplified with the 647-nm ArKr laser line. J Microsc, 190:298–304,
Jun 1998.
[28] Denk W.; Strickler J.H. and Webb W.W. Two-photon laser scanning fluorescence
microscopy. Science, 248:73–76, Apr 1990.

Chapter 2 53
[29] Born M. and Wolf E. Principles of optics. Cambridge University Press, Cambridge.,
1999.
[30] Nakamura O. Fundamental of two-photon microscopy. Micr. Res. Tec., 47(3):165–
171, 1993.
[31] Cannell M.B. and Soeller C. High resolution imaging using confocal and two-photon
molecular excitation microscopy. Proc. R. Microsc. Soc., 32:3–8., 1997.
[32] Williams et al. Nature Biotechnology, 21(11):1369–1377., 2003.
[33] Sheppard and Shotton. Springer-Verlag New York Inc., New York, 1997.

Chapter 3
The techniques
3.1 Theory
3.1.1 General insight
One of the first and most elegant experimental applications of the concept of time correlation
of a signal can be found in the study of the fluctuations of the light scattered by a
polymeric suspension. The Brownian diffusion of the polymers induces a tiny widening
of the light spectrum. As soon as the multi-bit digital electronics become available, the
measurements of the widening of the light scattering spectrum, due to the polymeric
motion, was made through the computation of the auto-correlation of the scattered
light, soon known as Dynamic Light Scattering (DLS)[1]. This technique made possible
the study of the free diffusion of colloidal particles in solution in order to evaluate
the diffusion coefficient of the scattering particles. DLS exploits a coherent process; the
auto-correlation function in DLS is dominated by the phase change of the field due to
the motion of a large number of diffusing particles. The fluctuations due to the number
of particles are negligible since N >> 1.
DLS concept led to the invention of the Fluorescence Correlation Spectroscopy (FCS),
by Webb and co-workers in the ’70s [2]. In this case the process is not coherent and the
fluctuations of the signal is a direct consequence of the fluctuations in the number of
observed molecules.
FCS and DLS differ from relaxation techniques such as pH or temperature jump, because
it looks at the statistical variations of a physical parameter in a system at the thermal
equilibrium: there is no perturbation, i.e. the pulse that generates the fluctuations is
not imposed from the outside and the fluctuations are spontaneous and determined by
the nature of the sample under investigation [3].
54

Chapter 3 55
3.2 Dynamic Light scattering
In a light scattering experiment a monochromatic and coherent beam of light (typically
lasers are used) impinges on a sample and is scattered into a detector placed at an angle
θ with respect to the transmitted beam. The intersection between the incident beam
and the scattered beam (determined by the collection optics) defines a volume V, called
scattering volume [4][5]. In an idealized light-scattering experiment the incident light is
a plane electromagnetic wave:
E i (r, t) = n i E 0 expi[k i · r − ω i t] (3.1)
of wavelenght λ, frequency ω i , polarization n i , amplitude E 0 , and wave vector k i ,
where k i is:
k i =
( ωi
)
ˆki (3.2)
c
and ˆk i is a unit vector specifying the direction of propagation of the incident wave.
E i (r, t) is the electric field at the point in space r at time t. The monochromatic beam
that impinges on a molecule, induces a dipole moment
µ = α · E(t) (3.3)
where α is the polarizability tensor. When the molecules in the irradiated volume are
subjected to this incident electric field their constituent charges experience a force and
are thereby accelerated. According to classical electromegnetic theory, an accelerating
charge radiates light. The radiated (or scattered) light field at the detector at a given
time is the sum (superposition) of the electric fields radiated from all of the charges in the
illuminated volume and consequently depends on the exact positions of the charges. The
molecule in the illuminated region are perpetually translating, rotating, and vibrating
by virtue of thermal interactions. Because of this motion the positions of the charges are
constantly changing so that the total scattered electric field at the detector will fluctuate
in time.
The fluctuations can be quantified in their strength and duration by temporally autocorrelating
the recorded intensity signal, a mathematical procedure that gave the technique
its name. Autocorrelation analysis provides a measure of the self-similarity of the signal
time evolution and therefore measures the persistence time.
DLS measurements involve the analysis of the time autocorrelation function of scattered
light as performed by a digital correlator. The normalized time autocorrelation

56 The techniques
function of the intensity of the scattered light g (2) (τ) for a given delay time τ is given
by:
g (2) (τ) =
〈I(t)I(t + τ)〉
〈I(t)〉 2 (3.4)
where I(t) and I(t+τ) are the intensities of the scattered light at times t and t+τ,
respectively, and the brackets indicate averaging over t.
In most cases of practical interest the intensity time autocorrelation function may
also be expressed in terms of the field time autocorrelation function g (1) (τ) as
with g (1) (τ) given by
g (2) (τ) = B + β[g (1) (τ)] 2 (3.5)
g (1) (τ) = 〈E(t)E∗ (t + τ)〉
〈E(t)E ∗ (t)〉
(3.6)
where E(t) and E(t+τ) are the scattered electric fields at times t and t+τ, respectively,
and β is a factor that depends on the experimental geometry (essentially the
numbers of coherence areas detected by the collection optics). Equation 3.5 is known
as the Siegert relation [6]. The factor B, commonly referred to as the baseline, is the
long-time value of g (2) (τ).
Although B should be equal to one, in practice, a small
amount of noise in the measurement can result in values that differ from unity by small
(≈ 10 −4 ) amounts that can however affect the best fit parameters in the g(τ) analysis.
Larger deviations of the baseline from one can indicate that the sample contains rare
large aggregates.
At short time delays the correlation is high because the particles do not have a
chance to move to a great extent. The two signals, I(t) and I(t + τ), are thus essentially
unchanged over very short lag time. As the time lags become larger, the correlation
starts to exponentially fall off to zero, meaning that there is no correlation between
I(t) and I(t + τ) after a long time lag. The reference relaxation time in this process is
the molecular diffusion time that depends on the molecule diffusion coefficient and the
scattering angle.
In details, for monodisperse particles in solution the field correlation function decays
exponentially, g (1) (τ) = exp(−Γτ), with a decay rate of Γ = Dq 2 , where D is the
diffusion coefficient of the particles and q is the magnitude of the scattering wave vector.
The scattering wave vector q is defined as the difference between the incident and the
scattered wave vectors, and its magnitude q is given by
q = 4πn ( ) θ
sin
λ 0 2
(3.7)

Chapter 3 57
where n is the refractive index of the solvent, λ 0 is the wavelength of the laser in
vacuum, and θ is the scattering angle. The Stokes−Einstein relation, D=k B T/6πηR h ,
where k B is Boltzmann’s constant, T is the temperature, and η is the dynamic viscosity,
relates the diffusion coefficient to the hydrodynamic radius R h of the particles.
For a polydisperse sample, g (1) (τ) can no longer be represented as a single exponential
and must be written as the sum or the integral over a distribution of decay rates G(Γ)
by[4]
g (1) (τ) =
where G(Γ) is normalized so that
∫ ∞
0
∫ ∞
0
G(Γ)exp(−Γτ)dΓ (3.8)
G(Γ)dΓ = 1 (3.9)
Not always a continous distribution is suited to the g (1) function fit. Alternatively,
the g (2) function may be expressed as a discrete sum of exponential decays with decay
rates Γ k =D k q 2 :
g (2) (τ) = 1 + f coh ( ∑ k
A k exp[−D k q 2 t]) 2 (3.10)
where D k is the translational diffusion coefficient of the kth species of the NPs or NP
aggregates, q is the wave vector. The parameter f coh is an indication of the ratio of the
detector to the coherence area. The pre-exponential factors, A k , which are proportional
to the product of the square of the molecular mass M k times the number concentration
n k , can be used as an estimate of the relative concentration of the particles according to
where V k is the hydrated volume of the kth species.
A k ≈ n k M 2 k ≈ n kV 2
k (3.11)
3.3 Depolarized light scattering
When an electromagnetic wave of a given polarization impinges on a molecule, it induces
a dipole moment that subsequently radiates. The magnitude and the direction of
the induced dipole moment depend in general on the orientation of the molecule with
respect to the incident electric field of the light. Because molecules in solution suffer
roto-Brownian diffusion, the magnitude and direction of their induced dipole moment
fluctuates. This leads to a change in the polarization of the total field emitted by the
ensemble of fluctuating induced dipole moments. The light scattered from an assembly of

58 The techniques
molecules therefore contains information also about their tumbling. In general molecules
are optically anisotropic; that is, the polarizability tensor α αβ (with α, β=x,y,z) contains
in general non-zero off-diagonal elements. This means that when such a molecule is
placed in an electric field, the component of the dipole moment induced by the field[4]
µ α = α αβ E β (3.12)
will not generally be parallel to the applied field. A set of axes in the molecule can
always be found in which the polarizability tensor is diagonal. These axes are called the
principal axes of the polarizability. Along these axes, µ and E have the same direction,
while for other choices of the body fixed axes this is generally not the case. These axes
define an ellipsoid with the principal axes acting as the major and minor axes.
The scattered radiation is analized through a polarizer oriented either parallel or
perpendicular to the polarization of the incident field 1 . When one measures the correlation
of the light scattered with polarization perpendicular to that of the incident light
(depolarized), the correlation function decays with a relaxation rate, Γ D :
where Γ T is:
Γ D = Γ T + 6Θ (3.13)
and, for a sphere of radius (hydrated) R h ,
Γ T = Dq 2 (3.14)
D =
k BT
6πηR h
(3.15)
Θ =
k BT
8πηR 3 h,R
(3.16)
are the translational and the tumbling rotational diffusion coefficients, used to derive
the average hydrodynamic radius R h .
In general, the field autocorrelation function g(t) for the polarized case is [7]:
g V V (t) ∝ α 2 exp[−tΓ T ] + 16π
45 β2 exp[−tΓ D ] (3.17)
When crossing the two polarizers one finds
1 in order to ensure a good control of the polarization the laser beam (tipically already polarized by
the Brewster windows of the cavity) is filtered by a polarized perpendicular to the scattering plane.

Chapter 3 59
with
g V H (t) ∝ β 2 exp[−tΓ D ] (3.18)
and
α = α ‖ + 2α ⊥
3
(3.19)
β = α ‖ − α ⊥ (3.20)
The polarizability α is 1/3 the trace of the molecule-fixed polarizability tensor, is also
called the isotropic part of the polarizability tensor, since it is independent of molecular
orientation 2 . The terms α ‖ and α ⊥ indicate the polarizabilities along the longer and the
shorter axis of the anisotropic object. The parameter β, however, measures the optical
anisotropy of the molecule and it is called the molecular optical anisotropy of the polarizability.
These two parameters determines the intensities of the different components
of the light-scattering spectrum.
In chapter....(mettere riferimento NR) the scattering of the nanoparticles (NPs) has
been measured through a vertical polarizer, i.e. parallel to the direction of the polarization
of the laser light. If we assume that we can define in the NP a direction along which
the polarizability α ‖ of the electrons is much larger than in the other two orthogonal
directions (α ⊥

60 The techniques
α ∝ 1 (L + 2d) (3.25)
3
The ratio of the amplitudes of the translational to the rotational exponential components
obtained from the best fit, R fit , is then given by:
R fit = 4 (α ‖ − α ⊥ ) 2
45 α 2 = 4 δα
2
45 〈α〉 2 (3.26)
From Eq. 5.8 it is straightforward to derive the (effective) axial ratio, d/L, as:
3.4 Method of Cumulants
( ) (
L
d = 3 2δalpha
3 〈α〉 − 1 1 − δ ) −1
alpha
(3.27)
3 〈α〉
Finding the precise functional form for the distribution of decay rates G(Γ) is problematic
because the correlation function is measured discretely only over an incomplete range of
lag times τ and there is always Poisson noise associated with the data. There are several
ways of using DLS data to characterize G(Γ), but one of the simplest is the method
of cumulants first proposed by Koppel[8]. This method is based on two relations: one
between g (1) (τ) and the moment-generating function of the distribution, and one between
the logarithm of g (1) (τ) and the cumulant-generating function of the distribution. It is
appropriate only for use in cases in which G(Γ) is monomodal [8][9]. In fact, as was
discussed by [8], the form of g (1) (τ) as given in Eq.3.8 is equivalent to the definition of
the moment-generating function M(−τ, Γ), of the distribution G(Γ) [10]:
M(−τ, Γ) =
∫ ∞
0
G(Γ)exp(−Γτ)dΓ ≡ g (1) (τ) (3.28)
The mth moment of the distribution m m (Γ) is given by the mth derivative of M(−τ, Γ)
with respect to τ:
m m (Γ) = dm M(−τ, Γ)
d(−τ) m | −τ=0 =
∫ ∞
0
G(Γ)Γ m exp(−Γτ)dΓ| −τ=0 (3.29)
Similarly, the logarithm of the field-correlation function is equivalent to the definition
of the cumulant generating function 7 K(−τ, Γ)
K(−τ, Γ) = ln[M(−τ, Γ)] ≡ ln[g (1) (τ)] (3.30)
where the mth cumulant of the distribution κ m (Γ) is given by the mth derivative of
K(−τ, Γ):

Chapter 3 61
κ m (Γ) = dm K(−τ, Γ)
d(−τ) m | −τ=0 (3.31)
By making use of the fact that the cumulants, except for the first, are invariant under
a change of origin, one can write the cumulants in terms of the moments about the mean
as
κ 1 (Γ) =
∫ ∞
0
G(Γ)ΓdΓ ≡ Γ (3.32)
κ 2 (Γ) = µ 2 (3.33)
where µ m are the moments about the mean, as defined by
µ m =
∫ ∞
0
κ 3 (Γ) = µ 3 (3.34)
G(Γ)(Γ − Γ) m dΓ (3.35)
The first cumulant describes the average decay rate of the distribution. The second
and the third cumulants correspond directly to the appropriate moments about the
mean: the second moment corresponds to the variance, and the third moment provides
a measure of the skewness or asymmetry of the distribution. The first two cumulants
must be positive, but the third cumulant can be positive or negative. The basis of the
cumulant expansion that is tipically used in the analysis of DLS data lies in expanding the
logarithm of g (1) in terms of the cumulants of the distribution. This relation follows from
the fact that the mth cumulant is the coefficient of (−τ) m /m! in the Taylor expansion
of K(−τ, Γ), about τ=0, as given by
ln[g (1) (τ)] ≡ K(−τ, Γ) = −Γτ + κ 2
2! τ 2 − κ 3
3! τ 3 ... (3.36)
To take advantage of this form and use linear least squares methods to fit this function
to the data requires that a key assumption be made about the data: the baseline must
be assumed to be exactly one. Then a fit can be made to
ln[g (2) (τ) − 1] = ln β 2 − Γτ + κ 2
2! τ 2 − κ 3
3! τ 3 + ... (3.37)
Equation 3.37 is the traditional fitting function that is described in many DLS texts.
[11][?][1] Although most modern correlators do an excellent job of measuring the baseline,
small amounts of noise can lead to small deviations from unity. Nonlinear fitting routines
permit the possibility of fitting the data to g (2) directly. From Eq. 3.37, we obtain

62 The techniques
g (2) = B + βexp(−2Γτ + κ 2 τ 2 − κ 3
3 τ 3 ...) (3.38)
where Γ is the average decay rate and κ 2 /Γ 2 is the second order polydispersity index
The first cumulant is directly proportional to some average diffusion coefficient, D,
called z average diffusion coefficient,
with
D = Γ/q 2 = D z (3.39)
D z ≡
∑
i c im i D i
∑i c im i
(3.40)
where m i is the molecular weight and c i the weight concentration.
3.5 MemExp program
The program MemExp uses the maximum entropy method (MEM) and either nonlinear
least squares (NLS) or maximum likelihood fitting to analyze a general time-dependent
signal in terms of distributed and discrete lifetimes. One or two distributions of effective
log-lifetimes, g(logτ) and h(logτ), plus an optional polynomial baseline (up to a cubic)
can be extracted from the data [12]. The h distribution is used to account for signals opposite
in sign to those described by the g distribution when analyzing data that rise and
fall. Both distributions are obtained numerically from the data and are not restricted
to any functional form. Simultaneously, MemExp performs a series of fit by discrete
exponentials in which exponentials are added one at a time and are initialized based
on the emerging structure in the MEM distribution. The amplitude and log-lifetime of
each exponential, plus any optional baseline parameters utilized, are varied using either
NLS (for Gaussian noise) or ML (for Poisson noise) fitting. MemExp automatically recommends
one distributed and one discrete description of the kinetics as optimal. The
graphical summary plotted by MemExp permits a through evaluation of the results.
Multiple MEM ’prior models’ are supported, facilitating a comprehensive analysis of the
kinetic data[12].
The hybrid MEM/NLS analysis is applicable to a general time-dependent signal. A
continuous description that evolves according to the MEM is used to guide a series of
discrete NLS fits during which one exponential is added at a time. Thus, the hybrid algorithm
provides an automated and objective approach to the multiple-minimum problem

Chapter 3 63
that plagues conventional parametric fitting (NLS and ML) when many kinetic processes
are present. Consequently, MemExp is particularly useful when a large number of
discrete processes is an appropriate kinetic description. By interpreting kinetics simultaneously
from the complementary perspectives of discrete exponentials and continuously
distributed lifetimes, a comprehensive assessment of the data can be performed conveniently.
Also, the use of several different priors is supported to improve the fidelity of the
distributions recovered. In the next section, the MemExp algorithm is described[12][13].
3.6 MemExp:a MEM/NLS algorithm
Kinetics measured at times t i can be fit by
∫ ∞
F i = D 0 dlogτ[g(logτ) − h(logτ)]e −ti/τ +
−∞
j=1
k=0
3∑
(b k − c k )(t i /t max ) k (3.41)
where g(log τ) and h(log τ) are the lifetime distributions that correspond to decaying
and rising kinetics, respectively, and a polynomial accounts for the baseline. The constant
t max is approximately the longest time measured and prevents baseline coefficients from
differing greatly in magnitude. Formally, Laplace transforms involve an integral over the
rate coefficient k = 1/τ, but when analyzing data that span several decades in time, it is
more convenient to express the underlying distribution as a function of log τ [14]. Upon
discretizing Eq. 4.1 for a computer, the g and h distributions become vectors. All g, h,
b, and c parameters may be restricted to positive values without loss of generality. The
normalization constant D 0 can be estimated from the data, provided that the temporal
window spanned by the measurement is sufficient to include all kinetic processes. In the
presence of experimental uncertainties, the unconstrained numerical inversion of Eq. 4.1
is known to be an ill-posed problem; infinitely many solutions exist that fit the data to
within the noise [15]. Alternatively, a fit can be done by n e exponentials:
∑n e
3∑
F i = D 0 A j e −t i/τ j
+ B k (t i /t max ) k (3.42)
where A j and τ j are the amplitude and lifetime of the jth exponential respectively,
and B k is the kth coefficient in the polynomial approximation of the baseline. The MEM
part of MemExp is an extension of previous work [12] and is based on the algorithm
of Cornwell and Evans [16]. It fits Eq. 4.1 to kinetics by an iterative, second-order
optimization of the entropy S [17],
k=0

64 The techniques
S( P ⃗ , F ⃗ M∑
) = [P j − F j − P j ln(P j /F j )] (3.43)
j=1
subject to constrained values of χ 2 :
χ 2 = 1 N
and, optionally, the normalization I,
N∑
i=1
( F i − D i
σ i
) 2 (3.44)
∑M 1 ∑M 2
I = ∆( g j − h j ) (3.45)
j=1
The constrained optimization is achieved by maximizing the function Q,
j=1
Q ≡ S − λχ 2 − αI (3.46)
where λ and α are Lagrange multipliers. The natural logarithm in Eq. 4.3 requires
that all elements of the desired image P be positive. Here, P is the concatenation of up
to four vectors: g, h, b, and c (Eq. 4.1). M 1 and M 2 pixels are used to discretize the
g and h distributions, respectively, and ∆ is the spacing in log τ. Implicit in the use
of χ 2 is the assumption that the standard errors, σ i , in the measured data, D i , can be
assumed Gaussian. Note also that, if P is normalized ( ∑ M
j=1 P j= constant) and if all F j
are set equal to a constant, then maximizing S (Eq. 4.3) is the same as maximizing a
more familiar expression for the entropy
S = −
M∑
P j ln(P j ) (3.47)
j=1
Unconstrained maximization of S (λ = α=0) with respect to P j yields P j = F j . That
is, F is the prior image that is defaulted to in the absence of good data. The ability to
manipulate F lends flexibility to MEM inversions not available to other regularization
methods. In contrast, this added flexibility requires that F be chosen wisely.
The MEM calculation is iterative, starting with an initial image having maximum entropy:
P = F (initially, a constant); λ ≈ α ≈ 0. NewtonRaphson optimizations of Q at
fixed values of λ and α are followed by automatic adjustments of the Lagrange multipliers.
The multipliers are changed gradually enough to ensure that the gradient of Q
remains sufficiently small [16]. Thus, P evolves from a flat distribution (with very large
χ 2 ) into a series of maximum-entropy distributions that become increasingly structured
as χ 2 and I approach the desired values. Whenever the prior F is derived from P, P is

Chapter 3 65
set equal to the new prior and the Lagrange multipliers are reset to near zero. The optimization
of Q and multiplier updating are iterated until χ 2 stops changing appreciably.
When χ 2 reaches a specified value (e.g., 1.2), the current P distribution is written to a
file, and MemExp performs the first NLS fit (Eq. 4.2) with one exponential included for
each MEM peak having an appreciable area and a mean within a specified log τ range.
Because distributions produced by the MEM are quite smooth and noiseless, simple
identification of minima and maxima in P is sufficient to detect peaks in the lifetime
distribution. Let i − and i + denote the location of the minima immediately preceding
and following the ith local maximum in P, respectively. Then the area and mean of this
MEM peak are obtained by numerical integration, as in [13]
Area i = ∆
i + ∑
j=i −
g j (3.48)
i + ∑
Mean i =
∆ g j logτ (3.49)
Area i
j=i −
For the outermost MEM peaks (fastest and slowest processes), the range of integration
extends to the distribution edge. The initial amplitude and log lifetime of the
ith exponential in the NLS fit are then taken as Area i and Mean i , respectively. During
NLS optimization of the discrete parameters, the lifetimes are free to stray outside the
specified log τ range. This range is used to exclude exponentials from the NLS fit that
correspond to undesirable ripples in the MEM distribution at either end of the log τ axis.
The analysis of local maxima performed by MemExp also serves to identify relatively
sharp, well-resolved peaks. In addition to the area and mean, two ratios are calculated
for each local maximum in the lifetime distribution. The value of P at the local maximum
is divided by each of the values of P at the adjacent minima. For a small ripple
superimposed on a broad peak, at least one of these ratios is approximately unity. For
a well-resolved peak, both ratios are large, perhaps exceeding 100. For a narrow peak
partially overlapping a broad peak, one ratio will be rather large (e.g., 100) and the
other will be smaller (e.g., 2), depending on the extent of overlap. If the ith MEM peak
is found to be sufficiently large and sufficiently well resolved, then it is subtracted from
P in the range from i − to i + . After all such peaks are subtracted, the remainder of P
is blurred uniformly. Then each subtracted peak is accounted for, either by adding it
to F unchanged or by adding a Gaussian with the same area and a reduced FWHM.
As the MEM convergence continues, this analysis (and storage) of the P distribution
is repeated periodically to determine the current number of MEM peaks n, the peak
means and areas, and the maximum-to-minimum ratios. If n has not changed since the
last time the image was analyzed, the MEM calculation is resumed. If n has increased

66 The techniques
by one, the areas and means of the current MEM peaks are characterized (Eqs. 4.11
and 4.12) and a new NLS fit is performed with one more exponential than was used
in the previous NLS fit. If n has increased by more than one, the n+1 MEM peaks of
largest area are introduced as exponentials in an NLS fit. Then, the next largest MEM
peak is accounted for in an additional NLS fit, and so on. Thus, the MEM is used to
introduce one exponential at a time into a series of NLS fits; as features are resolved in
the continuous kinetic description, exponentials are added to the discrete description.
In chapter (mettere riferimento) the autocorrelation functions were analized by means
of the MEM method: the acquired second order autocorrelation functions (ACFs) of
the scattering light were first converted into the first order ACFs, G (t), and the first
order ACFs were analyzed by means of the Maximum Entropy method obtaining the
distribution of relaxation times according to the relation[13]:
∫ ∞
G(t) = A dlog(τ)P τ (log(τ))exp[−t/τ] (3.50)
−∞
The relaxation time τ is inversely proportional to the particle diffusion coefficient,
D, and the exchanged wave vector, Q, by the relation τ = 1/(Dq 2 ).
The diffusion coefficient is related to the particle average hydrodynamic radius, R h ,
as, D = K B T/(6πηR h ). Therefore the linear relation between the relaxation time and
the hydrodynamic radius, τ = R h (6πη)/(K B T Q 2 ) = R h /ρ, can be used to compute the
number distribution of particles with radius R h by fitting the P R (R h ) = P τ (log(τ)) 1 τ
distributions to a sum of log-normal functions of the type:
P R (log(R h ))| Rh =ρτ = ρA ∑ j
α j 〈R〉 5 j exp[−(log(R h) − log(〈R〉 j
)) 2
2σ 2 j
(3.51)
where 〈R〉 j
and σ j are the average values of the hydrodynamic radius and the width
of the distribution component and α j is the number fraction of the j-th component
( ∑ j α j=1) in the distribution.
3.7 Fluorescence Fluctuation Spectroscopy (FFS)
Fluorescence Fluctuation Spectroscopy (FFS) is a branch of spectroscopy based on the
analysis of the statistical fluctuations of the fluorescence intensity obtained from small illuminated
sample volumes [18] [19] [20]. The fluctuations may be caused by spontaneous
changes in the number of the fluorescent molecules or/and in the molecular properties
of the molecules passing through the probed region of the sample [21]. Chemical reactions
[22], inter and intra-molecular dynamics [23], protein-protein and DNA-protein
interactions [24], enzyme activity [25] [26] may affect the fluorescence quantum yield of

Chapter 3 67
the molecules.
In order to analyze the statistical properties of the detected photons, two mathematical
approaches are used: the computation of the auto-correlation function, G(τ), of
the photon rate [19] [27] [28], and of the probability distribution, (p d ), of the number
of photon-counts per detection interval [20] [29] [30]. The former is the basis for Fluorescence
Correlation Spectroscopy (FCS) [19] [27] [28] and the second for the Photon
Counting Histogram (PCH) method [31] [22].
FCS compares the fluorescence intensity measured at time t, with the intensity measured
at a later time t+τ performing the product F(t+τ)F(t). This information averaged
over all the possible values of the lag time τ is stored in the auto-correlation function,
G(τ), making it a quantitative description of the temporal behaviour of the fluorescence
signal fluctuations [21]. The principal features of G(τ) [32] are the amplitude at lag time
zero, G(0), and the characteristic decay time, τ D , that are related, respectively, to the
average number of molecules in the excitation volume and to the characteristic diffusion
time of the molecules through the excitation volume as we report in section 3.8.
PCH is based on the analysis of the histogram of the photon counts measured over times
τ >> τ D [20].The photon-counting distribution is described by a Poisson function only
when the intensity of light is non-fluctuating [31]. The presence of fluctuations gives rise
to the deviation from Poissonian nature of the photon-counts and contain information
about the processes they have been generated from. PCH determines the average number
of molecules and the molecular brightness of the species under investigation looking
at the deviation of p d from the Poissonian statistics or through the two lowest moments
of p d [33], 〈k〉 and 〈 k 2〉 . Within the limit τ >> τ D PCH misses the dynamical information
but describes the amplitude of the fluctuations of the photon counts. The Mandel’s
factor Q together with 〈k〉 fully describes the properties of fluorophores as discussed in
section 3.9.
PCH and FCS are considered as complementary techniques because the first is centered
on the strength of the fluctuations while the second describes mainly the dynamical behaviour
of the system under investigation, two aspects of the same dynamic. Often both
the methods can be employed in order to analyze the same system and one can be used
to partially validate the results from the other.
3.8 Fluorescence Correlation Spectroscopy
Fluorescence Correlation Spectroscopy (FCS) is one of the first methods for high spatial
and temporal resolution analysis of extremely low concentrated biomolecule solutions.
In contrast to other fuorescence techniques, the parameter of primary interest is not the
emission intensity itself, but rather spontaneous intensity fluctuations caused by the de-

68 The techniques
viations of a finite molecular system from thermal equilibrium. In principle, all physical
parameters that give rise to fuctuations in the fuorescence signal are accessible by FCS.
It is, for example, rather straightforward to determine local concentrations, diffusion and
rotational coeffcients, flow rates, aggregation number, kinetic rate constants.
When the fluorophore diffuses into a focused light beam, there is a burst of emitted
photons due to multiple excitation-emission cycles from the same fluorophore. If the
fluorophore diffuses rapidly through the volume, the fluorescence burst is short lived. If
the fluorophore diffuses more slowly , the fluorescence burst displays a longer duration.
The fluctuations can be quantified in their strength and duration by temporally autocorrelating
the recorded intensity signal, a mathematical procedure that gave the technique
its name [34]. The main requirement to perform FCS is the reduction of the concentration
or the observation volume such that only few molecules are simultaneously detected.
A further essential requirement is a high fluorescence yield [32]. Therefore, improvement
in the signal/noise of FCS have been searched by using efficient fluorescent dyes to label
the molecules of interest, bright and stable light sources like lasers, and ultrasensitive
detectors, e.g. avalanche photodiodes with single-photon sensitivity. Since confocal and
TPE microscopy allow to limit the detection volume to less than one femtoliter, concentrations
in the nanomolar range are optimal for FCS measurements and correspond to
an average of 1 molecule per V exc . Under these circumstances, the signal fluctuations
induced by molecules diffusing into or out of the focal volume are large enough to yield
good signal-to-noise ratios (≈ 7 − 30%). During the time a particle spends in the focus,
chemical or photophysical reactions or conformational changes may alter the emission
characteristics of the fluorophore and give rise to additional fluctuations in the detected
signal.
3.8.1 The auto-correlation function G(τ): Brownian diffusion
We discuss now the mathematical details of the FCS autocorrelation function. FCS is
aimed at maximizing the fluctuations of fluorescence. At any given time, the number of
molecules moving in the excitation volume have a poissonian distribution, with average
value of 〈N〉=N and a standard deviation of σ(N)= √ N; therefore, the root mean square
fluctuation of the particle number N is given by [35]:
√
〈(δN) 2 〉
〈N〉
=
√
(N − 〈N〉)
2
〈N〉
= 1 √
〈N〉
(3.52)
Since the fluctuations become smaller at increasing 〈N〉, FCS requires small focal
volume V exc to minimizeand 〈N〉 and naxinize S/N. Dye concentration ≈ 10 −9 M in
V exc ≈ 1 femtoliter (1 fl = 10 −15 l) guarantees a good compromise between the average

Chapter 3 69
number of molecules in V exc and a good Signal to Noise Ratio, (S/N), as measured from
the zero lag time ACF, G(0) [36].
Assuming constant excitation power, the fluctuations, δF (t), of the fluorescence signal
are defined as the deviations from the time average of the signal[36]:
δF (t) = F (t) − 〈F (t)〉 (3.53)
where
∫ T
〈F (t)〉 = 1 F (t)dt (3.54)
T 0
and T is the total duration of the experiment.
Figure 3.1: Left: bright molecules that freely diffuse in or out the excitation volume selected in the sample
by the laser beam. Right: example of a typical auto-correlation function, G(τ), where are underlined
the most important parameters of G(τ): the amplitude at τ=0 linked to average number of molecules
present in V ex and the characteristic decay time τ D linked to the motion properties of the specie under
observation.
The fluctuations are related to the excitation intensity profile I ex (r) and the molecular
physical parameters by:
∫
δF (t) = κ I ex (r)S(r)δ[σ(r, t)q(r, t)c(r, t)]d 3 r (3.55)
V
In this equation κ is the detection efficiency and accounts for the quantum efficiency
of the detector employed, filter transmission curves, objective numerical aperture (N.A.);
S(r) is the optical collection efficiency function: a dimensionless quantity linked to the
collection properties of the set-up; σ(r, t), q(r, t) and c(r, t) the absorption cross-section,
the fluorophores quantum yield and the local concentration respectively. Usually S(r)
and I ex (r)/I 0 are unified into a single function that describes the spatial distribution of
the fluorescence signal that would be collected by a uniform and stable emitter: usually
a three-dimensional, (3D), Gaussian that decays at 1/e 2 of its maximum at ω 0 in the
radial direction and at z 0 in the axial dimension [37]:

70 The techniques
as [38]:
W (r) = I exr
S(r) ≈ exp(− 2(x2 + y 2 )
I 0 ω0
2 ) exp(− 2z2
z0
2 ) (3.56)
The factors σ(r, t) and q(r, t) can be grouped together to give a parameter defined
η 0 = I 0 κσq (3.57)
that accounts for the photon count rate per molecule per second while the only factor
responsible for the fluorescence fluctuations δF (t) is c(r, t) that represents the fluctuations
in the local concentration inside the V exc . On the basis of these considerations and
approximations we obtain:
∫
δF (t) =
substituting this expression in the definition of G(τ):
we have:
G(τ) =
G(τ) =
W (r)δ[η 0 c(r, t)]dV (3.58)
〈δF (t)δF (t + τ)〉
(〈F (t)〉) 2 (3.59)
∫ ∫ 〈
〉
′
W (r)W (r ) δ(ηC(r, t))δ(ηC(r ′ , t + τ)) d 3 rd 3 r ′
( ∫ W (r) 〈δ(ηC(r, t))〉 d 3 r) 2 (3.60)
the term that accounts for the fluctuation can be expressed as:
δ(ηC(r, t)) = C(r, t)δη + ηδ(C(r, t)) (3.61)
If the properties of the fluorophores do not change during the permanence of the
molecules in V exc , i.e. δη = 0 and the system is stationary:
G(τ) =
∫ ∫ 〈
〉
′
W (r)W (r ) δ(ηC(r, 0))δ(ηC(r ′ , τ)) d 3 rd 3 r ′
〈C〉 2 ( ∫ W (r)d 3 r) 2 (3.62)
〈
〉
The fluctuating term δ(C(r, 0))δ(C(r ′ , τ)) for a particle freely diffusing in an open
volume with diffusion coefficient D can be explicitly expressed as [39]:
δ(C(r, 0))δ(C(r ′ 1
, τ)) = 〈C〉
(4πDτ) − 3 2
and the auto-correlation function becomes:
exp(− (r − r′ ) 2
) (3.63)
4Dτ

Chapter 3 71
γ 1
G(τ) =
V ex 〈C〉 1 + τ
1
τ D (1 + ( ω 0
z 0
) 2 τ
τ D
) 1 2
where τ D and and the translational coefficient D are related as:
(3.64)
τ D = ω2 0
αD
(3.65)
where α is a numerical constant dependent on the experimental set-up (α=4 for
confocal geometry and α=8 for two-photon excitation experiments)
The excitation volume V ex [38] is defined as:
V ex = (∫ W (r)d 3 r)
∫ 2 ( ∫ exp(− 2(x2 +y 2 )
)exp(− 2z2 )) 2
W 2 (r)d 3 r = ω0
2 2z0
2
∫
exp(−
4(x 2 +y 2 )
)exp(− 4z2 )
ω0
2 2z0
2
Finally, the ACF, G(τ), can be cast in the form [36]:
= π 3 2 ω
2
0 z 0 (3.66)
G(τ) =
∫ ∫
W (r)W (r
′
) 〈C〉
1
(4πDτ) − 3 2
exp(− (r−r′ ) 2
4Dτ
)d 3 rd 3 r ′
(〈C〉 ∫ W (r)d 3 r) 2 (3.67)
By fitting the theoretical expression for G(τ) to the experimental data it is possible to
evaluate the local fluorophores concentration [32] from the zero lag time auto-correlation
function:
G(0) ≈
γ
V ex 〈C〉 =
γ
〈N〉 ⇒ 〈C〉 =
γ
V ex G(0)
(3.68)
where γ is a constant depending on the adopted experimental set-up (γ = 0.35 or
0.076 for confocal and TPE set-ups respectively) and the diffusion coefficient D can be
derived from the characteristic decay time τ D and the eq.5.14.
3.8.2 The auto-correlation function G(τ): beyond diffusion
In the previous section we derived the expression of G(τ) assuming that the only relevant
physical process that takes place in the system under investigation was the free Brownian
diffusion described by the condition δη = 0. This is rarely the case: the properties of the
fluorophores often changes during the time they spend in V ex leading to a phenomenon
called flickering [40]. The most common origin of flickering are the intra and intermolecular
reaction such as the intersystem-crossing S 1 → T 1 or drug-protein binding.
Since the transition T 1 → S 0 is almost forbidden, the molecule is dark during the time
it spends in T 1 with the result that the fluorescence signal incoming from each single
molecule is interrupted by dark intervals. If these intra or inter-molecular interactions

72 The techniques
give rise to fluorescence fluctuations on a time scale smaller than that of diffusion and if
D is not altered by the interaction processes [40] [41], the two dynamics can be separated
and the total auto-correlation function can be written as the product of two factors: the
first describing the interaction dynamics and the second related to the Brownian motion:
G total (τ) = G chem (τ)G BD (τ) (3.69)
where G chem (τ) is a simple exponential decay that accounts for the flickering and
G BD (τ) assumes the form of eq.3.64:
G chem = 1 +
γ 1
G BD (τ) =
V ex 〈C〉 1 + τ
T
1 − T exp(− t
τ chem
) (3.70)
1
τ D (1 + ( ω 0
z 0
) 2 τ
τ D
) 1 2
(3.71)
The flickering exponential appears as an additional shoulder in the total auto-correlation
function profile (Figure 3.2 in a time scale of few µs, typically, faster than that of diffusion
[36]. The S 1 → T 1 conversion can be considered as a prototype of any intra-molecular
process that results in a reversible switching between two states of which one is bright
(B) than the other dark (D):
B kB → ←kD D (3.72)
Within this model, the exponential characteristic decay time of G chem (τ) is related
to the forward and backward rate k D and k B as:
τ chem =
1
k D + k B
(3.73)
The fraction of molecules in the dark state, T, depends on the brightness of the states
and on the chemical kinetic rates:
⎧
⎨
T =
⎩
k D
k D +k B
if η D = 0
k D k B (ηB 2 −η2 D )
k D +k B (k D ηD 2 +k BηB 2 ) if η D ≠ 0
(3.74)
At least other two steps are necessary if we want to describe the auto-correlation
function in the more complex case of chemical reaction in a biomolecular solution. First:
reactions may alter the diffusion coefficient of the labelled particles leading to a situation
where there are at least two species simultaneously diffusing in the excitation volume.
Second: there are situations where the motion is not a simple free Brownian diffusion
[32] but it is confined to two dimensions or to a fractal shape. These situations can be
accounted for by writing the diffusional term of the auto-correlation function as:

Chapter 3 73
Figure 3.2: Auto-correlation function in presence of the two main components: the diffusive one (in
red) and the kinetic intra-molecular one (in blue); the last one appears as a shoulder in the ACF and is
characterized by the fraction F and the caracheristic time τ react.
G BD (τ) = 1 ∑ n
i=1 η i 〈C i 〉 M i (τ)
V ex ( ∑ n
i=1 η i 〈C i 〉) 2 (3.75)
where the i indicates the i − th species in the volume and M i (τ) is the decay whose
shape changes depending on the type of motion that the molecules undergo. The function
M i (τ) can be written as:
⎧
⎪⎨
(1 + τ
τ D,i
) −1 (1 + ( ω2 0
) 2 τ
z 2 τ
0 D,i
) − 1 2
M i (τ) = (1 + τ
τ D,i
) −1
⎪⎩
exp(− τν i
ω 0
) 2
(a)
(b)
(c)
(3.76)
Eq.3.76a applies to freely diffusing particles [42] in 3D; eq.3.76b to particles freely
diffusing on a surface (i.e. the case of cells or vescicule membranes) [43] and eq.3.76c
describes molecular drifts [44] with drift velocity magnitude v i . The description of the
diffusion on biological membranes in terms of a free two dimensional motion is not
realistic. Due to the presence of cellular compartments and different micro-domains
(called rafts) the diffusion becomes anomalous and the mean square displacement does
not depend linearly on the time but rather [45] as:
〈
r
2 〉 = t α (3.77)
with α ≤ 1. In the diffusional part of the auto-correlation function we have then to
make the change:
τ
→ (
τ ) α (3.78)
τ D,i τ an,i

74 The techniques
where α ≤ 2 and ”an” stands for ”anomalous”. Depending on the process that takes
place in the V ex the characteristic time of the G(τ) changes and thus the shape of G(τ);
for example active transport and the anomalous sub-diffusion decays appear respectively
faster and slower than that for the three dimensional Brownian diffusion (Figure 3.3).
Figure 3.3: Auto-correlation function shape in the case of three dimensional Brownian motion (red),
two-dimensional Brownian motion (yellow) and direct flow (blue).
3.8.3 Pseudo cross-correlation function
In the computation of the FCS auto-correlation function a series of electronic phenomena
may produce correlated output in the detectors. To this regard an important parameter
is the detector dead-time, t d , defined as the time span over which a detector is blind after
the collection of a photon [46]. For a Single Photon Avalanche Diode (SPAD) at high
(hundreds of volts) reversed bias, the dead time is 350 ps and it mostly affects the minimum
time resolution for applications regarding the excited state lifetime measurements.
However, most of the best available correlator boards provide 10 − 12 ns time resolution
for the computation of the auto-correlation function and this limitation is actually more
due to the cost of the GHz digital electronics than to the detector time jitter. The deadtime,
t d ≈ 50 ns, affects the linearity of the detector response for rates of the order of
few MHz and reduces the effective collection efficiency. Afterpulses, or spurious pulses
after a true photon counting event, also affect the auto-correlation function for lag times
≤ 1 µs.
To overcome these artefacts on the short lag times in the auto-correlation functions, one
can split the fluorescence signal on two different detectors and look at the pseudo crosscorrelation
3 . In this geometry the photon counting statistics is no more affected by the
3 In this case we use the term pseudo cross-correlation because the signal coming from both the
detectors is due to the same fluorophore instead of two different kind of dyes as in the case of proper

Chapter 3 75
electronic distortions of the detector signal and the experimental temporal resolution
is improved to the nanosecond time scale allowing the observation of two phenomena
that take place in this temporal window: the rotational diffusion and the so called antibunching.
Aragon [37] in 1970 derived the rotational correlation function in the case of a rotational
diffusion time, τ rot , larger than the fluorophore life-time [47]:
( ) ( ))
τ
3 τ
G rot (τ) = A rot
(c 1 exp + c 2 exp
τ rot 10 τ rot
(3.79)
where c 1,2 are experimental parameters depending on the polarization of the exciting
light and on the polarization direction selected in the detection channel. Most of the
fluorophores have a life-time ≤ 10 ns while the typical rotational decay time τ rot ranges
between 100 ps (for uncomplexed dyes) and few tens of ns (for labelled proteins). Therefore,
at least in the case of fluorescent proteins, the pseudo cross-correlation function,
eq.3.79, describes the rotational diffusion.
The fastest process that can be observed by means of auto-correlation in terms of pseudo
cross-correlation function is the anti-bunching [48] process. this is linked directly to the
statistics of the fluorescence emission and describes the probability that a chromophore
emits a photon at the time t given the last one was emitted at t = 0:
(
G AB (τ) = 1 + βexp − τ )
τ AB
(3.80)
where β is a constant < 0. The antibunching correlation function shows an initial
growth due to the fact that a chromophore spends a time t = life-time in the excited
state before undergoing the relaxation to the ground state through the emission of a
fluorescence photon [36].
Summarizing, G(τ) with the introduction of the pseudo cross-correlation acquisition
mode, allows to explore a time window that ranges from 10 −9 to 10 −2 second getting
information about a whole set of processes that affect the emission of the fluorophores
contained in the excitation volume.
We can imagine the ideal auto-correlation function as composed by four independent
factors (Figure 3.4) each one described by its characteristic time and related to a
particular process; the shape of G(τ) shows an initial growth due to the antibunching
followed by two rapid decays due to, respectively, rotational dynamics and intra- and
inter-molecular processes and, finally, the slow diffusive motion:
cross-correlation function.
G total (τ) = G AB (τ)G rot (τ)G fl (τ)G motion (τ) (3.81)

76 The techniques
Figure 3.4: A graphical representation of the total auto-correlation function G total (τ) where we have
underlined each component of G total (τ) in order to give a visual relation of the phenomena and their
characteristic times.
3.9 Photon Counting Histogram (PCH)
3.9.1 Theory
General insight
The Photon Counting Histogram (PCH) analysis was first introduced at the end of the
90s [20]. It can be considered the precursor of the burst analysis, that is now used in
many single molecule studies [49] [50], where one looks at the temporal (duration and
repetition rate) and amplitude (i.e. number of photons per burst) of the bursts generated
from particles diffusing through and undergoing reaction inside a tiny open volume V ex
(see Figure 3.5). PCH does not consider the single bursts but rather the statistical
analysis of the density distribution p(k) of the number of detected photons within a time
span T .
Practically one collects the fluorescence or scattering signal from the sample as a
function of time, F (t), with a sampling period T ; each interval is characterized by the
number of detected photons, k, of which one computes to the distribution p(k). Applying
to p(k) the moment analysis introduced by Elson [51] it is possible to calculate the
physical properties of the particles under investigation as described hereafter. Due to
the statistical considerations written above it is quite obvious that the description of the
p(k) through its moments allows one to get information about the physical processes that
take place in the excitation volume. Usually PCH analysis recovers two parameters from
p(k): the average number of molecules, 〈N〉, and the molecular brightness, ɛ, defined as
the number of photons emitted per molecule per second. As described by Gratton and

Chapter 3 77
Figure 3.5: left panel: particles freely diffusing in a tiny open volume defined by the cube at the center
of the image. Right panel: number of detected event as a function of time, 〈N〉 represents the mean
number of events during the total observation time; the vertical arrow indicates that the fluctuations are
governed by a Poissonian statistic.
co-workers [20] in the historical introductive article on PCH, this technique is complementary
to Fluorescence Correlation Spectroscopy (FCS): in fact it concerns with the
amplitude of the fluctuation distribution and misses the dynamical aspects, while FCS
describes the time behaviour of the fluorescence fluctuations, δF (t). Theoretically described
in [20], PCH is nowadays a powerful method to describe heterogeneous mixture
of labelled particles [52] [53] [54] [22] and to characterize the molecular behaviour in vitro
and in vivo [?] [?].
3.9.2 Freely diffusing particles
Suppose that a detector reveals the signal incoming from a perfect (i.e. stable in power
and steady in time) light source, then the output of the detector will not be constant
because of the statistics of the detection process [55]: this results in a Poissonian distribution:
P oi(k, 〈k〉) = (η II D ) k exp(−η I I D )
(3.82)
k!
where η I is a constant factor that accounts for the time duration, ∆t s , of the sampling
interval, supposed much less than the characteristic time of the fluctuations, and the
detection efficiency of the system and represents the proportionality constant between
the average number of photon, 〈k〉, and the intensity at the detector, I D :
〈k〉 = η I I D (3.83)
The presence of intensity fluctuations 4 changes the photon counting statistics that
4 Practically one observes the energy fluctuations rather than the intensity fluctuations obtained by

78 The techniques
results:
p(k) =
∫ ∞
0
P oi(k, η I I D )p(I D )dI D (3.85)
In this expression, introduced by Mandel [57], p(I D ) represents the intensity distribution
at the detector that describe the particular experimental conditions (i.e. in the
case of a constant light source p(I D ) is a delta-function and p(k) reduces at a Poisson
distribution). The Mandel’s formula describes p(k) as the superposition of a Poisson
distribution with the probability of intensity distribution p(I D ). The distribution of the
photon countings, p(k), detected from solutions is characterized by 〈 ∆k 2〉 ≻ 〈k〉; a property
typical of super Poissonian statistics [58]; this means that fluctuations broaden the
tails of the Poissonian distribution: the variance of the intensity distribution determines
the variance of the photon counting statistics:
〈
∆k
2 〉 = 〈k〉 + η I
〈
∆I
2
D
〉
(3.86)
the larger are the fluctuations of I D , the wider is p(k) with respect to a Poissonian
distribution and the more reliable are the information on the molecular photo-dynamics,
that can be extracted from the analysis of the histogram of the photon countings.
The components of the experimental set-ups (microscope, objective and the collection
optics) affect the optical Point Spread Function (PSF) whose shape determines the
photon counting statistics. The Mandel’s formula for the photon counting distribution
becomes[20], for a single particle freely diffusing in a volume V ex larger than that
defined by the PSF (V P SF ):
∫
p 1 (k, V 0 , ɛ) = P oi(k, ɛP SF (r))p(r)dr (3.87)
Since p(r) is the probability of finding the molecule in the position r while it is
diffusing trough V 0 and is defined as:
p(r) =
{
1
V 0
if r ∈ V 0
0 otherwise
(3.88)
then we can write:
integrating the intensity at the detector’s photo-cathode over the whole sensitive area, A, and on the
finite short time interval, ∆t s, needed for the detection:
W (r) =
∫ t+∆ts
t
∫
A
I(r, t)dAdt ≈ ∆t sP SF (I) (3.84)
However if ∆t s is shorter than the characteristic time of the process under investigation the energy
fluctuation carry on the information present in the intensity fluctuations; therefore it is equivalent to
speak in terms of intensity or energy distribution [56].

Chapter 3 79
p 1 (k, V 0 , ɛ) = 1 V 0
∫V 0
P oi(k, ɛP SF (r))dr (3.89)
for a single molecule.
The photon counting statistics depends on ɛ and on the shape of the PSF [56].
example:
For
〈k〉 = ɛ ∫
P SF (r)dr = ɛ V P SF
(3.90)
V 0 V 0
V 0
the average number of detected photon in the sampling period depends on the molecular
brightness ɛ and on the volume defined by the PSF. Usually the properties of a
particular species are expressed in terms of:
ɛ ′ =
ɛ
∆t s
(3.91)
that is the number of photons collected per molecule per second, a quantity independent
from the sampling period and thus useful for the comparison between different
measurement conditions and different set-ups.
If in the volume under investigation there are N ≻ 1 identical and independent particles,
the PCH is given by the convolution of N replicas of p 1 (k) [59]:
p N (k, V 0 , ɛ) = p 1 (k, V 0 , ɛ) ⊗ .......... ⊗ p 1 (k, V 0 , ɛ)
} {{ }
N times
(3.92)
The choice of describing the problem in terms of a finite number N does not reflects
the real situation; it is necessary to consider a small open sub-volume of V 0 in which
the molecules can freely diffuse, while keeping the average value < N ′
>=CV 0 (where
C is the concentration). The number of particles inside the sub-volume fluctuates with
a Poissonian statistics [?]:
〈
p(N) = P oi(N, N ′〉 ) (3.93)
where < N ′ > is the number of particles in V 0 . The PCH for an open sub-volume can
be defined as the average of the N-particles distribution over the Poissonian distribution
of the molecule number statistics in V 0 :
p(k, V 0 , N ′ , ɛ) =
〈 〈
p(k, V 0 , N ′〉 〉
, ɛ)
N
(3.94)
The properties of the open sub-volume are independent from the particular choice of
V ex [20], as soon as V 0 ≥ V P SF . Adopting the convention used in FCS, V 0 ≈ V P SF , the

80 The techniques
average number of photon counts 〈k〉 is related to the average number of molecules in
V 0 , 〈N〉, by the brightness ɛ:
〈k〉 = ɛ 〈N〉 (3.95)
PCH depends on the PSF shape: here we give the relation that describes the photon
counting statistics for the case of TPE for which the PSF is described by a Gaussian-
Lorentzian approximation (see Figure ??) that in cylindrical coordinates is [60]:
( )
P SF 2GL (ρ, z) = 4ω4 0
π 2 ω0 4(z) exp − 4ρ2
ω 2 (z)
( ) ) 2
where ω 2 (z) = ω0 (1 2(z) + z
z R
with z R = πω2 0
λ
on the wavelength. In this case the PCH is given by:
p (1)
2GL (k, V 0, ɛ) = 1 V 0
πω 4 0
2λk!
∫ ∞
0
(1 + x 2 )γ
(3.96)
and describes the PSF dependence
[
]
4ɛ
k,
(1 + x 2 ) 2 dx (3.97)
the integral contains the incomplete γ function and can be evaluated numerically.
Another wide spread approximation of the PSF shape is the choice of a three dimensional
Gaussian function, usually adopted in the case of confocal detection [61] [62]:
(
P SF 3DG (x, y, z) = exp − 2(x2 + y 2 )
)
exp
(− 2z2
p (1)
3DG (k, V 0, ɛ) = 1 V 0
πω 2 0
k!
ω0
2 ∫ ∞
0
γ
z 2 0
)
(3.98)
[
k, 2e −2x2] dx (3.99)
An interesting situation is also that of surface movements in an area A 0 approximated
by a two dimensional Gaussian function:
(
P SF 2DG (x, y) = exp − 2(x2 + y 2 )
)
ω0
2
(3.100)
p (1)
2DG (k, A 0, ɛ) = 1 πω0
2 γ(k, ɛ) (3.101)
A 0 2k!
where A 0 is the area of reference that plays the role of the V 0 in the case of two
dimensional motion. Once collected the photon counts, the data analysis has to be
performed by fitting the appropriate model to the experimental histogram; this method
is very accurate but time consuming. Alternatively it is possible to take advantage from
the fact that the probability distribution can be characterized by its low order moments
[33] [46] [63] as described hereafter.

Chapter 3 81
3.9.3 Photon Moment Analysis (PMA)
In a mono-disperse solution of bright molecules the distribution p(k) (eq.3.97) can be
characterized by the first two moments. These two quantities are incorporated in a unique
parameter that represents the normalized variance [?]. Two choices can be adopted. The
correlation function at zero lag time is:
G(0) =
〈
∆k
2 〉 − 〈k〉
〈k〉 2 (3.102)
G(0) is also the definition of the τ = 0 amplitude of the correlation function [?]
where γ is defined in section 3.8.1.
G(0) =
γ
〈N〉
(3.103)
A more conventional choice for PCH is the use of Mandel’s factor Q to characterize
the distribution:
[?]:
Q =
〈
∆k
2 〉 − 〈k〉
〈k〉
(3.104)
the parameter Q gives the deviation of the distribution from the Poissonian statistics
Q ≺ 0 sub Poissonian statistics
Q = 0 Poissonian statistics
Q ≻ 0 super Poissonian statistics
The presence of intensity fluctuations (eq.3.84) in V ex implies a super Poissonian
statistic. It can be proved that the parameter Q is proportional to the molecular brightness
as:
〈
∆k
2 〉 − 〈k〉
Q =
〈k〉
moreover comparing Q and G(0) one obtains:
= γɛ (3.105)
Q = G(0) 〈k〉 or 〈N〉 = 〈k〉
(3.106)
ɛ
therefore the determination of the first two moments of p(k) describes completely
the system by means of the molecular brightness and the average value of molecules
present in the excitation volume. Since the amplitude of the fluctuation for a super
Poissonian distribution scales as √ N and since Q = γɛ is a quantification for the entity

82 The techniques
of the deviation of p(k) from the Poissonian shape, 〈N〉 and ɛ are the most relevant
parameters that describe the broadening of the photon counting statistics: the larger is
〈N〉, the smaller is the amplitude of the fluctuations, the lower is ɛ, the more difficult is
to calculate Q and so to discriminate the real contribution due to the fluctuations from
the background noise.
3.10 Optical pathway
In Figure 3.6 is reported an overview of the two-photon excitation setup used to perform
FCS, time-domain lifetime measurements and spectra using nano-molar concentration.
It is based on a mode-locked Ti:Sapphire laser (Tsunami 3960, Spectra Physics, CA)
pumped by a solid state laser at 532 nm (Millennia V, Spectra Physics) coupled to
a Nikon (Japan) TE300 inverted microscope. The laser provides 280 fs pulses on the
sample plane ([?]) at a repetition frequency of 80 MHz in the range of 700-1000 nm.
Mirrors and beam steerings are used to direct light to the back door of the microscope.
Here, the fluorescence signal is collected by the objective (Plan Apochromat 60 x water,
NA=1.2, Nikon, Japan), is separated from the excitation beam by the dichroic mirror
to be detected by a CCD camera (spectra measurements) or APDs. Except for spectra
measurements, light is further selected by emission filters and split by a non-polarizing
50% cube splitter to perform cross corralation functions. The minima values of the radial
and axial FHWM of the microscope point spread function (PSF) are 240 ± 40nm and
780 ±50 nm, respectively, at a wavelength of 800 nm [64]. With the introduction of
a beam expander on the optical patway, it is possible to change the excitation volume
reducing the laser beam diameter at the entrance pupil of the objective lens. Typical
excitation volume employed in this work ranges from 0.5 to 0.8 µm 3 . An estimate of
the excitation intensity on the sample is obtained as the average power divided by the
area of the PSF in the focal plane (1 mW corresponds for the set-up described above to
about 80kW/cm 2 ).
Fluorescence signals and autocorrelation functions are acquired by an ALV5000E
(ALV, Langen, D) board and then analyzed by means of the nonleast-squares routine of
the Origin 7.0 software (OriginLab Inc.,Northampton, MA).
3.11 Setup calibration with a reference dye
Fluorescence spectroscopy and lifetime measurements require very high accuracy both in
sample preparation and in setup calibration. The samples have to be freshly preparated,
with filtered buffer, in order to eliminate possible impurities visible in the ACFs in the
diffusive range. The system is calibrated using a reference dye, with a well determined

Chapter 3 83
Figure 3.6: Schematic representation of the experimental setup.
diffusion coefficient: we have used a Rhodamine 6G dissolved in extra pure ethanol (spectroscopic
grade) or fluorescein in H 2 O or buffer at high pH, at nanomolar concentration.
The acquisition card we used (ALV5000E) offers the possibility to record and analyze
simultaneously fluorescence signals coming from the two SPAD units cross-correlating
them. When using a single fluorophore solution, the pseudo-cross correlation function
gives the same information that the autocorrelation function, without the spurious contribution
of the detectors. In order to have a good calibration of the system, the auto
and cross- correlation function must have the same behaviour (Figure 3.7 ).
Figure 3.7: Overlapping of autocorrelation functions (recorded for each channel) and the cross correlation
function for a sample of Rhodamine 6G.
A good overlapping of the autocorrelation function (recorded for each of the two
channels) and the cross correlation function, indicates that the excitation volume seen
by each detector is identical. In fact, both the diffusion time and the G(0) depend on
V exc . Measurements with a reference dye are useful not only to see if setup is properly
aligned, but also to determine the value of the excitation volume (remember in fact that

84 The techniques
the diffusion coeffcient is known, D = 280 µm 2 /s at room temperature). The calibration
procedure is performed daily.

Bibliography
[1] Chu B. In Laser Light scattering: Basic Principles and Practice. Academic Press.
New York., 1992.
[2] Magde D. Elson E.L. and Webb W.W. Fluorescence correlation spectroscopy. II.
An experimental realization. Biopolymers, 13(1):29–61., 1974.
[3] Schwille P. Haupts U. Maiti S. and Webb W.W. Molecular dynamics in living
cells observed by fluorescence correlation spectroscopy with one- and two-photon
excitation. Biophys. J., 77(4):2251–2265., 1999.
[4] Bruce J. Berne and Robert Pecora. Dynamic Light Scattering: With Applications
to Chemistry, Biology, and Physics. Dover, N.Y. (USA), 1976.
[5] Chirico G; Beretta S.; Baldini. J. Chem. Phys., 110:2297–2304., 1999.
[6] Schatzel. Single-photon correlation techniques. Dynamic Light Scattering,W.
Brown, ed. Oxford University, Oxford, UK, page 76148., 1993.
[7] Aragon S.R. Pecora R. Fluorescence correlation spectroscopy and Brownian rotational
diffusion. Biopolymers, 14(1):119–137., 1975.
[8] Dennis E.Koppel. Analysis of Macromolecular Polydispersity in Intensity Correlation
Spectroscopy: The Method of Cumulants . The Journal of Chemical Physics,
57, 1972.
[9] Barbara J. Frisken. Revisiting the method of cumulants for the analysis of dynamic
light-scattering data. The Journal of Chemical Physics, 40, 2001.
[10] A. Stuart and J. K. Ord. Kendalls Advanced Theory of Statistics. Wiley, New York,
1994.
[11] P. Stepanek. Data analysis in dynamic light scattering. Dynamic Light Scattering,
W. Brown, ed. Oxford University,Oxford, UK,, page 177240., 1993.
[12] Peter J. Steinbach; Roxana Ionescu; C. Robert Matthews. Analysis of Kinetics
Using a Hybrid Maximum-Entropy/Nonlinear-Least-Squares Method: Application
to Protein Folding. Biophysical Journal, 82:22442255, 2002.
[13] Steinbach P. J. Two-dimensional distributions of activation enthalpy and entropy
from kinetics by the maximum entropy method. Biophys. J., 70:15211528., 1996.

86 Bibliography
[14] Livesey A. K.; J.-C. Brochon. Analyzing the distribution of decay constants in pulsefluorimetry
using the maximum entropy method. Biophysical Journal, 52:693706,
1987.
[15] Provencher S. W. Low-bias macroscopic analysis of polydispersity. In Laser Light
Scattering in Biochemistry. S. E. Harding, D. B. Sattelle, and V. A. Bloomfield,
editors. The Royal Society of Chemistry, Cambridge, U.K., page 92111., 1992.
[16] Cornwell T. J.; K. F. Evans. A simple maximum entropy deconvolution algorithm.
Astron. Astrophys., 143:7783., 1985.
[17] Skilling J. Classic maximum entropy. In Maximum Entropy and Bayesian Methods.
J. Skilling, editor. Kluwer Academic, Norwell, MA., page 4552., 1989.
[18] Novikov E. Analytical model of the fluorescence fluctuation spectroscopy experiment.
J. Chem. Phys., 114:1745–1753, 2001.
[19] Magde D. ; Elson E. L. ;Webb W. W. Fluorescence correlation spectroscopy. I.
Conceptual basis and theory. Biopolymers 13: 1-27. Biopolymers, 13:1–27, 1974.
[20] Chen Y. Muller J.D. So P.T.C. and Gratton E. The photon counting histogram in
fluorescence fluctuation spectroscopy. Biophys. J., 77:553, 1974.
[21] Van Orden A. Fogarty K. and Jung J. Fluorescence Fluctuation Spectroscopy: A
Coming of Age Story. Appl. Spec., 58:122A–137A, 2004.
[22] E. Muller J. D. ; Chen Y.; Gratton. Resolving heterogeneity on the single molecular
level with the photon-counting histogram. Biophys. J., 78:474–486, Jan 2000.
[23] Edman L.; Mets U.; Rigler R. Conformational transitions monitored for single
molecules in solution. Proc. Natl. Acad. Sci. U.S.A., 93:6710–6715, Jun 1996.
[24] Rauer B.; Neumann E. ;Widengren J.; Rigler R. Fluorescence correlation spectrometry
of the interaction kinetics of tetramethylrhodamin alpha-bungarotoxin with
Torpedo californica acetylcholine receptor. Biophys. Chem., 58:3–12, Jan 1996.
[25] Collini M.; Caccia M.;Chirico G.;Barone F.;Dogliotti E.; Mazzei F. Two-photon
fluorescence cross-correlation spectroscopy as a potential tool for high-throughput
screening of DNA repair activity. Nucleic Acids Res., 33:e165, 2005.
[26] P. Jahnz M.; Schwille. An ultrasensitive site-specific DNA recombination assay
based on dual-color fluorescence cross-correlation spectroscopy. Nucleic Acids Res.,
33:e60, 2005.

Chapter 3 87
[27] Eigen S.M. and Rigler R. Eigen S.M. and Rigler R. 1994. Sorting single molecules
- Application to diagnostic and evolutionary biotechnology. Proc. Natl. Acad. Sci.
U.S.A., 91:5740, 1994.
[28] Maiti S.; Haupts U.;Webb W. W. Fluorescence correlation spectroscopy: diagnostics
for sparse molecules. Proc. Natl. Acad. Sci. U.S.A., 94:11753–11757, Oct 1997.
[29] Kask P. Palo K. Ullmann D. and Gall K. Fluorescence-intensity distribution analysis
and its application in biomolecular detection technology. Proc. Natl. Acad. Sci.
U.S.A., 96:13756, 1999.
[30] Enderlein J. Robbins D.L. Ambrose W.P. Goodwin P.M. and Keller R.A. Statistics
of single-molecule detection. J. Phys. Chem. B, 101:3626, 1997.
[31] Saleh B. In photoelectron Statistics. Springer, Berlin., 1978.
[32] Schwille P. Haupts U. Maiti S. and Webb W.W. Molecular dynamics in living
cells observed by fluorescence correlation spectroscopy with one- and two-photon
excitation. Biophys. J., 77(4):2251–2265, 1999.
[33] Chen Y. Muller J.D. Ruan Q. and Gratton E. Molecular brightness characterization
of EGFP in vivo by fluorescence fluctuation spectroscopy. Biophys. J., 82(1):133–
144, 2002.
[34] P. Haustein E. ; Schwille. Fluorescence correlation spectroscopy: novel variations
of an established technique. Annu Rev Biophys Biomol Struct, 36:151–169, 2007.
[35] Chandrasekhar S. Dynamical Friction. I. General Considerations: the Coefficient of
Dynamical Friction. Astrophysical J., 97:255, 1943.
[36] P. Haustein E.; Schwille. Fluorescence correlation spectroscopy: novel variations of
an established technique. Annu Rev Biophys Biomol Struct, 36:151–169, 2007.
[37] Aragon S.R. Pecora R. Fluorescence correlation spectroscopy and Brownian rotational
diffusion. Biopolymers, 14(1):119–137, 1975.
[38] Muller J.D. Chen Y. and Gratton E. Fluorescence correlation spectroscopy. Methods
Enzymol., 361:69–92, 2005.
[39] Doi M. and Edwards S.F. Fluorescence correlation spectroscopy. In the theory of
polymer dynamics.Oxford, New York., 1981.
[40] Widengren J. and Rigler R. Fluorescence correlation spectroscopy as a tool to
investigate chemical reactions in solutions and on cell surfaces. Cell. Mol. Biol.
(Noisy-le-grand), 44:857–879, Jul 1998.

88 Bibliography
[41] Palmer A. G. and Thompson N. L. Molecular aggregation characterized by high
order autocorrelation in fluorescence correlation spectroscopy. Biophys. J., 52:257–
270, Aug 1987.
[42] Magde D.;Elson E. L.;Webb W. W. Fluorescence correlation spectroscopy. II. An
experimental realization. Biopolymers, 13:29–61, Jan 1974.
[43] Magde D., Elson E. L., and Webb W. W. Fluorescence correlation spectroscopy. II.
An experimental realization. Biopolymers, 13:29–61, Jan 1974.
[44] Magde D. Webb W.W. and Elson E.L. Fluorescence correlation spectroscopy.III
Uniform translation and laminar flow. Biopolymers, 17(2):361–376, 1978.
[45] Feder T. J., Brust-Mascher I., Slattery J. P., Baird B., and Webb W. W. Constrained
diffusion or immobile fraction on cell surfaces: a new interpretation. Biophys. J.,
70:2767–2773, Jun 1996.
[46] Hillesheim L. N. and Muller J. D. The dual-color photon counting histogram with
non-ideal photodetectors. Biophys. J., 89:3491–3507, Nov 2005.
[47] Eggeling C., Fries J. R., Brand L., Gunther R., and Seidel C. A. Monitoring conformational
dynamics of a single molecule by selective fluorescence spectroscopy. Proc.
Natl. Acad. Sci. U.S.A., 95:1556–1561, Feb 1998.
[48] Felekyan S. Kuhnemuth R. Kudryavtsev V. Sandhagen C. Becker W. and Seidel C.A.
Full correlation from picoseconds to seconds by time-resolved and time-correlated
single photon detection. Rev. Sci. Instrum., 76:83104, 2005.
[49] Widengren J. Kudryavtsev V. Antonik M. Berger S. Gerken M. and Seidel C.A.M.
Single-molecule detection and identification of multiple species by multi-parameter
fluorescence detection. Anal. Chem., 78(6):2039–2050, 2006.
[50] Eggeling C. Widengren J. Brand L. Schaffer J. Felekyan S. and Seidel C.A.M. Analysis
of photo-bleaching in single-molecule multi-color excitation and Forster resonance
energy transfer measurement. J. Phys. Chem., 110(9):2979–2995, 2006.
[51] Qian H. and Elson E. L. Distribution of molecular aggregation by analysis of fluctuation
moments. Proc. Natl. Acad. Sci. U.S.A., 87:5479–5483, Jul 1990.
[52] Terada N., Shimozawa T., Ishiwata S., and Funatsu T. Size distribution of linear and
helical polymers in actin solution analyzed by photon counting histogram. Biophys.
J., 92:2162–2171, Mar 2007.

Chapter 3 89
[53] Chen Y., Tekmen M., Hillesheim L., Skinner J., Wu B., and Muller J. D. Dual-color
photon-counting histogram. Biophys. J., 88:2177–2192, Mar 2005.
[54] Huang B., Perroud T. D., and Zare R. N. Photon counting histogram: one-photon
excitation. Chemphyschem, 5:1523–1531, Oct 2004.
[55] Snyder D.L. In random point processes. Wiley-Interscience, New York., 1975.
[56] Muller J.D. Chen Y. and Gratton E. In Fluorescence correlation spectroscopy.
Springer Berlin, 2001.
[57] Mandel L. Fluctuations of Photon Beams: The Distribution of the Photo-Electrons.
. SProc. Phys. Soc., 74:233–243, 1959.
[58] Teich M.C. and Saleh B.E.A. In progress in optics. . North Holland, Amsterdam,
1988.
[59] Feller W. An introduction to probability theory and its applications. . John Wiley,
New York., 1957.
[60] Berland K. M., So P. T., and Gratton E. Two-photon fluorescence correlation
spectroscopy: method and application to the intracellular environment. Biophys.
J., 68:694–701, Feb 1995.
[61] Qian H. and Elson E. L. Analysis of confocal laser-microscope optics for 3-D fluorescence
correlation spectroscopy. Appl Opt, 30:1185–1195, Apr 1991.
[62] Rigler R. Mets U. Widengren J. and Kask P. Fluorescence correlation spectroscopy
with high count rate and low-background - Analysis of translational diffusion. Eur.
Biophys. J. Biophys. Letters, 22(3):169–755, 1993.
[63] Caccia M., Camozzi E., Collini M., Zaccolo M., and Chirico G. Photon moment
analysis in cells in the presence of photo-bleaching. Appl Spectrosc, 59:227–236, Feb
2005.
[64] Malengo G.; Milani R.; Krol S.; Diaspro A.; Chirico G. ReV. Sci. Instrum., 75:2746–
2752., 2004.

Chapter 4
Nano-bio sensors for protein
detection
The interaction of the surface plasmons of gold nanoparticles (NP) a few nanometers
in size with fluorophores can be used to engineer their fluorescence properties. This
possibility can be exploited in principle to obtain nanodevices for protein-protein recognition.
We studied different types of constructs based on gold NPs on which derivatives
of fluorescein were bound. The interaction of this fluorophore with the gold surface
plasmon resonances, mainly occurring through quenching, affects its excited-state lifetime
that is measured by fluorescence burst analysis. The binding of proteins to the
gold NPs through antigen-antibody recognition further modifies the dye excited-state
lifetime. This change can therefore be used to measure the protein concentration.
In particular, we have tested the nanodevice measuring the change of the fluorophore
excited-state lifetime after the binding of the model protein bovine serum albumin (BSA);
then we have applied the nanoassay in order to recognize the p53 protein, whose detection
in the body is highly valuable as marker for early cancer diagnosis and prognosis,
both in standard solutions and in total cell extracts.
The contents of this chapter are also reported in the publications:
• S.Freddi, L.D’Alfonso, M.Collini, M.Caccia, L.Sironi, G.Tallarida, S.Caprioli, G.Chirico
”Excited-state lifetime assay for protein detection on gold colloids-fluorophore complexes”
J.Phys.Chem. C, 113:2722-2730, Jan 2009
• L.Sironi, S.Freddi, L.D’Alfonso, M.Collini, T.Gorletta, S.Soddu, G.Chirico
”P53 detection by fluorescence lifetime on a hybrid fluorescein-isothiocyanate gold
nanosensor”
J.Biomedical nanotechnology, 5:683-691, 2009
90

Chapter 4 91
• L.Sironi, S.Freddi, L.D’Alfonso, M.Collini, T.Gorletta, S.Soddu, G.Chirico
”In vitro and in-vivo detection of p53 by fluorescence lifetime on a hybrid FITCgold
nanosensor”
Proceedings of SPIE, 7574:757403, 2010
4.1 Introduction
Biotechnology is pushing the minimum amount of detectable material toward lower and
lower values [1]. The limiting values of detection depend on the experimental method,
and they are of the order of 1-10 pM that, for a 60 000 Da molecular weight protein,
correspond to about 6-60 ng/L [2] [3]. One of the most promising properties that can be
exploited in this field is related to the plasmonic resonances of noble-metal nanoparticles
(NPs). These resonances, present also in bulk metals but shifted toward higher energy
gaps with respect to the nanoparticle case, lie in the visible range of the spectrum and
are due to excitation of surface waves of electrons on the nanoparticles or in a thin (

92 Nano-bio sensors for protein detection
of dyes and substantially modify their brightness and excited-state lifetime. Depending
on the fluorophores-NP distance and the NPs anisotropy one can obtain fluorescence
enhancement or quenching [9],[11]. Emission enhancement effects can be due, for metal
colloids, to local field enhancement and coupling between the molecular dipole and the
plasmonic excitations that can then radiate in the far field. Quenching, on the other
hand, is related to the nonradiative coupling of the high spatial frequency components
of the dye excitation to the plasmonic resonances that cannot radiate into the far field
[12] and dominates the dipole-metal interactions for a reciprocal distance of 2-4 nm. Few
examples of fluorescence based assays have been reported [13],[14], which exploit either
the metal-induced fluorescence enhancement [15],[7] (MEF), also due to field enhancement
on aggregated anisotropic particles [7] or the fluorescence quenching [16]-[17]. In
particular a relevant effort has been done for exploiting fluorescence quenching induced
by nonradiative resonant energy transfer to quantum dots and gold nanoparticles for detection
of proteins [18]-[19]. Plasmonic resonances depend on the difference between the
metal dielectric permittivity and that of the surface layer [12],[20]. In the same way the
energy-transfer mechanism is maximum at the frequency at which the real part of the
metal dielectric permittivity equals that of the layer on the interface [16]. Therefore, in
the case of both fluorescence enhancement and quenching we expect that any change in
the dielectric constant of the NP surface, induced by a biorecognition process that occurs
on the surface itself, can produce a change in the emission properties of the fluorophores
[15], [9].
With this putative mechanism in mind, we studied the effect of the binding of proteins
to the surface of gold NPs on the fluorescence emission properties (brightness and
excited-state lifetime) of fluorescein bound to the NP itself.
This process can be in principle applied to any protein. Due to its relevance in cancer
diagnosis, we have focused here on the p53 protein. The aim of the experiments reported
hereafter is therefore to assess the possibility to detect traces of p53 protein by exploiting
the fluorescence emission of the dye bound to the NPs.
The tumor suppressor p53 has been indicated as a tumor antigen, as an oncogene product
or as a tumor marker [21]. It is one of the most commonly mutated proteins found
in human cancers and its mutated forms may have specific functions [22],[23] alternative
to original one. This property is the main reason why to p53 have been assigned widely
different roles in the past years [24]. Wild type p53 (wt-p53) acts as a transcription
factor that activates the expression of proteins that regulate the cell cycle, either halting
the cycle until damage can be repaired, or in more extreme cases, causing the cell to

Chapter 4 93
die. Mutations in the TP53 gene are among the most common genetic alterations in
human cancer. These alterations are usually associated with accumulation of mutant
p53 forms in cancer cells, due to the longer half-life of the p53 mutants with respect to
the wtp53. Thus, the early detection of p53 in cell extracts and its quantification can be
an extremely valuable tool in cancer prevention [25]. Since antibodies for an enormous
variety of p53 mutated forms are available [23],[26], immuno-recognition is the candidate
method to accomplish this goal. The limitations of the actual detection methods
(changes in the optical response of thin gold layers [4],[9] or in the fluorescence signal on
antibody functionalized surfaces by total internal reflection methods [27], [28], or colorimetric
assays based on immuno-driven aggregation of gold colloids [29],[30]) are mainly
related to the implementation of immuno-recognition on surface immobilized molecules,
a process that is intrinsically slow, being diffusion limited. It would then be interesting
to devise all-solution methods to enhance and fasten the recognition process which occurs
on the nano-constructs and that could be used also for intracellular detection.
A hybrid nanodevice is proposed here, in which the protein detection occurs on the
surface of an organic-metal hybrid complex: a gold nanocrystal functionalized by a fluoresceine
derivative and the specific protein antibody. The detection of the protein is
based on the measure of the dye excited state lifetime during the short fluorescence
bursts that are due to the diffusion of the complexes through the excitation volume and
it appears to be fairly insensitive to the dye photo-bleaching. This construct would be
suitable for intracellular detection by choosing a capping layer that fosters the internalization
of the NPs.
First the new metal-organic system has been tested for the recognition of a protein model
system (bovine serum albumine, BSA). Then the feasibility of the detection of p53 in
solutions and in total cell extracts (TCEs) has been addessed together with the selectivity
of the proposed nano-constructs for p53 with respect to other globular proteins.
Before treating the experimental details and the results we briefly discuss the known
properties of p53 protein.
4.2 p53 protein properties
Cancer is a serious disease caused by defective control of cell proliferation. The inactivation
of tumor suppressor genes and deregulated expression of oncogenes is often the
cause of cellular transformation. The multistep process of cancer development requires
several genetic changes over a long period of time. Each cell contains the genetic information
that has to be replicated and passed to the next progeny in the process of cell
cycle. This hereditary code is, however, altered constantly due to external pressure and
the DNA in most of the cells experience numerous mutations every day. To support the

94 Nano-bio sensors for protein detection
precise genetic code from one cell generation to the next one, the cells have developed a
number of regulatory pathways to monitor the entire process. Despite the high fidelity
of this machinery, some occasional mistakes can be passed by this system and be further
transferred to the progeny. Errors in the control of the damage response may lead to the
accumulation of genetic lesions and multiple phases of clonal selection eventually results
in uncontrolled growth and predisposition to cancer.
One of the key proteins in the regulation of the genomic integrity is tumor suppressor
protein p53. p53 can prevent accumulation of harmfull mutations, inhibiting tumorpromotion.
Upon exposure to various kind of damage, p53 protein is activated and halts
the cell cycle to give the repair machinery some time to solve the errors in the hereditary
material. In case of excessive damage, however, the DNA may be in an unrepairable condition
and the cell may have to choose a cell death pathway instead of the growth arrest
to ensure maintenance of the genome. The ability of p53 to induced this programmed
cell death, apoptosis, is probably its major function in preventing the neoplastic transformation.
The early events leading to cellular p53 response are not totally understood,
even though they have been studied extensively over the past two decades. Inactivation
of the p53 pathway is very common in cancers and reactivation a potential key factor in
killing tumor cells. Although preventing the incidence of cancer by eliminating the risk
factors would probably be the most effective way of reducing the number of cancer cases,
new therapeutic possibilities are required. Activation of the p53 pathway may have an
important role in this process. Thus, knowing the factors that affect the function of p53
are critical to understand in detail.
Activities that have been attributed to p53 include: regulation of gene expression,
DNA syntheis and repair, control of DNA replication, DNA damage response and cell
cycle control. p53 acts also in cellular differentiation, senescence , inhibition of angiogenesis,
and in programmed cell death. p53 has gained special interest due to its activation
in response to cellular stress to mediate a variety of anti-proliferative processes. p53
protein is a sensor of diverse forms of stress such as genotoxic stress (UV and IR, cytotoxic
drugs, carcinogens), various non genotoxic stress (hypoxia, temperature changes,
redox changes) and oncogenic stress. Disruption of p53 function promotes checkpoint
defects, cellular immortalization, genomic instability, and appropriate survival, allowing
the continued proliferation and evolution of damaged cells.
4.2.1 P53 structure
Human p53 is a nuclear phosphoprotein of MW 53 kDa encoded by a 20 Kbp gene containing
11 exons interrupted by 10 introns, which is located in a single copy on the short

Chapter 4 95
arm of chromosome 17. This gene belongs to a highly conserved gene family containing
at least two other members, p63 and p73. Wild-type p53 protein contains 393 amino
acids and is composed of several structural and functional domains: a N-terminus containing
an amino-terminal domain (residues 1-42) and a proline rich region with multiple
copies of the PXXP sequence (residues 61-94, where X is any amino acid), a central core
domain (residues 102-292), and a C-terminal region (residues 301-393) containing an
oligomerization domain (residues 324-355), a strongly basic carboxylterminal regulatory
domain (residues 363-393), a nuclear localization signal sequence and 3 nuclear export
signal sequence.
Figure 4.1: Schematic representation of the p53 structure. p53 contains 393 amino acids, consisting if
three functional domains, i.e. an N-terminal activation domain, DNA binding domain and C-terminal
tetramerization domain. The N-terminal domain includes transactivation subdomain and a PXXP region
that is a proline-rich fragment. The central DNA binding domain is required for sequence-specific DNA
binding and amino acid residues within this domain are frequently mutated in human cancer cells and
tumor tissues. The C-terminal region is considered to perform a regulatory function. Residues on this
basic C-terminal domain undergo posttranslational modifications including phosphorylation and acetylation.
Numbers indicate residue number. NLS, nuclear localization signal sequence; NES, nuclear export
signal sequence.
The amino terminal region
The first 42 amino acids of p53 make up an acidic region called transcription-activation
domain, because this region regulates gene expression and interacts with various transcription
factor including acetyltransferases and Mdm2 (murine double minute 2, which
in humans is identified as Hdm2). Later, second transcription activation domain has been
described (between amino acids 43 and 73). p53 has a proline rich region (amino acids
63-97) necessary for transcriptional repression by p53 and is required for (transcriptionindependent)
growth suppression by p53-mediated apoptosis. The activity of p53 is
modulated by the interaction of the N-terminal part with other proteins and several
posttranslational modifications take place in this part of p53.

96 Nano-bio sensors for protein detection
The core domain
The sequence specific DNA binding domain is an antiparallel β sheet that serves as scaffold
for 2 α helix loops that physically bind DNA. The DNA-binding domain or core
domain spans form amino acid 102 to 292 and forms a separate independently folding
structure that binds to the DNA sequence specifically. In human tumors, the p53 protein
is often mutated and mutant proteins have principally lost the growth suppression functions.
95% of tumor-related mutations map to the residues of the DNA binding region
and among them certain ”hot spots” of mutations have been described. These mutations
occur as residues essential for DNA-binding and therefore inactivate the transcriptional
activation function of p53, giving an example how p53 is inactivated in tumors. The
core domain contains also region necessary for the binding of p53 with the negative regulators
of apoptosis. Altogether, these findings confirm how important this region is for
the function of p53 as a tumor suppressor.
The C-terminal region
The central core domain is connected to the C-terminal region with the flexible linker
region (amino acids 300-318).
p53 requires nuclear localization for the function, which is ensured by three nuclear localization
sequences in the C-terminal region (amino acids 316-325, 369-375, 379-384).
The p53 export from the nucleus has been shown to be mediated by two nuclear export
signal sequences located in the activation domain (amino acids 11-27) and in the
tetramerization domain (amino acis 339-352) of p53. Full length p53 protein forms stable
tetramers and the tetramerization domain including amino acids 324-355 is responsible
for the oligomerization of p53. Although monomeric p53 is able to bind DNA, activate
transcription, and suppress growth , p53 is believed to function much more efficiently as
a tetramer probably due to the greater DNA binding affinity.
Normal p53 funcion also relies on proper conformation and oligomerization of the
protein, a process that is regulated by the tetramerization domain. Two monomers associate
to form an antiparallel double stranded sheet, known as a dimer. The tetrameric
form of p53, a dimer of dimers, has enhanced ability to interact with either DNA or various
proteins. Although relatively uncommon, mutations in the tetramerization domain
(4%) prevent tetramerization, DNA binding and tumor cell growth inhibition. Finally,
the C-terminus contains 9 basic amino acids that specifically bind DNA and RNA, and
appear important in p53 regulation of DNA repair processes.
Studies of the regulation of p53 tertiary structure have provided ideas for the conformation
model for the functioning of p53. According to this model, mutations that
deregulate the normal control of p53 conformation may lead to cancer. It has been

Chapter 4 97
shown that one of the two alleles of the p53 gene is mutated in the cell, the mutant p53
protein can inactivate the wild type p53 in a dominant-negative manner. This dominantnegative
activity has been explained with the ability of the mutant p53 protein having
different conformation to form mixed tetramers with wild type p53 driving the latter into
the mutant conformation. These kinds of hetero-oligomers lack the growth suppression
function. Due to such dominant negative effect over wild type protein, the wild type p53
cannot avoid malignant growth if it co-expresses with the mutant p53 protein.
Next to the oligomerization domain is the basic region (amino acids 363-393), named
regulatory domain, which is required for regulation of the p53 activity. p53 is unusual
among transcription factors, because it has also non-specific DNA-binding ability in
addition to the sequence-specific DNA binding.
A model for the activation of p53 has been proposed according to which the C-
terminus of p53 interacts with the core domain. This interaction locks the core domain
into a conformation keeping p53 in a latent, low-affinity DNA-binding form that is inactive
for DNA binding. The core domain is able to adopt an active conformation (efficient
DNA-binding) after the modification or deletion of C-terminus, or protein-protein interaction.
The binding of ssDNA ends or short ssDNA fragments to C-terminal domain
may also stabilize p53 in a conformation, active for binding to target DNA sequences.
Furthermore, the proline-rich region at the N-terminus of p53 has been suggested to
cooperate with C-terminus for the maintenance of the latent, low affinity DNA-binding
conformation of p53. Thus, p53 requires a structural change for the activation of sequence
specific DNA binding and this occurs through boh N-terminal and the basic
C-terminal domain.
4.2.2 Control of the p53 protein half-life
The p53 protein level in the cell is determined by the rates of its synthesis and degradation.
p53 has a short half-life of only about 20 minutes in normal cells due to rapid
degradation. When active as a transcription factor, p53 encourages the synthesis of
Mdm2- the agent of its own destruction. This creates a negative-feedback loop that
usually functions to ensure that p53 molecules are degraded soon after their synthesis,
resulting in the very low steady-state levels of p53 protein observed in normal, unperturbed
cells. The operations of this p53-Mdm2 feedback loop explain a bizzarre aspect
of p53 behavior. In human cancer cells that carry mutant, defective p53 alleles, the
p53 protein is almost invariably present in high concentrations, in contrast to its virtual
absence from normal cells. At first glance, this might appear paradoxical, since high
levels of a growth-suppressing protein like p53 would seem to be incompatible with malignant
cell proliferation. The paradox is resolved by the fact, mentioned above, that the

98 Nano-bio sensors for protein detection
great majority of the mutations affecting the p53 gene cause the p53 protein to lose its
transcription-activating powers. As a direct consequence, p53 is unable to induce Mdm2
transcription and thus Mdm2 protein synthesis. In the absence of Mdm2, p53 escapes
degradation and accumulates to very high levels. This means that many types of human
cancer cells accumlate high concentrations of essentially inert p53 molecules.
4.2.3 p53 induction
Three well-studied monitoring systems have been found to send alarm signals to p53
in the event that they detect damage or signaling imbalances. The first of these responds
to double-strand breaks in chromosomal DNA, notably those that are created
by ionizing radiation such as x-rays. Indeed, a single dsDNA break occurring anywhere
in the genome seems sufficient to induce a measurable increase in p53 levels. A second
signaling pathway is activated by a wide variety of DNA damaging agents, including certain
chemotherapeutic drugs and UV radiation; certain inhibitors of protein kinases also
stimulate this pathway. A third pathway leading to p53 activation is triggered by aberrant
growth signals. The mechanisms by which other physiologic stresses or imbalances,
such as hypoxia, trigger increases of p53 levels remain poorly understood. These converging
signaling pathways reveal a profound vulnerability of the mammalian cells. The
funnelling of these diverse signals to a single protein would seem to represent an elegant
and economic design of the cellular signaling circuitry. But it also put cells at a major
disadvantage, since loss of this single protein from a cell’s regulatory circuitry results in
a catastrophic loss of the cell’s ability to monitor its own well-being and respond with
appropriate countermeasures in the event that certain operating systems malfunction.
In one stroke (actually, the two strokes that cause successive inactivation of the two p53
gene copies) , the cell becomes blind to many of its own defects. It thereby gains the
ability to continue active proliferation under circumstances that would normally cause
it to call a halt to proliferation or to enter into apoptotic death. In addition, loss of the
DNA repair and genome-stabilizing functions promoted by p53 will make descendants of
a p53 −/− cell more likely to acquire further mutations and advance more rapidly down
the road of malignancy. All three pathways inhibit the degradation of p53 protein, thus
stabilizing p53 at high concentration. The increased concentration of p53 allows the protein
to carry out its major function: to bind to particular DNA sequences and activate
the expression (transcription) of adjacent genes. These genes, directly or indirectly, lead
ultimately to cell death or the inhibition of cell division. In the event that DNA is successfully
repaired, the signals that have protected p53 from destruction will disappear.
The consequence is that the levels of p53 collapse. This allows cell cycle progression
to resume and the DNA replication to proceeds. By preventing cell cycle advance and

Chapter 4 99
DNA replication while chromosomal DNA is damaged and by inducing expression of
DNA repair enzymes, p53 can reduce the rate at which mutations accumulate in cellular
genomes. Conversely, cells that have lost p53 function may proceed to replicate
damaged, still-unrepaired DNA, and this can cause them, in turn, to exhibit relatively
mutable genomes, that is, genomes that accumulate mutations at an abnormally high
rate per cell generation.
4.2.4 P53 often ushers in the apoptotic death program
p53 can opt, under certain conditions, to provoke a response that is far more drastic
than the reversible halting of cell cycle advance. In response to massive, essentially
irreparable genomic damage, anoxia (extreme oxygen deprivation), or severe signaling
imbalances, p53 will trigger apoptosis. The cellular changes that constitute the apoptotic
program proceed according to a precisely coordinated schedule. Within minutes, patches
of the plasma membrane herniate to form structures known as blebs; indeed, in timelapse
movies, the cell surface appears to be boiling. The nucleus collapses into a dense
structure- the state termed pyknosis- and fragments as the chromosomal DNA is cleaved
into small segments. Ultimately, usually within an hour, the apoptotic cell breaks up
into small fragments, sometimes called apoptotic bodies, which are rapidly injested by
neighboring cells in the tissue or by itinerant macrophages. In the specific context of
cancer pathogenesis, the organism uses p53-triggered apoptosis as a means of weeding
out cells that have the potential to become neoplastic, including some cells that have
sustained certain types of growth-deregulating mutations and others that have suffered
widespread damage of their genome.
To summarize, the various observations cited here indicate that the biological actions
of p53 fall into 2 major categories. In certain circumstances, p53 acts in a cytostatic
fashion to halt cell cycle advance. In other situations, p53 activates a cell’s previously
latent apoptotic machinery, thereby ensuring cell death. The choice made between these
alternative modes of action seems to depend on the type of physiologic stress or genetic
damage, the severity of the stress or damage, the cell type, and the presence of other
pro and anti-apoptotic signals operating in a cell. At the biochemical level, it remains
unclear how p53 decides between imposing cell cycle arrest and triggering apoptosis.

100 Nano-bio sensors for protein detection
4.3 Experimental details
4.3.1 Sample preparation
Chemicals, Proteins, and Nanoparticles
Gold colloids 10 and 5 nm in size, functionalized by streptavidin, were purchased from
Ted Pella Inc. (Redding,CA, code 15840 and 15841) and purified by two steps of 13000
rpm centrifugation. As reported on the product data sheets, the 10 nm and 5 nm gold
NPs have an average number of 20 and 5 streptavidins and a concentration of 28 and 280
nM, respectively (stock solutions). The p53 protein, its monoclonal antibody and the
monoclonal antibody of BSA protein were obtained from Abcam (Cambridge, UK, P53,
ab43615-50, mAb to p53 (biotinylated) ab27696-100, RbpAB to BSA (BIOTIN) ab7636-
1) and the biotinylated fluorescein isothiocyanate (FITC) dye from Pierce (Rockford,
IL, code 22030). Bovine serum albumin (BSA), beta-lactoglobulin (BLG) and lysozyme
proteins were purchased from Sigma-Aldrich (codes A-0281, L-8005, L-6876) and used
without further purification. After the NP-fluorophores binding reaction, all the solutions
of gold colloids were dialized (15000 D molecular weight) against phosphate buffer in
order to get rid of the unlabeled fraction of the fluorescent molecules. The solutions
were then centrifuged twice at 13000 rpm before use. For the fluorescence experiments,
the stock solutions were diluted 1:1000 in phosphate buffer at pH 7.5.
Nanoparticle-Dye Complexes
In order to understand the changes in the fluorescence emission of the dye when coupled
to the gold NPs we devised a series of constructs all based on the strong streptavidinbiotin
interaction. The basic sensor construct (NP-FITC-Ab) is obtained by binding
biotin-FITC and the protein biotinilated antibody (Ab) to the gold NP functionalized
with streptavidin (Sav) as shown in Figure 4.2.
As reported on the product data sheets, the 10 and 5 nm gold NPs have an average
number of 20 and 5 Sav, respectively. The binding of the two components, FITC and
Ab, to the Sav-NP was performed in a single step in order to avoid saturation of the
binding sites on the NP with either the Ab or the FITC. All binding reactions were
performed overnight in the dark under mild stirring conditions in phosphate buffer. The
ratio, R F IT C = [Ab] : [F IT C], of the sites on the gold NPs occupied by the antibodies
to those occupied by FITC was tuned by changing the stoichiometric ratio of the two
components during labeling. We performed preliminary tests on various R F IT C ratios
and found that the best sensitivity in terms of the relative change in the FITC excited
state lifetime and limited reduction of the FITC fluorescence signal when bound to the
gold NP (see par. 4.9) corresponds to a ratio [Ab]:[FITC] = 3:1 (construct A in Figure

Chapter 4 101
4.2). The ratio R = [Ab]:[protein] was varied in the experiments as detailed in par. 4.9.
Figure 4.2: (A-C) Basic constructs used to investigate the possibility of fluorescence detection of proteins
at picomolar concentrations. The symbols refer to Streptavidin (Sav), gold nanoparticles (NP),
fluorescein isothiocyanate (FITC). Relative sizes are not drawn to scale. The constructs are (A) the gold
NP conjugated to the biotinilated FITC through the Sav-biotin binding (NP-FITC); (B) gold NP coupled
to the biotin-FITC and the protein antibody at a ratio [Ab]:[FITC]=3:1 (NP-FITC-Ab); (C) construct B
after reaction with the protein BSA or P53 (NP-FITC-Ab-protein).
Cell Lines, Culture Conditions and Treatments
Human HCT116 colon cancer cells (wild type p53, wtp53) and H1299 lung cancer cells
(homozygous partial deletion of the TP53 gene 1 [31]) were maintained as monolayers in
DMEM supplemented with 10% heat-inactivated fetal calf serum, 2 mM L-Glutamine,
PenicillinStreptomycin (EuroClone) at 37 ◦ C in a 5% CO 2 atmosphere. The HCT116
line expresses the wild-type p53 (wtp53) protein. The H1299 cells have a homozygous
partial deletion of the TP53 gene and as a result do not express the tumor suppressor
p53 protein, which in part accounts for their proliferative propensity [31]. For DNA
damage induction, sub-confluent cells were UV irradiated with 20 J/cm 2 for 30 seconds
and harvested 18 h later. Cell pellets were stored at -80 ◦ C until lysed for further analysis.
TCEs were obtained by incubating the cells for 30 min with lysis buffer [50 mM
Tris HCl (pH 7.5), 5 mM EDTA, 250 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate,
1 kind gift of Dr. Silvia Soddu, Experimental Oncology Department, Molecular Oncogenesis Laboratory,
Regina Elena Cancer Institute, Rome

102 Nano-bio sensors for protein detection
0.1% SDS] supplemented with a protease inhibitor mix.
4.3.2 Experimental methods
Spectral characterization
Spectrophotometric measurements were performed on a Jasco V570 spectrophotometer
(Jasco, Japan). The fluorescence emission spectra were acquired on a Varian Eclipse
spectrofluorimeter (Varian, U.K.).
Dynamic light scattering characterization
For Dynamic Light Scattering (DLS) a home-made setup [18] for variable angle measurement
of the scattered light autocorrelation function has been used with a He-Ne 30 mW
polarized laser source. The correlator board was an ISS (Urbana Champaign, IL) single
photon counting acquisition board and the data were analyzed as described below.
For the constructs based on model protein BSA, the normalized intensity autocorrelation
functions (ACFs) were fit to a multiexponential decay
G(τ) = 〈I(t + τ)I(t)〉 t
〈I〉 2
= 1 + f coh ( ∑ k
A k exp[−D k q 2 t]) 2 (4.1)
where D k is the translational diffusion coefficient of the k-th species of the NPs or NP
aggregates, q is the wave vector, q =(4πn)/(λ)sin(θ/2), n is the solution (water) index
of refraction, and λ is the laser light wavelength (633 nm). The parameter f coh is an
indication of the ratio of the detector to the coherence area and was left as a free fitting
parameter, typically in the range 0.15-0.35. The pre-exponential factors, A k , which are
proportional to the product of the square of the molecular mass M k times the number
concentration n k , can be used as an estimate of the relative concentration of the particles
according to
where V k is the hydrated volume of the k-th species.
A k ≈ n k M 2 k ≈ n kV 2
k (4.2)
The cumulant analysis of
the data was performed by fitting the ACFs (maximum lag time = 600-800 µs) to a
third-order cumulant function
G(τ) = 〈I(t + τ)I(t)〉 t
〈I〉 2 = Aexp[−2(tD cum q 2 − t2 2 (D cumq 2 ) 2 σ + t3 6 (D cumq 2 ) 3 µ 3 )] (4.3)

Chapter 4 103
The polydispersity of the samples, σ, depends on the NP construct as discussed later.
The analysis of p53 protein based constructs was performed both with cumulant
analysis and, in a more exhaustive way, with the Maximum Entropy method, through
the procedure described in the following. The second order autocorrelation functions
(ACFs) of the scattering light were first converted into the first order ACFs, G(t), and
the first order ACFs were analyzed by means of the Maximum Entropy method [32]
obtaining the distribution of relaxation times according to the relation:
∫ ∞
G(t) = A dlog(τ)P τ (log(τ))exp[−t/τ] (4.4)
−∞
The relaxation time τ is inversely proportional to the particle diffusion coefficient,
D, and the exchanged wave vector, q, by the relation τ = 1/(Dq 2 ).
D is related to the particle average hydrodynamic radius, R h , as D = K B T/(6πηR h ).
We assume here an average globular shape of the bare and the protein coated nanoparticles.
Therefore the linear relation between the relaxation time and the hydrodynamic
radius, τ = R h (6πη)/(K B T q 2 ) = R h /ρ, can be used to compute the number distribution
of particles with radius R h by fitting the P τ distributions to a sum of log-normal
functions of the type:
P R (log(R h )) = 1 τ P τ (log(τ)) (4.5)
P R (log(R h ))| Rh =ρτ = A ∑ j
α j 〈R〉 6 j exp [
− (log(R ]
h) − log(〈R〉 j
)) 2
2σj
2
(4.6)
where 〈R〉 j
and σ j are the average values of the hydrodynamic radius and the width
of the distribution component and α j is the number fraction of the j-th component
( ∑ j α j=1) in the distribution. The number distribution of size shown in section 4.7is
given by the computation of the function:
P n,R (log(R h ))| Rh =ρτ = A ∑ j
α j exp[− (log(R h) − log(〈R〉 j
)) 2
2σ 2 j
(4.7)
with the parameters obtained by the best fit of the experimental distribution of the
relaxation times.
4.4 Fluorescence Spectroscopy
The fluorescence spectroscopy and lifetime measurements were performed on a custommade
micro-spectrometer based on a Nikon TE300 microscope equipped with a mode-

104 Nano-bio sensors for protein detection
locked Ti:Sapphire laser (Tsunami, Spectra Physics, Mountain View, CA) with pulses
of 230 fs full width at half-maximum on the sample and 80 MHz repetition frequency.
In all experiments on FITC the fluorescence emission (band pass filter HQ515/30) was
primed by Two Photon Excitation (TPE) at λ exc =800 nm, in order to achieve a small
excitation volume that was typically 0.8 µm 3 , as estimated from Fluorescence Correlation
Spectroscopy measurements on reference dyes [33]. The average laser power was 40 mW
for 10 nm and 80 mW (before the entrance of the microscope) for 5 nm NP samples. The
fluorescence emission was acquired by a Single Photon Avalanche Diode (SPCM-AQR15
Perkin-Elmer, USA) whose signal was fed to a digital TimeHarp 200 (Picoquant, Berlin,
D) board. The analysis was performed either with a dedicated software written in CVI
(National Instruments, USA) or with the SymphoTime program by Picoquant in order to
compute the rate trace at the desired sampling time and the lifetime histograms. For the
lifetime histogram analysis the setup impulse response function (IRF), close to 350 ps,
was measured by collecting the signal scattered from the pure solvent. The IRF and the
decays were then deconvoluted by means of the SymphoTime program. The normalized
ACFs were acquired by a ALV5000E (ALV, Langen, D) board and analyzed by means
of the non-least square routine of the Origin 7.0 (OriginLab Inc., Northampton, MA)
software.
In the fluorescence traces, distinct photon bursts approximately 10-200 ms wide were
detected and ascribed to the passage of the NP constructs through the excitation beam
waist (fig. 4.3). The diffusion coefficient of the constructs can be evaluated as D ∼ =
ω 2 0 /8τ D, where the diffusing time, τ D , is taken here as the FWHM of the burst. The
lifetime histograms were computed on the selected bursts and, for comparison, on the
background, as detailed in the par. 4.9.
The lifetime histograms have been fitted to a multiexponential decay according to
I(t) = ∑ i
α i exp(−t/τ i ) (4.8)
where α i are the pre-exponential factors of the i-th component with lifetime τ i . The
prexponetial factors are related to the fractional intensities f i by
f i = α i τ i / ∑ i
α i τ i (4.9)
For a double-exponential decay the average lifetime can be defined as
since f 2 = 1 − f 1 .
〈τ〉 = f 1 τ 1 + (1 − f 1 )τ 2 (4.10)

Chapter 4 105
The fitting of the FCS autocorrelation function, G(t), was performed according to a
3D diffusion model with a Gaussian-Lorentzian beam profile described by the function
G(t) = 0.076
〈N〉
(
1 + t ) ( −1 ( ) )
λ
2
−0.5
t
1 + √ (4.11)
τ D 2πω0 τ D
where 〈N〉 is the average number of molecules in the excitation volume, τ D is the
diffusion time, and ω 0 is the beam waist, typically 0.67 µm.
The typical excitation
volume, at an excitation wavelength λ= 800 nm, was V exc = πω 4 0 /λ= 0.8 ± 0.1 µm3 . The
translational diffusion coefficient, D=(ω 2 0 )/(8τ D), is related, for a spherically symmetric
diffusing object, to the hydrodynamic radius by
〈R h 〉 = k BT
6πηD
(4.12)
Figure 4.3: The passage of the constructs, which recognized the protein, through the excitation beam
waist produces fluorescence burst approx 10-200 ms wide (panel A). Panel B shows a zoom of the burst in
panel A. The lifetime histogram in panel C is computed by SymphoTime sofware and analized as described
in the text.
4.5 Photon Counting Statistics
In order to estimate the particle brightness we have computed the first two moments of
the photon counting distribution according to the Photon Counting Histogram (PCH)
method [34] and its subsequent modification to Photon Counting Moment Analysis
(PCMA) [35], [36]. The photon counting was computed over 50 µs sampling time, much

106 Nano-bio sensors for protein detection
shorter than the diffusion time both of the free FITC dye and the gold NPs, and the
average number of particles per observation volume, N, and the average brightness, ɛ,
were computed according to the relations:
〈k〉 = 〈N〉 ɛ (4.13)
〈
k
2 〉 − 〈k〉
〈k〉
− 1 = 0.076ɛ (4.14)
Dead-time and afterpulsing corrections were applied as is [36].
4.6 Measurements
The aim is to exploit changes of the dye excited-state lifetime and brightness induced by
its interaction with the gold surface plasmons for detection of tiny amounts of protein in
solution under physiological conditions. The system we investigated is based on 10 and 5
nm diameter gold NPs coupled (via a biotin-streptavidin linker) to the FITC dye and to
a specific protein antibody (Figure 4.2, panels A-C). The interaction of the fluorophore
with the gold surface plasmon resonances, mainly occuring through quenching, affetcs
the excited state lifetime that is measured by fluorescence burst analysis in standard
solutions. The binding of protein to the gold NPs through antigen-antibody recognition
further modifies the dye excited-state lifetime. This change can therefore be used to
measure the protein concentration.
At first the system composed of a gold NP 5 or 10 nm in size, functionalized with
a fluoresceine derivative (FITC), the specific protein antibody and the protein (BSA or
P53) has been characterized through Dynamic Light Scattering (constructs NP-FITC-
Ab BSA -BSA and NP-FITC-Ab P 53 -P53). In particular the construct NP-FITC-Ab BSA -
BSA was analized through cumulant analysis, whereas the assay NP-FITC-Ab P 53 -P53
was also characterized in a more exhaustive way through Maximum Entropy analysis.
Then the nanodevice has been analized through Fluorescence Spectroscopy and used as a
test for the recognition of the model protein BSA. At last, the feasibility of the detection
of the p53 protein in solution and in total cell extracts (TCEs) has been addressed
together with the selectivity of the construct with respect other globular proteins.
4.6.1 Dynamic Light Scattering: cumulant analysis
In order to know the degree of aggregation of the constructs in detail, dynamic light
scattering studies have been performed.
The effective radii of the constructs can be estimated by taking into account the size of

Chapter 4 107
the streptavidin (4.5 x 4.5 x 5.0 nm),[37] of the BSA specific antibody (hydrodynamic
radius of 5.5 nm) [19], and of the BSA itself (hydrodynamic radius of 3.4 nm) [20]. Since
the Sav completely covers the gold nanoparticle surface we can estimate the NP-Sav
radius as R mon
∼ = (5 + 4.5 + 4.5)/2 ∼ = 7 and (10 + 4.5 + 4.5)/2 = 9.5 nm for the 5 and
10 nm diameter NPs, respectively. Upon antibody addition the maximum radius that
can be obtained at the highest degree of saturation is R mon = 12.5 and 15 nm for the
5 and 10 nm diameter NPs, respectively. When BSA is also added these values become
R mon = 15.9 and 18.4 nm, at most, for the 5 and 10 nm diameter NPs, respectively. We
notice that the size of the monomer NP with its Sav and antibody layer cannot be easily
obtained from DLS analysis due to the polydispersity of the solutions.
The high scattering cross-section of the gold NPs [4] allowed us to perform experiments
on the same diluted solutions (n 10nm
∼ = 100 pM; n5nm ∼ = 50 pM) used in the fluorescence
experiments. Typical ACFs of the scattered light collected from 5 and 10 nm NP-FITC-
Ab-BSA at R = 1:1 are reported in Figure 4.4. The cumulant analysis (shown in the
inset) of the ACFs according to eq 4.3 yields a polydispersity σ ∼ = 0.6 for the 5 nm
NPs and σ ∼ = 1.6 for the 10 nm NPs.
For the smaller constructs a continuous mass
distribution is present, whereas for the 10 nm NPs the contribution of larger aggregates
gives rise to a separate component in the ACF decay. The average hydrodynamic radius
obtained by the cumulant analysis is R cum = D cum /(k B T ) = 55 ± 5 and 100 ± 10 nm for
the 5 and 10 nm gold NPs solutions, respectively. The aggregation number N agg scales
according to the metal colloids fractal dimension [[38],[30]] d f =1.9 and is related to the
aggregate and monomer size by:
N agg = ( R cum
R mon
) d f
(4.15)
Therefore from the R h data reported in table 4.1 we can estimate N DLS
agg
and 25± 5 for the 5 and 10 size gold NPs, respectively (Table 4.1).
∼= 10 ± 3
NPs A 1(nm) D 1 (µm 2 /s) R h1 (nm) A 2 (nm) D 2 (µm 2 /s) R h2 (µm) R cum (nm)
5 nm 0.22±0.02 52±5 82±8 0.76±0.02 9.3±0.3 0.45±0.02 55±5
10 nm 0.30±0.05 19.5±0.7 215±15 0.59±0.07 1.4±0.1 3±0.2 100±10
Table 4.1: Analysis of the DLS ACFs. Fitting parameters of the DLS ACFs. The scattering angle was
90 ◦ , and the laser wavelength was λ=633 nm. The binding ratio was R=[Ab]:[BSA]=1:1. The ACFs
were analyzed by fitting the data to eq 4.1 with two average components. The radii R h1 and R h2 are
obtained through eq 4.12. The aggregation number is evaluated by comparing the average hydrodynamic
radius to the estimate of the monomeric gold NP (eq 4.18). The first cumulant analysis was performed
according to eq 4.3, and it provides the value of the first cumulant particle radius, R cum.
A fit of the whole ACFs decay can be performed by a double exponential analysis according to eq

108 Nano-bio sensors for protein detection
4.1, and it provides the average diffusion coefficients for each component together with their amplitudes,
summarized in Table 4.1. The faster component corresponds to an average hydrodynamic radius of
〈R h1 〉 ∼ = 82 nm for the 5 nm NPs and 〈R h1 〉 ∼ = 215 nm for the 10 nm NPs, while the slower component
corresponds to a size 〈R h2 〉 ∼ = 0.45 µm for the 5 nm NPs and 〈R h2 〉 ∼ = 3 µm for the 10 nm NPs. As derived
from eq 4.2 the molar concentration of each species is proportional to n i = (A i)/(〈R hi 〉 2d f
. Therefore,
the concentration ratio of the two populations can be estimated as n 1/n 2
∼ = 200 for the 5 nm gold NPs
and n 1/n 2
∼ = 10 4 for the 10 nm gold NPs. This finding indicates the presence of less than 1% in number
of very large aggregates. Since n 1
∼ = 100 pM, the corresponding number concentration of these large
aggregates should be n 2
∼ = 0.5-0.01 pM. It must be noted that the size derived from DLS is related to
the mass average through the relation
∑
1
〈R〉 = i ( 1 R i
n iMi 2 )
∑
i (niM (4.16)
i
2
Therefore, more massive aggregates give a larger contribution to the ACFs, thus explaining the more
marked presence of the larger aggregate component for 10 nm NPs ACFs in spite of their lower number
concentration compared to the 5 nm constructs.
Figure 4.4: ACFs of the light scattered by the 5 (open squares) and 10 nm (full squares) NP-FITC-Ab-
BSA constructs at [Ab]:[BSA]= 1:1. Lines represent best fit to a two-component decay as discussed in the
text. The inset shows the cumulant analysis expressed by a log-lin plot of the field correlation function
≈ √ G(τ) together with a third-order polynomial fit (solid lines). The symbols are as in the main panel.
The best fit polynomials are as follows: -0.25 -135t + 11243t 2 - 570842t 3 (5 nm NPs) and -0.61 -47.4t
+ 3801t 2 -129334t 3 (10 nm NPs). DLS parameters are reported in Table 4.1.
4.7 Dynamic Light Scattering of p53 construct: MEM analysis
The estimate of the average size of the antibody and the streptavidin is not straightforward. From the
pdb data bank (www.pdb.org) we evaluate radii of about 4 and 8 nm for the Streptavidin (entry 2IZJ,
http://www.rcsb.org/) and the antibody (entry 1 IGT, http://www.rcsb.org/), respectively. We then
estimate an average radius R (5) ∼ mon = 2.5 + 8 + 16 ∼ = 26.5 nm for a 5 nm diameter gold colloid and R (10) ∼ mon =

Chapter 4 109
29 nm for the 10 nm size gold colloid.
The MEM analysis of the first order scattered light ACFs (Figs. 4.5) indicates the presence of at least
two populations with widely different number concentrations and sizes (Table 4.2). For the NP-FITC-
Ab p53 construct in the absence of p53, the major population in terms of number concentration has an
average radius ∼ = 45± 20 nm, independently of the size of the NP used for the construct (Table 4.2). This
is consistent with the observation that most of the size of the NP-FITC-Ab p53 construct is due to the
streptavidins and to the antibodies. The MEM analysis indicates also the presence of tiny amounts (<
0.1% in number concentration) of larger aggregates in the sample, that correspond to sizes ≥ 500-600 nm.
If we assume that the NP-FITC-Ab p53 aggregates have a fractal shape and that the aggregation number
scales as N agg = (R ave/R mon) d f
, where d f
∼ = 1.9 is the fractal dimension of the aggregates, we can
estimate, from the average size 〈R h 〉 ∼ = 45 nm, aggregation numbers of the order of 2.8 ± 0.8 and 2.3 ±
0.8 for the 5 and the 10 nm NP constructs in the presence of p53. When the NP-FITC-Ab p53 constructs
interact with the p53, we measure, through the MEM analysis of DLS autocorrelation functions, that the
major component of the size distribution has an average hydrodynamic radius, 〈R 5nm〉 = 200± 60 nm
and 〈R 10nm〉 = 170± 70 nm, for a 1:1 stochiometric ratio between the protein and the NP-FITC-Ab p53
construct. In the computation of the aggregation number we can also take into account the small increase
of the monomer size due to the p53 binding, ∼ = 2 nm (entry 2J0Z, http://www.rcsb.org/), and evaluate
N agg= 5 ± 0.8 and 7 ± 0.6 for the 5 and 10 nm NP constructs. It should however be noted that for such
low aggregation numbers the fractal aggregation assumption may fail.
Figure 4.5: Left: the ACFs of the light scattered by solutions of 5 nm NP-FITC-Ab p53 constructs
without (filled symbols, R=∞) and with (open symbols, R=5:1) p53, together with their MEM best fit
functions (red solid lines). Right: the number distributions of the hydrodynamic radii, Rh, computed
from the best fit parameters of the relaxation time distribution functions (MEM analysis). The solid and
dashed lines refer to the R=∞ and R = 5:1 cases (NP size 5 nm), respectively. The relaxation time
distribution functions are reported in the inset with the same symbols as in panel on the left.

110 Nano-bio sensors for protein detection
10 nm 10 nm 5 nm 5 nm
Construct R=∞ R=5:1 R=∞ R=5:1
〈R h,1 〉 [nm] 130± 30 166±70 45±24 210±60
〈R h,2 〉 [nm] 1100± 400 1400±600 700±400 1200±450
α 2 (%) ∼ =0.003 0.01±0.005 ∼ =0.001 0.5±0.3
Table 4.2: MEM analysis of the photon correlation spectroscopy autocorrelation functions. The hydrodynamic
radii, 〈R h1 〉 and 〈R h2 〉, and the number fractions, α 2, were obtained by fitting the MEM
distributions of the relaxation times to Eq.4.6 as described in the text. R = [Abp53]:[p53]=∞ indicates
that no p53 was added to the sample.
4.7.1 Absorption spectra
The absorption spectrum of both the 10 and 5 nm size colloids is well superimposed on the dye absorption
and emission spectra (Figure 4.6), and therefore, we expect a substantial interaction of the plasmon
resonance with the dye transitions [34] in both cases. The absorption spectra of the 5 and 10 nm gold
NPs can be fit to a single Lorentzian peak superimposed on a smooth 1/λ 4 scattering law according to:
y = b + 2A1
π
Γ 1
(4(λ − λ 1) 2 + Γ 2 1 ) + Bscatt
λ 4 (4.17)
as shown in Figure 4.6 and Table 4.3. The plasmon peak occurs at slightly different wavelengths,
530 and 536 nm (Table 4.3) for the two size of NPs, due to the dependence of the resonance on the
particle size and the effect of the surface dielectric constant (here affected by the presence of Sav) on the
plasmon peak wavelength [39],[40].
Figure 4.6: Absorption spectra of the solutions of Sav-NP. Blue squares and green triangles refer to the
10 and 5 nm diameter gold NPs, respectively. The solid lines superimposed on the data are the best fit
functions to eq 4.17. The thick dashed and solid lines refer to the biotin-FITC absorption and emission
spectrum.

Chapter 4 111
A 1 Γ 1 (nm) λ 1 (nm) B scatt (nm −4 ) b
10 nm NPs 8.3±0.5 116±1 530±0.2 1.7 10 9 0.001
5 nm NPs 4.0±0.1 112±2 536±0.2 5.7 10 8 0.001
Table 4.3: Analysis of the Sav-NP Absorption Spectra.
4.7.2 Characterization of complexed FITC in solution
We characterized the fluorescence response of biotinilated FITC (pH ∼ = 7.5) in solution in terms of its
fluorescence brightness and excited-state lifetime. In order to reach low limit of detection values, highly
diluted (nanomolar to picomolar) solutions were investigated by performing single-particle detection.
The excitation mode used here, based on two-photon absorption, allows reducing the contribution of
the Raileigh and Raman scattering to the dye fluorescence emission. Negligible direct absorption of the
spherically symmetric gold NPs in the near-infrared (800-1000 nm) is expected for isotropic gold NPs 2
[41], though the nonlinear cross-section of fluorophores bound to the gold surface [9] may be enhanced
by the plasmon-fluorophore interaction [42]. The fluorescence traces of 10 nM solutions of FITC coupled
to biotin at pH=7.5 are relatively uniform, and the excited-state lifetime histogram can be fit to a
single-exponential decay finding τ F IT C = 3.50± 0.05 ns (computed over 10 measurements of 10 5 photons
each) as shown in Figure 4.8. The brightness (fluorescence rate per molecule) can be obtained, by FCS
measurements, from the average fluorescence rate divided by the number of molecules in the excitation
volume, ɛ = 〈F (t)〉 / 〈N〉. The diffusion time and average number of molecules per excitation volume
are derived from the fitting of the fluorescence ACFs to eq 4.11 [43],[44]. This procedure provides the
value τ D
∼ = 85 ± 15µs that corresponds to a translational diffusion coefficient D = 300 ±30µm 2 /s, in
agreement with literature results [18]. The FITC brightness at an average excitation power of 100 mW
is ɛ ∼ = 4.2± 0.9 KHz/molecule.
Hereafter the characterization of the constructs based on gold NPs decorated with the FITC dye
and the specific protein antibody is reported. The constructs have been first tested for the detection of
the model protein BSA and then it has been applied to the recognition of the p53 protein.
4.8 Detection of the model protein BSA
4.8.1 Gold NP-FITC complexes in the absence of BSA
When the biotin-FITC is bound to the gold NP (NP-FITC-Ab, Figure 4.2) the fluorescence fluctuations
correlate over relaxation times systematically longer than that of free FITC diffusion (Figure 4.7) that
correspond to the best fit values D=9 ± 4 µm 2 /s (5 nm gold NP constructs) and 11 ± 4 µm 2 /s (10
nm gold NP constructs). In the case of monomeric NP constructs the diffusion coefficients should be
D 5nm
∼ = 13.3µm 2 /s and D 10nm
∼ = 11.5µm 2 /s.
3 If we additionally take into account some aggregation
(N agg
∼ = 10) as suggested by DLS data, we can estimate the average aggregation number from the average
size by assuming a fractal shape for the NP aggregates 4 . The aggregation number N agg scales according
2 This is not true for non spherically symmetric gold NPs, see chapter..
3 The dimensions of the NPs, FITC and BSA protein are reported in section 4.6.1.
4 AFM data (not shown) substantially confirm this hypothesis

112 Nano-bio sensors for protein detection
to the metal colloids fractal dimension [[38],[30]] d f =1.9 and is related to the aggregate and monomer
size by:
N agg = ( Rave
R mon
) d f
(4.18)
According to the relation D ∼ = (k BT )/(6πηR ave) = (k BT )/(6πηR mon)(Nagg) −1/d f
, we obtain
〈D 5nm〉 agg
∼ = 4.2µm 2 /s and 〈D 10nm〉 agg
∼ = 3.7µm 2 /s. The ACFs fitting (Figure 4.7) is then compatible
with a distribution of sizes ranging from the monomer to small aggregates with N agg = 10-15.
The histograms of the excited-state lifetime of NP-FITC-Ab constructs (Figure 4.8) can be fit by two
exponential decays (see Table 4.4) where the weight of the shortest component (0.6-0.8 ns) is f 2 = 4−6%.
The corresponding average lifetime values are 〈τ〉 ∼ = 3.1 ± 0.3 and 3.2± 0.3 ns for FITC complexed to
the 5 and 10 nm gold NPs, respectively.
Construct τ 1 (ns) f 2 τ 2 (ns) 〈τ〉 (ns)
Biotin-FITC 3.50± 0.05
NP-FITC-Ab (10 nm) 3.3± 0.1 0.06±0.04 0.8±0.1 3.2±0.3
NP-FITC-Ab (5 nm) 3.2± 0.1 0.04±0.02 0.6±0.1 3.1±0.3
Table 4.4: Excited-State Lifetime Values for the FITC and NP-FITC-Ab Constructs.
Figure 4.7: (Normalized ACFs of the fluorescence fluctuations measured on a diluted solution of biotin-
FITC (open squares) and NP-FITC-Ab constructs for the 5 (open triangles) and 10 nm (filled squares)
size in PBS pH 7.5. The lines are the best fit to eq 4.11 with D=300 (dotted line), 13 (dashed line), 11
µm 2 /s (line). In the inset the fluorescence traes are shown.

Chapter 4 113
Figure 4.8: (Left: Fluorescence traces of the biotin-FITC (upper trace) and NP-FITC-Ab for the 5
nm NP (lower trace)). Right: Excited-state lifetime decays of the biotin-FITC (upper curve) and NP-
FITC-Ab for the 5 nm NP (lower curve). Solid lines represent fits obtained with eq 4.8, and the fitting
parameters are reported in Table 4.4.
4.8.2 Protein interactions with the FITC gold NP complexes
We investigated the effect of the binding of BSA to the gold NPs through its specific antibody (Ab) on
the FITC lifetime and brightness. These parameters of FITC were monitored as a function of the ratio
R = [Ab]:[BSA], in the range 5:1 to 1:1, and with an average concentration of antibodies ∼ = 1500 and
180 pM for the 10 and 5 nm gold NP, respectively. The concentration of protein was varied in the range
60-1500 pM, and the ratio R F IT C between the antibodies and the biotinilated FITC dyes on the gold
NPs was kept at 3:1 in all experiments reported hereafter unless otherwise explicitly stated.
4.9 Fluorescence Spectroscopy: Burst analysis
4.9.1 Aggregates size estimate
The time fluorescence traces of (NP-FITC-Ab-BSA) (construct C in Figure 4.2) are shown in figure 4.9
and 4.10 for the 5 and 10 nm NPs at the maximum [Ab]:[BSA] ratio R = 1:1. Several fluorescence
bursts are evident. The average width of the fluorescence bursts during the BSA titration experiments
can give information on the average aggregate size, and from each burst, the number of photons emitted
and the lifetime histogram can be computed. For the maximum ratio R = 1:1 typical burst width
values are ≈ 150 and 90-100 ms for the 10 (under 40 mW excitation power) and 5 nm (under 100 mW
excitation power) gold NPs, respectively. The diffusion coefficients for NPs in the presence of BSA are
D 5nm
∼ = 13.3µm 2 /s and D 10nm
∼ = 11.5µm 2 /s. These values imply average diffusion times of the NPs
through the excitation volume, V exc
∼ = 0.8µm 3 , of approximately τ D,5nm= 4.2 ms and τ D,10nm 4.9 ms,
much smaller than the observed value of the burst width ∼ = 100-60 ms. If we additionally take into account
the aggregation number (Table 3) we compute values of the diffusion coefficients 〈D 5nm〉 ∼
agg = 4.0µm 2 /s
and 〈D 10nm〉 ∼
agg = 2.3µm 2 /s, yielding diffusion times τ D,5nm = 13.4± 3 ms and τ D,10nm = 25± 4 ms, still
smaller than the experimental data. However, the burst width depends on the excitation power, as shown
in Figure 4.9, suggesting the occurrence of optical trapping for the gold NPs. In fact, by performing
experiments at decreasing laser excitation power on the 5 nm gold NPs constructs we found that the
average burst width decreases linearly with the excitation power (inset of Figure 4.9) and reaches, in
the limit of vanishing excitation power, a value of the width ∼ = 18 ± 3 ms, in good agreement with the

114 Nano-bio sensors for protein detection
theoretical estimate made above for an average aggregate size of approximately 10 NPs.
A further indication in this sense comes from measurements of the recurrence frequency of the fluorescence
bursts, γ R
∼ = 0.1± 0.06 Hz (Figure 4.10). A similar value can be recovered by employing the expression
γ R
∼ = 4πω0((n)/(N agg))((k BT )/(6πηR mon))(N agg) −1/d f
, with a number concentration n ∼ = 70-100 pM,
a laser beam waist w 0 = 0.67 µm, and an aggregation number N agg = 10-20. Therefore, DLS and burst
width analysis indicate that, in the presence of BSA, the aggregation is similar to that found for the NP-
FITC-Ab constructs, as probed also by FCS. Besides the burst width we can also consider the average
value of the total number of photons collected per burst, 〈N BP 〉. In the case of 5 nm NP constructs,
we obtain an increase of the average value of N BP when raising the BSA concentration. The value
〈N BP 〉 ∼ = 3000± 300 photons/burst found for the ratio R=[Ab]:[BSA]= 5:1 increases to 〈N BP 〉 ∼ = 7200±
900 photons/burst for the ratio R=[Ab]:[BSA]=1:1 (100 mW of laser power). The value of 〈N BP 〉
for the 10 nm gold NPs, on the contrary, is almost independent of the stoichiometric ratio R giving
〈N BP 〉 ∼ = 6350± 500 photons/burst at a power of 40 mW.
Figure 4.9: (Zoom of fluorescence bursts in a time trace for the NP-FITC-Ab-BSA constructs at R=1:1,
excitation power P=100 (B1) and 40 mW (B2). (Inset) Power dependence of the peak FWHM width for
the same constructs.
4.9.2 Excited state lifetime in the presence of BSA
When BSA is added to the NP-FITC-Ab constructs, sharp burst are observed in the fluorescence traces
(Figure 4.10, red boxes), indicating that this protein is recognized by the constructs. The excited-state
lifetime of FITC was then measured on these fluorescence bursts. We checked that the presence of the
bursts should not be ascribed to photons coming from scattering bleeding through the green band-pass
filter since control experiments, performed on unlabeled gold NPs observed through the same emission
filter (band-pass 535/50 nm), have not shown any fluorescence burst (data not shown). The excited-state
fluorescence decay appears to be affected by the protein binding to the NP-FITC-Ab constructs, as seen
by comparing Figure 4.8 and Figure 4.10(right panels).
Upon BSA addition the fractional intensity of the faster component of the lifetime values derived from
the histograms calculated on the bursts increases to f 2 = 30% depending on the [Ab]:[BSA] ratio. On the
other hand, the lifetime values computed on the histograms obtained from the fluorescence background
signal detected in between the bursts are close to those obtained without BSA. The lifetime decay has

Chapter 4 115
been systematically investigated for the NP-FITC-Ab constructs based on the 5 and 10 nm gold NPs as
a function of the BSA concentration in standard buffer solutions, and the average time distributions are
shown in Figure 4.11. For comparison, the distribution in the absence of protein has been reported on
the same plot.
Figure 4.10: Left: Fluorescence traces of the NP-FITC-Ab-BSA constructs at a molar ratio
R=[Ab]:[BSA]=1:1 for the 10 (upper trace, 40 mW excitation power) and 5 nm (lower trace, 100 mW
excitation power) NPs. Right: Excited-state lifetime decays calculated on the bursts of NP-FITCAb-BSA
5 nm NPs at R=1:1, bottom curve, and 5:1, middle curve, and calculated on the background (top curve).
Figure 4.11: Distribution of FITC average excited-state lifetimes for the NP-FITC-Ab-BSA based on
the 5 (A) and 10 nm (B) gold NPs (at the ratio [Ab]:[FITC]=3:1). The solid lines correspond to the
Gaussian best fit of the histograms. The stoichiometric ratio R=[Ab]:[BSA] is indicated in the figures.
The distribution center shifts toward shorter average lifetimes at increasing BSA concentration for
both construct sizes due to the increase in the fractional intensity of the shorter component. The result
of the data analysis reported in Table 4.5 and figure 4.12 indicates that the 5 nm gold NP construct
displays larger sensitivity to the protein-antibody binding than the 10 nm constructs. The change of the
protein concentration from ≈ 40 to ≈ 180 pM induces a decrease in the mean excited-state lifetime for
the 5 nm gold NP constructs from 2.7 ± 0.1 to 0.9 ± 0.2 ns, mostly due to metal induced quenching

116 Nano-bio sensors for protein detection
of FITC dyes. These are in fact bound to an estimated distance from the gold surface of the order of
3-4 nm and therefore lie in a region where the quenching mechanism due to nonradiative energy transfer
is dominant though metal-enhanced fluorescence (MEF) may begin to play a role [45],[46] as indicated
by the slight anticorrelation between 〈N P B〉 and 〈τ〉 (Figure 4.13), which can be considered as a MEF
fingerprint.
For the 10 nm sized NPs, contrary to the 5 nm constructs, the average number of photons emitted
does not show appreciable changes, in agreement with the constant value found for the mean lifetime
that reaches a ’saturation’ value of 〈τ〉 = 2.3 ± 0.2 ns, already at the lowest protein to antibody ratio
explored (Table 4.5). It should be noticed that a change in the lifetime is indeed occurring also for
the 10 nm constructs upon BSA binding for [BSA] < 100-150 pM though with a reduced sensitivity
since the maximum change in the lifetime is ca. -30% compared to the ca. -70% decrease found for
the 5 nm constructs. Therefore, the parameter that appears to be the most robust and sensitive to the
protein concentration is the average lifetime of FITC on the 5 nm constructs that is linearly dependent
on the BSA concentration in solutions (Figure 4.12). This behavior can be exploited as a calibration
for detection of unknown BSA concentration in solution. The experimental uncertainty of 0.1 ns on
the mean lifetime limits the sensitivity of the technique to a protein concentration of ≈ 5 pM, thereby
allowing detection of traces of protein in solution.
〈τ〉 (ns) 〈N BP 〉 R=[Ab]:[BSA] [BSA] (pM)
10 nm diameter
2.3±0.1 6700± 400 1:1 1500
2.3±0.1 8400± 1200 2:1 1000
2.5±0.2 6050±350 3:1 500
5 nm diameter
0.9±0.2 7200± 900 1:1 180
2.3±0.1 7200± 1000 3:1 60
2.7±0.1 3000±300 5:1 36
Table 4.5: Excited-State Lifetime Values for the NP-FITC-Ab-BSA Constructs. Effect of the binding
of BSA to NP-FITC-Ab on the excited-state lifetime of FITC. The data reported refer to the constructs
C in Figure 1 with [FITC]:[Ab]=1:3. The binding of BSA is performed at different stoichiometric ratios
R=[Ab]:[BSA]. The reported excited-state lifetimes of FITC are the mean of the average lifetime
distributions.

Chapter 4 117
Figure 4.12: Left: Mean lifetime of FITC in NP-FITC-Ab-BSA constructs for 5 nm NPs versus
the concentration of protein in solution. The solid line is a linear fit of the data leading to 〈τ〉 =
3.2(±0.1) − 0.026(±0.002)nspM −1 * [BSA]. The point at [BSA]=0 corresponds to the value measured
on the NP-FITC-Ab construct (Table 4.4). Right: Mean lifetime for the 10 nm constructs versus BSA
concentration.
Figure 4.13: Average number of fluorescence photons per burst versus the average excited-state lifetime
of NP-FITC-Ab-BSA constructs for the 5 (A) and 10 nm (B) NPs. The solid line is a linear fit to the
data. The stoichiometric ratio R=[Ab]:[BSA] is indicated in the figures.

118 Nano-bio sensors for protein detection
A possible rationale for the observed phenomenon, i.e., the marked sensitivity of the FITC lifetime
on the BSA bound to the 5 nm constructs (larger than that found for the 10 nm constructs), can
be drawn starting from a few considerations of the metal-dipole interaction theory [45],[46],[47] and the
experimental observations reported in the literature [48],[11]. The interaction between a metal surface and
a molecular dipole is a complex interplay of different phenomena. First, we recognize the enhancement of
the absorption and emission due to the increase in the local field, related to the field reflection on the gold
surface [49], and to the surface roughness.[50] In this case we expect a substantial increase in the emission
rate with a limited change in the excited-state lifetimes. Actually a slight increase of the lifetime with
the amplitude of the surface electric field has been reported [48]. Second, metal-dye interactions also
include quenching of the molecular emission due to dipole energy transfer to the surface plasmons and
electron-hole couples within the metal (nonlocal effects) which results in a decrease of the dye lifetime
and brightness. Third, the dye emission may be enhanced due to the radiation in the far field of a fraction
of the energy transferred to the surface plasmons, basically that corresponding to high spatial frequencies
in the near field emission of the molecule.[45] In this case we expect to observe an anticorrelation of the
excited-state lifetime and the number of emitted photons. In those cases when the chromophores and
gold NPs are separated by bulky spacers, such as antibodies that correspond approximately to a distance
> 7 nm, the fluorescence is actually less quenched [51]. Quenching is dominating for distances up to a
few nanometers [52],[11],[49], and it is largely determined by the shape of the metal structure, by the
dipole orientation with respect to the surface, by the size of the metal particle [46] and by the difference
between the dielectric permittivities of the metal and the surface layer [45],[53]. Since in our case the
distance between the metal and FITC is only ca. 3-4 nm, we believe that the main result reported here,
namely, the change in the FITC lifetime upon BSA-NP binding, is mostly due to a change of quenching
efficiency rather than to an effective emission enhancement. To partially support this hypothesis, we can
bring the observation of the initial decrease in the lifetime of FITC upon binding to the gold surface
(Table 4.4) and of the additional larger decrease induced by the BSA binding (Table 4.5). The tiny
increase of the emitted photons per burst, 〈N BP 〉, at rising BSA concentrations (Figure 4.13), on the
other hand, can be taken as an indication of the presence of high (and heterogeneous) local electric fields
on the surface of the nanoclusters that produces a concomitant increase in the molecular brightness.
The gold-induced quenching of FITC is directly related to the Fresnel coefficients at the metal-surface
boundary [45]. Their change upon binding of proteins to the surface is determined then by the change
in the surface layer dielectric permittivity related to the protein relative concentration on the surface
layer. For this reason, we tried to keep larger the concentration of BSA on the surface, while keeping
the FITC signal per gold NP cluster at measurable levels. This reasoning has also driven our choice
of the ratio [Ab]:[FITC]= 3:1 reported above. As cited in the section 4.3.1, the choice of the ratio
[Ab]:[FITC] = 3:1 has provided us with the better sensitivity. FITC bound to constructs prepared at the
lower value of [Ab]:[FITC] )= 1:1 displays a reduced average excited-state lifetime already at [BSA]:[Ab]
= 0 with respect to the case [Ab]:[FITC] = 3:1. For example, we found 〈τ〉 = 2.1± 0.3 ns for the
10 nm constructs prepared at [Ab]:[FITC]= 1:1, more than 30% lower than the [Ab]:[FITC]=3:1 case.
Moreover, upon BSA addition the FITC average lifetime increases in the [Ab]:[FITC] = 1:1 constructs
up to 30% but with very large uncertainty (± 22%). This behavior, which makes the construct prepared
at [Ab]:[FITC] = 1:1 less adequate than the case [Ab]:[FITC] = 3:1 for protein assay applications, is due
to the interplay of the different mechanisms that have been discussed above. In particular, we find that
the increase in the FITC lifetime measured in the [Ab]:[FITC] = 1:1 case is due to a marked increase of
the long lifetime component which is also affected by a large variability (data not shown). This behavior
is probably related to the increase in the local surface field [48] and its inhomogeneity on the surface of
the larger gold NP nanoclusters [54] present in the 10 nm gold-dye constructs. These issues may be also

Chapter 4 119
at the basis of the observed reduced sensitivity of the 10 nm gold NPs constructs with respect to the 5
nm constructs. In fact, small colloids are expected to quench fluorescence more efficiently than larger
ones since the absorption component of the extinction coefficient is dominant over the scattering one,
which is responsible for the plasmon-induced fluorescence enhancement [46],[15] and this would result in
a reduced lifetime. On the other hand, the local field enhancement is expected to be larger and more
inhomogeneous on larger, possibly aggregated, structures, such as those expected for the 10 nm NPs
constructs, and this would result in an increase of the FITC lifetime [48] for these larger constructs. The
fine balancing of these two opposite behaviors would then determine the reduced sensitivity of the 10
nm particles constructs compared to those based on the 5 nm gold NPs described here.
4.10 Basic gold nanocrystal protein sensor
Up to this stage of the project, we reported a detailed analysis of the effect of binding the FITC dye
to gold NPs of size in the 5-10 nm range and explored the possibility of using the fluorescence emission
of FITC bound to their surface in order to monitor traces of proteins in standard solutions at pH = 7.
The data presented and discussed here indicate that the FITC excited-state lifetime is a very sensitive
parameter in order to detect tiny amounts of protein in solution with an estimated limit of detection of ca.
5 pM, mostly determined by the statistical accuracy of the lifetime measurement. This value, compatible
with the sensitivity of most nanotechnology-based assays, encourages application of this method to other
protein systems.
For a direct application of the devised sensor to a protein detection in cellular extracts we should also
evaluate its specificity. We expect that the degree of specificity is largely determined by the aspecific
protein binding to the gold NP surface with respect to the antibody-protein recognition. It is clear that
a better understanding of the fine balance of metal-induced quenching and enhancement that occurs on
constructs of different sizes and extent of aggregation would be invaluable for the possibility to design
specific protein or DNA assays.
The next steps will then be to devise a sensor sensitive to p53 and to test its specificity against this
protein. We first investigate the possibility to sense tiny amount of p53 proteins in buffer solutions and
move then to experiments in Total Cell Extracts (TCEs). Finally we test specificity against several
globular proteins beside BSA. In fact BSA sensing and specificity may be affected by the high stickyness
of all serum proteins.
4.11 p53 protein detection
A modification of the protein sensor construct, called hereafter NP-FITC-Ab p53, has been created on a
gold nanoparticle on whose surface we have bound the specific anti-p53 antibody and the FITC dye with
R F IT C= [Ab]:[FITC]=3:1 (Figura 4.2)
In the fluorescence assay presented here picomolar concentration of the constructs was used and
therefore we could discern distinct fluorescence bursts on a much lower background. The diffusion time
of the construct through the laser beam waist fall in the tens of millisecond range allowing us to collect
several thousands of photons per burst, enough to estimate the lifetime of the fluorophore with a few
percent of uncertainty on a single burst. This is actually found in the fluorescence traces acquired on the
NP-FITC-Ab p53 constructs as shown in Figure 4.14 (inset of panel B). The bursts that occur as multiple
peaks were analyzed by multi-component Gaussian fit and the most likely value of their FWHM, assumed
here as the diffusion time, was 〈τ D〉=60 ± 10 ms. These values correspond to an average hydrodynamic

120 Nano-bio sensors for protein detection
radius 〈R h 〉 = 230± 50 nm and to a most likely value R h
∼ = 170 nm, substantially larger than the
estimated monomer radius, indicating that some degree of aggregation must be present at least when
the protein is interacting with the gold constructs.
4.12 Fluorescence burst analysis in solution
The FITC fluorescence emission can be characterized on the fluorescence bursts in terms of the particle
brightness and the fluorophore lifetime. As described in the section 4.3.1, the ratio R F IT C between the
Abs and the FITC dyes on the gold NPs was kept constant at 3:1 in all experiments, with an average
constant concentration of Ab ∼ = 1000 and 510 pM for the 5 and 10 nm NP, respectively. The fluorescence
brightness can be computed by means of PCH methods [35], [36] and the average lifetime is evaluated
through the analysis in terms of one or two of exponential components of the FITC fluorescence decay.
The size of the constructs was further characterized by the average burst width, ∆t, analyzed as a
Gaussian function (Figs. 4.14).
Figure 4.14: Fluorescence traces collected from R = 1:1 solutions of p53 and 5 nm (panel A) or 10 nm
(panel B) NP-FITC-Ab p53 constructs. Inset of panel A reports the average particle brightness measured
on the bursts for the 5 (open squares) and the 10 nm (filled squares) NP-FITC-Ab p53 constructs. Inset
of panel B reports the distribution of the bursts full width measured through a multi-component Gaussian
fit to the bursts.
It is important to notice that when no protein is added to the NP-FITC-Ab p53 constructs solutions,
rare if any fluorescence bursts were detected (figure 4.15A) and the lifetime of FITC was determined on
the background that we ascribed to the NP-FITC-Ab p53 constructs. The average fluorescence brightness
is in this case 4.4 ± 1.7 photons/ms. The average brightness of the NP-FITC-Ab p53 constructs in the

Chapter 4 121
presence of p53, computed by PCH method on the fluorescence bursts, increases to ∼ = 50 photons/ms
(Fig. 4.14(A), inset), that would correspond to approximately an aggregate of 8 to 10 NPs. This result
is in qualitative agreement with the light scattering analysis of the size of the NP constructs reported in
section 4.7. The uncertainty in the size measurement and therefore aggregation number, does not allow
to draw quantitative conclusions on any possible change in the FITC molecular brightness upon protein
binding to the NP surface.
Figure 4.15: A: Fluorescence traces of the FITC solution. Panels B and C show the fluorescence traces
of the constructs NP-FITC-Ab for the 5 and 10 nm size nanoparticles respectively. D: Fluorescence trace
in case protein p53 is added to the constructs NP-FITC-Ab for 5 nm size nanoparticles. The lifetime
decays are determined on the area selected: 〈τ〉 F IT C
is computed on the green, 〈τ〉 F IT C−NP −Ab
on the
red and 〈τ〉 F IT C−NP −Ab−p53
on the blue area.
We have then analyzed the effect of the protein-antibody recognition on the value of the average
FITC lifetime measured on the fluorescence bursts. In the control case in which no protein is added to the
solutions the fluorescence decay is well described by a single exponential component (Fig. 4.16(A)) and we
measure τ=3.4 ± 0.2 and 3.5 ± 0.2 ns for the 5 and 10 nm NP-FITC-Ab p53 constructs, respectively. This
value is close to that found in solution and reported above, for the biotynilated form of FITC, τ= 3.5 ±
0.05 ns (Table 4.6). When p53 is added to the solutions of NP-FITC-Ab p53 constructs, the FITC lifetime
decreases substantially. Actually, when fitting the histogram decay, a faster component in addition to
the ∼ = 3.5 ns one is required (Fig. 4.16 (B)). Since the fitting procedure involves the deconvolution of
the decay with the system IRF and is therefore affected by some uncertainty, we describe the overall
fluorescence decay by the average lifetime 〈τ〉, as defined in Eq. 4.10. The average of 〈τ〉 over a sample
of fluorescence bursts will be indicated as 〈τ〉.
Construct < τ > (ns)
Biotin-FITC (uncomplexed) 3.50± 0.05
NP-FITC-Ab p53 (10 nm) 3.5± 0.2
NP-FITC-Ab p53 (5 nm) 3.4± 0.2
Table 4.6: Excited-state lifetime values for the FITC and NP-FITCAb p53 constructs.

122 Nano-bio sensors for protein detection
4.13 Dependence of the FITC Lifetime on the p53 Concentration
The effect of p53 concentration on the FITC lifetime has been investigated by changing the ratio R
=[Ab p53]/[p53] in the range 5:1 to 1:2 (figure 4.17). The protein concentration for these experiments
was varied in the range 180-1000 pM for 5 nm NPs and 90-510 pM for 10 nm NPs. The result of the
data analysis reported in Figures 4.16(C, D) indicates that both constructs display large sensitivity to
proteinantibody binding and that a range of protein concentrations, up to 200-250 pM, can be selected
in which the FITC lifetime varies linearly with the protein concentration. The sensitivity of the p53
detection is related to the slope of the linear dependence of 〈τ〉 upon the p53 concentration, that is
and
− δ 〈τ〉 /δ[p53] = 0.0014 ± 10 −4 [ns/pM] (4.19)
− δ 〈τ〉 /δ[p53] = 0.005 ± 0.001[ns/pM] (4.20)
for the 5 and the 10 nm NP-FITC-Ab p53 constructs, respectively. The 10 nm NP constructs appear
therefore to be at least five times more sensitive to the p53 detection.
Figure 4.16: Panel (A): lifetime decays of the NP-FITC-Ab p53 construct for the 5 nm NPs in the
absence of p53. The solid line corresponds to the best fit of the data to a single exponential decay, τ= 3.5
ns. Panel (B): lifetime decays of the NP-FITC-Ab p53 construct for the 5 nm NPs in the presence of p53
at a stochiometric ratio R = 1:1 (upper curve, open symbols) and R = 5:1 (lower curve, filled symbols).
The solid lines represent the double exponential decay fit of the data. The fit to the R = 1:1 data gives an
average lifetime 〈τ〉= 2.2 ns. Panels (C) and (D) report the trend of the average lifetime as a function
of the p53 concentration and refer to the 5 nm and 10 nm constructs, respectively. The solid lines are
the linear best fit of the data.

Chapter 4 123
Figure 4.17: The figure reports the trend of the average lifetime as a function of the p53 concentration
and refer to the 5 nm (red symbols) and 10 nm (blue symbols) constructs, respectively.
As discussed above, due to the uncertainty on the measurement of the FITC molecular brightness
and the degree of aggregation of the NP constructs, it is not clear whether the mechanism leading to
the observed decrease of the excited state lifetime is related to fluorescence enhancement or rather to
aggregation induced quenching. The signature of such a mechanism, related to the metal-mediated
increase of the radiative rate of the dye, would be a parallel increase in the molecular brightness and
decrease of the lifetime of FITC. As for the data presented here, the molecular brightness is not decreasing
upon the dye binding to the metal surface, though we are not in the position to obtain a quantitative
evaluation of the possible, metal induced, brightness increase. The presence of two decay components
seem to be related to a tight interaction of the dye with the metal or with the streptavidin and antibody
layer, with a consequent perturbation of the electronic levels of the dye. These are indeed affected by
the subsequent binding of the protein. In fact, the relative fraction of the faster component show only a
slight increase when R decreases and the observed decrease in the average lifetime is closely related the
decrease of both the relaxation times (Fig.4.18).

124 Nano-bio sensors for protein detection
Figure 4.18: Analysis of the two relaxation components in the fluorescence decay of FITC dyes as
a function of the protein concentration in solution. Panels A and B report the lifetimes of the two
components with which the fluorescence decay has been analyzed (open symbols τ 1, filled symbols τ 2) for
the 5 nm and 10 nm gold NP constructs respectively. Panel C reports the distribution of the fraction,f2, of
the faster component (5nm NP construct) for the stechiometric ratio R=1:1 (sparse pattern) and R=5:1
(dense pattern). Panel D reports the average value of the faster component fraction, f2, as a function of
the protein concentration. The solid lines in all the panels are the best linear fit to the data.
4.14 In Vitro Selectivity of the p53 Assay
In order to verify the possible use of the proposed p53 assay for in-vivo screening, we must check first for
possible false positive results induced by recognition of other proteins present in cell extracts. Regarding
the issue of the protein selectivity, we have tested the 5 nm NP-FITC-Ab p53 constructs for recognition
of small globular proteins that may compete with p53 by specific or aspecific binding to the NPs.
p53 is a mainly nuclear protein that can be also found in the cytoplasm [55]. Therefore the possible
competitiveness of serum proteins with p53 should not hinder the application of this NP based p53 assay
to in vivo tests. We have then tested the competitive binding of BLG and lysozyme, used as here as
reference globular proteins, for the p53 antibody on the NPs. As a reference for the serum proteins we
have taken BSA. Indeed when we add BSA to the NP-FITC-Ab p53 constructs we observe large and wide
( ∼ = 3 s) bursts on which sharper bursts are superimposed (Fig. 4.19 (upper panel)). This behavior can
be taken as an indication of a massive aggregation of the constructs. However, the average FITC lifetime
computed on these bursts is definitely smaller (〈τ〉 ∼ = 2.7± 0.2 ns) than the reference value of τ ∼ = 3.5
ns (Fig. 4.19 (upper panel)). BSA can therefore interfere with p53 detection by the NP-FITC-Ab p53
construct. On the contrary, we were not able to observe fluorescence bursts in the case of lysozyme and
BLG over repeated 30 s long traces (Figs. 4.19 (middle and lower panel)), indicating that there is little or
no recognition of these proteins by the NP-FITC-Ab p53 construct. Moreover, from the single exponential
analysis of the fluorescence decays (Figs. 4.19 (middle and lower panel)), we obtained an average FITC
lifetime τ = 3.4± 0.2 ns for the case of both BLG and lysozyme. These observations indicate that, apart

Chapter 4 125
from probably aspecific binding of BSA, the NP-FITC-Ab p53 constructs are highly specific for p53 in
vitro.
Figure 4.19: Fluorescence time decay of FITC bound to the NP-FITC-Ab p53 constructs (5 nm in size)
in the presence of BSA , lysozyme and BLG. All proteins were added to the solution in order to obtain
R=1:1. The laser excitation power was 40 mW. The panels on the right report exemplary fluorescence
traces for the three proteins.
4.15 In Vivo Test of the p53 Assay
The ability of the NP-FITC-Ab p53 constructs to recognize p53 in vivo has been tested on extracts from
HC116 (p53 positive) and H1299 (p53-null) cell lines. We have first checked the pure TCEs for fluorescence
emission at 515 nm finding an average emission of 7.5 ± 3 photons/ms and a single exponential
decay of the fluorescence with average time τ = 2.6± 0.1 ns (Fig. 4.20). Also in this case we detected
no evident burst over 120 s of measurement (Fig. 4.20 (inset)). This signal is probably due to the
auto-fluorescence of the protein bound coenzyme NADH present in the TCEs [56], [57]. When we mixed
the p53-null TCEs (p53 −/− ) with the NP-FITC-Ab p53 constructs at a 1:1 volume ratio, we found rare
bursts with width ∆t = 200 ms or larger (Fig. 4.21 A, filled triangles). The distribution of the excited
state lifetime indicates the presence of two populations with 〈τ〉 ∼ = 3.2± 0.2 ns and 〈τ〉 ∼ = 2.3± 0.15 ns
(Fig. 4.21 A, patterned bars) with a clear anticorrelation to the burst size (Fig. 4.21 A). We ascribe
the two lifetime populations to the emission of FITC bound to the NP-FITC-Ab p53 constructs that have
not recognized p53 proteins (τ ∼ = 3.1 ns, as found in the in-vitro experiments) and to the cell proteins
auto-fluorescence (τ ∼ = 2.6 ns). When monitoring the FITC lifetime in a 1:1 (volume ratio) solution of
p53 +/+ TCEs and NP-FITC-Ab p53 constructs (Fig.4.21 B), sharp and large bursts appear in the fluorescence
traces showed in the inset of Figure 4.20 together with an example of the fluorescence decay.
We detect a population of fluorescence bursts characterized by even shorter lifetime values, τ ∼ = 1 ns,
that we ascribe to the complexes of p53 with NP-FITC-Ab p53 nano-constructs. Actually, the analysis of
the lifetime histogram in terms of a sum of Gaussian functions (Fig.4.21 B, histograms) corresponds to

126 Nano-bio sensors for protein detection
the three values 〈τ〉 = 1.0± 0.3 ns, 2.0 ± 0.3 ns and 2.7 ± 0.3 ns. The two longest ones agree quite well
with the values obtained in the p53 −/− TCEs (Fig.4.21 A, patterned bars). This conclusion is supported
by the control experiments performed by adding free p53 protein to the solution of p53 −/− TCEs in the
presence of NP-FITC-Ab p53 gold constructs. Also in this case we observed sharp and intense fluorescence
peaks (Fig. 4.20). As seen in Fig.4.21 A (open symbols), we find three components at 〈τ〉 = 1.1± 0.2
ns, 1.7 ± 0.5 ns and 3.2 ± 0.1 ns (Fig.4.21 A, dashed curves, white bars).
Regarding the possibility to develop a quantitative assay for p53 detection in cell extracts, we
suggest that, instead of searching a linear relationship between the average FITC lifetime and the p53
concentration, we should focus on the fraction of bursts that corresponds to the smallest FITC lifetime
values ( ∼ = 1 ns). In the data reported in Fig. 4.21 A (p53 −/− with addition of p53 in a 1:1 stoichiometry
between the antibodies and the p53 protein) and Fig. 4.21 B (p53 +/+ in a 1:1 ratio between TCE and
the NP-FITC-Ab p53stock ), the lifetime component that corresponds to cell auto-fluorescence falls at an
average value of FITC lifetime 〈τ〉 autof
= 1.9± 0.2 ns. We can therefore count the number of events
(fluorescence bursts) that correspond to FITC lifetime τ < 1.5 ns, that corresponds to a 2σ standard
deviations from 〈τ〉 autof
. The corresponding fraction is F = 43±13 % and F = 45 ± 10 % for the p53 −/−
(with addition of p53) and for the p53 +/+ TCEs, respectively. The uncertainty on F is only indicative
since it has been evaluated by computing the fractions of events with deviation of one, two and three
standard deviations from 〈τ〉 autof
and taking these as independent measurements. The relatively small
uncertainty on the F parameter suggests moreover that F can be used as a measure of the amount of
p53 in the TCEs. The close values found in the two experiments indicates that the concentration of p53
in the p53 +/+ TCEs is of the order of the anti-p53 antibodies in the TCEs, and therefore ∼ = 500 pM.
Figure 4.20: Experiments on TCEs. Fluorescence decay measured from pure TCEs (open squares, red),
p53 positive TCEs in the presence of 5 nm NP-FITC-Ab p53 constructs (filled squares, black), p53-null
TCEs in the presence of 5 nm NP-FITC-Ab p53 constructs and with the addition of p53, R = 1:1 (open
circles, green). The solid lines are the best fit decays convoluted with the IRF and they correspond to
a single exponential decay with τ= 2.6 ns (pure TCEs) and to double exponential decays with 〈τ〉=1.5
ns and 〈τ〉=1.7 ns for the p53 positive and the p53-null plus p53 cases, respectively. Inset: fluorescence
traces for the three cases reported in the main panel: pure TCEs (red), p53 positive TCEs (black) and
p53-null TCEs with the addition of p53 protein (green). For sake of clarity two traces have been displaced
by 50 and 100 units (photons/ms), respectively.

Chapter 4 127
Figure 4.21: Panel A: number of photons collected per burst as a function of the lifetime for the solutions
(1:1 volume ratio) of NP-FITCAb p53 constructs and p53 −/− TCEs in the absence (filled triangles) and
in the presence of (1:1 antibody to p53 stechiometry) p53 protein (open circles). The solid and dashed
lines on the lifetime distributions are best fit Gaussian functions of the components. The up-right panel
reports a short stretch of fluorescence traces (photons/ms) collected in the absence (lower trace) and in
the presence of p53 protein (upper trace). Panel B: number of photons collected per burst as a function
of the lifetime for the NP-FITC-Ab p53 constructs in p53 +/+ TCEs (open squares). The solid lines on
the lifetime distribution are best fit Gaussian functions of the three components.
4.16 Conclusion
We have discussed the possibility to use a hybrid nanodevice composed of spherically symmetric gold
NPs decorated with specific antibodies to detect tiny amounts of proteins. Focusing on the p53 wt
recognition and using FITC dyes (ratio [Ab]:[FITC] = 3:1), we have shown that p53 in vitro detection
can be performed by this device by exploiting the linear decrease of the FITC excited state lifetime,
measured on the fluorescence bursts, with the p53 concentration. The nanodevice proposed has been
also tested for specificity with respect to some exemplary globular proteins and applied to p53 detection
in p53 positive and p53 null cell lines. The analysis reported here suggests that the gold NP-FITC-
Ab p53 nano-constructs can be used to detect the presence of traces of p53 proteins in TCEs opening the
way to their application to in-vivo studies. In particular, we suggest to take into account the observed
anti-correlation between the burst size and the FITC lifetime value in order to single out the p53-( NP-
FITC-Ab p53) complexes from auto-fluorescence and un-bound NP-FITC-Ab p53 crystals, and to measure
the fraction of the small lifetime events to get an estimate of the p53 concentration in the TCEs.

128 Nano-bio sensors for protein detection

Bibliography
[1] Anker J. N.; Hall W. P.; Lyandres O.; Shah N. C.; Zhao J.; Van Duyne R. P. Nat. Mater., 7:442453,
2008.
[2] Pita M.; Cui L.; Gaikwad R. M.; Katz E.; Sokolov I. Nanotechnology, 19:375502, 2008.
[3] Bizzarri A. R.; Cannistraro S. Nanomed. Nanotechnol. Biol. Med., 3:306310, 2007.
[4] A. J. Haes; R. P. van Duyne. A unified view of propagating and localized surface plasmon resonance
biosensors. Anal. Bioanal. Chem., 379:920–930, 2004.
[5] Malmqvist M. Nature, 361:186–187, Jan 1993.
[6] He L.; Musick M. D.; Nicewarner S. R.; Salinas F. G.; Benkovic S. J.; Natan M. J.; Keating C. D.
J. Am. Chem. Soc., 122:90719077., 2000.
[7] Aslan K.; Lakowicz J. R.; Geddes C. D. Anal. Bioanal. Chem., 382:926933., 2005.
[8] Hutter E.; Fendler J. H. AdV. Mater., 16:16851706., 2004.
[9] Lakowicz J. R.; Gryczynski I.; Malicka J.; Gryczynski Z.; Geddes C. D. J. Fluoresc., 12:299302.,
2002.
[10] Stuart D. A.; Haes A. J.; Yonzon C. R.; Hicks E. M.; Van Duyne R. P. IEE Proc. Nanobiotechnol.,
152:1332., 2005.
[11] Dulkeith E.; Morteani A. C.; Niedereichholz T.; Klar T. A.; Feldmann J.; Levi S. A.; van Veggel F.
C. J. M.; Reinhoudt D. N.; Moeller M.; Gittins D. I. Phys. ReV. Lett., 89:1203002., 2002.
[12] Barnes W. L. J. Mod. Opt., 45:661699., 1998.
[13] Aslan K.; Lakowicz J. R.; Szmacinski H.; Geddes C. D. J. Fluoresc., 14:677679., 2004.
[14] Gueroui Z.; Libchaber A. Phys. ReV. Lett., 93:166108., 2004.
[15] Aslan K.; Lakowicz J. R.; Geddes C. D. Curr. Opin. Chem. Biol., 9:538544., 2005.
[16] Ruppin R. J. Chem. Phys., 76:16811684., 1981.
[17] Rachel A. O. Rep. Prog. Phys., 71:076501., 2008.
[18] Chirico G.; Beretta S. Phys. ReV. E., 60:21482153., 1999.
[19] Prosperi D.; Morasso C.; Tortora P.; Monti D.; Bellini T. ChemBioChem., 8:10211028., 2007.
[20] Flecha F. L. G.; Levi V. Biochem. Mol. Biol. Educ., 31:319322., 2003.
[21] A. J. Levine; J. Momand; C. A. Finlay. The p53 tumour suppressor gene. Nature, 351:453, 1991.
[22] C. Whibley; P. D. P. Pharoah; M. Hollstein. p53 polymorphisms: Cancer Implications. Nature,
9:95, 2009.
[23] J. C. Bourdon. p53 and its isoforms in cancer. Brit. J. Cancer, 97:277, 2007.
[24] B. Vogelstein; D. Lane; A. J. Levine. Surfing the p53 network. Brit. J. Cancer, 408:307, 2000.
[25] S. Kumar; A. Mohan; R. Guleria. Prognostic implications of circulating anti-p53 antibodies in lung
cancera review. Eur. J. Cancer Care, 18:248, 2009.
[26] T. Soussi. p53 Antibodies in the sera of patients with various types of cancer: A review. Cancer
Res., 60:1777, 2000.

130 Bibliography
[27] E. M. Tan; S. J. S. Smolen; J. S. Mcdougal; B. T. Butcher; D. Conn; R. Dawkins; M. J. Fritzler;
T. Gordon;J. A. Hardin; J. R. Kalden; R. G. Lahita; R. N. Maini; N. F. Rothfield; R. Smeenk;
Y. Takasaki; W. J. Van Venrooij; A. Wiik; M. Wilson; and J. A. Koziol;. A critical evaluation of
enzyme immunoassays for detection of antinuclear autoantibodies of defined specificities. Arthritis
and Rheumatism, 42:455, 1999.
[28] N.-S. Lai; C.-C. Wang; H.-L. Chiang; L.-K. Chau. Detection of antinuclear antibodies by a colloidal
gold modified optical fiber: Comparison with ELISA. Anal. Bioanal. Chem., 388:901, 2007.
[29] B. A. Du; Z. P. Li;Y. Q. Cheng. Homogeneous immunoassay based on aggregation of antibodyfunctionalized
gold nanoparticles coupled with light scattering detection. Talanta, 75:959, 2008.
[30] Y. Liu; Y. Liu; R. L. Mernaugh; X. Zeng. Single chain fragment variable recombinant antibody functionalized
gold nanoparticles for a highly sensitive colorimetric immunoassay. Biosens. Bioelectron.,
24:2853, 2009.
[31] Lin D. L.; Chang C. p53 is a mediator for radiation-repressed human TR2 orphan receptor expression
in MCF-7 cells, a new pathway from tumor suppressor to member of the steroid receptor superfamily.
J. Biol. Chem., 271:14649–14652, Jun 1996.
[32] P. J. Steinbach; R. Ionescu;C. R. Matthews. Analysis of kinetics using a hybrid maximumentropy/nonlinear-least-squares
method: Application to protein folding. Biophys. J., 82:2244, 2002.
[33] S. Freddi; L. DAlfonso; M. Collini; M. Caccia; L. Sironi; G. Tallarida; S. Caprioli; G. Chirico.
Excited-state lifetime assay for protein detection on gold colloids-fluorophore complexes. J. Phys.
Chem. C, 113:2722, 2009.
[34] Y. Chen; J. D. Mueller; P. T. C. So; E. Gratton. The photon counting histogram in fluorescence
fluctuation spectroscopy. Biophys. J., 77:553, 1999.
[35] J. D. Mueller. Cumulant analysis in fluorescence fluctuation spectroscopy. Biophys. J., 86:3981,
2004.
[36] M. Caccia; E. Camozzi; M. Collini; M. Zaccolo; G. Chirico. Photon moment analysis in cells in the
presence of photo-bleaching. Appl. Spec., 59:227, 2005.
[37] Arakaki A.; Hideshima S.; Nakagawa T.; Niwa D.; Tanaka T.; Matsunaga T.; Osaka T. Biotechnol.
Bioeng., 88:543546., 2004.
[38] O. Malcai; D. A. Lidar; O. Biham; D. Avnir. Scaling range and cutoffs in empirical fractals. Phys.
Rev. E, 56:28172828., 1997.
[39] Link S.; El-Sayed M. A. Int. ReV. Phys. Chem., 19:409453., 2000.
[40] Ung T.; Liz-Marzan L.; Mulvaney P. J. Phys. Chem. B, 105:34413452., 2001.
[41] N. J. Durr; T. Larson; D. K. Smith; B. A. Korgel; K. Sokolov; A. Ben-Yakar. Two-photon luminescence
imaging of cancer cells using molecularly targeted gold nanorods. Nano Letters, 7:941945.,
2007.
[42] Gryczynski I.; Malicka J.; Shen Y.; Gryczynski Z.; Lakowicz J. R. J. Phys. Chem. B, 106:21912195.,
2002.
[43] Berland M.; So P. T.; Gratton E. Biophys. J., 68:694701., 1995.
[44] Berland K. M.; So P. T. C.; Chen Y.; Mantulin W. W.; Gratton E. Biophys. J., 71:410420., 1996.
[45] Maxwell D. J.; Taylor J. R.; Nie S. J. Am. Chem. Soc., 124:96069612, 2002.
[46] Gersten J.; Nitzan A. J. Chem. Phys., 75:11391152., 1981.

Chapter 4 131
[47] Das P.; Metiu H. J. Phys. Chem., 89:46804687., 1985.
[48] Pibrik R.; Aslan K.; Zhang Y.; Geddes C. D. J. Phys. Chem., 112:1796917973, 2008.
[49] Amos R; Barnes W. L. Phys. ReV. B, 55:72497254., 1997.
[50] Liao P. F.; Wokaun A. J. Chem. Phys., 76:751752., 1982.
[51] Powell R. D.; Halsey C. M. R.; Hainfeld J. F. Microsc. Res. Technol., 42:212., 1998.
[52] Huang C. C.; Chiang C. K.; Lin Z. H.; Lee K. H.; Chang H. T. Anal. Chem., 80:14971504, 2008.
[53] Hohenau A.; Leitner A.; Aussenegg F. R. In Surface Plasmon Nanophotonics, Springer: The
Netherlands, 2007.
[54] Podolskiy V. A.; Shalaev V. M. Laser Phys., 11:2630., 2001.
[55] M. Li; C. L. Brooks; F. Wu-Baer; D. Chen; R. Baer; W. Gu. Mono-versus polyubiquitination:
differential control of p53 fate by Mdm2. Science, 302:1972, 2003.
[56] H. Schneckenburger; M. Wagner; P. Weber; W. S. L. Strauss; R. Sailer. Autofluorescence lifetime
imaging of cultivated cells using a UV picosecond laser diode. J. Fluor., 14:649, 2004.
[57] K. Konig; P. T. C. So; W. W. Mantulin; B. J. Tromberg; E. Gratton. Two-photon excited lifetime
imaging of autofluorescence in cells during UVA and NIR photostress. J. Micros., 183:197, 1996.

Chapter 5
Anisotropic nanoparticles
5.1 Introduction
In the last two decades several groups have investigated the changes of chemical and
physical properties of materials whose dimensions are reduced to nanometric scales.
These studies highlighted a number of possible applications for nanostructures, which
are now employed in biology and medicine for imaging, diagnosis, and therapy, owing to
the combination of their unique shape/size-dependent properties, strong absorption/scattering
of light with stability and low citotoxicity. Among all, gold nanorods and
anisotropic NPs are found to be more popular and useful for potential applications such
as biochemical sensing, biomedical diagnostics, and therapeutics due to possible tuning
of their surface plasmon resonance (by increasing their aspect ratio)in the visible and
near-infrared region, which is the potential window of the electromagnetic spectrum for
in-vivo applications. Gold nanoparticles not spherically symmetric can efficiently convert
optical energy into heat via nonradiative electron relaxation dynamics (2224), endowing
them with intense photothermal properties (2535). Such localized heating effects can be
directed toward the eradication of diseased tissue, providing a noninvasive alternative to
surgery (36).
Gold is not the only nobel metal with such properties when reduced to nanoparticle
size. Silver and Pd are also used for similar reason though the spectral window is blue
shifted. Moreover, a wide field of applications in thermal therapy of cancer involves the
use of paramagnetic (superparamagnetic) nanoparticles excited by radiofrequency waves.
Engineering of gold nanorods with novel properties by controlled synthesis for potential
applications is very important, particularly because of the NP toxicity and interaction
with cells. The most popular method for synthesizing gold nanorods involves the seedmediated
approach using cetyltrimethylammonium bromide (CTAB) surfactant as the
shape-directing agent.
132

Chapter 5 133
In this chapter the characterization of asymmetric branched gold nanoparticles obtained
using for the first time in the seed growth method approach a zwitterionic surfactant,
laurylsulphobetaine (LSB), is reported. LSB concentration in the growth solution
allows to control the dimension of the NPs and the SPR position, that can be tuned in
the 700-1100 nm Near Infrared range. The synthesis procedure has been devised by the
group of Prof. P.Pallavicini of the University of Pavia (General Chemistry Department).
The samples synthesized with CTAB and LSB have been analized with several techniques
to obtain a complete characterization: from the data obtained through the absorption
spectra in the UV-Visible region, the TEM images of the solutions, FCS and
DLS experiments, we reached information on the nanoparticles shapes and dimensions.
The second part of the spectroscopic characterization has been performed by employing
two photon excitation (TPE), which affects greatly the luminescence quantum
yield of the nanorods. In fact, due to non-linear phenomena such as the TPE, the luminescence
intensity is enhanced by many orders of magnitude with respect to the single
photon excitation of flat surfaces, improving the usefulness of these nanoparticles for cell
imaging.
Gold NRs exhibit a high photon to thermal conversion efficiency, with a larger absorption
cross-section at NIR frequencies per unit volume than most other types of
nanostructures. Based on this characteristic, anisotropic nanoparticles can be used for
diagnostic purposes, by optical imaging, and for therapeutic purposes, by thermal phototherapy.
Is clear that in order to apply thermal phototherapy in medical field is
essential to know the temperature at which colloids interact with cells.
In order to measure the local temperature’s rise on the particles surface , induced by the
absorption of infrared radiation, a novel sensor was designed based on the conjugation
of Rhodamine-B fluorophores with anisotropic gold nanoparticles.
The Rhodamine-B lifetime decreases gradually when the value of the temperature is
changed by the laser irradiation showing that the sensor is able to play the role of molecular
thermometer and it can be applied in vivo for imaging and therapeutic purposes.
5.2 Gold nanoparticles luminescence
The gold metal photoluminescence in the visible range was first reported in 1969 by
Mooradian [1]. Although suggested as a band-structure probe, the photoluminescence
of noble metals remained relatively unexplored for nearly two decades until 1986, when
Boyd [2] reported multiphoton induced luminescence on roughened noble metal surfaces.
In recent years, there has been a renewed interest in photoluminescence from metal surfaces,
primarily in the photoluminescence from noble metal nanoparticles, generated by

134 Anisotropic nanoparticles
their potential biomedical applications.
This may seem counter-intuitive at first, as gold is well known to quench the emission
of nearby fluorophores due to back-electron transfer (60,61). However, gold itself is able
to produce a weak photoemission via interband transitions, and this can be enhanced
by many orders of magnitude when it couples with an appropriate plasmon excitation
(62,63). For example, linear photoluminescence from gold NRs can be enhanced by a
factor of over a million compared with bulk gold (64), and NRs subjected to pulsed laser
excitation are able to emit a strong TPL (18,41,65).
In the above studies the origin of the photoluminescence was attributed to the radiative
recombination of an electron-hole pair. Visible photoluminescence process in gold metal
begins when an electron in the d band is excited to an unoccupied state in the conduction
band (figure 5.2). This creates a hole in the d band, which after some time will
recombine with an electron. The recombination occurs generally through nonradiative
mechanisms, but the hole can also radiatively recombine with electrons from the conduction
band. Since the photon absorption will necessarily create d holes at points in
the Brillouin zone where the conduction band is vacant, interband radiative relaxation
is only efficient when intraband scattering processes move the holes closer to the Brillouin
zone coordinate that corresponds to the Fermi level in the conduction bans. The
peak energy of emitted photons is therefore strongly connected to the energy separation
between d holes and the Fermi surface, which near X is roughly 1.8 eV, and near L,
2.4 eV. Due to these separations, visible photoluminescence is generated when the hole
recombines with electrons near the Fermi surface (figure 5.2). Although the quantum efficiency
of the photoluminescence from bulk noble metal is very low, tipically of the order
of 10 −10 , the luminescence was found to be enhanced by several order of magnitude on
rough or curved metal surfaces due to the lightning rod effect(conventionally described
as the crowding of the electric field lines at a sharp metallic tip). Even in this cases the
quantum yield is likely to be low. Moreover the overall brightness (number of photons/s
particle) is high due to the ∼ =10 6 times larger σ of GNR compared to conventional dyes.
The rough metal surface can in fact be regarded as a collection of randomly oriented
hemispheroids of nanometre size dimension on a smooth surface. These hemispheroids
show a surface plasmon resonance and therefore the exciting and radiating electric fields
are amplified by the local field induced around the hemispheroids by the plasmon resonances.
A similar enhancement has recently been found for the luminescence of gold
nanorods. The luminescence efficiency is six orders of magnitude greater than that found
in the bulk because of the lighning rod effect. According to the theoretical studies of
the photoinduced luminescence from rough surface of noble metals, the incoming and
outgoing fields are proposed to be enhanced via coupling to the local plasmon resonances.

Chapter 5 135
Figure 5.1: Left:Symmetry points in the first Brillouin zone of gold. Right:Regions of the band structure
near X and L close to the Fermi surface.
5.2.1 Two-photon luminescence (TPL)
Luminescence may also be excited by multiphoton absorption and can be enhanced
by many orders of magnitude when coupled with an appropriate plasmon excitation.
Multiphoton-induced luminescence was first observed as a broadband background in
second-harmonic-generation (SHG) measurements on roughened noble metals.
Because the multiphoton luminescence was observed from a roughened metal, the
effects of the localized plasmon resonances must be included in the interpretation of the
spectra. In particular, the multiphoton luminescence is more sensitive to the local fields
than the single-photon luminescence. Since the local fields are strongest just outside the
metal surface, multiphoton-induced emission from the surface atoms may dominate over
that from the bulk.
For two-photon excitation processes, two different schemes have been proposed in
litterature: a sequence of one-photon excitations and a coherent two-photon excitation.
These two cases are distinguished by different behavior on the incident polarizations.
In the first theory, two-photon absorption in gold is a two-step process consisting of two
successive one-photon steps, as recently proposed by Imura et al.,8 with the lifetime of
the intermediate state, after the first photon absorption, ruling TPL dynamics. The
analysis by Imura et al. was supported by the TPL polarization dependence, which was,
however, questioned by later experimental results,12,13,24 so that a strong support to
their interpretation is still missing. The excitation mechanism can be hence described as
follows (Figure 5.2): upon the optical excitation, the first photon excites an electron in
the sp conduction band located below the Fermi surface to the sp conduction band above
the Fermi surface via an intraband transition. At the same time, a hole is created in

136 Anisotropic nanoparticles
the sp conduction band below the Fermi level. This transition is resonant with photons
with a polarization along the long axis of the nanorod. After the excitation, the memory
of the polarization is rapidly lost. Then, the second photon excites an electron in the d
band to the sp conduction band, where the hole was created by the first photon. The
transition by the second photon is not sensitive to the polarization. The second photon
generates a hole in the d band. As a consequence, an electron-hole pair is generated,
which can recombine to radiate later.
Figure 5.2: Excitation schemes of sequential one-photon absorptions near the X and L symmetry points.
Open and closed circles denote holes and electrons, respectively.
In the second scheme, two-photon induced photoluminescence in gold is generally
considered as a three-step process. Electrons from occupied d-bands are excited by twophoton
absorption to unoccupied state of the sp-conduction band. Subsequent intraband
scattering processes move the electrons closer to the Fermi level. The relaxation of
the electron-hole pair can then recombine either through non-radiative processes or by
emission of luminescence. Radiative relaxation energies are therefore strongly connected
to the interband separation, and for bulk material these energies are ∼ =1.5-2.4 eV and
occurs around the X and L points of the Brillouin zone.14
Much of the interest is generated by the potential biomedical applications of nanoparticles.
TPL can be spectrally separated from tissue autofluorescence and the power densities
required for TPL imaging are orders of magnitude below the damage threshold of
biological tissue.
The nanorods’ intrinsic TPL properties are useful for real-time imaging in vitro and in
vivo, as demonstrated by characterization of their uptake into cells 13 and by monitoring
their blood residency in animal models.11 Most recently, it has been found that gold
nanorods are potent agents for mediating the photothermal destruction of tumor cells,
and the TPL can be used to obtain insights into their ability to inflict photoinduced
injury.14

Chapter 5 137
5.3 Experimental details
5.3.1 Nanoparticles synthesis
In a typical preparation 1 gold nanoparticles (NPs) seed solution is obtained by mixing 5
mL of 5·10 −4 M HAuCl 4· 2H 2 O with 5 mL 0.20 M LSB (Laurylsulphobetaine N-dodecyl-
N’, N”-dimethyl-3-ammonio-1-propane-sulphonate) in water, and by adding 600 µL of
NaBH 4 0.01 M, obtaining the typical brownish colour of a few-nm sized Au NP dispersion.
TEM (Transmission Electron Microscopy) reveals the formation of spherical NP
with d

138 Anisotropic nanoparticles
Dynamic light scattering characterization
For Dynamic Light Scattering (DLS) a home-made setup for variable angle measurement
of the scattered light autocorrelation function has been used with a He-Ne 30 mW
polarized laser source. The correlator board was an ISS (Urbana Champaign, IL) single
photon counting acquisition board and the data were analyzed as described in section
5.5.
Fluorescence Spectroscopy: FCS measurement
A brief description of the set-up is reported in section ....
The excitation polarization was modified using half and quarter waveplates. The fluorescence
signal was split by a polarizing or non-polarizing beamsplitter cube and detected
by two SPAD. Using linearly or circularly polarized excitation light (denoted as X or C),
we derived cross-correlation functions from two channels that detect linearly polarized
or non-polarized emission light (denoted as XX, XY, or NP, depending on the configuration
of the polarization for each channel). We measured FCS curves for all investigated
samples. They all displayed two distinct decays: one correlation contribution with a
characteristic relaxation time on the order of milliseconds, which is attributed to translational
diffusion through the confocal observation volume; and a second contribution
decaying on shorter time scales, and which is strongly dependent on the polarization
configuration.
The TPL emission spectra were recorded by means of a CCD (DV420A-BV, Andor,
IRL) based spectrometer (MS125, Lot-Oriel,UK), connected to the backport of the
microscope. The excitation wavelenght is variable in the range 730-920 nm with a constant
average power on sample of about 2 mW. Each of the spectra is the result of the
accumulation of 10-20 1 s time acquisitions.
Fluorescence Microscopy
The laser source is a mode-locked Ti:Sapphire laser (Mai Tai, Spectra Physics, CA)
pumped by a solid state laser at 532 nm that produces pulses with ∼ = 100 fs FWHM,
repetition rate = 80 MHz and average power ∼ = 2.8 W at λ= 800 nm at the output
of the laser source. The optical set-up is built around a confocal scanning head (FV300,
Olympus, Japan) mounted on an optical microscope (BX51, Olympus, Japan) modified
for direct (non de-scanned) detection of the signal. The laser beam is sent to the scanning
head and is driven on the sample by means of two galvanometric mirrors and a scan
lens: the galvos are responsible for the specimen scan while the lens assures that the

Chapter 5 139
back aperture of the objective (N.A. = 0.95, 20X, water immersion, Olympus, Japan) is
overfilled during the entire scan process. The objective simultaneously focuses the laser
beam on the sample and collects, in epifluorescenc geometry, the harmonic or fluorescence
signal through either the descanned (FV300) or the non-descanned (ND-unit described
in section ...) detection unit. TPL emission is filtered through a short-pass 670 nm filter
(Chroma Inc., Brattelboro, VT) and selected by a band-pass filter at 535 nm (Chroma
Inc., Brattelboro, VT, full width = 50 nm).
In order to study the dependence of the TPL signal on the excitation polarization an
half-wave plate was inserted in the optical path right below the objective in order to
rotate the polarization of the incident radiation.
5.4 Results
5.4.1 UV-Vis characterization
The extinction spectra of the NPs are shown in figure 5.3 : seven sample were synthesized
according to the method described in section 5.3.1 with LBS concentration variable in
the range 0.2-0.6 M, as shown in the panel.
Figure 5.3: Left: UV-Vis spectra of samples synthesized with LSB surfactant. Right: TEM images; the
color of the frames correspond to the absorption spectra.
Observation of UV-Vis spectra revealed presence of three bands:
1. a major band at λ max 700-1100 nm, depending on LSB concentration (”long” band)
2. a less intense band at 520-530 nm (”short” band)
3. a third band with λ max positioned in the 650-750 nm range, depending on LSB
concentration (”intermediate” band)
The ’intermediate band’ appears as single peak or a shoulder, depending on the
position of the long band. Moreover, its absorbance varied randomly from negligible to

140 Anisotropic nanoparticles
comparable to that of the long band. Attribution of the three bands was obtained by
TEM images, as descibed in the following paragraphs.
The position of the long LSPR band scales linearly with the LSB concentration (Figure
5.4) and its origin can be assigned by investigating the NR spectrum.
Figure 5.4: Variation of different parameters as a function of the LSB molarity added to the growth
solution. A: LSPR position. B: L/B for type C (blue squares) and type B (red circles) nano-objects.
C:pictorially represents the ideal shape features of the nanoobjects (L=branch length, B = branch base
width)
In figure 5.5 the extinction spectrum of the sample obtained with 0.2 M CTAB is
reported; in this case the solution is homogeneous and presents a single kind of NPs,
named nanorods and the spectrum presents only the long and short bands. Therefore we
can ascribe the ”long” and ”short” band to the SPR parallel and perpendicular (which
is close to the nanosphere spectrum SPR) to the rod axis.
Figure 5.5: Left: UV-Vis spectra of samples synthesized with 0.2 M CTAB surfactant. Right: TEM
image.

Chapter 5 141
5.4.2 Transmission Electron Microscopy (TEM) characterization
For LSB growth the situation is more complex therefore the structure of the nanoparticles
was studied by TEM and correlated to the visible spectrum. Figure 5.6 shows
the comparison between the TEM images acquired for nanoparticles synthesized with
CTAB and variable concentration of LSB surfactant. The CTAB growth method provides
nanorods with well defined aspect ratio; instead, the constructs obtained with LSB
exhibit a prevalent star-like structure with thin fingers.
Figure 5.6: The panels show the comparison between NPs synthesized with LSB or CTAB surfactants.
In particular, in the image (i) in Figure 5.7, taken on a dispersion from a 0.6M LSB
growth solution, three kinds of objects are present: A) nanospheres (< 20 nm diameter);
B) nanostars, with large trapezoidal branches (intermediate band); C) branched
asymmetric NP, with narrow, long branches (long band) of high AR (Figures 5.7(iii)-(v)
display isolated and magnified images of the three typologies).
Image 5.7(ii) has been taken on a dispersion obtained with a 0.5 M growth solution
with negligible intermediate band in the UV-Vis spectrum: in this case objects of tipology
B are almost absent.
Based on TEM images, it is possible to assign the intermediate band to the B typology
objects (nanostars) and of the long band to the C typology objects, in analogy to the
GNR case (figure 5.5). The 520-530 nm LSPR band is attributed to the nanospheres,
although a contribution to this band may also be due to the transverse LSPR of the
larger nanoobjects.

142 Anisotropic nanoparticles
Figure 5.7: TEM images. (i): obtained from a growth solution prepared with 0.6 M LSB, with a
three band absorption spectrum. (ii): obtained form a growth solution 0.5 M LSB, with a negligible
intermediate band. (iii) detail of nanospheres (from a 0.6 M LSB growth solution), 30% magnified. (iv)
detail of nanostar, (v) detail of branched asymmetric nanoobjects 30% magnified (both from a 0.2M LSB
growth solution).

Chapter 5 143
For anisotropic branched NP, the LSPR position has been reported to be proportional
to the length/base ratio of the branches (L/B, Figure 5.4) or, similarly, to the
length/aperture angle ratio of the branches, while it is independent on the number of
branches grown on the core. This agrees with the assignment of the intermediate band
to the B typology objects (nanostars) and of the long band to the C typology objects. A
plot of the ratio of the length to base width (L/B) of the branches, calculated from TEM
images, shows a similar dependence on the LSB concentration for the type C objects
(Figure 5.4B, blue squares).
The L/B ratio for type B objects has been calculated only on images from 0.2 to 0.45M
solutions (with higher LSB concentrations the tendency of the nanobjects to crop in
dense assemblies on the TEM grids prevented graphical evaluations of sizes on the star
nanoobjects). In each synthesis a ≈ 3 units lower L/B ratio is found for the B population
with respect to type C objects, with a slight tendency to increase with LSB concentration
(Figure 5.4, red circles).
From TEM images, the dimensions of the different objects can be determined; the
lenght L, the width B and the aspect ratio (L/B), as defined in figure 5.4C, were measured
and reported in table 5.1. The index 1 and 2 refer to C and B object tipology respectively;
F1 is the relative fraction of C type nanoparticles and F2 of B type NPs.
[LSB] [M] L 1 [nm] D 1 [nm] AR 1 L 2 [nm] D 2 [nm] AR 2 F 1 (%) F 2 (%)
0.2 54.5±3 8.0±0.8 6.9±0.8 35.8±1.9 9.1±1.3 3.9±0.4 49.4% 37.6%
0.35 65.4±5.3 8.7±0.3 7.5±0.6 41.0±10.9 8.5±1.5 4.9±1.3 43.3% 47.1%
0.45 60.9±1.9 7.5±1.1 8.1±1.1 38.7±5 8.7±0.7 4.5±0.8 58.8% 30.4%
0.5 98.9±5.1 7.9±1.2 12.5±1.8 - - - 71.8%
0.6 64.3±11.1 8.2±3.7 7.8±1.4 39.9±4.7 11.4±1.9 3.5±0.5 30% 36%
Table 5.1: Values of the lenght, width and aspect ratio of the different tipology of nanoparticles
synthesized with LSB surfactant. AR is the axial ratio.
Also the samples synthesized with the more conventional surfactant CTAB were
characterized though trasmission electron microscopy (figure 5.5). In these cases only
one typology of nanoparticles is observed; the values of the length and the width are
reported in table 5.2.

144 Anisotropic nanoparticles
[CTAB] [M] L [nm] D [nm] AR [nm]
0.2 47.9±4.5 17.2±2.1 2.8±0.3
0.2 PEG2000 41.8±3.1 11.2±2.9 3.7±0.9
Table 5.2: Values of the length and width parameters for the three sample synthesized
with the surfactant CTAB. AR is the axial ratio.
5.4.3 Z-potential characterization
The nanoparticles solutions are characterized by a ζ-potential value of (-27.2±6.1) mV
for 0.3 M LSB concentration and of (-13.4±4.55) mV for 0.45 M LSB: this numeric value
indicate that the nanoparticles have a prevalent negative surface charge; the first solution
is moderatly stable and the second one is in an incipient instability condition.
The sample synthesized with 0.2 M CTAB surfactant have a ζ-potential of (-14.0±10.5)
mV; in this case the prevalent surface charge is negative but the charge distribution is
wide and the solution is in an incipient instability form.
5.5 Dynamic Light Scattering characterization
Dynamic light scattering (DLS) measurements substantially validate the picture obtained
through TEM. Contribution to scattering of LSB micelles is negligible (pure LSB solutions
revealed spherical micelles with diameter increasing from 3.2±0.24 nm to 6.0±0.4
nm when increasing LSB concentration from 0.2 ot 0.6M). In all the suspensions of NPs
small, spherical objects are observed (type A NPs, hydrodynamic radius, R H , 6-10 nm)
together with larger ones. The latter display both a rotational and a translational decay
of the correlation function, whose values allow to evaluate the amount of depolarized
scattering to the correlation function and therefore shape anisotropy.
5.5.1 Solutions of LSB (LSB micelles)
Solutions of LSB were prepared in water at increasing concentrations from 0.2 to 0.6 M.
The AutoCorrelation Function (ACF) of the intensity of the light scattered at the angle
θ= 90 ◦ can be fit to a single exponential decay according to the relation:
G(t) = 〈I〉 2 ( 1 + f 2 exp [ −2DQ 2 t ]) ∣ = 〈I〉
(1 2 + f 2 ∣∣g (1) (t) ∣ 2) (5.1)
where Q is the scattering vector, Q 2 = (4nsin(θ/2)/λ) 2 ∼ =3.49x10 10 cm −2 , where η
is the viscosity, T = 297.15 K, is the solution temperature. The translational diffusion

Chapter 5 145
coefficient D is then translated into the hydrodynamic radius, R h , by means of the
Stokes-Einstein relation:
D =
k BT
6πηR h
(5.2)
The signal to noise ratio f (Eq.5.1) is typically ∼ =0.03. The ACFs reported in Fig. 5.8
indicates clearly that the relaxation time increases with the LSB concentration. Indeed
the average hydrodynamic radius changes from R h
∼ =1.6 nm at [LSB] = 0.2 M, to Rh ∼ =3
nm for [LSB] = 0.6 M, with a linear trend.
[LSB] [M]
R h [nm]
0.2 1.6±0.12
0.3 2.1±0.2
0.4 2.3±0.2
0.5 2.6±0.2
0.6 3.0±0.2
Table 5.3: Average hydrodynamic radii of micelles in solutions as obtained from polarized
VV dynamic light scattering.
Figure 5.8: ACFs of the light scattered by the LSB micelle solutions. The symbols refer to [LSB] =
0.2 M (open squares); [LSB] = 0.3 M (filled circles); [LSB] = 0.4 M (open triangles); [LSB] = 0.5 M
(filled triangles); [LSB] = 0.6 M (open circles). The dashed lines are the best fit of Eq.1 to the data.
A small baseline has been added to the data fitting and is probably due to residual dust contribution
to the scattering intensity.
The top inset shows the linear trend of the Rh as a function of the LSB
concentration. The bottom inset shows the ACFs in log-linear scale.

146 Anisotropic nanoparticles
5.5.2 Dynamic light scattering of the NPs
The scattering of the solutions of the NPs has been measured through a vertical polarizer,
i.e. parallel to the direction of the polarization of the laser light. In this case two
exponential decays are needed to fit the data with relaxation times of the order 10 and
200 µs approximately (see Fig. 5.9). The presence of the faster component is taken as an
evidence of the polarizability anisotropy of the NPs. In the view of the TEM analysis,
the anisotropy of the polarizability is probably entirely due to the shape anisotropy of
the NPs. The two relaxation components are then ascribed to the rotational and the
translational diffusion of the NPs. According to the TEM image analysis the shape of
the NPs may vary from spherical to branched and no clear cylindrical symmetry can
be pointed out. The polarizability of the particle is in general a rank two tensor and
can be diagonalized to give three eigenvalues that have the meaning of the polarizability
along three orthogonal axes in the particle frame of reference. For sake of simplicity, we
assume here that we can define in the NP a direction along which the polarizability (α ‖ )
of the electrons is much larger than in the other two orthogonal directions (α ⊥

Chapter 5 147
Figure 5.9: ACF of the light scattered by a solution of NPs prepared at [LSB] = 0.2 M observed through
a polarizer whose axis is parallel to the polarization direction of the excitation beam (vertical). The dashed
line is the best fit to the function 5.4. The solid line represent the translational component of the ACF.
A small baseline accounts for the presence of larger aggregates as also indicated by the TEM analysis.
We assume that the anisotropy along a particle axis is proportional to the length of
the axis and therefore:
α ‖ ∝ L (5.5)
α ⊥ ∝ D (5.6)
α ∝ 1 (L + 2D) (5.7)
3
The ratio of the amplitudes of the translational to the rotational exponential components
obtained from the best fit, A fit , is then given by:
A fit = 4 (α ‖ − α ⊥ ) 2
45 α 2 = 4 δα
2
45 〈α〉 2 (5.8)
From Eqs. 5.5-5.8 it is straightforward to derive the (effective) axial ratio, D/L, as:
( ) (
L
D = 3 2δalpha
3 〈α〉 − 1 1 − δ ) −1
alpha
(5.9)
3 〈α〉
The best fit values of the ratio A fit are shown in Fig.5.10 and reported in Table
5.4 together with the estimates of the axial ratios, L/D. The comparison to the L/B
(length/base) values measured on the branches of asymmetric NP (type C objects, TEM C
in Table 5.4) and on the branches of nanostars (type B objects, TEM B in Table 5.4)

148 Anisotropic nanoparticles
from TEM images indicates that the analysis reported here underestimate the shape
anisotropy. This fact is expected since the axial ratio values are measured on the single
branches on the TEM images, while Eq. 5.9 makes use of an overall polarizability
anisotropy.
[LSB] [M] A fit R h,T [nm] R h,R [nm] (L/D) (L/D) T EMC (L/D) T EMB
0.2 1.8±0.2 24±2 21±3 5.5±0.9 6.9±0.8 3.9±0.4
0.35 1.9±0.2 29±2 28±4 6.2±0.9 7.5±0.6 4.9±1.3
0.45 2.2±0.3 21±3 27±3 8.9±3.0 8.1±1.1 4.5±0.8
0.5 2.4±0.2 19±3 27±3 13.2±5.0 12.5±1.8
0.6 2.4±0.2 21±3 25±3 13.6±5.0 7.8±1.4 3.5±0.5
Table 5.4: DDLS amplitude analysis:R fit from eq. 5.8, R h,T and R h,R are the hyroynamic
radii, L/D is from eq. 5.9 and (L/D) T EM from TEM measurement.
The nanostars and the branched asymmetric NP (i.e. type B and C) do not give separate
contributions in the DLS autocorrelation functions, this meaning that they have
similar hyrodynamic raius R H . Though the average R H is in fact 23±4 nm, independent
of the LSB concentration, the ratio of the rotational to the translational scattering
amplitude (A fit ) does increase linearly with the LSB concentration (Figure 5.10). Although
this cannot be put in direct relation with the L/D ratio, it indicates that the
increase of surfactant concentration promotes the formation of objects with increasing
shape anisotropy.
Figure 5.10: Ratio of the rotational to the translational scattering amplitude in DLS.
The best fit values of the ratio A fit are obtained from the analysis of the ACFs

Chapter 5 149
measured through a vertical polarizer. The hydrodynamic radius, R h,T , is obtained from
the slower relaxation rate (Γ T ) and Eq. 5.2. The hydrodynamic radius, R h,R , is obtained
from the faster relaxation rate (Γ R ) and Eq. 5.10. The ratios (L/D) are estimated from
Eq. 5.9. The ratios (L/D) T EM are estimated from the analysis of the TEM images on
the C and B families of NPs.
The best fit values of the average hydrodynamic radius, reported in table 5.4, indicate
that the average encumbrance of the NPs is not changing with the concentration of
the LSB surfactant, though a marked increase in the anisotropy is observed through the
A fit parameter (figure 5.10).
Finally, regarding the information that can be obtained from the rotational relaxation
rate, Γ R = 6Θ, we observe that to the first approximation we can estimate an average
hydrodynamic radius from the relation:
Θ =
k BT
8πηR 3 h,R
(5.10)
From Eq.5.10 a very rough estimate of the overall encumbrance can be done, reported
in Table 5.4, that, however, appears to be consistent with the estimate obtained from
the analysis of the translational relaxation rate. The average values of the hydrodynamic
radii obtained from the analysis of the translational and rotational components of the
ACFs are in fact: 〈R h,T 〉 = 22.8±4 nm and 〈R h,R 〉 = 26±3 nm. An more refined
analysis of the rotational relaxation component can be done according to a number of
approaches. We assume here the formulation given by Tirado and Garcia de la Torre
[cit] for the rotational diffusion coefficient of a rod:
with
Θ = 3k [ ( ) ]
BT L
πηL 3 ln + σ
D
(5.11)
σ = −0.662 + 0.917 D ( ) D 2
L − 0.05 (5.12)
L
From the measurement of the rotational diffusion coefficient, Θ= Γ R /6, and of the
axial ratio, obtained above from the amplitudes of the two components of the ACFs, we
can gain an additional estimate of the ”long” and ”short” axes of the effective ellipsoid
with which we have approximated the NPs.
The table reports the comparison of the estimate of the sizes of the axes of the NPs
obtained from TEM and from DDLS analysis. The estimates of the long (L R ) and short
(D R ) axes are obtained from the rotational diffusion coefficient according to Eq. 5.18.

150 Anisotropic nanoparticles
(L/D) Θ [kHz] σ L R [nm] D R [nm] L T EM [nm] D T EM [nm]
0.2 5.5±0.9 16.6±3 -0.497 65±4 12±2 54.5±3 8.0±0.8
0.35 6.2±0.9 6.9±0.4 -0.515 89±2 14±2 65.4±5.3 8.7±0.3
0.45 8.9±3.0 8.1±0.4 -0.559 91±1 10±4 60.9±1.9 7.5±1.1
0.5 13.2±5.0 8.2±0.5 -0.593 97±2 7.3±3 98.9±5.1 7.9±1.2
0.6 13.6±5.0 10.0±1 -0.595 92±3 6.7±2 64.3±11.1 8.2±3.7
Table 5.5: Evaluation of the long and short effective axes of the NPs. L/D is taken from
A fit and eq. 5.8 and 5.9
When trying to compare the values of length of the long and short axes of the NPs
obtained from DDLS and from TEM, one should consider that in the analysis of the
TEM images, the lengths are taken on the single branches while in the assumptions
made for the analysis of the DDLS ACFs we have described the whole NP as revolution
ellipsoid. In this sense the overestimation of the length of the long axis observed in Table
5.5, can be understood and the DDLS data can be considered consistent with the TEM
data.
5.6 FCS experiment
A precise characterization of the particles dimensions is fundamental for in vivo applications
and it has been performed here by means of the Fluorescence correlation
spectroscopy (FCS) technique. FCS is a versatile, noninvasive high-resolution technique
that takes advantage of the spontaneous intensity fluctuations of the fluorescence signal
emitted by extremely low concentrated objects in a focal volume of about one femtoliter.
The strength and duration of the fluorescence fluctuations are quantified by temporally
autocorrelating the recorded signal: the autocorrelation function is a measure of the selfsimilarity
of the signal after a lag time. In our project, the use of FCS has been aimed
at determining the size of anisotropic gold nanoparticles of different shapes: (ellipsoidal
nanoparticles, called nanorods, and star-shaped nanoparticles, called nanostars) in order
to compare them to the DLS and TEM measurements. Moreover FCS has been employed
to measure the concentration of NPs stock solution, a basic information needed
to use NPs as probes for cellular imaging and to evaluate their toxicity. NPs are excited
by means of TPE: the source of two-photon excitation is a pulsed infrared laser,
with a repetition rate of 80 Hz, a pulse width of 100 fs and an emission band between
700 nm and 1000 nm; the excitation light is focalized on the sample by the objective
of an inverted optical microscope and the emitted fluorescence is collected by the same

Chapter 5 151
objective in epifluorescence geometry and sent to the detectors. A correlation board
allows the calculation of the autocorrelation function. Once the autocorrelation functions
have been collected, nanoparticles size has been determined by the characteristic
time of translational and rotational diffusion; in fact, the nanoparticles are such as to
to exhibit well distinct time scales for translational and rotational diffusion, which can
be separated in the analytical expression of the autocorrelation function: the former is
described by a decay with a characteristic relaxation time on the order of milliseconds,
while the latter leads to a second decay, expressed by a single exponential function, on
the microsecond time scale. Auto-correlation functions have been collected by using linearly
and circularly polarized excitation light; a strong dependence on the polarization
configuration was found in the shape and amplitude of the function on the microsecond
time window and made it possible to attribute the decay in this temporal interval to
rotational diffusion. Circular polarization also appeared to be particularly appealing
since it allowed to enhance the contribution of rotational diffusion. Translational and
rotational diffusion constants were extracted from the diffusion times; by making use of
the Stokes-Einstein relations, an estimate of the effective hydrodynamic radius has been
obtained, which provides a measure for the size of the nanoparticles.
Results similar to that obtain by TEM measurements, in which the rotational contribution
is even more pronounced can be obtained by means of two-photon (TPE)
luminescence correlation spectroscopy.
NPs luminescence is excited by means of TPE. The autocorrelation or cross-correlation
functions were computed and the nanoparticles size was determined from the characteristic
time of translational and rotational diffusion obtained from the ACF fitting. In
fact, the nanoparticles studied here exhibit well distinct time scales for translational and
rotational diffusion, which can be separated in the analytical expression of the autocorrelation
function: the former is described by a decay with a characteristic relaxation
time of the order of milliseconds, while the latter leads to a second decay, expressed by
a single exponential function, on the microsecond time scale. Auto-correlation functions
have been collected by using linearly and circularly polarized excitation light; a strong
dependence on the polarization configuration was found in the shape and amplitude of
the function on the microsecond time window and made it possible to attribute the decay
in this temporal interval to rotational diffusion.
Circularly polarized excitation light has been obtained through a λ/4 waveplate positioned
below the entrance pupil of the objective. In general it is difficult to position
the waveplate in order to convert the linear into fully circularly polarized excitation,
so the rotational component τ R is not completely extincted/lost (see for example figure

152 Anisotropic nanoparticles
5.13) Translational and rotational diffusion constants were extracted from the diffusion
times and, by making use of the Stokes - Einstein relations, an estimate of the effective
hydrodynamic radius has been obtained. The hydrodynamic radius provides a measure
for the size of the nanoparticles, approximated to spheres, and, when evaluated from the
translational diffusion coeffcient, it has been found in good agreement with the estimate
obtained by means of a transmission electron microscope (TEM). In order to take into
account the larger relevance of the difference from the spherical shape, further analysis
was conducted for nanorods: their translational and rotational diffusion coefficients have
been compared to those obtained through Tirado and Garcia de la Torre’s theory [Ref.].
5.7 Results
Extensive ACFs’ analysis has been undertaken to separate the contributions of rotational
diffusion and translational diffusion.7,14-16 With typical translational diffusion times being
of the order of milliseconds, the rotational diffusion, on the order of microseconds,
contributions are well separated in the ACFs. According to Kask et al.16 and Widengren
et al.,31 the rotational diffusion term of the total auto-correlation function can be
separated from the translational diffusion term and, to a first approximation, be written
as a single exponential function:
G(τ) =
(
1 + A )
1 − A e−τ/τ R
(
1 + 8D T
ω 2 0
G T (0)
) √ 1 + 8Dt
ω0
2
(
λ √
2πω0
) 2
(5.13)
This analytical expression of the correlation function has been derived assuming
spherical diffusors (having isotropic polarizability) and fluorescence lifetimes much shorter
than the rotational correlation times. The effective hydrodynamic radius R h was derived
by extracting τ D and τ R from the correlation function, determining the diffusion constant
D and the rotational diffusion contant Θ, and using the Stokes-Einstein relation:
D =
k BT
6πηR h
∝ 1 L
(5.14)
Θ =
k BT
8πηRh
3 ∝ 1 L 3 (5.15)
where k B is the Boltzmann constant, T is the temperature, η is the solvent viscosity,
R h is the particle radius and L is the length of nanoparticle’s branches.
FCS ACFs were acquired for the samples synthesized with both LSB and CTAB
surfactant.

Chapter 5 153
The fitting of FCS curves acquired with both circular and linear polarized excitation
polarization provides the translational and rotational diffusion coefficients D and Θ.
From these and eqs. 5.14 and 5.26 we obtain the hydrodynamic radius, which is a first
estimate of the nanopaticle’s branch length. Figures 5.11,5.12 and 5.13 show the FCS
curves and the distribution of hydrodynamic radii obtained for the samples synthesized
with 0.2, 0.45 and 0.6 M LSB surfactant.
Figure 5.11: Left:FCS curves obtained with linear and circular polarization. Right: Distribution of the
hydrodynamic radii determined through relation 5.14 and 5.26 for sample obtained with 0.2M LSB.
Figure 5.12: Left:FCS curves obtained with linear and circular polarization. Right: Distribution of the
hydrodynamic radii determined through relation 5.14 and 5.26 for sample obtained with 0.45M LSB.

154 Anisotropic nanoparticles
Figure 5.13: Left:FCS curves obtained with linear and circular polarization. Right: Distribution of the
hydrodynamic radii determined through relation 5.14 and 5.26 for sample obtained with 0.6M LSB.
The best fitting results are summarized in table 5.6:
for each sample the mean
translational and rotational diffusion times τ D and τ R , the diffusion coefficients D and
Θ and the mean hydrodynamic radii R T H
R and RROT
H
are reported (table 5.6). For
comparison, the branch length derived from TEM images, L T EM , and the aspect ratio
(A.R.) T EM are shown in table 5.7 (see also table 5.1).
Sample τ R [µs] Θ [µs −1 ] D [µm 2 /s −1 ] τ D [ms]
0.2 M LSB 33.1±4.2 0.008±0.001 7.0±0.3 7.8±0.3
0.45 M LSB 22.0±1.7 0.009±0.001 3.5±0.2 15.6±0.8
0.6 M LSB 25.2±2.1 0.008±0.001 4.2±0.3 13.2±0.9
Sample
R T h R [nm] R ROT
h
0.2 M LSB 34±2 30±1
0.45 M LSB 67±4 27±1
0.6 M LSB 57±4 29±1
[nm]
Table 5.6: The tables show the mean translational and rotational diffusion times τ D and
τ R , the diffusion coefficients D and Θ and the mean hydrodynamic radii R T H
R
for anisotropic NPs obtained with variable LSB concentration.
and RROT
H

Chapter 5 155
Sample Population (%) R T EM [nm] (A.R.) T EM [nm]
0.2 M LSB 49.4 54.5±3.0 6.9±0.8
0.2 M LSB 37.6 35.8±1.9 3.9±0.4
0.45 M LSB 58.8 60.9±1.9 8.1±1.1
0.45 M LSB 30.4 38.7±5.0 4.5±0.8
0.6 M LSB 30 64.3±11.1 7.8±1.4
0.6 M LSB 36 39.9±4.7 3.5±0.5
Table 5.7: The table shows the values of the branch length derived from TEM images
R T EM and the aspect ratio (A.R.) T EM .
The hydrodynamic radii obtained from the translational and rotational diffusion
coefficient for the sample synthesized with 0.2 M LSB are in good reciprocal agreement
with DLS measurements (see table 5.4). In particular R T R
∼ =RROT demonstrating that
spherical approximation is valid in this case. On the contrary, the radii obtained from D
and Θ are different from DLS measurements for samples synthesized with 0.45 and 0.6
M LSB. This reflects the inhomogeneity of NPs populations; in fact, as shown in section
5.4.2, the samples consist of sherical, star-like and NPs with high aspect ratio branches.
Because Θ ∝L −3 , Θ is largely affected by high aspect ratio branched NPs.
We pass now to the control NPs synthesized with CTAB. The FCS autocorrelation
functions for the sample synthesized with 0.2 M CTAB (nanorods) are shown in figure
5.14 and the translational and rotational diffusion coefficients are reported in table 5.8.
Figure 5.14: Left:FCS curves obtained with linear and circular polarization. Right: Distribution of the
hydrodynamic radii determined through relation 5.14 and 5.26 for sample obtained with 0.2M CTAB.

156 Anisotropic nanoparticles
Sample D F CS [µm 2 /s] Θ F CS [µs −1 ]
0.2 M CTAB 8.5±0.8 0.021±0.002
0.2 M CTAB+PEG 2000 4.3±0.2 0.031±0.004
Table 5.8: Translational and rotational diffusion coefficients obtained from FCS data.
The spherical approximation for this kind of NPs is not valid and it was not possible
to value their dimensions. However, due to TEM results (figure 5.6), we can assume for
this sample the formulation given by Tirado and Garcia de la Torre for the translational
and rotational diffusion coefficient of a rod:
D = k [ ( ) ]
BT L
ln + ν
3πηL D
(5.16)
with
ν = 0.312 + 0.565 D L − 0.1 ( D
L
) 2
(5.17)
and
Θ = 3k [ ( ) ]
BT L
πηL 3 ln + σ
D
(5.18)
with
σ = −0.662 + 0.917 D L − 0.05 ( D
L
) 2
(5.19)
where L and D are the nanorod length and width respectively, η is the viscosity and
T the temperature. By replacing L and D with the values found in TEM measurement,
we computed the values of D and Θ reported in table 5.9 If the theoric and experimental
values (reported in table 5.8) are compared, the NPs aggregation can be inferred through
χ 2 minimization, which is of 4-5 NPs for the sample considered.
Sample D T G [µm 2 /s] Θ T G [µs −1 ]
0.2 M CTAB 17.4±0.1 0.044±0.001
0.2 M CTAB+PEG 2000 17.8±0.1 0.039±0.001
Table 5.9: Translational and rotational diffusion coefficients obtained evaluated from
Tirado-Garcia de la Torre relations (eqs. 5.18-5.19).

Chapter 5 157
5.8 Concentration evalutation
As we said, the interest of FCS is in the NP counting. The concentration of the samples,
in terms of the number of NPs per excitation volume, was calculated from the G(0)
values (see also section...):
G(0) = γ N = 0.076
N
(5.20)
Because the excitation volume is known, the number of particles/L and also the molar
concentration can be derived, as reported in table 5.8:
Sample C [nM] C [mg/ml]
0.2 M CTAB+PEG 2000 8.06±1.73 2.06±0.44
0.3 M LSB 8.6±0.6 1.7±0.2
0.35 M LSB 10.3±2.2 1.5±0.3
0.45 M LSB 6.2±0.4 1.03±0.06
0.5 M LSB 1.05±0.07 0.32±0.02
5.9 TPL dependence on excitation power
The nonlinear nature of the TPL signal emitted by the gold nanoparticles was confirmed
by measuring the dependence of the luminescence intensity I T P L as a function of the
average excitation power P exc . Two different samples were evaluated:
• gold nanoparticles synthesized with the surfactant CTAB 0.2 M
• gold nanoparticles synthesized with the surfactant LSB 0.4 M
The excitation wavelenght was set equal to the wavelength of surface plasmon resonance
λ SP R , which is λ exc =λ SP R =780 nm for the first sample (CTAB 0.2 M) and
λ exc =λ SP R = 900 nm for the second one (LSB 0.4 M) (figure 5.3). The measure was
performed on a drop of gold nanoparticles solution dried on top of a coverslip; the signal
is obrained averaging 3 images through a 535/50 nm pass band filter. The evaluation
of the signal level on the images was accomplished by computing an average on at least
5 different regions of interest (ROIs) taken from bright regions of the image. Signal
intensities were collected for increasing incident power from 0 to 2 mW. Figure 5.15
show the log-log plot of the emitted TPL intensity for the samples analysed; a quadratic
dependence of the signal intensity I T P L on the input power was observed, with slope

158 Anisotropic nanoparticles
values of 2.15 ± 0.16 in the range 0.1-0.8 mW and 1.95 ± 0.04 in the range 0.1-1 mW
for the gold nanoparticles synthesized with CTAB and LSB, respectively.
Figure 5.15: Quadratic dependence of the signal intensity I T P L on the input power for samples obtained
with 0,45M LSB(A) and 0.2M CTAB (B).
To confirm also qualitatively that the emission signal is due to a two-photon excitation
process, the laser was switched in the contiuous wave (CW) operation mode. The same
excitation power was set for both the CW and the 80 MHz-pulsed laser configurations:
in the first case (figure 5.16) no signal was emitted by the samples: irradiation of the
nanorods in a wide range of power levels reduced the signal to background noise level,
confirming the two-photon nature of the nanorod luminescence.
Figure 5.16: Irradiation of nanorods in CW laser mode (left) and 80 MHz-pulse laser configuration
(right).
5.10 Emission Spectra
TPL spectra were obtained by using a solution of gold nanorods at different excitation
wavelengths (see Fig.1 c-e), variable in the range 730-920 nm with P exc ≈2 mW. In figure
5.17 the emission spectra of the different sample synthesize with CTAB (image D) and
LSB surfactant are reported (A,B,C).
The emission spectra is a broad band in the visible region (400-650 nm); the cut-off
at 670 nm is due to the presence of a dichroic filter in the optical path which avoid the
incident radiation reaching the CCD.

Chapter 5 159
Figure 5.17: Emission spectra of NPs synthesizez with 0.35M LSB (A), 0.5M LSB (B),0.35M LSB
(C), 0.2M CTAB (D).
The emission spectrum includes the 6s-5d (L) transition (electron-hole recombination
near L point of the Brillouin zone) and 6s-5d (X) transition (recombination near X point
of the Brillouin zone), which occurs for bulk gold at 518 and 654 nm, respectively (3).
The peak positions in the TPL spectra are independent on the excitation energy and on
LSB concentration (and the NP aspect ratio as consequence).
5.11 Excitation spectra
The overall TPL intensity correlated strongly with excitation wavelength and reached a
maximum at plasmon resonance, indicating a plasmon-enhanced two-photon absorption
cross section.
In order to reconstruct the excitation spectra and to determine the relation between the
TPL and the surface plasmon resonance (SPR), the area subtanded under the emission
TPL spectra was calculated for each excitation wavelenght λ exc and the values were
reported as a function of λ exc , as shown in figure 5.18. We find (figure 5.19) that the
excitation spectrum (open squares) overlaps well with the longitudinal plasmon band,
indicating that the TPL intensity is governed by the local field enhancement from the
plasmon resonance.

160 Anisotropic nanoparticles
Figure 5.18: Two-photon excitation spectra for the samples synthesized with 0.45M (red) and 0.5M
LSB (black) and 0.2M CTAB (blue).
Figure 5.19: Superposition of extinction and excitation spectra for samples obtained with 0.45 M LSB
(left) and 0.2 M CTAB (right).

Chapter 5 161
5.12 Dependence of TPL on the Polarization of the exciting
field light beam
The TPL intensity of the nanorods was examined as a function of polarization angle θ
between the incident laser beam polarization and the NPs principal axis for the samples
synthesized with 0.3, 0.4 M LSB and 0.2 M CTAB.
A drop of nanoparticles solution was dispersed and immobilized onto glass cover slips such
that isolated particles could be irradiated by fs-pulsed excitation at their longitudinal
plasmon resonance wavelength with an average excitation power of 1 mW on the sample.
In the optical path of the experimental set-up reported in figure.. and described in
chapter..., an half-wave plate was inserted in order to rotate the polarization of the
incident radiation on the whole 360 ◦ range. The TPL image was obtained by averaging
3 images collected through a 535/50 nm pass band filter. The evaluation of the TPL
signal of the NPs was accomplished by computing the average signal on different regions
of interest (ROIs) on the images, taken around single bright spots ascribed to isolated
particles. The signal corresponding to each particle was then graphicated versus θ.
We assume that TPL is maximized when the incident field, tuned at the plasmon
resonance, has its polarization parallel to the long axis of nanorod or the principal axis
of a branche nanostar (25, 26).
Two typical results are shown in Fig. 5.20 and 5.21; in both cases, the data can be
fitted to a cos 4 function.
Figure 5.20: Dependence of TPL emission on incident polarization for NPs synthesized with 0.40M
LSB.

162 Anisotropic nanoparticles
Figure 5.21: Dependence of TPL emission on incident polarization for NPs synthesized with 0.2M
CTAB.
The polarization anisotropy ν defined as
ν = I max − I min
I max + I min
(5.21)
is a good measure of the sensitivity to the polarization vector where I max and I min
are the TPL intensity for the polarization excitation oriented along the longitudinal and
transverse axis, respectively.
For every spot of each separated sample the ν parameters were calculated and their
frequency count histograms are shown in figure 5.22. From a gaussian fit of the histograms
it was possible to obtain the results reported hereafter:
• ν=0.70±0.15 for sample obtained with an LSB concentration of 0.3 M
• ν=0.9±0.1 for 0.4 M LSB
• ν=0.9±0.1 for 0.2 M CTAB
The anisotropy parameter increases with the increasing of the LSB concentration and
therefore with the aspect ratio (and the anisotropy) of the nanoparticles synthesized.
The polarization anisotropy can then be used as simple method to obtain a rough evaluation
of nanoparticles anisotropy.
Having characterized the NPs from structural an spectroscopic point of views, we now
pass to analyze their application in the biomedical research. Cellular uptake of gold
nanoparticles is a relevant issue in this field that should be explored when considering
their use in diagnostic and therapeutic applications. Additional issues involve cytotoxicity
and will be treated later in this chapter.

Chapter 5 163
Figure 5.22: Histograms of the anisotropy distribution for NPs obtained with 0.3 and 0.4 M LSB (Left)
and 0.2 M CTAB (Right). The anistropy increases with the increasing of LSB concentration.
5.12.1 Citotoxicity
The toxicity of the NPs was tested in order to verify their potential application in in-vivo
cell imaging. The cells viability was tested after the treatment for 24h with an increasing
concentration of the NPs, as shown in figures for NPs obtained with 0.2 M CTAB-PEG
2000, 0.35 M LSB and 0.5 M LSB, respectively. The histograms show the percentage
of cells in vitality conditions in dependence of an increasing concentration of gold NPs.
The concentration of the NP suspension was obtained by FCS using V=(4/3)πR 3 h and
ρ Au
∼ =19.3 g/cm 3 .
Samples were rinsed three times with MilliQ; no aggregation effect was present (measured
by DLS; data not shown).
Figure 5.23: Cell viability in dependence of increasing NPs concentration. NPs were obtained with 0.2
M CTAB; C is the control case in which no NPs were added to cells.

164 Anisotropic nanoparticles
Figure 5.24: Cell viability in dependence of increasing NPs concentration. NPs were obtained with
0.35 M LSB; C is the control case in which no NPs were added to cells.
Figure 5.25: Cell viability in dependence of increasing NPs concentration. NPs were obtained with 0.5
M LSB (figure ?? C is the control case in which no NPs were added to cells.
From these data we can infer that the nanoparticles synthesized with 0.2 M CTAB and
functionalized with PEG polymer are not toxic for the cells also at high concentration.
Otherwise the number of dead cells increase rising the concentration of the NPs obtained
with LSB surfactant; in particular at concentration used in the experiments reported
above the 80% of the cells are in vitality conditions for NPs obtained with 0.35 M LSB
(figure 5.24), and this percentage lower to ≈70% for NPs synthesized with 0.5 M LSB
(figure 5.25). The toxicity is probably due to the LSB surfactant: in fact also after 3
centrifugation and resuspension in Milli-Q water cycles, a tiny amount is present in the
solution which may be the origin of the cells death. In order to reduce the toxicity, it
is necessary to shield the surfactant with PEG or polyelectrolyte layers (Viability ∼ =95%,
figure 5.23.

Chapter 5 165
5.13 Cellular uptake
Their strong TPL signal, resistance to photobleaching, chemical stability, ease of synthesis,
simplicity of conjugation chemistry, and biocompatibility make gold anisotropic
NPs an attractive contrast agent for two-photon imaging of cells.
We have therefore investigated the ability of the NPs to permeate cell membranes,
by measuring the cellular uptake from HEK-293 cells and mice macrophages in tissue
plated cells. We used HEK cells because they are easy to grow and representative of
epithelial cells; macrophages can instead be used to study the effect of gold NPs on the
immune system, in fact they act in both non-specific defense (innate immunity) as well
as to help initiate specific defense mechanisms (adaptive immunity). In particular the
uptake of four different samples was analized, as shown in table:
Number
Sample
1 NPs synthesized with 0.5 M LSB
2 NPs synthesized with 0.35 M LSB
3 NPs synthesized with 0.2 M CTAB
4 NPs synthesized with 0.2 M CTAB and functionalized with the polymer PEG-2000
Images have been recorded 30 min after the addition of the NPs by exploiting two
photon excitation at 800 nm. The TPL emission was detected in a wide spectral range,
through 485/30, 535/50 and 600/40 band pass filters. Images shown in the following
panels are the result of 5 kalman average scans with 10 µs of residence time per pixel.
The absence of relevant bleeding through of autofluorescence has been verified on non
stained cells by measuring the fluorescence emission with and without the band pass filter
used to select the TPL emission of the NPs. Measurements performed under the same
laser power in the absence of NPs showed that cells autofluorescence was negligible both
for HEK and macrophages cells. Figure 5.26 shows that autofluorescence of macrophages
(left) and HEK cells (right) with an excitation power P exc =50 mW is negligible; otherwise
with P exc =150 mW a strong autofluorescence signal is measured through the 485/30,
535/50 and 600/40 nm band pass filters (see figure 5.27).

166 Anisotropic nanoparticles
Figure 5.26: Autofluorescence of macrophages and HEK cells with λ=800 nm and P exc=50 mW.
Figure 5.27: Autofluorescence signal of macrophages cells selected through 485/30, 535/50 and 600/40
nm band pass filters, from left to right, with P exc=150 mW
By restricting P

Chapter 5 167
Figure 5.29: NPs obtained with 0.35 M LSB with C=40 µg/ml internalized in HEK cells. The three
images refer (from left to right) to the NPs emission selected through 485/30, 535/50 and 600/40 nm
band pass filters with P exc=20 mW
Figure 5.30: NPs, obtained with 0.2 M CTAB and functionalized with PEG-2000, with C=100 µg/ml
internalized in HEK cells. The three images refer (from left to right) to the NPs emission selected through
485/30, 535/50 and 600/40 nm band pass filters with P exc=15 mW
HEK cells are representative of epithelial cells. We have also investigated cells of the
immuno response, since NPs tipically switch on immunoreaction.
An early study of the interaction of NPs and cells of the immune system, such as
macrophages, was then accomplished. Not only do macrophage-based misfunctions play
a central role in auto-immune diseases such as artherosclerosis, diabetes or rheumatoid
arthritis, but these cells are also frequently the host for parasitic organisms such as Toxoplasma
gondii, Mycobacterium tuberculosis and Listeria monocytogenes (Moghimi et
al., 2005). Therefore the concept of designing nanoparticle vehicles that will attach to
and/or be selectively ingested by macrophages appears to hold considerable potential.
Figures 5.31 and 5.32 show that macrophages internalize NPs obtained with 0.2 M CTAB
(C=100 µg/ml) and 0.35 M LSB (C=40 mug/ml) respectively; the NPs are distributed
in cytoplasmatic vesicles but not in the nuclei and are visible with the excitation power
P exc ≈ 15 mW. In order to demonstrate this fact, the nuclei were stained with the DAPI
dye: in figure 5.32 the nuclei are shown in blue (DAPI emission was selected through the
band-pass filter 485/30 nm) and the emission of NPs in red (band-pass filter 600/40 nm).
Instead, macrophages cells don’t internalize the PEGylated nanorods (figure 5.33),

168 Anisotropic nanoparticles
in fact coating the particle with thiolated polyethylene glycol (PEG) renders it invisible
to the immune system ; PEGylation is known to reduce interaction of particles with
cells due to the formation of a hydrophilic stealth coating around the particles leading
to reduced uptake (Hamblin et al., 2003).
The cells, stained with sample 4, are visible only with P exc ≈100 mW, which is the
power that produces the autofluorescence signal. No NP signal was detected even of this
power.
Figure 5.31: NPs obtained with 0.2 M CTAB with C=100 µg/ml internalized in macrophages cells.
The three images refer to the NPs emission selected through 485/30, 535/50 and 600/40 nm band pass
filters with P exc=15 mW ; NPs are internalized in cytoplasmatic vesicles.
Figure 5.32: NPs obtained with 0.35 M LSB with C=40 µg/ml internalized in macrophages cells.
The nuclei, stained with the DAPI dye and are shown in blue (DAPI emission was selected through the
band-pass filter 485/30 nm) and the emission of NPs in red (band-pass filter 600/40 nm). P exc=15 mW

Chapter 5 169
Figure 5.33: Autofluorescence signal from macrophages cells incubated with pegylated NPs. P exc=100
mW
These preliminary results indicate that NPs are viable as a fluorophore for cell imaging,
although more detailed experiments are needed in order to establish the kinetics
and mechanism of the process, the influence of the kind of cells used, concentration of
gold nanoparticles on cellular uptake, adsorption of proteins and toxicity. The positive
outcome of the investigation reported here and in the following is that they can serve as
multifunctional imaging and therapeutic agents for specific targeted cells.
Moreover, in order to verify the possibility to detect gold NPs also in explanted
organs, a concentration of C=1 mg/ml was inhaled by mice 2 , the lungs were extracted,
dissected and fixed on a coverslip.
Figure 5.34 shows the autofluorescence signal of alveolar tissue selected with the bandpass
filters 485/30, 535/50 and 600/40 nm with P exc =20 mW: it is negligible in the green
range of emission spectrum. In figure 5.35 and 5.36 the emission of the NPs is broadband
and can be seen through the 3 band-pass filters with an emission signal higher than the
autofluorescence, especially in the green channel; in the superposition of the three images
the NPs are shown in white.
Figure 5.34: Autofluorescence signal of alveolar tissue through 485/30, 535/50 and 600/40 nm (from
left to right) band pass filters with P exc=20 mW)
2 in collaboration with Dr. P.Mantecca, Dr. G.Soncini and Dr.Maurizio Gualtieri, Dipartimento di
Scienze dell’Ambiente e del Territorio (Polaris Project), Universit di Milano-Bicocca, Milano

170 Anisotropic nanoparticles
Figure 5.35: Signal due to autofluorescence of alveolar tissue and NPs luminescence through 485/30
(up, left), 535/50 (up, right) and 600/40 nm (bottom, left). In the superposition of the three channel
(bottom, right) the NPs luminescence is shown in white.
Figure 5.36: Signal due to autofluorescence of alveolar tissue and NPs luminescence through 485/30
(up, left), 535/50 (up, right) and 600/40 nm (bottom, left). In the superposition of the three channel
(bottom, right) the NPs luminescence is shown in white.

Chapter 5 171
5.14 Photothermal effects of gold NRs
The appeal of gold NRs as contrast agents for imaging is ”corroborated” (enhanced)
by their additional capability to serve as photothermal agents, with as high as 96% of
the absorbed photons converted into heat by nonradiative processes (98). The in vitro
photothermal effects of NRs have been reported by a number of groups on cultured tumor
cells (31,33,91,92), as well on parasitic protozoans (93), macrophage (94) and bacterial
pathogens (95,96). The potential of using NRs for in vivo photothermal therapy has
recently been demonstrated by Dickerson et al. who exploited the enhanced permeation
and retention effect for the accumulation of PEG-NRs in tumor xenografts in mice (97).
Gold NRs exhibit a high optothermal conversion efficiency, with a larger absorption crosssection
at NIR frequencies per unit volume than most other types of nanostructures
(98). The absorption of light energy is essentially instantaneous, and faster than the
relaxation processes which mediate its release as heat. While initial photon absorption
rates are on the fs timescale, the electron-phonon transition is on the order of a few ps,
and heat diffusion to the surrounding media generally requires tens to hundreds of ps
(82). At low powers or photon densities, heat diffusion is efficient and the result is a
measurable increase in temperature, in agreement with the widely appreciated concept
of local hyperthermia. At high photon densities, the repetitive absorption of photons by
gold NRs can exceed the rate of heat diffusion and lead to an extremely rapid rise in
local temperature and superheating, resulting in cavitation effects.
It is clear that in order to apply thermal phototherapy in medical field is essential
to know the temperature at which the NPs interact with cells. In order to measure the
local temperature’s rise on the particles surface, induced by the absorption of infrared
radiation, we have devised a novel sensor based on the conjugation of Rhodamine-B
fluorophores with NPs and tested it on anisotropic gold nanoparticles (nanorods).
Our aim is to exploit changes of the dye excited-state lifetime induced by the temperature
increasing of gold anisotropic nanoparticles due to laser irradiation, as reported
in the following.
5.15 Experimental details
5.15.1 Sample preparation
Nanoparticle-Dye complexes
Gold nanorod-dye complex is based on electrostatic bond among/between negative and
postive charged polyelectrolyte, and the NPs. Since gold nanorods stabilized with CTAB

172 Anisotropic nanoparticles
show strong cytotoxicity, polyelectrolyte coating was used/developed in order to enable
in-vivo application of the sensor. Moreover multiple polyelectrolyte layers avoid a ”direct”
interaction between gold nanoparticles and fluorophores, which could result in
fluorophore damage or quenching of its fluorescence emission by resonant energy transfer.
The functionalization was performed as reported in the following; about 10 mL of
as prepared NRs was centrifuged twice at 5,000 rpm for 10 min, the supernatant was
discarded, and the precipitate was redispersed in 5 mL CTAB (0.2 M). Subsequently, it
was added dropwise to 5 mL of the negative charged polyelectrolite PSS (2 g L −1 ) aqueous
solution. After 1 h adsorption time, it was centrifuged twice at 5,000 rpm to remove
excess polyelectrolyte and dispersed in 5 mL deionized water. Finally, the PSS-coated
GNRs were added dropwise to 5 mL of the positive charged polyelectrolyte PAH (2 g
L −1 ) aqueous solution. After 1 h, it was centrifuged twice at 5,000 rpm to remove excess
polyelectrolyte and dispersed in 5 mL of deionized water. The procedure was iterated
and a second layer of PSS was added. In order to obtain the NP-dye complex, the NPs
were added dropwise to 5 mL Rhodamine-B solution (≈ 5 nM). Finally, the solution was
centrifuged at 5,000 rpm to remove excess unbound dye molecules.
Zeta potentials (ζ) were measured to follow the formation of the NPs-dye bioconjugates
3 . The ζ-potential of the coated GNRs was measured after deposition of each layer,
as shown in Fig. 5.37. The ζ-potential after the adsorption of each layer is reported in
table 5.10 and in figure 5.37.
Layer ζ potential [mV]
PSS (first) -36 ± 4
PAH (first) 12± 7
PSS (second) -26±2
Rhodamine-B 49.7±0.5
Table 5.10: ζ-potential as a function of the number of PSS-PAH layers adsorbed to the
GNRs.
3 Measures were taken on a Malvern Zeta sizers at the Biotechnology Department.

Chapter 5 173
Figure 5.37: ζ-potential after the adsorption of PSS-PAH layers.
5.15.2 Spectral characterization
The fluorescence emission spectra of Rhodamine-B dye were acquired on a Varian Eclipse
spectrofluorimeter (Varian, U.K.).
5.15.3 Fluorescence Spectroscopy
The description of the system is reported in section ... The lifetime histograms were
computed on the selected bursts for Np-dye complexes and on the background for free
Rhodamine-B dye.
The free Rhodamine-B lifetime histograms were fitted to a single exponential function,
while the NP-dye complexes fluorescence decays were fitted to double exponential functions
with fractional intensities, f 1 and f 2 =1-f 1 , and relaxation times, τ 1 and τ 2 .
5.15.4 Dye-Lifetime measurement
Lifetime measurements in frequency domain were performed by the multifrequency crosscorrelation
phase and modulation fluorometer K2 (ISS, Illinois).
The beam, coming from Argon ion laser, is reflected by a series of mirrors on a two waypolarizer
and later directed into a Pockel cell, which is responsible for modulating the
excitation light. A radio-frequencies synthesizer (Marconi 2022C) regulates the voltage
- in the range of 0.6-330 MHz - given to the birefringent crystal (Potassium dihydrogen

174 Anisotropic nanoparticles
phospate) of the Pockel cell, causing a shift between the ordinary and the extraordinary
beam. A beam splitter divides the output beam from Pockel cell in two parts: a beam is
detected by a photomultiplier tube (PMT) and represents the reference signal for excitation
phase and modulation measurements; the other reaches the thermostatting sample
compartment, a two cuvettes rotating chamber, computer controlled, for the sample and
for a reference dye necessary in lifetime measurements. Light from sample and reference,
selected by a pass band filter in order to eliminate Rayleigh and Raman scattering,
is orthogonally detected by a second PMT powered by 3W amplifier (Model 525LA,
EN). PMT (R928 Hamamatsu) voltage is manually controlled, with the maximum gain
of 900V. Moreover, the PMT is coupled with a second frequencies synthesizer, which
generates frequencies synchronized with Pockel cell synthesizer but different by 80 Hz
in order to perform the Cross Correlation Detection (see next Paragraph). A dedicated
board processes the signal from PMT and allows to change amplifier gain depending
on sample signal intensity. The data analysis is performed by a proper software (Vinci
Analysis) that includes lifetime calculations, analysis of the fluorescence anisotropy and
phase and modulation resolved spectra.
5.16 Results
5.17 Rhodamine-B characterization
5.17.1 Fluorescence emission measurement
The dependence of the Rhodamine-B emission spectra in solution was first measured
while increasing the temperature from 10 to 60 ◦ C. As shown in figure 5.38A and in
table 5.11 the peak emission intensity decreases increasing the temperature.
Figure 5.38: A: Rhodamine-B emission spectra while increasing the temperature. B: peak emission
intensity vs. temperature. The linear fit is described by the relation 〈τ〉=[2.4-0.029T( ◦ C)] ns

Chapter 5 175
Temperature [ ◦ C] Emission intensity [a.u.]
10 882±8
15 794±8
20 704±7
25 631±6
30 552±7
35 487±8
40 429±7
45 379±7
50 331±6
55 291±6
60 255±6
Table 5.11: Dependence of Rhodamine-B emission intensity on temperature.
The quantum yield, defined as:
φ =
k R
k R + k NR
(5.22)
where k R and k NR are the rate of radiative and non-radiative decay of Rhodamine-
B, respectively. It can be assumed that the decrease in the Rhoamine-B fluorescence
emission (and as consequence of φ) of the fluorophore is due to the increasing of k NR ,
i.e. the number of non-radiative de-excitation, as k R is typically not affected when no
resonant energy transfer is present.
5.17.2 Lifetime measurement
We characterized the fluorescence response of Rhodamine-B in solution also in terms of
its excited-state lifetime τ increasing the temperature.
The results are reported in table 5.12.
As shown in table 5.12 and in figure 5.39 τ also decreases by incresing the temperature.
Since τ is:
τ = 1/K =
1
k R + k NR
(5.23)
in agreement with the emission, we can infer from the data that k NR increases leading
to the excited-state lifetime decrease.

176 Anisotropic nanoparticles
Temperature [ ◦ C] Lifetime [ns]
20 1.85±0.09
27.8 1.68±0.08
32.3 1.51±0.07
35 1.44±0.07
39 1.27±0.06
43.5 1.15±0.05
46.6 1.09±0.05
50.7 0.95±0.04
58 0.80±0.04
Table 5.12: Dependence of Rhodamine-B excited state lifetime on temperature.
Figure 5.39: Emission intensity (red) and excited-state lifetime (blue) of Rhodamine-B dye decrease as
a function of the temperature

Chapter 5 177
In figure 5.39 the fluorescence intensity and lifetime of Rhodamine-B as a function
of T are reported. The trend observed is very similar confirming our hypothesis that k R
stays ≈ constant with T.
As shown in figure 5.38B data are well approximated by a linear fit, described by the
following relation:
〈τ〉 = [2.4 ± 0.3 − 0.029 ± 0.005T ( ◦ C)]ns (5.24)
which enables to evaluate the temperature increase T from the dye excited state lifetime
〈τ〉.
These results show that Rhodamine-B is a good sensor of the temperature in the
range 10-60 ◦ C. We can then conjugate it with gold NPs in order to create a novel assay
for the online monitoring of the local temperature on GNRs to be used for photothermal
therapy.
5.18 Assay characterization
A solution with a concentration ≈300 pM of gold nanoparticles-fluorophores complexes
was irradiated with an infrared laser with power ranging from 10 mW to 125 mW and the
wavelength set at the surface plasmon resonance λ=λ SP R = 800 nm which excites both
the T sensitive dye and the NPs (with the consequent fluorescence and TPL emission),
and heats the NPs.
The time fluorescence traces of the assays show several fluorescence burst which are
ascribed to the transit of the complexes through the excitation volume. The lifetime
histograms were computed on the selected bursts and the fluorescence decays were fitted
to double exponential functions with fractional intensities, f 1 and f 2 =1-f 1 , and relaxation
times, τ 1 and τ 2 . From this analysis the average lifetime was also computed as:
〈τ〉 = f 1 τ 1 + f 2 τ 2 (5.25)
τ 1 ≤400 ps was ascribed to the NPs lifetime, while τ 2 ≈1-3 ns is due to the Rhodamine-
B dye lifetime.
< τ > decreases while rising the laser power and presumably the NPs surface temperature,
as shown in figure 5.41 and in table 5.13. The histograms of < τ > distribution
are also reported in figures ?? and 5.40.

178 Anisotropic nanoparticles
Figure 5.40: Superposition of mean excited state lifetime histograms
Figure 5.41: Dependence of mean excited-state lifetime on incident power.

Chapter 5 179
P [mW] < τ > [ns]
10 1.74±0.2
20 1.66±0.2
40 1.52±0.16
75 1.35±0.07
100 1.25±0.01
125 1.0±0.1
Table 5.13: The table shows the < τ > values of the NPs-Rhodamine B assay obtained
rising the laser power.
If we assume that the decrease of < τ > is due to a T increase on the surface we can
infer the local temperature by writing equation 5.24 written in the form:
T =
2.4 − 〈τ〉
0.029
(5.26)
This links the local temperature due to surface NPs heating to < τ > of the
Rhodamine-B dye.
In table 5.14 the temperature values corresponding to the Rhodamine-B mean lifetime
are reported as a function of the T.
Figure 5.42: Dependence of surface NPs temperature, measured through Rhodamine-B excited state
lifetime, on incident power.

180 Anisotropic nanoparticles
P [mW] < τ > [ns] T [ ◦ C]
10 1,74±0,25 22,6±0,9
20 1,66±0,2 25,5±1,1
40 1,52±0,16 30,3±1,2
75 1,35±0,07 36,0±1,4
100 1,25±0,01 39,6±1,6
125 1,0±0,1 47,6±1,9
Table 5.14: The temperature values corresponding to each Rhodamine-B mean lifetime
are shown.
Data reported in these sections show that, once the link between the dye excited state
lifetime and the temperature is known, the increase in the NP local surface temperature,
due to laser heating, can be deduced from the measurement of the dye τ (figure 5.42).
Moreover, the fluorescence-excited state lifetime of the free dye in solution was measured
increasing the laser radiation power in the same range (10-125 mW) in order to
verify that the temperatute rising was due to NPs surface heating. It was found that
the Rhodamine-B lifetime τ=1.85 ns doesn’t depend on the laser power in the range
explored. This prove that the decrease in dye excited state lifetime is due to NP surface
heating and as consequence, the assay is able to play the role of molecular thermometer
to measure local tempeature.
5.19 Conclusion
In this chapter we reported the characterization of asymmetric branched gold nanoparticles
obtained by using for the first time a zwitterionic surfactant, laurylsulphobetaine
(LSB), in the seed growth method approach 4 . LSB concentration in the growth solution
allows to control the dimension of the NPs and the LSPR position, that can be tuned
in the 700-1100 nm Near Infrared range. The samples have been analized with several
techniques to obtain a complete characterization: from the data obtained through the
absorption spectra in the UV-Visible region, the TEM images of the solutions, FCS and
DLS experiments, we reached information on the nanoparticles shape and dimensions.
In particular, in the case of LSB surfactant three different populations have been found:
nanospheres with diameter lower than 20 nm, nanostars characterized by large trapezoidal
branches, and asymmetric branched nanoparticles with high aspect ratio.
4 The synthesis was proposed and performed by the group of Prof. P.Pallavicini, at University of Pavia
(General Chemistry Department)

Chapter 5 181
The dependence of the luminescence intensity on the incident light intensity for samples
realized with LSB and CTAB was studied, checking the quadratic dependence and
therefore the two photon nature of the observed luminescence (TPL). By measuring the
luminescence spectra for different excitation wavelengths with a spectral CCD camera,
it has been possible to obtain the excitation spectra of the NPs, very similar to the extinction
spectra: this result is indicative of the role played by surface plasmons in TPL.
As a further characterization, the dependence of the TPL on polarization angle of the
incident radiation was checked.
The use of gold anisotropic nanoparticles as bright contrast agents for two-photon
luminescence (TPL) imaging of cells was also explored: NPs synthesized with both LSB
and CTAB surfactant are internalized in the cytoplasm of HEK and macrophages cells;
otherwise pegylated gold NPs are invisible to the immune system and none cellular uptake
was detected.
Finally we have developed a novel hybrid metal-organic system, based on NPs complexed
to RhB by electrostatic adsorption on multiple poly-electrolyte layers, in order to
detect the local temperature around the NP induced by the laser heating. Rhodamine-
B-NPs assay is able to play the role of molecular thermometer and can be tested in-vivo
to evaluate the damage temperature of tumoral cells. The hybrid sensor in fact can be
used for imaging and photothermal purposes at the same time: the infrared laser radiation
can excite both Rhodamine-B and TLP from the gold NRs in order to visualize the
intarnalization of the sensor in cells and the same radiation can heat the NRs, damaging
the tumor cell.

Chapter 6
In-vivo microscopy
Engineered nanoparticles are at the forefront of the rapidly developing field of nanomedicine
as they can be injected into the body to perform specific medical applications: fluorescent
agents for imaging, drug delivery carriers, or therapeutic agents for the destruction
of cancer cells (for instance in thermolysis), just to name a few.
Because of their size range, 10-100 nm, nanoparticles are particularly suitable for manipulations
at the molecular level, for example cell-receptor binding for site-selective
imaging and targeting, localization of encapsulated therapeutics for delivery, and decoration
of expression systems for substrate-based nanosensing.
Thanks to the considerable increase in the brightness of some NPs (gold NPs and quantum
dots) relative to conventional organic dyes, nanoparticles can be exploited for the
detection of pathological processes in-vivo, with improved sensitivity and resolution, at
their earliest stage and, for monitoring in real time cellular and molecular processes in
vivo and the effectiveness of therapies.
Moreover, NPs can stimulate and/or suppress the immune responses. Their compatibility
with the immune system is largely determined by their surface physico-chemical
properties (size, shape, charge, surface groups, and so on). The modification of these factors
can significantly reduce the immunotoxicity of nanoparticles and make them useful
platforms for drug delivery. For example, nanoparticles may keep drugs away from blood
cells and normal tissues, releasing them only at targeted sites; they may also decrease
the immunotoxicity of a drug by improving their solubility.
Nanoparticles can also be engineered to serve as vaccine carriers and adjuvants
(agents added to a vaccine to augment immune responses toward antigens). Adjuvants
can act in several ways to increase both native and adaptative immune responses that
will generate an effective immunological memory. Ideally, an adjuvant should improve
antigen presentation and increase co-stimulatory molecules and cytokine production.
182

Chapter 6 183
Currently, only aluminum salts have been used in licensed human vaccines (Rabinovich,
N.R., McInnes, P., Klein, D.L., Hall, B.F., 1994. Vaccine technologies: view to the
future. Science 265, 1401-1404.). As in previous chapter we have focused here on gold
NPs. In this scenario, the use of Au NP conjugates as adjuvants may offer several natural
advantages: rational design, low toxicity when pegylated (Shukla, R.; Bansal, V.;
Chaudhary, M.; Basu, A.; Bhonde, R. R.; Sastry, M. Langmuir 2005, 23, 10644-10654.),
low-cost and modifiable biodistribution (molecules bound to an NP travel through different
routes through the body than those that are unbound).
In addition to an intrinsic immunogenicity of small (20 nm) gold NPs, conjugation
of an antigen to an Au NP can modify the antigen delivery mechanism and other immunogenic
properties.
Using nanotechnology, particles can potentially be engineered to possess certain properties
so they resemble pathogens and are dealt with in the body in a similar way. For
example, key molecules that alert the immune system to pathogen infection are the Tolllike
receptor (TLR) ligands. Modifying particles with TLR ligands essentially creates
a pathogen-like particle that can be recognized by dendritic cells (DCs). When TLRmodified
nanoparticles are loaded with antigens, they can stimulate DCs to activate T
or natural killer (NK) cells. Such regulated delivery of an antigen to the DCs can control
the outcome of any immune response. The fundamental question, however, is whether
manipulation of nanoparticle surfaces can direct particle-bound antigens through particular
pathway for more efficient processing, presentation and clearance of the antigens.
In this scenario and in order to employ in future the peculiar emission (TPL) properties
of nanoparticles in order to visualize, with improved sensitivity, real time cellular
and molecular processes in vivo and to evaluate their immune properties, an initial study
of cell dynamics in-vivo in standard condition (with cells labelled with traditional fluorescent
dyes) has been performed in parallel with experiments described in chapters ..
and .... These experiments have brought into evidence the most critical issues in optical
in-vivo imaging, brightness of the probe, efficiency of scattering from the tissue, etc. The
overcoming of these problems could be achieved by the future use of NPs as stainer or
immune activator.
The immunological sense of these experiments is the study of the interaction between
two cell lines belonging to the immune system, dendritic(DC) and natural killer (NK). In
this field I have developed a novel analysis method that enables the characterization of
NK dynamic behavior, based on the supervising of a set of parameters (mean and istantaneous
velocity, confinement ratio, NK-DC distance, root-mean square displacement).

184 In-vivo microscopy
Moreover, I have verified that immune NK cells properties can be activated following/as
a consequence of a direct interaction with DC cells.
In this chapter, the technology of two-photon microscopy as applied to immunoimaging
is briefly discussed together with its applications to live-cell imaging in the lymph
node. The immunological problems concerning the application of two photon laser scanning
system to intravital microscopy, the details about the parameters related to the
cells motility and the description of the experimental plan are also reported.
In Appendix A, a brief introduction to the immune system and the lymph-nodes, with a
particular regard to the DC and NK cells, which are the object of this study, is reported.
6.1 Intravital microscopy
During the last 30 years the edge between optical fluorescence based microscopy and
the world of biomedical research has become thinner since Two Photon Laser Scanning
Microscopy (TPLSM) is, nowadays, one of the most powerful tool for immunological and
medical research [1] [2] [3].
Since the beginning of their collaboration the common aim of physicists and immunologists
was to shed some light on the mechanisms that drive the immune responses. In
particular in the last 30 years fluorescence based experiments took advantage from the
introduction of new techniques such as FCS, PCH, Confocal microscopy, FRET and so
on. But it was only in the ’90s with the introduction of Two-Photon Excitation (TPE)
[4] and finally with the coupling between IR excitation lasers and scanning microscopy,
that the sensitivness of fluorescence met succesfully the immediacy of imaging. Nowaday
TPE laser scanning microscopes are commonly found in laboratories and the literature
is a great bloom of publications dealing with IntraVital Microscopy (IVM) [?] [5].
Intravital microscopy (IVM) affords a view into the lives and fates of diverse immune
cell populations in lymphoid organs and peripheral tissues [6] [7] [8] [9]. The practice
of IVM was first described in the nineteenth century (Cohnheim, 1889; Wagner, 1839)
and has led to numerous discoveries, especially regarding the molecular and biophysical
mechanisms of leukocyte adhesion to endothelial cells [10]. However, advances in molecular
and genetic tools and optical equipment as well as immunological understanding
during the past few years have dramatically enhanced microscopists’ ability to observe,
quantify, and probe the complex behaviors of units such as T cells, B cells, granulocytes,
and dendritic cells (DC) in their native anatomical context. Traditional IVM
approaches to observe immune cell adhesion and migration in situ have employed twodimensional
imaging methods such as brightfield transillumination or epifluorescence
videomicroscopy [10]. These microscopy technologies and molecular tools for IVM, in-

Chapter 6 185
tact organ studies, and live cell interactions have been extensively reviewed elsewhere
[11] [12][10]. The relatively recent development of TPLSM and in general of multiphoton
microscopy (MPM) employing infrared pulsed laser excitation to generate optical
sections of fluorescent signals hundreds of micrometers below the surface of solid tissues
[4]. The combination of MPM technology with IVM (MP-IVM) enables the analysis
of cell migration in time-lapse recordings of 3D tissue reconstructions [11]. Nonlinear
microscopy differs fundamentally from linear techniques in that the elementary process
involves near simultaneous interactions of two (or more) photons, so that the signal varies
as the square (or higher power) of incident light intensity, rather than linearly.
In addition to multiphoton absorption, another nonlinear interaction that becomes prominent
at very high light intensities is that of optical-harmonic generation, in which two
(or more) photons are almost simultaneously scattered to generate a single photon with
exactly twice (or higher multiples of) the incoming energy [13]. Second harmonic generation
is produced by spatially ordered non-centrosymmetric molecules and has proved
especially useful for imaging ordered structural proteins such as collagen fibers [14] and
microtubules [15] by detecting emitted light through a bandpass filter of twice the wavelength
of the excitation light without the need for fluorescent labeling.
MP-IVM studies can take advantage of excellent tissue penetration afforded by the
infrared excitation, which allows much deeper imaging within solid organs than with
conventional single-photon excitation for confocal imaging (in lymphnodes-LN- up to 450
µm versus less than 100 µm respectively [11]). Modern laser scanheads allow the rapid
acquisition of stacks of 2D optical tissue sections, which are then digitally reassembled
into 3D rendering of the original sample. Iterative generation of multiple image stacks
at defined time intarvals can then be used to produce 3D time-lapse movies, which can
be analyzed by various means. In addition to advances in photonics, IVM has gained
momentum recently by innovations in other areas: the availability of refined fluorescent
probes, improved hardware and software for three-dimensional image analysis, and a
diverse array of drugs, antibodies, recombinant proteins, and gene-targeted mice have
contributed to the increasing acceptance and utilization of IVM as a powerful instrument
for immunological research.
6.2 Problems in IVM
The use of Two-Photon Excitation (TPE) for fluorescence and Second Harmonic Generation
(SHG) imaging of tissues is nowadays well established on a variety of model systems
[16] [10] [17] [18] [19]. The low Rayleigh scattering of infra-red (IR) radiation (for small
isotropic particles scattering scales as the inverse fourth power of the wavelength, λ −4 ) is
usually reported as the factor that mostly determines the possibility to perform optical

186 In-vivo microscopy
tomographic imaging at increased depths in tissues, with respect to single photon confocal
microscopy [20] [16] [1]. However it must be recognized that scattering from large
anisotropic particles decreases with a much smaller exponent, ∼ = λ −2 . Moreover large
deformation and micro-focusing of the laser beam are induced by micron-size objects
within biological tissues: cells and collagen fibers are the most important sources of the
laser beam scattering [16].
A second issue regarding deep tissue imaging is the need to increase exponentially the
excitation power when reaching distances ≻ 20 − 30 µm below the specimen surface, due
to short mean free paths, l S
∼ = 50 − 100 µm. In fact, most of the time, the full power
( ∼ = 1 W) of the laser beam must be employed when reaching hundreds of micrometers
in tissues, such as neocortex [?]. We must consider then that, although TPE usually
limits the out-of-focus photo-bleaching of the chromophores and photo-damage of cells
and tissues [21] [4], on the focal plane a considerable TPE induced photo-damage has
been reported, mainly due to thermal load of the sample [1].
Due to these considerations, the first requirement to be satisfied in order to obtain
sensible results in biological and medical experiments [22] is the optimization of the excitation
and detection efficiency, since it allows to minimize the photo-damage, to extend
the observation time and to reach deeper planes in the tissue. In order to overcome in
part these problems, anisotropic gold nanoparticles can be used as novel probes thanks
to their high excinction cross section (≥ 10 5 -10 6 ), larger than that of organic dye and
their high photostability in contrast to common molecular chromophores. Moreover, the
luminescence (TPL) induced by TPE is enhanced (when coupled with an appropriate
plasmon resonance) by many orders of magnitude in non spherically symmetric NPs of
noble metal with respect to the single photon excitation on smooth noble metal surfaces.
These properties promise to improve the usefulness of these nanoparticles for in-vivo
imaging in the NIR region of the electromagnetic spectrum.
Another problem is the fact that most of the losses in the detection efficiency are due to
the use of complex and inessential (for TPE) optical paths that are typical of commercial
confocal scanning systems. These, having been developed for single photon excitation
confocal microscopy, are often not optimized for IR excitation. Even if the coupling
between IR pulsed lasers and commercial confocal scanning heads is quiet easy, experimental
setups suffer of poor detection efficiency and they do not allow to control the
parameters needed to maintain the samples at physiological conditions. Our setup is
designed to overcome these problems. Thanks to an home made non descanned detection
unit (Caccia2008) it limits the looses of fluorescence photons due to the complex
and inessential path the photons have to follow back toward the detectors lodged in the
scanning heads. Moreover a custom designed plexiglass thermostatic chamber equipped

Chapter 6 187
with liquid flow channels surrounds the entire microscope allowing to control all the parameters
useful to maintain the samples at physiological conditions. The combination
of these two aspects makes our system particularly suitable for IVM measurements on
explanted organs. Further details can be found in [Caccia2008].
6.3 The biological problem
6.3.1 NK-DC interaction
Dendritic cells (DC) are specialized in the presentation of antigens and the initiation
of specific immune responses. They have been involved recently in supporting innate
immunity by interacting with various innate lymphocytes, such as natural killer (NK),
NKT or T cell receptor (see also Appendix B). The functional links between innate
lymphocytes and DC have been investigated widely and different studies demonstrated
that reciprocal activations follow on from NK/DC interactions. The cross-talk between
innate cells and DC which leads to innate lymphocyte activation and DC maturation
was found to be multi-directional, involving not only cell-cell contacts but also soluble
factors. The final outcome of these cellular interactions may have a dramatic impact on
the quality and strength of the down-stream immune responses, mainly in the context
of early responses to tumour cells and infectious agents.
The first evidence for innate immunity stimulated by DC was provided by the work of
Fernandez et al. [23], which showed a direct activation of NK cells antitumor effector
functions by DC in vivo and in vitro.
Starting from the pioneering work of Fernandez et al. [23], many groups have investigated
the role of DC derived cytokines and membrane-bound molecules in the activation of NK
cells. These studies have generally reported a major role for DC-derived type I IFN in
NK cell cytotoxic activity [24][25] [26], but two main pathways have been identified as
accounting for DC mediated NK cell interferon-γ (IFN-γ) 1 release in both humans and
mice: i) DC-mediated NK cell activation is dependent on interleukin-12 (IL-12) 2 and
other DC-derived cytokines/molecules [27] [28]; and ii) NK cell functions are strongly
boosted by DC-derived IL-2 in association with other cytokines/molecules [29] [30].
IL-12 appears to be essential for the induction of IFN-γ by NK cells in different experimental
settings [27][28]. It is required for NK cell activation and is released after DC
1 Interferons (IFNs) are proteins made and released by lymphocytes in response to the presence of
pathogenssuch as viruses, bacteria, or parasitesor tumor cells. They allow communication between cells
to trigger the protective defenses of the immune system that eradicate pathogens or tumors
2 Interleukins are a group of cytokines (secreted proteins/signaling molecules). The function of the
immune system depends in a large part on interleukins, and rare deficiencies of a number of them have
been described, all featuring autoimmune diseases or immune deficiency.

188 In-vivo microscopy
and NK cells come into contact [31], following the stimulation of DC derived with LPS.
DC-derived IL-12 seems to be presented to NK through the formation of stimulatory
synapses, ensuring the efficient presentation of low doses of IL-12 to both human and
murine NK cells [32]. In some experimental settings, IL-12 has also been shown to be
a key regulator of NK cell cytotoxicity [33]. In other studies based on peripheral blood
DCs, IL-12 and cell-cell contacts were found to play only a marginal role, suggesting that
other factors inducing NK cell cytotoxicity may be present. Other cytokines released by
DC, such as type I IFN, IL-15 and IL-18, may also affect NK cell functions such as IFNγ
production, migration, cytotoxic function and proliferation [34]. In particular, IL-18
seems to play an important role in enabling NK cells to migrate to secondary lymphoid
organs, where they can interact with DCs [35]. A membrane-bound form of DC-derived
IL-15 seems to be required to trigger activation, or at least proliferation, of NK cells
[29]. It is possible that the two different modes of NK cell activation by DCs described
in published studies reflect different conditions for DC culture and the heterogeneity of
DC populations differentiated in vivo (Fig. 6.2). Indeed, mouse and human DCs secrete
bioactive IL-12 efficiently only if previously exposed to IL-4 [?]. DCs exposed to the
semi-maturation stimulus IL-4 acquire the capacity to activate NK cells independent of
microbial stimuli and IL-2, although microbial stimuli do enhance this process. IL-4
also inhibits microbe-induced IL-2 production by DCs. Thus, IL-12 may contribute to
DC-mediated NK cell activation if DCs have previously been exposed to IL-4 in vitro or
in vivo. Moreover, blockades of IL-2 activity in vivo or in vitro with DCs derived ex vivo
result in strong inhibition of the activities ofNKcells, but not in the complete inhibition
of NK cell functions, suggesting that the heterogeneity of in vivo differentiated DCs may
result in two different pathways of NK cell activation, as previously described [?].
Many studies have suggested that cell-cell contact, in addition to soluble factors, may
play a role in the DC mediated NK cell activation process [24][36] [37]. On the one
hand, cell-cell contact probably reflects a need for the formation of activating synapses
between DCs and NK cells, facilitating the local delivery of high concentrations of known
or unknown cytokines.

Chapter 6 189
Figure 6.1: DC in close contact with an autologous NK confirmed the mobilization of IL-12 toward the
DC/NK synapse. DCs, stimulated by LPS, were admixed with autologous resting NK cells; after fixation,
cells were permeabilized and stained with antiIL-12 antibody conjugated with a fluorochrome.
forming tight conjugates with NK cells, IL-12 selectively concentrated at the DC-NK interface [32]
In DCs
Figure 6.2: Mediators of NK cell activation produced by activated DCs differentiated in the presence of
GM-CSF (Granulocyte-macrophage colony-stimulating factor) alone or GM-CSF and IL-4 [?].
In the mouse, different groups have demonstrated that few NK cells can be found
in non-inflamed LN but they can be rapidly recruited to this site during inflammation.
Martin-Fontecha et al. [26] first showed the accumulation of NK cells after injection
of adjuvants (such as R848 and Ribi) or activated DC. It was recently published [38]
that pre-activated DC alone, without in vivo injection of any soluble molecule, strongly
induced accumulation of NK cells at the draining lymph node. In particular, LPS (see
section 6.10.1 in Appendix B) and CpG 3 pre-activated DC induce a strong mobilization
3 CpG sites are regions of DNA where a cytosine nucleotide occurs next to a guanine nucleotide in the

190 In-vivo microscopy
of NK cells towards the draining LN. Other groups demonstrated the accumulation of
NK cells in the draining lymph node after the footpath inoculation of Leishmania major
[39] or after the injection of DC coltured in GM-CSF 4 and IL-4 [40].
In this scenario, the aim of the present project is to shed some light on the mechanism
of the response of the immune system in the presence of pathologic agents. In particular,
the behavior of two kind of cells (dendritic and natural killer) was studied, whose mechanism
of interaction is not completely understood but is suspected to play an important
role in increasing the efficiency of the immune system. The Dendritic (DC) and Natural
Killer (NK) cells [41] are part of the immune system and are essential in recognizing and
destroying cancer cells or cells that have been infected by a virus [42]. Recent studies
of Intra Vital Microscopy (IVM) have demonstrated the existence of an interaction between
DC and NK at the level of the secondary lymphnodes and have shown that the
presence of particular pathogenic agents modifies the interaction between DC and NK
cells [43] [39]. In light of this evidence, the aim of the present work was to understand if
and how much the ”communication” between DC and NK cells amplifies and increases
the efficiency of the immune response of the NK cells. The dynamic proprieties of the
NK cells (average velocity , instantaneous velocity v i (t), Root Mean Square RMS,
confinement ratio) have been studied through IVM in lymphnodes from mice systems
at the stationary state (no pathogenic agents), and they have been compared with the
values found in subjects where the DC were brought to maturity through contact with
a specific stimulus, LPS, a lipopolysaccharide that is a characteristic component of the
external walls of the negative Gram bacteria.
This project is a part of ENCITE - European Network for Cell Imaging and Tracking
Expertise - program that consists of 29 project partners from 10 countries with leading
expertise in the field of cell imaging, with the European Institute for Biomedical Imaging
Research as the coordinating partner.
The 4-year project started in June 2008 and comprises the following objectives:
• New imaging methods to improve the spatio-temporal tracking of labelled cells
• Dual- and multimodality imaging procedures to cross-validate each individual approach
linear sequence of bases along its length. Unmethylated CpG sites can be detected by Toll-Like Receptor
9 on dendritic cells and B cells. This is used to detect intracellular viral, fungal, and bacterial pathogen
DNA
4 Granulocyte-macrophage colony-stimulating factor, often abbreviated to GM-CSF, is a protein secreted
by macrophages, T cells, mast cells, endothelial cells and fibroblasts.

Chapter 6 191
• New contrast agents and procedures that will improve the sensitivity and specificity
of cellular labelling
• Combining of molecular biology for the generation of molecular and cellular imaging
reporters with multimodal imaging techniques
6.3.2 Sample preparation methods
Currently, two methods are mainly being used that try to mimic the situation in vivo
as closely as possible [44]. One ’semiintravital’ method is the preparation of explanted
intact organs. The excised lymph nodes, thymus or spleen are imaged while being perfused
in warm medium with or without oxygenation [45][46][18][47]. This preserves the
structural integrity of the natural tissue, but normal vascular and lymphatic circulation
are severed. A second approach is the intravital imaging of lymph nodes. In these experiments
animals are anesthetized and lymph nodes are surgically prepared [46][11][10].
Easily accessible are the inguinal lymph node of the mouse, by folding back a broad flap
of abdominal skin, or the popliteal lymph nodes of the feet [10]. A rubber ring is glued
on the inner surface of the exposed skin flap with tissue adhesive. Thereby a watertight
chamber filled with phosphate-buffered saline is formed into which a water immersion
objective is lowered for imaging. Both mouse and chamber are warmed and kept at 35
to 36 ◦ C.
In principle, both methods have certain limitations. One concern with explanted tissues
has been the maintenance of physiological oxygen tension. Whereas some investigators
have studied explanted lymph nodes in culture medium perfused with 95% O 2 and 5%
CO 2 [45], others have argued that normal oxygen tension in lymph nodes may be low
[19,26,27] and culture conditions perfused with 95% O2 might represent unphysiological
conditions causing abnormal lymphocyte motility. However, recent experiments [46]
have reported similar T cell mobility in lymph nodes of living animals breathing either
room air or 95% O 2 5% CO 2 .
In contrast, manipulations involved in the intravital approach including the trauma associated
with anesthesia and surgery could also introduce considerable artefacts, and
data are still too limited to estimate the impact on subtle cell-cell interactions. Finally,
the anatomical situation of certain tissues itself may limit the amount of data one can
collect because the available field of view is sometimes diminished in comparison with
the explant method, in which the isolated tissue can be analyzed from multiple imaging
angles [44]. In general, however, the results reported so far have shown a remarkable
concordance for both approaches with respect to the motility rates of different cell types,
the dynamics of cell movement and antigen-dependent T cell-DC contacts.
These results suggest that explant and intravital imaging techniques, at least for lymph

192 In-vivo microscopy
nodes, can provide conditions that are physiologically appropriate.
Although clearly a superior method in terms of physiology, one limitation of intravital
imaging is that anatomical constraints can prevent effective data collection due to restrictions
on the field of view within the target lymphoid structure. Explant methods,
on the other hand, circumvent this problem by allowing multiple imaging angles and
the collection of more representative data throughout the sample, while also minimizing
motion artifacts that prevent effective dynamic data collection [4][45]. The combination
of the two methods provides a mean of obtaining more complete and physiologically
relevant data sets.
6.4 Data acquisition and analysis
6.5 Results
6.5.1 Materials and methods
Fluorescence Microscopy
The laser source is a mode-locked Ti:Sapphire laser (Mai Tai, Spectra Physics, CA)
pumped by a solid state laser at 532 nm that produces pulses with ∼ = 100 fs FWHM,
repetition rate = 80 MHz and average power ∼ = 2.8 W at λ= 800 nm at the output
of the laser source. The experimental set-up is built around a confocal scanning head
(FV300, Olympus, Japan) mounted on an optical microscope (BX51, Olympus, Japan)
modified for direct (non de-scanned) detection of the signal. The laser beam is sent to
the scanning head and is driven on the sample by means of two galvanometric mirrors
and a scan lens: the galvos are responsible for the specimen scan while the lens assures
that the back aperture of the objective (N.A. = 0.95, 20X, water immersion, Olympus,
Japan) is overfilled during the entire scan process. The objective simultaneously focuses
the laser beam on the sample and collects, in epifluorescenc geometry, the harmonic
or fluorescence signal through either the descanned (FV300) or the non-descanned (NDunit,
described in section ...) detection unit. The emission is filtered through a short-pass
670 nm filter (Chroma Inc., Brattelboro, VT), splitted by two dichoic mirrors (490 and
540 nm) and selected by band-pass filters centered at the emission maximum of the
fluorophores used: 400 nm (Chroma Inc., Brattelboro, VT, full width = 20 nm, SHG),
460 nm (Chroma Inc., Brattelboro, VT, full width = 30 nm, CMAC emission), 515 nm
(Chroma Inc., Brattelboro, VT, full width = 30 nm, GFP emission), 600 nm (Chroma
Inc., Brattelboro, VT, full width = 40 nm, CMTPX emission).
The entire microscope is surrounded by a custom made thermostatic cabinet in which
the temperature is kept at 37 ◦ C and physiological conditions are guaranteed during the

Chapter 6 193
experiments by flowing 37 ◦ C buffer solutions saturated with a mixture of 95% O 2 5%
CO 2 .
Data acquisition and analysis
Practically, a two-photon imaging experiment is carried out by acquiring sequential images
of a three dimensional volume selected in a lymph node that contains fluorescently
labelled cells. This is achieved by recording the fluorescence signals at successive focal
planes and repeating this process every 10-30 seconds for periods of up to an hour and
half (Fig. 6.3B) [48] [49]. The experimental data were stored as multi-tiff time series of
volumes (typically 460x460x50 µm 3 ) collected at 100-250 µm depth below the external
capsule of the DLN where we expect to observe DC-NK interactions. The wt-GFP transgenic
DCs emits at λ wtGF P = 515 nm, while the NK and T cells are visible at λ CMT P X
= 600 nm and λ CMAC = 460 nm, respecitvely. The fluorescence signals were detected
through band pass filters at 515/20 nm (wt-GFP), 460/20 nm (CMAC) and 600/40 nm
(CMTPX). The motion and the interactions of the cells were followed in explanted DLNs
kept in physiological conditions as described above. During the observation of the NK
and DC cells one of the channel of the ND unit was devoted to the detection of the
SHG signal in order to obtain information on the DLN structure and to define the depth
of acquisition.
The multidimensional data set that is generated can be displayed as
a time-lapse movie, and important information can then be extracted through manual
or automated analyses (Fig. 6.3c). Sequences of image stacks were transformed into
volume-rendered four-dimensional movies using Volocity software (Improvision), which
was also used for semi-automated tracking of the cell in three dimensions. From the x,
y and z coordinates of the cell centroids, the principal dynamic parameters of the cells
(the mean 3D velocity, < v >, the instantaneous velocity, v(t), the Root Mean Square
displacement, < |∆x| 2 > and the confinement ratio) can be calculated using custom
scripts in Matlab (MathWorks).

194 In-vivo microscopy
Figure 6.3: Steps Involved in the Analysis of In Vivo Cell Motility by MP-IVM (A) A pulsed beam of
infrared (IR) laser light is raster scanned via a microscope objective with high numerical aperture onto
a thick specimen by rapid, synchronized movement of a pair of steering mirrors. Fluorescent signals are
generated by the multiphoton effect in a ≈1 µm deep section in the objectives focal plane (Denk et al.,
1990). (B) Incremental vertical motion of the objective relative to the specimen yields stacks of optical
sections, which are serially reacquired at defined time intervals (in this example, 15 s are required to
acquire one z stack composed of six optical sections). (C and D) In (C), each z stack is rendered as a
three-dimensional volume, and in (D) two-dimensional projections of image stacks are exported as movie
files for demonstration and further off-line analysis.
(E) Determination of cell centroids to represent
cell position allows automated tracking of migration paths of three or more cell populations recorded in
separate color channels. This is achieved by defining a track from a series of images of individual cells.
(F and G) In (F), tracks consisting of serial sets of xyzt coordinates of single cell centroids are exported
as numerical databases, and in (G) are used to compute parameters of cell motility (see also Table 2).
Specialized software for automated 2D or 3D cell tracking and computing of the acquired tracks is essential
for in-depth, large-scale analysis of migratory behavior at the single-cell level. Ref. [6]
Volocity software
Volocity is a software that follows the cells trajectories in three dimensions and it is based
on two steps: the segmentation and the recontruction of the trajectories (tracking).
The segmentation step is based on the ’edge detection’ algorithm which aim is the identification
of regions in a digital image at which the image brightness changes sharply (i.e.

Chapter 6 195
it has discontinuities). The result of applying an edge detector to an image lead to a
set of connected curves that indicate the boundaries of objects. Thus, applying an edge
detection algorithm to an image allows the identification of the cells and so significantly
reduce the amount of data to be processed and therefore filter out information that
should (it is our assumption) be less relevant, while preserving the important structural
properties of an image. If this step is successful, the subsequent task of interpreting the
information contents in the original image may be substantially simplified.
From a practical point of view, masks (named classifier) are built, based on some characteristics
the objects have to fulfill to be considered as cells (such as their volume and
brighness). The classifier is applied to each time point of the movie and the centroids of
the objects can be retrieved.
The trajectories reconstruction (or tracking) is based on the ’feature matching’ algorithm,
which joins the points obtained from the segmentation on the basis of some characteristics:
the object area, its diameter or the maximum distance between two subsequent
images. The final output is a track that identifies the movement of each object recognized
as a cell.
From the x, y and z coordinates of the cell centroids as a function of time (frames),
hereafter called the ”trajectory”, the parameters related to the cellular motility can
be calculated by using custom scripts in Matlab (MathWorks) and analized by using
OriginLab 7.5.
Sample preparation
Transgenic mice (BALB/c) with DCs expressing wtGFP (wild type Green Fluorescent
Protein) were injected with 5 million of NK and/or T cells labeled with fluorescent
dyes (CMTPX and CMAC, respectively) 24h before the experiment. The day of the
experiment the mice were sacrificed, the draining lymph nodes (DLNs) extracted, fixed
on a Petri dish and kept in in-vivo conditions as described above. For the acquisition
in inflammatory conditions mice were injected with LPS (Sigma Aldrich, USA) 30 min
before explanting the DLNs (figure 6.5).

196 In-vivo microscopy
Figure 6.4: Scheme of the sample preparation procedure. Transgenic mice (BALB/c) with DCs expressing
wtGFP (wild type Green Fluorescent Protein) were injected intra-venous (i.v.) with 5 million
of NK and/or T cells labeled with fluorescent dyes (CMTPX and CMAC, respectively) 24h before the
experiment. T cells were use only for the validation of the experimental setup performances. LPS was
injected sub-cutaneous (s.c.) 30 min before explanting the DLNs.
Figure 6.5: Scheme of the experimental procedure.
Motility parameters
Several motility parameters are used in the literature to quantify cell migration. Such
analysis starts after the experiments and then the cell tracking are carried out and
trajectrories are obtained.
• Plotting tracks: by plotting the trajectories of the tracked centres of mass (centroids)
of immune cells over time in two or three dimensions, the migratory behavior
of the cell population can be qualitatively examined. One option is to simply plot
the unshifted coordinates in the image volume, possibly overlayed on the acquired
images of the cells. This can help to see whether the visualized cells prefer particular
regions of the tissue. The other option is to shift the first position of each
individual cell to the same starting point in space while maintaining its orientation
(Fig. 6.6a), allowing us to roughly see whether cells are travelling in a preferred di-

Chapter 6 197
P t.
rection. If all possible directions are approximately equally covered, this indicates
that motion is more or less random on the timescale of the duration of the plotted
tracks. It is noteworthy that plotting tracks only gives qualitative information.
• Root Mean Square 〈 displacement (RMS). As reported in literature [50][6] the RMS
displacement, |∆x| 2〉 〈
= |x(t + τ) − x(τ)| 2〉 , can be a good descriptor of the cell
motility [Sumen 2004]. In literature three trends have been highlighted:
– If the cells are moving randomly in the lymph-node, apart from the duration
of the eventual interaction, one would expect an initial linear trend of
〈|∆x| 2〉
as a function of the observation time t (see figure 6.14blue). In this condition,
the slope, M, can be taken as a measure of the cell diffusion coefficient, M=6D
(figure 6.7);
– in absence of specific interactions with the environment, a directional motion
is observed. The cell follows a uniform linear trend: the displacement is
proportional 〈 to time t and, as consequence, a quadratic increase is observed
in the plot of |∆x| 2〉 vs t (figure 6.14green) and the motion is defined balistic.
In the lymph-nodes this trend is observed in the initial time interval (tipically
few minutes), for 0 < t < P t 5 , as shown for the experimental RMS data in
figure 6.7. If this trend persists also for t> P t , the cell follows a preferred
direction (persisten motion);
– otherwise, at the increasing of t, if the cell is confined due to interaction with
other cells or physical and biological 〈 restriction (e.g., due to the influence
of long- range chemoattractants), |∆x| 2〉 trend in function of t shows a
transition of the linear segment to a plateau (figure 6.14red).
• Speed: the mean speed over the brief interval between two sequential time frames
can be approximated by dividing the distance the cell travels by the time period
between the frames. Because in reality a trajectory will not be exactly straight,
this is an underestimate of the true speed of the cell. The longer the time period
between the frames, the larger this error will be.
However, this error remains
limited providing the time period between frames to be shorter than that for which
cells tend to move in a persistent direction.
• Confinement ratio: this is a measure of the extension or confinement of the cell
tracks. It is the ratio of the displacement of a cell to the total length of the path that
the cell has travelled (FIG. 6.6b). Because the path length is always at least the
5 The time interval to the transition from the exponential to the linear segment is the persistence time,

198 In-vivo microscopy
distance of the displacement, the confinement ratio varies between 0 (a completely
condensed track, so the cell returns to the exact position where it started) and 1 (a
perfectly straight track). An issue to overcome with the confinement ratio is that
its value tends to zero as the track duration goes to infinity (Fig.6.6b uncorrected
ratio). This can be seen by noting that the confinement ratio is closely linked to a
mean displacement analysis; thus a comparison of cell tracks of different durations
and of different experiments is problematic [51][50]. One way to solve this is to
refer the confinement ratio to fixed duration, but this means discarding shorter
cell tracks as well as parts of cell tracks that exceed the chosen duration. Another
way to circumvent the problem of dependency on the cell track duration is to
multiply the confinement ratio by the square root of time; this simple correction
removes the dependency on track duration (Fig. 6.6b, corrected ratio). To use
this approach to directly compare cell tracks of different durations, the total track
duration should be longer than the typical duration of persistent motion, and
the time interval between sequential images should be shorter than the typical
duration of persistence. The disadvantage of this method is that the values of
the confinement ratio are not unitless and are not restricted to between 0 and 1.
However, the confinement ratio is not a sufficient condition to infer the constraint
of the motion, in fact it does not provide information about the temporal step
between the start and the end of the track. An example of this behavior is shown
in figure 6.8: the confinement ratio is 0.05 ∼ =0, so a confined motion could be
inferred; instead, the NK comes back to the initial point after a long trajectory.
In this light it is necessary to evaluate the instantaneous value of the confinement
ratio in order to evaluate deeply his trend.
• Contact time analysis: in addition to analysing immune cell migration, a highly
significative investigated parameter is, when present, the duration of interactions
between immune cells. Usually, contact durations in different experimental conditions
are compared and correlated to functional readouts of an immune response
[52] [53]. The contact duration, or the distribution of contact times, is a difficult
parameter to analyze and to determine because imaging is limited both in time and
space. This means that for many contacts the initiation and/or the termination
is not observed, and that the observed contact duration often underestimates the
true duration.

Chapter 6 199
Figure 6.6: Commonly calculated migration parameters. a) A track plot in which each track has been
shifted such that it starts at the origin of the x and y axes. b) The confinement ratio is calculated
by dividing the displacement of the cell by its total path length. In the corrected confinement ratio the
resulting value is multiplied by the square root of the cell track duration. The right panel shows how the
confinement ratio and its corrected version depend on the track duration. The uncorrected ratio tends
to zero for large track durations, whereas the corrected ratio reaches a constant number (for random
migration).
Figure 6.7: Mean Square Displacement Plots (MDP).

200 In-vivo microscopy
Figure 6.8: (A)shows a DC cell with a trajectory characterized by a low confinement ratio (0.05) but
not interacing; (B)instantaneous confinement ratio of the NK cell in panel A.
6.5.2 Lymph nodes topography
In order to verify the NK and DC cells distribution in the lymph-nodes, transgenic mice
(BALB/c) with DCs expressing wtGFP were injected with 10/15 million of NK and
B cells labelled respectively with fluorescent dyes CMAC (λ em =460 nm) and CMTPX
(λ em =600 nm) respectively. As reported in the literature [Scholer 2008] T cells patrol
the paracortical area of the DLNs looking for a mature DC able to activate their specific
immune action [Cahalan 2003]. NK cells were particularly enriched in the most
peripheral regions of the T-cell zone [Garrod 2007]. DCs are present in the perifollicular
zone and T cell area [Lindquist 2004], and occasional cells are present in the B cell
follicle and subcapsular space. The DC networks surrounded the B cell follicles on all
sides and extended into the T cell zones, but are particularly dense in the border zone
between the T and B cell follicle, where T cell-dependent immune responses are initiated.
We report in Fig.6.9 the 3D images of armpit and brachial lymph nodes in order to
verify the NK, B and DC cells distribution. The distribution of T cells was assumed to
be known.
Image 6.9A shows the external structure of the DLN, in particular the capsule and
the pericapsular zone. In figure 6.9B the cortical zone is reported: the B cell follicle
is visible in red, together with the distribution of DC (green) and NK cells (blue), also
shown in images 6.9C and 6.9D.

Chapter 6 201
Figure 6.9: 3D and 2D section of armpit and brachial lymph-nodes,as described in the text: green=DC,
red=B cells, blue=NK cells. The excitation wavelenght was λ exc = 780 nm with the excitation power
P=100 mW.
6.5.3 Validation of the experimental setup performances: T and DC
cells
The behavior of the T and DC cells in steady state conditions within the DLN are taken
here as a test of the reliability of the physiological conditions in which the DLNs are
kept.
From the instantaneous velocity and cells trajectories (figure 6.10) analysis we can infer
that T are highly motile and span a large area within the DLN with a 3D mean velocity
of 6±1 µm/min and in good agreement with the literature [Miller 2002]. DC cells are
instead characterized by a slower 3D mean velocity, 3±1 µm/min and follow much more
confined trajectories (figure 6.10B).
All together these observations seem to ensure that our experimental setup allows us to
observe the behavior of lymphocytes close to in vivo conditions.

202 In-vivo microscopy
Figure 6.10: (A)-(B) T and DC cells instantaneous velocity, (C)-(D) T and DC cells trajectories as
they are tracked by Volocity software on the 4D volumes collected in DLNs on the TPE microscope.
6.5.4 NKs in steady state conditions
Natural Killer cells in explanted DLNs show a behavior very similar to that observed for
the T lymphocytes. They span large area within the DLN (figure 6.11) with a 3D mean
velocity of 7±2 µm/min.
The surprising similarity of motion displayed by the T and NK cells seems to indicate
the possibility that also the NK cells are tailored for searching activating contacts with
the dendritic cells in a way similar to what already observed for the T cells[Scholer 2008].
We moved then to study any possible evidence of interactions between the NK and the
DC cells, that are, for the purpose of the present analysis, considered stationary within
the DLNs.

Chapter 6 203
Figure 6.11: (A) NK cells instantaneous velocity, (B) NK cells trajectories as they are tracked by
Volocity software on the 4D volumes collected in DLNs on the TPE microscope.
6.6 DC-NK cells dynamics
The mutual behavior between DC and NK cells were visualized and studied at the level
of the DLNs both in resting and in inflammatory conditions using the experimental
procedure illustrared in figure 6.4 and described in paragraph 6.5.1.
At the steady state condition, individual NK cells established sometimes short-lived
interactions with DCs (< 15 minutes) but were almost never observed forming stable
contacts (figure 6.12).
Figure 6.12: Interaction between a NK (red) and a DC (green) in steady-state condition. In the panels
B-E and N-Q the NK shows the short-lived contact with the same DC with ∆t ∼ =1 min in both contacts. In
panels F-L the NK moves according to a directional random motion, passing near two DC cells without
interaction.
After the maturation of the DC cells induced by the LPS injected i.v. in the mice, we

204 In-vivo microscopy
observed a marked change in the behavior of the NK cells that assumed longer and more
stable contacts (∆t>15 min) with the DCs with respect to the steady state conditions.
As an example we report in figure 6.25 the motion of a particular Natural Killer cell.
During its motion the NK seems to recognize a mature DC probably due to chemical
attraction (figure 6.25A and B), subsequently it contacts the DC (figure 6.25C, and D)
and it forms a temporarly stable contact with the dendritic cell (figure 6.25E and F).
Even if, once encountered the DC, the contact is dynamic with the NK rocking and
rolling around the DC, its motion becomes confined and, as consequence of the long
lasting of the contact (∆t>15 min), the dynamic parameters change significantly during
the interaction .
Figure 6.13: Interaction between a NK (red) and a DC (green) in stimulated condition. In the panels
A, B and C the NK moves according to a directional random motion while in the panels D and E the
NK moves to the DC. Panel F shows the stable contact DCNK: it is noteworthy the time lapse between E
(starting of the interaction) and F (established interaction) is 21 minutes. We consider stable this type
of contact.
Despite authors in literature report dynamic cell parameters, no one explains how
they can conclude that a cell is going down an interaction, depending on which physical
paramenters one can consider two or more cells interacting and how to evaluate the
contact time.
In order to obtain this information on a solid quantitative ground we have established
the routines reported in the following. At first, we identify〈
the NK probably intecting
cells, analizing the root mean square (RMS) displacement |∆x| 2〉 of the cells. As reported
above the RMS, |∆x| 2〉 〈
〈
(τ)= |x(t + τ) − x(τ)| 2〉 , is a good descriptor of the
〈 t
cell motility [Sumen 2004]. If this relationship between |∆x| 2〉 and t is linear, it means
that cells behave as randomly moving objects (figure 6.14, blue trace); a faster than
linear increase in the mean square displacement plot is reminiscent of directed motion
(figure 6.14, green trace). One factor that could cause directional motion is the presence
of a chemokine gradient that the immune cells have to follow. A slower than linear increase
(figure 6.14, red trace) of the mean square displacement plot means that the cells

Chapter 6 205
are somehow confined, for example because the interactions with other cell types keep
them within a specific region or due to physical or biological restriction.
Figure 6.14: RMS for a cell performing random walk motion (blue), directed motion (green) and
confined motion (red).
Through this first analysis step, based on the RMS trend, divides the traces of cells
presenting a plateau (indicating confinement of a cell population) from the other one.
In this way we have the first sceening able to select only NK cells that probably interact
with a DC or have a confined motion due to physical restriction. However, although the
mean displacement plot is a useful tool to investigate the type of motility involved, the
underlying mechanism of migration cannot be inferred from it. This is because multiple
underlying micro-processes can give rise to the same or very similar mean displacement
plots.
In order to select only the NK cells effectively interacting with a DC cells, a second
screening was required and therefore we created a procedure based on the following
parameters:
1. distance between NK and DC cells in NK neighborly
2. instantaneous and mean velocity of the NK
3. confinement ratio of the NK: this parameter is the ratio of the displacement of a
cell to the total length of the path that the cell has travelled and is calculated for
each timepont. This is a measure of the straightness or confinement of the cell
tracks.
We imposed that these parameters have to obey simultaneously some constraints
during interaction:

206 In-vivo microscopy
1. instantaneous velocity: the mean and the instantaneous velocity v i (t) during NK-
DC interaction should be slower than the one calculated for cells moving as free
objects. Then because even if the interaction is dynamic, the NK instantaneous
velocity shows a lot of zeros with respect the free motion part. As a consequence v i
should be lower than at least the 60% of the mean velocity < v > during DC-NK
interaction. When this request is satisfied a second graph v i (t) is obtained with
the following statement (figure 6.16A-B):
if v i =0 v i < 60% < v >
if v i =v i v i > 60% < v >
2. confinement ratio (CR): this parameter is a measure of the straightness or confinement
of cell tracks, if an NK cell interact with a DC cell it decreases or remains
constant (figure 6.16C) because when a cell interacts it is at least at resting and
so CR does not change.
3. distance: the interaction between NK and DC cells take places when the cells are
close each other, i.e. < 25 µm (figure 6.15) , which is the sum of the DC mean
radius (≈ 7 µm), the dendrites mean lenght (≈ 10 µm) and the NK mean radius (≈
6 µm). The trajectories of NK and DC cells are reconstructed and their distance
is calculated and plotted vs time, as shown for a couple of NK-DC cell in figure
6.16D. The second parameter, named T on T off (figure 6.16E)was calculated as:
if T on T off =-1 µm distance NK−DC 25
In the graph in figure 6.16E T on T off reveals when the contraint on the distance is
verified by the NK-DC couple examinated.
In figure 6.16 panels A and B show the values of the instantaneous velocity before
and after the threshold operation, panel C the confinement ratio and panels D and E
the distance and the NK-DC distance and the T on T off parameter, respectively.

Chapter 6 207
Figure 6.15: Distance between NK and DC cells.
Figure 6.16: The figure shows the parameters used to evaluate DC-NK interaction. (A) and (B) report
the plot of the NK instantaneous velocity before and after the threshold evaluation. (C) shows the value
of the NK confinement ratio as a function of time. (D) and (E) display the NK-DC distance value as a
function of time and the T onT off parameter, obtained as reported in the text.
In order to identify a contact, the parameters have to obey the constraints reported
above simultaneously; if at least one of the parameters don’t fulfill the statements, the
NK cell is not considered interacting with the DC.

208 In-vivo microscopy
Hereafter some examples in which one or more parameters don’t fulfill the constraints
and, as consequence, the NK is not considered interacting with the DC, are reported in
figures 6.17,6.18.
In particular in figure 6.17, the panel A shows the screenshot of the acquired movie where
a NK is near two DC cells. In the graph of panel B the four parameters, instantaneous
velocity (blue), confinement ratio (red), T on T off (magenta) and distance (green) are
reported together. In this case, even if the distance NK-DC2 is 25 µm; as consequence the NK
is not interacting with DC1. Also for the NK-DC2 couple (panel C), at the timepoints
93-94 and 110-113, the instantaneous velocity is and the confinement ratio
decreases, while the distance is above the threshold. We can then infer that also the NK
cell is not interacting with the DC2.
Figure 6.17: (A)The screenshot of the acquired movie of a NK a DC cell is reported. (B)The panel
shows the graph in which the four parameters (distance, instantaneous velocity, confinement ratio and
T onT off ) of the NK-DC1 couple are reported togheter. In order to visualize them simultaneously, the
instantaneous velocity and the distance are normalized and the T onT off is rescaled. In this case, even
if the distance NK-DC1 is

Chapter 6 209
Figure 6.18: (A)The screenshot of the acquired movie of a NK a DC cell is reported. (B)The panel
shows the graph in which the four parameters (distance, instantaneous velocity, confinement ratio and
T onT off ) of the NK-DC1 couple are reported togheter. In order to visualize them simultaneously, the
instantaneous velocity and the distance are normalized and the T onT off is rescaled. Even if the distance
NK-DC1 and the velocity respect the threshold (i.e. distance NK−DC1 is . In both
cases we infer that the NK doesn’t interact with the DC1 cell. Each timepoint corresponds to 24.8 s. The
image dimension is 155x135x45µm 3 .
Figure 6.19: (A)The screenshot of the acquired movie of a NK near two DC cells is reported. (B)The
panel shows the graph in which the four parameters (distance, instantaneous velocity, confinement ratio
and T onT off ) of the NK-DC1 couple are reported togheter. In order to visualize them simultaneously,
the instantaneous velocity and the distance are normalized and the T onT off is rescaled. At the timepoints
102-103 v i(t) and the confinement ratio decreases, but the distance is >25 µm; as
consequence the NK is not interacting with DC1. (C) Also for the NK-DC2 couple, at the timepoints
93-94 and 110-113, the instantaneous velocity is and the confinement ratio decreases, while
the distance is above the threshold. We can then infer that the NK cell is not interacting with the DC.
Each timepoint corresponds to 24.8 s. The image dimension is 80x80x45 µm 3

210 In-vivo microscopy
Otherwise, when the four parameters are verified simultaneously, the interaction can
be identified and the touch-time quantified.
If the constaints on v i (t), the confinement ratio, the distance and T on T off are verified
at the same time, as shown for example in figure 6.20, the cell is definitely assigned to
the interacting group and the ∆t during which the statements are fulfilled quantifies
the duration of the interaction between the NK-DC cells (green square in figure 6.20.
Other examples of DC-NK interacting cells are shown in figures 6.21 and 6.22 : panel
A shows the 3D reconstruction of the NK (black) and DC (red) trajectories as obtained
by Volocity; panel B is the screenshot of the acquired movie with NK trajectory superimposed;
the RMS parameters reported in panel 6.21C and 6.22C enabled us to classify
the NK cells as interacting bon the basis of the plateau trend; in panel D the parameters
v i (after the threshold evaluation), confinement ratio, NK-DC distance and T on T off are
overlapped in order to decide if the NK-DC cells are interacting and to quantify the
touch-time (as indicated by the green frame). Instead, figure 6.23 shows an example
of non interacting cell: the RMS in this case shows a faster than linear increase as a
function of time.
Figure 6.20: This plot shows the superposition of the parameters used to evaluate the NK-DC interaction:
confinement ratio (red), threshold instantaneous velocity (blue), NK-DC distance (green) and
T onT off
parameter (magenta). The green box highlights the time lapse in which the interaction between
NK and DC cells occurs because the parameters obey at the same time the constraints reported in the
text.

Chapter 6 211
Figure 6.21: The figure shows an example of an NK cell interacting with a DC cell. (A)3D reconstruction
of the NK (black) trajectory as obtained by Volocity; (B) screenshot of the acquired movie with
NK trajectory superimposed; (C) the RMS parameters enabled us to classify the NK cells as interacting
because of/thanks to the plateau trend; (D) the parameters v i(t) (after the threshold evaluation), confinement
ratio, NK-DC distance and T onT off are overlapped in order to decide if the NK-DC cells are
interacting and to quantify the touch-time (as indicated by the green frame).
Figure 6.22: The figure shows an example of an NK cell interacting with a DC cell. (A)3D reconstruction
of the NK (black) and DC (red) trajectories as obtained by Volocity; (B) screenshot of the acquired
movie with NK trajectory superimposed; (C) the RMS parameters enabled us to classify the NK cells as
interacting because of/thanks to the plateau trend; (D) the parameters v i(t) (after the threshold evaluation),
confinement ratio, NK-DC distance and T onT off are overlapped in order to decide if the NK-DC
cells are interacting and to quantify the touch-time (as indicated by the green frame).

212 In-vivo microscopy
Figure 6.23: The figure shows an example of a non-interacting NK cell. (A)3D reconstruction of the
NK (black) trajectory as obtained by Volocity; (B) screenshot of the acquired movie with NK trajectory
superimposed; (C) the RMS shows a faster than linear increase as a function of time showing that NK
cell is non-interacting.
The data analysis of cell trajectories has highlighted that the percentage of NK-
DC interaction increases from 14% (in standard condition) to 25% (on pathological
condition) of the total number of NK cells. Moreover, the NK cells dynamic behavior
changes radically after the LPS injection: the interactions are long-lasting and take
place for longer periods than in stationary conditions. In fact, in non-inflammatory
state, the touch time ∆t is shorter than 15 minutes (which in literature is regarded as
the boundary between long and short-lasting contacts) while, in inflammatory condition,
∆t>15 minutes in more than 50% DC-NK interactions. This analysis is supported by
the decrease of the average 3D NK interacting cells velocity < v > respect to the noninteracting
in pathological conditions, as shown in figure 6.24:

Chapter 6 213
Figure 6.24: Parameters.
6.7 Conclusion
The analysis method described in this chapter was applied to lymph-nodes from 15 mices
resulting in about a hundred of analized cells for both steady state and inflammatory
conditions. The NK cells show a well defined behavior change if one compares standard
and pathological condition.
First of all the percentage of interacting NK cells increases from the 14% to the 25%.
This increase is extremely meaningful because, also at first glance, it is the evidence that
the maturation of the DCs has an effect on the NK cells. They not only are recruited at
the DLNs (i.e. the place where the adaptive immune response is started), but they also
contact just right the DCs, the cells that are considered the bridge between the innate
and adaptive immunity.
Moreover the nature of the contacts strongly changes. Looking at the distribution
of the touch times (i.e. the time the contact lasts, Figure 6.25) it is clear thet the NK
cells, in inflammatory conditions, tend to establish longer contacts with the DCs with
respect to the stationary case. Using a cut off time of 15 minutes 6 in order to divide the
short and not stable interactions from the longer and stable ones, it is straightforward
to observe that the number of long lasting contacts increases after the LPS injection: if
6 The value of 15 minutes is choosen on the base of a simple thuoght: an experimental acquisition has
an extent of about 60 minutes during which the cells can diffuse in or out the observation volume with
the same probability, as a consequence we can infer that a single cell spends half of the experiment time
inside the observation volume (i.e. 30 minutes) and conclude that a contact lasting for more than 15
minutes (a duration that covers more than half a track) can be considered as stable.

214 In-vivo microscopy
only less than the 10% of the interacting NK cells establish stable contact in steady state
conditions, this percentage increases to more than the 50% in inflammatory conditions.
Figure 6.25: Distribution of touch times in non-inflammatory and inflammatory condition.
Finally, in independent cytofluorimetric experiments, it was calculated that LPS leds
the 28% of the NK cells stimulated to maturation. This means that 28% of the monitored
NK cells are able to give rise to an immune response, a percentage that matches
strikingly well the one founded during the TPE experiments for stable interacting NK
cells.
In summary, althuogh we don’t know which chemical or bioogical agent drives the
interactions and this analysis method cannot explain the physiological function of the
NK-DC contact, the protocol can clearly highlight and charcterize the contacts both in
steady and inflammatory contions allowing us to infer relevant conclusions from not only
a physical but also a bilogical point of view.
The existence of long lasting interactions between NK and DCs in inflammatory conditions
together with the agreement between cytofluorimetric and TPE data confirms
that the NK cells can undergo an activation mechanism driven by other cellular species.
This is very similar to what happens for cells involved in the adaptive immunity and
moreover seems to be parallel to the NK cells innate immune activity. Our conclusion
does not agree with the classical classification of the NK cells in the world of the immune
system where they are considered as a part of the innate immune hemisphere only. A
further confirmation of the obtained results would lead to a review of the collocation

Chapter 6 215
of the natural killer cells in the immune system. In particular it would be possible to
conclude definitively that besides their role in the innate immune response, they are able
to take a part in the adptive immunity too; a more meaningful classification of the NK
cells have to be done placing them at the edge between adaptive and innate immunity
rather than confine them in the innate immune system.
Moreover we can believe our analysis method as a complete novelty for interpreting
TPE data. In fact, despite the litterature is a bloom of paper depicting dynamic cell
parameters and how they have to be to reveal an interaction, no one explains how he can
conclude that a cell is going down an intercation. Here we give a protocol based on well
konwn and simple physical parameters that taken together allow not only to discerne
real contacts from the false positives but also to determine their temporal duration. In
addiction, since the protocol is based on simple parameters, it results intuitive, rapid,
unambigous and sets the user free form multiple movie visualizations resulting in a great
save of time 7 .
7 In the inflammatory case the time required by the analisys of the touch is also reduced by the first
screening of the cells based on the RMS.

216 In-vivo microscopy
6.8 Appendix A
6.9 Immune system
All living organisms, from bacteria through humans, have evolved strategies to counter
parasitic infections [54] [55]. In higher organisms, the varied and numerous strategies
involved in defense from parasitic microbes are collectively referred to as the immune
system. The mammalian immune system consists of two interrelated arms: the evolutionarily
ancient and immediate innate immune system, and the highly specific, but
temporally delayed, adaptive immune system (figure 6.27). The combination of innate
and adaptive immunity enables the mammalian immune system to recognize and eliminate
invading pathogens with maximal efficacy and minimal damage to self, as well as
to provide protection from re-infection with the same pathogen.
Figure 6.26: The innate and adaptive immune response. The innate immune response acts as the
first line of defence against infection. It consists of soluble factors, such as complement proteins, and
diverse cellular components including granulocytes (basophils, eosinophils and neutrophils), mast cells,
macrophages, dendritic cells and natural killer cells. The adaptive immune response is slower to develop,
but manifests as increased antigenic specificity and memory. It consists of antibodies, B cells, and CD4 +
and CD8 + T lymphocytes. Natural killer T cells and γδ T cells are cytotoxic lymphocytes that straddle
the interface of innate and adaptive immunity. [54]
The innate response includes soluble factors, such as complemet proteins, and several
cellular effectors, including granulocytes, mast cells, macrophages, dendritic cells (DCs)
and natural killer (NK) cells. Innate immunity serves as the first line of defence against
infection. By contrast, adaptive immunity, mediated by antibodies, B cells and CD4 +

Chapter 6 217
and CD8 + T cells, is slower to develop. This reflects the requirement for the expansion
of rare lymphocytes that harbour somatically rearranged immunoglobulin molecules, or
T-cell receptors that are specific for either microbial derived proteins or processed peptides
that are presented by Major Histocompatibility Complex (MHC) molecules 8 [54].
NKT cells and γδ T CELLS are cytotoxic T lymphocytes that act at the intersection of
innate and adaptive immunity.
T cells are lymphocytes, and play a central role in cell-mediated immunity. They can be
distinguished from other lymphocyte types, such as B cells and natural killer cells (NK
cells) by the presence of a special receptor on their cell surface called T cell receptors
(TCR), which is able to recognize a non-self peptide presented by the major histocompatibility
complex (MHC). The abbreviation T, in T cell, stands for thymus, since this
is the principal organ responsible for the T cell’s maturation.
B cells are lymphocytes which principal functions are to make antibodies against antigens,
perform the role of antigen-presenting cells (APCs) and eventually develop into
memory B cells after activation by antigen interaction. In mammals, immature B cells
are formed in the bone marrow, which is used as a backronym for the cells’ name [54].
A critical difference between B cells and T cells is how these lymphocytes recognize the
antigen. B cells, contrary to T cells, recognize their cognate antigen in its native form.
They recognize free (soluble) antigen in the blood or lymph using their B cell receptor
(BCR) or membrane bound-immunoglobulin.
The innate and adaptive immune systems use two fundamentally different strategies
to recognize microbial invaders specifically: the innate immune system detects infection
using a limited number of germ-line encoded receptors that recognize molecular structures
unique to classes of infectious microbes, while the adaptive immune system uses
randomly generated, clonally expressed, highly specific receptors of seemingly limitless
specificity. It is the combination of these two strategies of recognition that makes the
mammalian immune system highly efficacious.
A conceptual framework for the current understanding of the functioning of the innate
immune system and its control of adaptive immunity was proposed by Charles
Janeway Jr nearly 20 years ago [56]. Janeways hypothesis was essentially as follows: the
adaptive immune system, because of the use of randomly generated receptors for antigen
recognition, cannot reliably distinguish between self and non-self. Therefore, adaptive
immune cells must be instructed as to the origin of an antigen by a system that can
determine, with high fidelity, whether an antigen is derived from self, infectious (i.e. mi-
8 The major histocompatibility complex (MHC) is a large genomic region or gene family found in most
vertebrates that encodes MHC molecules. A MHC molecule inside the cell takes a fragment of microbial
invaders proteins and displays it on the cell surface.

218 In-vivo microscopy
crobial) non-self, or innocuous (i.e. non-infectious and non-microbial) non-self. Janeway
suggested that the evolutionarily ancient, and at that point severely understudied, innate
immune system might be able to provide such instruction. Furthermore, he proposed
a concrete mechanism by which the innate immune system could sense the presence of
infection and relay its conclusions to the adaptive immune system. Janeway posited
that the innate immune system would sense the presence of infection via recognition of
conserved microbial pathogen-associated molecular patterns (PAMPs), by means of the
germ-line encoded receptors he dubbed pattern recognition receptors (PRRs). These
PAMPs would possess certain qualities to be effective. First, they would have to be
unique to microbes and absent from eukaryotic cells so that they would accurately signal
infection. Second, they would be common to a broad class of microbes so that a limited
number of germ-line encoded receptors could detect all infections. Third, they would be
essential for the life of the microbe so that their detection could not be easily eliminated
via mutation. Maybe most importantly, Janeway also predicted that recognition of infection
by PRRs on cells of the innate immune system would lead to the induction of
signals involved in activation of the adaptive immune system and thus would result in
initiation of adaptive immunity. This conceptual framework has now been proven to be
correct, although new advances naturally require further development of the theory.
6.10 PRRs and control of adaptive immunity
All PRRs can detect the presence and type of microbial infection and activate the appropriate
innate immune response [57] [58]. Some PRRs can also control adaptive immunity.
This instruction of adaptive immunity occurs largely through triggering the maturation
of DCs from highly phagocytic, weakly immunogenic, tissue-resident cells into weakly
phagocytic, highly immunogenic, lymph node-homing cells that are competent to induce
tailored T-cell responses to non-self antigens acquired in the periphery. Not all PRRs
are equal in terms of their ability to trigger the adaptive immune response. Indeed,
while some PRRs are sufficient to induce both T- and B-cell responses, other PRRs (e.g.
the mannose receptor and scavenger receptors) are not competent to induce adaptive
immunity by themselves [57]. Presumably, the ability and sufficiency of a given PRR
to control adaptive immunity should correlate with the ability of that particular PRR
to accurately detect the presence, extent, and duration of infection, and the microbial
origin of the antigens, as well as to effectively relay this information to the adaptive
immune system.

Chapter 6 219
6.10.1 TLRs
TLRs are the canonical PRRs that fulfill all of the predicted qualities of a PRR that links
innate and adaptive immunity that is, TLRs sense infection through the recognition of
PAMPs and induce appropriately tailored innate and adaptive immune responses. TLRs
have also been shown to be critical for the induction of adaptive immunity in response
to various immunizations and infections [57]. TLRs have been shown to control adaptive
immune responses at multiple levels, including control of antigen uptake and antigen
selection for presentation in DCs, control of DC maturation and cytokine production.
TLR4 is the best characterized member of the TLR family; it is activated by lipopolysaccharide,
LPS [59] [60] [61] (i.e. the main constituent of the external membrane of the
Gram negative bacteria) and together with CD14 (a membrane protein belonging to
the glycosylphosphatidyl-inositol-anchored receptor, GPI-AR, family) and MD-2 (a little
protein formed by only 160 aminoacids) constitutes a receptor complex necessary and
sufficient for a full response to the LPS stimulus.
Lypo-Poly-Saccharide (LPS)
The external bacteria membrane determines the Gram color answer and allows the classification
of the various species as positive or negative Gram bacteria. The membrane of
the positive Gram bacteria is composed mainly of peptido-glycan: it is compact, complex
and it does not show high flexibility; instead, for negative Gram bacteria, the external
membrane is more flexible and allows to give specific responses to the external stimulations.
The external layer is thinner and presents a low number of internal link respect to
that of positive Gram bacteria. The outer part of the membrane is rich of amphypatic
molecules such as LPS, protein and lypo-proteins [62]; within those, LPS s the most
interesting due to its anti-genic role. LPS is composed by three different and distinct
domain. (1) The lipidic portion, called Lipid A, is formed by a dimer of glucosammine
phosphorilate: it is associated to the external side of the membrane where it substitutes
phospholipids. The lipid A is called endotoxin because it is involved in the genesis of
a wide number of infective diseases. It is responsible for the activation of the dendritic
cells. (2) The central portion, called core, constituted of a limited number of sugars, it is
expressed from all the member of the negative Gram bacteria family. It is characterized
by the presence of different structural variants within the same kind of bacteria species.
(3) The external lateral chain, called antigen-O, composed by a series of replicas of the
same tetra- or penta-saccharide unit. The characteristic of the antigen-O, that works as
a barrier against antibiotics, determines the serotype of the LPS; for example, if the side
chain is composed by long O-polysaccharidic tail, the LPS is called smooth LPS; instead
variants without the antigen-O show a rugous surface and are called rough LPS

220 In-vivo microscopy
The described structural variability of LPS, localized in the core and in the antigen-O,
translates in the great heterogeneity of the LPSs that compound the external negative
Gram bacteria membrane. The distinction between the different form of LPS is of great
importance for the dynamics of reaction of the cells [63].
Figure 6.27: Physiological functions of Toll-like receptors (TLrs). TLRs are involved in recognition of
microbial and endogenously derived molecular patterns. This occurs both at the plasma membrane and
at intracellular compartments. After ligation of TLR ligands either directly or with the help of accessory
molecules such as cD14, MD2 (also known as LY96) and cD36, TLRs dimerize and transmit signals
throughout the cell by means of adaptor molecules such as myeloid differentiation factor 88 (MYD88)
and TRiF (TIR-domain-containing adapter-inducing interferon-β. It mediates the rather delayed cascade
of two TLR-associated signaling cascades, where the other one is dependent upon a MyD88 adapter.).
This leads to the activation of multiple cellular phenomena, the best-described of which being the activation
signal transduction to the nucleus (such as through activation of nuclear factor-κB (NF-κB), MAPKs
(Mitogen-activated protein (MAP) kinases) and interferon regulatory factors (iRFs).
TLR activation
leads to regulation of innate and adaptive immune responses, inflammation and tissue repair.
lipopolysaccharide; the MD-2 is a protein that appears to associate with toll-like receptor 4 on the cell
surface and confers responsiveness to lipopolysaccaride (LPS), thus providing a link between the receptor
and LPS signaling; cDxx, Cluster of differentiation xx: group of cell surface marker proteins. [64]
LPs,
6.11 Dendritic cells
The main function of dendritic cells is to process antigen material and present it on the
surface to other cells of the immune system, thus functioning as antigen-presenting cells.
They act as messengers between the innate and the adaptive immunity. Dendritic cells

Chapter 6 221
arise from myeloid progenitors within the bone marrow, and emerge from it to migrate
in the blood to peripheral tissues. In these tissues, they have an immature phenotype
and are characterized by high endocytic activity and low T-cell activation potential.
Immature DC constantly sample the surrounding environment for pathogens such as
viruses and bacteria, through PRRs, such as TLRs. When a dendritic cell takes up
pathogen in infected tissue, it becomes activated, and travels to a nearby lymphnode.
On activation, the dendritic cell matures into a highly effective antigen-presenting cell
(APC) and undergoes changes that enable it to activate pathogen-specific lymphocytes
that it encounters in the lymphnode (a model known as Langerhans cell paradigm [65]
[66], figure 6.28 9 ). Activated dendritic cells secrete cytokines that influence both innate
and adaptive immune responses, making these cells essential gatekeepers that determine
whether and how the immune system responds to the presence of infectious agents.
Figure 6.28: The regeneration of Langerhans cells. Langerhans cells (LCs) are normally generated and
maintained locally in the steady state from precursors in the epidermis itself. This is sufficient to produce
the low-level, steady-state efflux of LCs to the draining lymph nodes. In mice, replenishment of LCs from
bone-marrow precursors can only be observed experimentally following depletion of the LC population by
local skin inflammation. Then, the LC precursors are replenished by inflammatory monocytes that enter
the epidermis from the bloodstream. This emergency replenishment of LCs might also be a model for the
origin of LCs from non-inflammatory blood monocytes during early development, and perhaps a model
for an ongoing, low-frequency event in the steady state [67].
9 According to this paradigm, DCs are present in peripheral tissues in an immature state that is
specialized for sampling the environment using various endocytic mechanisms, but is characterized by
low levels of expression of MHC molecules and T-cell co-stimulatory molecules. Immature DCs are
well equipped with a series of receptors for pathogen-associated molecular patterns and for secondary
inflammatory compounds, such as Toll-like receptors (TLRs). Signalling through these receptors triggers
DC migration towards the secondary lymphoid organs. On reaching these organs, DCs develop into
a mature state, which is characterized by high levels of expression of MHC and T-cell co-stimulatory
molecules, and the ability to present antigen captured in the periphery to T cells. According to this
pathway, DCs would provide the necessary link between the probable points of pathogen entry and the
lymph nodes, bringing in and presenting antigens that T cells would otherwise not be able to detect.

222 In-vivo microscopy
6.12 Natural Killer cells
Natural killer (NK) cells are effector lymphocytes of the innate immune system that
control several types of tumors and microbial infections by limiting their spread and
subsequent tissue damage. Recent research highlights the fact that NK cells are also
regulatory cells engaged in reciprocal interactions with dendritic cells, macrophages, T
cells and endothelial cells. NK cells can thus limit or exacerbate immune responses.
Although NK cells might appear to be redundant in several conditions of immune challenge
in humans, NK cell manipulation seems to hold promise in efforts to improve
hematopoietic and solid organ transplantation, promote antitumor immunotherapy and
control inflammatory and autoimmune disorders. Natural killer cells (NK cells) develop
in the bone marrow from the common lymphoid progenitor cell and circulate in the
blood. They are larger than T and B lymphocytes, have distinctive cytoplasmic granules,
and are functionally identified by their ability to kill certain lymphoid tumor cell
lines in vitro without the need for prior immunization or activation [68].
Their functions can be classified into three categories [69]:
• Cytotoxicity: NK cells possess large numbers of cytolytic granules, lysosomes containing
mainly perforins and various granzymes.
• Regulatory capabilities mediated by cytokines and chemokines release: NK cells
exert their activity by producing high amount of IFN-γ, that activates a strong
inflammatory response. Indeed, other than IFN-γ, NK cells are able to produce
many other important cytokines and chemokines, including myeloid differentiation
and activation factors such as IL-3, GM-CSF, CSF-1, TNF-; type 2 cytokines (IL-5,
IL-13); IL-10; and chemokines such as MIP-1, RANTES and IL-8.
• Contact-dependent co-stimulation: NK cells also express several costimulatory ligands,
allowing them to provide direct co-stimulation to T cells and B cells.
6.13 Lymph-nodes
The lymphoid organs are organized tissues containing large numbers of lymphocytes in
a framework of non-lymphoid cells. In these organs, the interactions lymphocytes make
with non-lymphoid cells are important either to lymphocyte development, to the initiation
of adaptive immune responses, or to the sustenance of lymphocytes. Pathogens
can enter the body by many routes and set up an infection anywhere, but antigen and
lymphocytes will eventually encounter each other in the peripheral lymphoid organs-the
lymph nodes, the spleen and the muscosal lymphoid tissues. Lymphocytes are continually
recirculating through these tissues, to which antigen is also carried from sites

Chapter 6 223
of infection, primarily within macrophages and dendritic cells. Within the lymphoid
organs, specialized cells such as mature dendritic cells display antigen to lymphocytes.
Lymph draining from the extracellular spaces of the body carries antigens in phagocytic
dendritic cells and macrophages from the tissues to the lymph node via the afferent
lymphatics [28][70].
In the event of an infection, lymphocytes that recognize the infectious agent are
arrested in the lymphoid tissue, where they proliferate and differentiate into effector
cells capable of combating the infection. Because they are not involved in initiating
adaptive immune responses, the peripheral lymphoid tissues are not static structures
but vary quite dramatically depending upon whether or not infection is present [28].
The LNs have several important functions in the immune system: to recruit large
numbers of naive lymphocytes from the blood [27]; to collect antigen and DCs from
peripheral tissues; to provide the environment for antigen-specific tolerance or productive
primary and secondary effector responses; to modulate the homing characteristics of
effector or memory T cells, targeting them to tissues that contain their cognate antigen;
and finally, to provide a look-out for central memory cells. To carry out these diverse
functions the cellular participants must migrate into the LNs and find their correct place
within them.
Figure 6.29: Scheme of the lymph-nodes in mice. 5)armpit LN; 6)brachial LN; 9)popliteal LN [27]

224 In-vivo microscopy
6.13.1 Lymph-node architecture
Two main regions can be distinguished histologically in LNs-the cortex and the medulla
(FIG. 6.30).The cortex is further divided into the paracortex (the T-cell area), and the
more superficial B-cell area that consists of primary follicles and (after antigen challenge)
germinal centres [9]. B-cell follicles are the main site of humoral responses, whereas the
paracortex is the site where circulating lymphocytes enter the LNs and where T cells
interact with DCs [71]. The medulla is a labyrinth of lymph-draining sinuses that are
separated by medullary cords, which contain many plasma cells, and some macrophages
and memory T cells. The function of the medulla is still poorly understood.
The fine architecture of the LN cortex is complex and variable [72]. Nevertheless, common
structural features have been identified(FIG. 6.30). On the basis of electron microscopy
studies, it has been proposed that the paracortex is arranged in paracortical
cords that originate between or below the B-cell follicles and extend towards the medulla
where they merge into medullary cords [73]. The cords are bordered by lymph-filled cortical
sinuses and permeated by reticular fibres.At the centre of each paracortical cord is
an HEV (High endothelial venules) that is surrounded by concentric layers of pericytes
known as fibroblastic reticular cells (FRCs). A narrow space between the basement
membrane of the HEV and the pericytes is known as the perivenular channel. It has
been proposed that this channel receives an ultrafiltrate of lymph from the FRC conduit.
Networks of FRCs that are often arranged in spiral layers around the HEVs enclose
10-15 µm wide corridors along which lymphocytes are thought to migrate [74]. Venous
blood flows through HEVs along a venular tree, the trunk of which is formed by a large
collecting venule in the medulla. This venule drains into a large vein at the hilus a
discrete region where the capsule is penetrated by efferent lymph and blood vessels. Adhesive
interactions between leukocytes and endothelial cells are absent in LN arterioles
and capillaries, but they occur frequently in venules. Recent work indicates that the
medulla-associated segments of the venular tree express unique adhesion molecules that
are distinct from those expressed by HEVs.
HEVs are normally only found in secondary lymphoid tissues (except for the spleen)3.
The composition and distribution of lymphocyte traffic molecules on HEVs differs between
lymphoid organs, and the presence and function of HEVs is regulated throughout
life.

Chapter 6 225
Figure 6.30: Lymph-node architecture. a) Schematic diagram showing the major structural components
of a lymph node. The main routes of lymph flow into and within lymph nodes are indicated by arrows.
Blind-ending afferent lymph vessels collect and channel interstitial fluid into the subcapsular sinus. From
here, the lymph is drained towards the hilus through the fibroblastic reticular cell (FRC) conduit and trabecular
sinuses that connect to medullary sinuses. b) Schematic depiction of a paracortical cord (modified
according to REFS 21,22). The T-cell-rich cord (light blue) is shown adjacent to a B-cell follicle (pink)
and demarcated lymph-filled sinuses (green). The cord is penetrated by reticular fibres consisting of type
1 and type 3 collagen that are contained within the sleeves of the FRCs forming a conduit. At the centre
of each cord is a high endothelial venule (HEV) that is surrounded by concentric layers of FRCs. The
FRC conduit drains lymph into the perivenular channel. [9]

226 In-vivo microscopy

Bibliography
[1] A. Diaspro, G. Chirico, and M. Collini. Two-photon fluorescence excitation and
related techniques in biological microscopy. Q. Rev. Biophys., 38:97–166, May 2005.
[2] C. Halin, J. Rodrigo Mora, C. Sumen, and U. H. von Andrian. In vivo imaging of
lymphocyte trafficking. Annu. Rev. Cell Dev. Biol., 21:581–603, 2005.
[3] D. W. Piston. When two is better than one: elements of intravital microscopy.
PLoS Biol., 3:e207, Jun 2005.
[4] W. Denk, J. H. Strickler, and W. W. Webb. Two-photon laser scanning fluorescence
microscopy. Science, 248:73–76, Apr 1990.
[5] A. Scholer, S. Hugues, A. Boissonnas, L. Fetler, and S. Amigorena. Intercellular
adhesion molecule-1-dependent stable interactions between T cells and dendritic
cells determine CD8+ T cell memory. Immunity, 28:258–270, Feb 2008.
[6] C. Sumen, T. R. Mempel, I. B. Mazo, and U. H. von Andrian. Intravital microscopy:
visualizing immunity in context. Immunity, 21:315–329, Sep 2004.
[7] M. D. Cahalan, I. Parker, S. H. Wei, and M. J. Miller. Two-photon tissue imaging:
seeing the immune system in a fresh light. Nat. Rev. Immunol., 2:872–880, Nov
2002.
[8] R. N. Germain, F. Castellino, M. Chieppa, J. G. Egen, A. Y. Huang, L. Y. Koo, and
H. Qi. An extended vision for dynamic high-resolution intravital immune imaging.
Semin. Immunol., 17:431–441, Dec 2005.
[9] U. H. von Andrian and T. R. Mempel. Homing and cellular traffic in lymph nodes.
Nat. Rev. Immunol., 3:867–878, Nov 2003.
[10] T. R. Mempel, M. L. Scimone, J. R. Mora, and U. H. von Andrian. In vivo imaging
of leukocyte trafficking in blood vessels and tissues. Curr. Opin. Immunol., 16:406–
417, Aug 2004.
[11] M. D. Cahalan, I. Parker, S. H. Wei, and M. J. Miller. Two-photon tissue imaging:
seeing the immune system in a fresh light. Nat. Rev. Immunol., 2:872–880, Nov
2002.
[12] J. B. Huppa and M. M. Davis. T-cell-antigen recognition and the immunological
synapse. Nat. Rev. Immunol., 3:973–983, Dec 2003.

228 Bibliography
[13] J. Mertz. Nonlinear microscopy: new techniques and applications. Curr. Opin.
Neurobiol., 14:610–616, Oct 2004.
[14] W. Mohler, A. C. Millard, and P. J. Campagnola. Second harmonic generation
imaging of endogenous structural proteins. Methods, 29:97–109, Jan 2003.
[15] M. J. Levene, D. A. Dombeck, K. A. Kasischke, R. P. Molloy, and W. W. Webb. In
vivo multiphoton microscopy of deep brain tissue. J. Neurophysiol., 91:1908–1912,
Apr 2004.
[16] F. Helmchen and W. Denk. Deep tissue two-photon microscopy. Nat. Methods,
2:932–940, Dec 2005.
[17] J. S. Salafsky. Detection of protein conformational change by optical secondharmonic
generation. J Chem Phys, 125:074701, Aug 2006.
[18] P. Bousso, N. R. Bhakta, R. S. Lewis, and E. Robey. Dynamics of thymocyte-stromal
cell interactions visualized by two-photon microscopy. Science, 296:1876–1880, Jun
2002.
[19] P. Stoller, B. M. Kim, A. M. Rubenchik, K. M. Reiser, and L. B. Da Silva.
Polarization-dependent optical second-harmonic imaging of a rat-tail tendon. J
Biomed Opt, 7:205–214, Apr 2002.
[20] V. E. Centonze and J. G. White. Multiphoton excitation provides optical sections
from deeper within scattering specimens than confocal imaging. Biophys. J.,
75:2015–2024, Oct 1998.
[21] P. So, H. Kim, and I. Kochevar. Two-Photon deep tissue ex vivo imaging of mouse
dermal and subcutaneous structures. Opt Express, 3:339–350, Oct 1998.
[22] J. M. Squirrell, D. L. Wokosin, J. G. White, and B. D. Bavister. Long-term twophoton
fluorescence imaging of mammalian embryos without compromising viability.
Nat. Biotechnol., 17:763–767, Aug 1999.
[23] N. C. Fernandez, A. Lozier, C. Flament, P. Ricciardi-Castagnoli, D. Bellet, M. Suter,
M. Perricaudet, T. Tursz, E. Maraskovsky, and L. Zitvogel. Dendritic cells directly
trigger NK cell functions: cross-talk relevant in innate anti-tumor immune responses
in vivo. Nat. Med., 5:405–411, Apr 1999.
[24] H. Groux, A. O’Garra, M. Bigler, M. Rouleau, S. Antonenko, J. E. de Vries, and
M. G. Roncarolo. A CD4+ T-cell subset inhibits antigen-specific T-cell responses
and prevents colitis. Nature, 389:737–742, Oct 1997.

Chapter 6 229
[25] E. O. Long. Regulation of immune responses through inhibitory receptors. Annu.
Rev. Immunol., 17:875–904, 1999.
[26] A. Martin-Fontecha, L. L. Thomsen, S. Brett, C. Gerard, M. Lipp, A. Lanzavecchia,
and F. Sallusto. Induced recruitment of NK cells to lymph nodes provides IFNgamma
for T(H)1 priming. Nat. Immunol., 5:1260–1265, Dec 2004.
[27] W. Van den Broeck, A. Derore, and P. Simoens. Anatomy and nomenclature of
murine lymph nodes: Descriptive study and nomenclatory standardization in BAL-
B/cAnNCrl mice. J. Immunol. Methods, 312:12–19, May 2006.
[28] R. Kim, M. Emi, K. Tanabe, and K. Arihiro. Immunobiology of the sentinel lymph
node and its potential role for antitumour immunity. Lancet Oncol., 7:1006–1016,
Dec 2006.
[29] G. Ferlazzo, M. Pack, D. Thomas, C. Paludan, D. Schmid, T. Strowig, G. Bougras,
W. A. Muller, L. Moretta, and C. Munz. Distinct roles of IL-12 and IL-15 in human
natural killer cell activation by dendritic cells from secondary lymphoid organs.
Proc. Natl. Acad. Sci. U.S.A., 101:16606–16611, Nov 2004.
[30] Janeway C.A.Jr. Immunobiology. 6th Edition, 2006.
[31] M. A. Cooper, T. A. Fehniger, and M. A. Caligiuri. The biology of human natural
killer-cell subsets. Trends Immunol., 22:633–640, Nov 2001.
[32] C. Borg, A. Jalil, D. Laderach, K. Maruyama, H. Wakasugi, S. Charrier, B. Ryffel,
A. Cambi, C. Figdor, W. Vainchenker, A. Galy, A. Caignard, and L. Zitvogel. NK
cell activation by dendritic cells (DCs) requires the formation of a synapse leading
to IL-12 polarization in DCs. Blood, 104:3267–3275, Nov 2004.
[33] R. S. Alli and A. Khar. Interleukin-12 secreted by mature dendritic cells mediates
activation of NK cell function. FEBS Lett., 559:71–76, Feb 2004.
[34] S. H. Kassim, N. K. Rajasagi, X. Zhao, R. Chervenak, and S. R. Jennings. In vivo
ablation of CD11c-positive dendritic cells increases susceptibility to herpes simplex
virus type 1 infection and diminishes NK and T-cell responses. J. Virol., 80:3985–
3993, Apr 2006.
[35] C. A. Stewart, E. Vivier, and M. Colonna. Strategies of natural killer cell recognition
and signaling. Curr. Top. Microbiol. Immunol., 298:1–21, 2006.
[36] D. Jiang, J. Liang, J. Fan, S. Yu, S. Chen, Y. Luo, G. D. Prestwich, M. M. Mascarenhas,
H. G. Garg, D. A. Quinn, R. J. Homer, D. R. Goldstein, R. Bucala,

230 Bibliography
P. J. Lee, R. Medzhitov, and P. W. Noble. Regulation of lung injury and repair by
Toll-like receptors and hyaluronan. Nat. Med., 11:1173–1179, Nov 2005.
[37] O. Wald, I. D. Weiss, H. Wald, H. Shoham, Y. Bar-Shavit, K. Beider, E. Galun,
L. Weiss, L. Flaishon, I. Shachar, A. Nagler, B. Lu, C. Gerard, J. L. Gao, E. Mishani,
J. Farber, and A. Peled. IFN-gamma acts on T cells to induce NK cell mobilization
and accumulation in target organs. J. Immunol., 176:4716–4729, Apr 2006.
[38] I. Zanoni, M. Foti, P. Ricciardi-Castagnoli, and F. Granucci. TLR-dependent activation
stimuli associated with Th1 responses confer NK cell stimulatory capacity
to mouse dendritic cells. J. Immunol., 175:286–292, Jul 2005.
[39] M. Bajenoff, B. Breart, A. Y. Huang, H. Qi, J. Cazareth, V. M. Braud, R. N.
Germain, and N. Glaichenhaus. Natural killer cell behavior in lymph nodes revealed
by static and real-time imaging. J. Exp. Med., 203:619–631, Mar 2006.
[40] M. Terme, E. Tomasello, K. Maruyama, F. Crepineau, N. Chaput, C. Flament, J. P.
Marolleau, E. Angevin, E. F. Wagner, B. Salomon, F. A. Lemonnier, H. Wakasugi,
M. Colonna, E. Vivier, and L. Zitvogel. IL-4 confers NK stimulatory capacity to
murine dendritic cells: a signaling pathway involving KARAP/DAP12-triggering
receptor expressed on myeloid cell 2 molecules. J. Immunol., 172:5957–5966, May
2004.
[41] E. Vivier, E. Tomasello, M. Baratin, T. Walzer, and S. Ugolini. Functions of natural
killer cells. Nat. Immunol., 9:503–510, May 2008.
[42] M. Bajenoff, B. Breart, A. Y. Huang, H. Qi, J. Cazareth, V. M. Braud, R. N.
Germain, and N. Glaichenhaus. Natural killer cell behavior in lymph nodes revealed
by static and real-time imaging. J. Exp. Med., 203:619–631, Mar 2006.
[43] H. Beuneu, J. Deguine, B. Breart, O. Mandelboim, J. P. Di Santo, and P. Bousso.
Dynamic behavior of NK cells during activation in lymph nodes. Blood, 114:3227–
3234, Oct 2009.
[44] C. Scheinecker. Application of in vivo microscopy: evaluating the immune response
in living animals. Arthritis Res. Ther., 7:246–252, 2005.
[45] S. Stoll, J. Delon, T. M. Brotz, and R. N. Germain. Dynamic imaging of T celldendritic
cell interactions in lymph nodes. Science, 296:1873–1876, Jun 2002.
[46] M. J. Miller, S. H. Wei, I. Parker, and M. D. Cahalan. Two-photon imaging of
lymphocyte motility and antigen response in intact lymph node. Science, 296:1869–
1873, Jun 2002.

Chapter 6 231
[47] S. H. Wei, M. J. Miller, M. D. Cahalan, and I. Parker. Two-photon imaging in
intact lymphoid tissue. Adv. Exp. Med. Biol., 512:203–208, 2002.
[48] E. A. Robey and P. Bousso. Visualizing thymocyte motility using 2-photon microscopy.
Immunol. Rev., 195:51–57, Oct 2003.
[49] P. Bousso. T-cell activation by dendritic cells in the lymph node: lessons from the
movies. Nat. Rev. Immunol., 8:675–684, Sep 2008.
[50] J. B. Beltman, A. F. Maree, and R. J. de Boer. Analysing immune cell migration.
Nat. Rev. Immunol., 9:789–798, Nov 2009.
[51] G. Bogle and P. R. Dunbar. Simulating T-cell motility in the lymph node paracortex
with a packed lattice geometry. Immunol. Cell Biol., 86:676–687, 2008.
[52] C. D. Allen, T. Okada, H. L. Tang, and J. G. Cyster. Imaging of germinal center
selection events during affinity maturation. Science, 315:528–531, Jan 2007.
[53] P. Mrass, H. Takano, L. G. Ng, S. Daxini, M. O. Lasaro, A. Iparraguirre, L. L.
Cavanagh, U. H. von Andrian, H. C. Ertl, P. G. Haydon, and W. Weninger. Random
migration precedes stable target cell interactions of tumor-infiltrating T cells. J.
Exp. Med., 203:2749–2761, Nov 2006.
[54] G. Dranoff. Cytokines in cancer pathogenesis and cancer therapy. Nat. Rev. Cancer,
4:11–22, Jan 2004.
[55] R. Medzhitov and C. Janeway. Innate immunity. N. Engl. J. Med., 343:338–344,
Aug 2000.
[56] C. A. Janeway. Approaching the asymptote? Evolution and revolution in immunology.
Cold Spring Harb. Symp. Quant. Biol., 54 Pt 1:1–13, 1989.
[57] N. W. Palm and R. Medzhitov. Pattern recognition receptors and control of adaptive
immunity. Immunol. Rev., 227:221–233, Jan 2009.
[58] G. R. Vasta. Roles of galectins in infection. Nat. Rev. Microbiol., 7:424–438, Jun
2009.
[59] A. Poltorak, X. He, I. Smirnova, M. Y. Liu, C. Van Huffel, X. Du, D. Birdwell,
E. Alejos, M. Silva, C. Galanos, M. Freudenberg, P. Ricciardi-Castagnoli, B. Layton,
and B. Beutler. Defective LPS signaling in C3H/HeJ and C57BL/10ScCr mice:
mutations in Tlr4 gene. Science, 282:2085–2088, Dec 1998.

232 Bibliography
[60] S. T. Qureshi, L. Lariviere, G. Leveque, S. Clermont, K. J. Moore, P. Gros, and
D. Malo. Endotoxin-tolerant mice have mutations in Toll-like receptor 4 (Tlr4). J.
Exp. Med., 189:615–625, Feb 1999.
[61] K. Hoshino, O. Takeuchi, T. Kawai, H. Sanjo, T. Ogawa, Y. Takeda, K. Takeda,
and S. Akira. Cutting edge: Toll-like receptor 4 (TLR4)-deficient mice are hyporesponsive
to lipopolysaccharide: evidence for TLR4 as the Lps gene product. J.
Immunol., 162:3749–3752, Apr 1999.
[62] Bergey D.H. Holt J.G. Krieg N.R. Sneath P.H.A. In Bergey’s Manual of Determinative
Bacteriology Lippincott Williams and Wilkins edition, New York, 162:3749–
3752, Apr 1994.
[63] M. G. Rittig, A. Kaufmann, A. Robins, B. Shaw, H. Sprenger, D. Gemsa, V. Foulongne,
B. Rouot, and J. Dornand. Smooth and rough lipopolysaccharide phenotypes
of Brucella induce different intracellular trafficking and cytokine/chemokine
release in human monocytes. J. Leukoc. Biol., 74:1045–1055, Dec 2003.
[64] S. Rakoff-Nahoum and R. Medzhitov. Toll-like receptors and cancer. Nat. Rev.
Cancer, 9:57–63, Jan 2009.
[65] Nussenzweig M.C. Steinman R.M., Hawiger D. Tolerogenic dendritic cells. Annual
Reviews Immunology, 21:685–711, 2003.
[66] R. M. Steinman. The dendritic cell system and its role in immunogenicity. Annu.
Rev. Immunol., 9:271–296, 1991.
[67] K. Shortman and S. H. Naik. Steady-state and inflammatory dendritic-cell development.
Nat. Rev. Immunol., 7:19–30, Jan 2007.
[68] G. Trinchieri. Biology of natural killer cells. Adv. Immunol., 47:187–376, 1989.
[69] I. Zanoni, F. Granucci, M. Foti, and P. Ricciardi-Castagnoli. Self-tolerance, dendritic
cell (DC)-mediated activation and tissue distribution of natural killer (NK)
cells. Immunol. Lett., 110:6–17, May 2007.
[70] F. M. Burnet. The concept of immunological surveillance. Prog Exp Tumor Res,
13:1–27, 1970.
[71] A. Mondino, A. Khoruts, and M. K. Jenkins. The anatomy of T-cell activation and
tolerance. Proc. Natl. Acad. Sci. U.S.A., 93:2245–2252, Mar 1996.

Chapter 6 233
[72] S. Fossum and W. L. Ford. The organization of cell populations within lymph nodes:
their origin, life history and functional relationships. Histopathology, 9:469–499, May
1985.
[73] R. H. Kelly. Functional anatomy of lymph nodes. I. The paracortical cords. Int.
Arch. Allergy Appl. Immunol., 48:836–849, 1975.
[74] J. E. Gretz, A. O. Anderson, and S. Shaw. Cords, channels, corridors and conduits:
critical architectural elements facilitating cell interactions in the lymph node cortex.
Immunol. Rev., 156:11–24, Apr 1997.

Appendix B: Instruments details
During the experiments basically two excitation sources and two types of photo-detectors
were employed, depending on the applications. Though much is known on these devices,
some of the innovations described in this PhD report are related to some technical details.
For clarity reasons and in order to ease the reading of some of the chapters, we briefly
discuss some of these technical details hereafter.
6.14 Laser sources
A larger portion of data discussed here (TPE, lifetime measurement) required pulsed
laser sources that are bare pulsed infra-red lasers from Spectra Physics (Mountain View,
CA): the Tsunami passive cavity and the Mai Tai infra-red laser. Both are femtosecond
pulsed lasers with a spectral range between 700 and 1000 nm and both are equipped with
an active modelocking system that guarantees the production of pulses nominally 100 fs
width at a repetition rate of 80 MHz (i.e. at the output of the cavity one measures a pulse
every 12.5 ns). The characteristics of the two optical resonant cavities are almost identical
because the Mai Tai is the compact and fully automatized version of the combination of
a Millennia solid state laser and a Tsunami cavity. Hereafter we review the description
of the main components of these excitation systems. In both cases a 532 nm CW solid
state laser (Millennia X, Spectra Physics, CA) is used as a pump.
6.14.1 Millennia
The millennia laser is based on two diode lasers whose emission is used to pump a
solid state laser based on Nd 3+ ions crystalline matrix doped with yttrium vanadate
Nd:YVO4). The output wavelength is 1064 nm and is converted to a 532 nm green
beam by means of a second harmonic generation process that takes place in a lithium
triborate (LBO) non-linear crystal. The output is composed by a unique transverse
mode (equivalent to the TEM 00 of a conventional laser) with a pseudo-Gaussian intensity
profile characterized by an high ellipticity that has to be corrected before entering the IR
234

Chapter 235
mode-locked cavity. This is made by means of a couple of anamorph prisms mounted at
an angle close to the Brewster angle with respect to incident beam (Figure 6.31 ). The
emission from the diode laser is perfectly superimposed to the absorption band of the
Nd 3+ ion allowing a good coupling between the pump and the active medium (Figure
6.32).
Figure 6.31: Correction of the beam ellipticity by means of two prism at the Brewster angle.
Figure 6.32: Panel A: Absorption spectrum for the Nd 3+ ion. Panel B: Emission spectrum of the diode
pump laser.
The neodymium bar (active medium) has the principal absorption band in the red
and near infra-red region of the electromagnetic spectrum, whereas the laser emission
is optimized at 1064 nm. The generation of the 532 nm green pump beam is made by
means of the second harmonic generation process that takes place in a LBO crystal. This
material is preferred to others with higher non-linear coefficient, because it is possible to
optimize the conversion efficiency of LBO simply varying the working temperature. A
dichroic mirror reflects back in the cavity the first harmonic from the solid state laser and
allows the 532 nm beam to be sent in the Tsunami or Mai Tai optical cavity. Since the
intensity of the second harmonic radiation depends on the square of the first harmonic
power it is possible to obtain a high conversion efficiency increasing the power of the

236 Instruments details
pump beam. Unluckily a solid state laser pumped by a diode laser gives rise to a chaotic
emission characterized by high intensity fluctuations that avoid the application of the
source to scientific experiments. These instabilities are mainly due to the non-linear
coupling of the axial modes in the sum frequency (i.e. second harmonic generation)
process. In the Tsunami and Mai Tai the problem is overcome by adopting the so called
QMAD (Quiet Multi Axial mode Doubling) solution: that employs many axial modes
allowing the oscillation of more than 100 longitudinal modes: in this way the power of
each axial mode is so low that none of them reaches the pick power needed to induce
high non-linear looses. The non-linear coupling terms are therefore mediated with the
effect that the second harmonic emission presents a very low noise (Figure 6.33).
Figure 6.33: Quiet Multi-Axial mode-Doubling (QMAD). Panel (a): intracavity frequency doubling in a
laser with a few axial modes produces large amplitude fluctuations in the second harmonic output resulting
from non-linear coupling of the modes through sum frequency mixing. Panel (b): single frequency solution
forces oscillation on a single axis mode to eliminate mode coupling. Panel (c): QMAD solution produces
oscillation on many axial modes, effectively averaging the non-linear coupling terms to provide highly
stable second harmonic output.
6.14.2 Titanium-Sapphire (Ti-Sa) optical resonant cavity
The 532 nm beam from the Millennia is used to pump a Titanium-Sapphire (Ti-Sa) rod
that gives rise to the IR output of the laser system. The Ti-Sa is a crystalline solid
obtained by introducing Ti 2 O 3 into a solution of Al 2 O 3 allowing the substitution of a
little amount of Al 3+ ions with Ti 3+ ions. The Ti 3+ electronic configuration can be
represented as two distinct energy levels with a wide broadening caused by the presence
of many vibrational levels (Figure 6.34). The result is a broaden absorption band between
400 and 600 nm. The fluorescence emission is within 600 and 1000 nm and is due to the
fact that the transition occur between vibrational levels whose energy lies between that
of the excited and that of the ground state.
However the laser coherent emission is possible only for wavelength λ > 670 nm since
forλ < 670 nm the emission band is superimposed to the absorption one giving rise to
auto-absorption processes that reduce the efficiency of the fluorescence emission (Figure
6.35).

Chapter 237
Figure 6.34: Energy level structure for Ti 3+ ion in Sapphire.
Figure 6.35: Normalized absorption and emission spectra of Ti-Sa.
Wavelength selection
Since the Ti-Sa rod is birefringent, lasing is obtained when the c-axis of the rod is aligned
coplanar to the polarization of the electric field in the cavity. The chamber that hosts
the rod orients the rod surfaces at Brewster’s angle and allows the c-axis to be coplanar
to the electric field vector. The output λ can be varied within 690-1000 nm by means
of a system of four prisms and a slit. The prisms create a region in the cavity where
the different λ are separated and provide also the compensation of the group velocity
dispersion. The slit, located in the central part of the four prisms arrangement, allows
to select the desired λ.
Pulse width
The temporal duration or the pulses depends mainly on three factors:
1. the properties of the Ti-Sa rod,
2. the size of the optical cavity,
3. the selected wavelength.
It is possible to control the pulse width by changing the Group Velocity Dispersion
(GVD) in the optical cavity. The material refraction index depends on λ, thus every

238 Instruments details
color component travels along a particular direction with slightly different speed, giving
rise to a temporal separation of the wavelength components present in the pulse: this
phenomenon is called GVD dispersion. For GVD ≥ 0 the red colors travel faster than
the blue ones (Figure 6.36).
Figure 6.36: Typical wavelength dependence of the refractive index of a material.
The GVD is related directly to the change of the group velocity versus the wavelength.
By a simple computation one can demonstrate that the GVD is directly proportional to
the second derivative of the refraction index versus the wavelength, d 2 n(λ)=dλ 2 . In the
more general case we must refer to the change in the optical path instead to compute
simply the change of the group velocity. For normal dispersion, the term d 2 n(λ)=dλ 2 is
positive and so is the GVD. The optical refractive components inside the resonant cavity
create a positive GVD giving rise to a spread of the pulse. A compensation for the pulse
spreading must be searched by thinking in terms of the optical path instead of the group
velocity. The GVD is then defined as d 2 L(λ)=dλ 2 , where L is the optical path of the
various wavelength components of the laser beam 10 .
The normal dispersion introduces a positive GVD, therefore, in order to keep the pulse
short, it is necessary to introduce a negative GVD. Usually the negative GVD is obtained
by means of an arrangement of two or four prisms. Varying the distance between the
prisms, while keeping at a minimum the thickness of glass passed through by the beam,
it is possible to have pulses with a time duration close to the minimal width ∼ = 100 fs
(Figure 6.37). The minimum pulse width is limited by the width of the laser spectrum.
For the Tsunami optical resonant cavity the selection of output wavelength and
the temporal characteristics of the pulses have to be checked by the operator (through
a spectrum analyzer and an optical autocorrelator) 11 , while for the Mai Tai they are
10 It must be noticed that the pulses are further broadened by the Self Phase Modulation (SPM)
processes that occur in the Ti-Sa rod.
11 The pulse width must be checked on the sample. This is performed by an homemade optical autocorrelator;
the typical pulse width on the sample is 280 fs on the Nikon and 180 fs on the Olympus (Mai

Chapter 239
Figure 6.37: Prisms sequence used for dispersion compensation. An input pulse with a positive chirp
(red frequencies at the leading edge of the pulse) experiences a positive GVD (red frequencies have longer
group delay time) in the prisms sequence. The net effect is that the prisms sequence compensates for the
positive GVD and produces a pulse that has no chirp.
driven by a CVI written control program. This is the main difference between the two
laser sources.
6.15 Microscopes
The Tsunami and Mai Tai optical resonant cavity are respectively coupled to Nikon
TE300 and Olympus BX51 microscopes. The former is principally devoted to the analysis
of the fluorescence fluctuations (from solutions or in cells) and it has been used during
the PCH and FCS measurements. The second microscope, equipped with a commercial
confocal scanning head, has been employed basically for MPM experiments.
6.15.1 Nikon TE300
The Nikon TE300 is an inverted optical microscope that collects the signal in epifluorescence
geometry; the sample is placed on an (x,y) translational stage equipped with
two DC electrical motors (M230.25 Physik Instrumente, Milano, I) that allow macroscopic
movements until 25 mm with a minimum step of 50 µm and an accuracy of 0.1
µm. The sample stage has been modified in order to install these DC motors and a
piezo-electric translator (x,y,z) cube (Figure 6.38) that allows nano-positioning on the
sample with an accuracy of 50 nm and a maximum excursion of 100 µm.
While conventional instruments (i.e. a spectrometer) collect the signal at 90 ◦ with
respect to the direction of the excitation beam, in an epi-fluorescence microscope the
objective has the double function of focusing the excitation beam and of collecting the
fluorescence signal arising from the sample (Figure 6.39). In details: the excitation IR
Tai-Deep See)

240 Instruments details
Figure 6.38: Nikon TE300 microscope. In the figure are underlined the modifications took in the
experimental set-up: the position of the piezo electric translator cube, the translational stage and the
support for the cover-slip.
light enters the back aperture of the microscope and encounters a dichroic mirror 12 ,
mounted at 45 ◦ respect to the objective, (see Figure 6.39) that reflects the light toward
the objective that focuses it on the sample.
Figure 6.39: This simplified scheme explains the role of the dichroic mirror mounted in the Nikon
TE300 microscope.
The fluorescence signal from the sample, once collected by the objective, passes
through the dichroic and is reflected to the detectors. In order to remove unwanted
residual components of the IR excitation beam (due to reflections on the mechanical
12 An ideal dichroic mirror is characterized by a cutoff wavelength, λ co, and it transmits the components
with λ ≤ λ co while it reflects the components with λ ≥ λ co, thus enabling the separation between the
excitation and fluorescence light. An ideal dichroic mirror has reflectivity = 50% = transmittivity for
λ = λ co. Moreover it changes slightly the value of λ co if tilted at angles different by 45 ◦ with respect to
the beam.

Chapter 241
part of the TE300 and to the non-ideal behaviour of the dichroic mirror) and to select
the desired detection wavelength, λ em , the signal from the sample passes through a short
pass filter with absorbance 4-6 OD for λ ≻ 670 nm and a pass-band filter centered at the
emission wavelength of the dye under investigation (for example employing Rhodamine
6G that presents an emission maximum at 560 nm it is necessary to mount a filter
centered at 560 nm with a full width at half maximum not larger than 40 nm, i.e. a
560±40 nm). The most important component of a microscope is the objective that is
characterized by three parameters: the magnification, M, the immersion medium and
particularly by the Numerical Aperture, N.A., defined as:
N.A. = n sin(θ) (6.1)
where n is the refraction index of the immersion liquid and θ is the semi-angle defined
by the collection lens of the objective. The larger is the N.A., the smaller is the beam
waist on the focal plane and higher is the collection efficiency of the signal. This means
that the excitation volume V exc results smaller and the S/N ratio is enhanced. The
immersion medium, usually water or oil, (that present respectively n = 1.33 and 1.5)
is used to guarantee the correct index matching between the cover-slip and the sample
in order to avoid total reflection of the fluorescence light along the optical path (Figure
6.40 ). If total reflection occurs, the collection efficiency results degraded because the
effective numerical aperture is lower than the nominal N.A..
Figure 6.40: Role of the immersion medium in order to achieve the index matching condition. In this
figure are compared an air and an oil immersion objectives.
TE300 can be equipped with an oil immersion objective Nikon M = 100X, N.A. =
1.3 or with a water immersion Nikon objective M = 60X, N.A. = 1.2. Both guarantee a
good collection efficiency. A better index matching with the cover-slip (n = 1.5) and the
sample is obtained with the water immersion objective since we use water solutions or
biological samples that present an index of refraction close to n = 1.33. Moreover it must
be considered that, in order to minimize V exc , it is of crucial importance to (correctly)

242 Instruments details
fill the back aperture of the objective; in fact the divergence of a Gaussian beam apart
from the beam waist is (Figure 6.41):
( ) ω(z)
Θ = 2 arctan
z
( ) λ
∼= 2 arctan
πω 0
(6.2)
will match 2θ = 2N.A./n only if the back aperture is completely filled. Otherwise Θ
will be smaller since the N.A. of the objective is not completely used, and V exc will be
larger with a loss of S/N ratio.
Figure 6.41: Axial profile of a Gaussian beam in the proximity of the beam waist.
6.15.2 Olympus BX51
The Olympus BX51 is an upright microscope that, as the Nikon TE300, collects the
signal in epi-fluorescence (see Figure 6.39) geometry. In our experimental set-up, the
BX51 is equipped with a confocal scanning head (FV300) (see Figure 6.42) modified
in order to allow TPE. It mounts low pass filters in front of its PMT in order to clear
the signal from the residuals of the excitation beam and the larger pin-hole was removed
from the pin-hole wheel because TPE is intrinsically spatially confined and thus does not
require the spatial filtering provided by a pin-hole. The scanning head can be used both
in confocal and TPE mode with detection through the PMT units internal to the FV300
unit or through the external PMT devices mounted in the non-descanning unit. The IR
incoming laser light enter the FV300 at right angle with respect to the visible lasers. All
the laser beams are then sent to the galvo scanning mirrors that are mounted in front
of the side aperture of the BX51. Before entering the main BX51 housing through the
scanning lens, a visible/IR dichroic mirror can be mount to discriminate between the
visible (confocal) and the IR (TPE) scanning mode.
Once reflected in the microscope by the galvos, the excitation light may:
1. be directly focused onto the sample by means of the objective lens or
2. pass through a second visible/IR dichroic mirror (used for the non-descanning
detection mode) and focused onto the sample by the objective lens.

Chapter 243
Figure 6.42: The Olympus BX51 microscope equipped with the FV300 confocal scanning head.
The signal arising from the focus plane is collected by the same objective and may
follow these two routes:
1. in the descanned detection mode the emission light does not pass through the
second dichroic mirror and it is sent to the FV300 galvo-mirrors to be filtered by
the pin-hole and detected by the PMT internal to the FV300 unit,
2. in the non-descanned detection mode the emission light is reflected by the second
dichroic mirror and sent to the non-descanned unit where it is detected by the
PMT devices external to the FV300 unit.
The BX51 is equipped with two water immersion Olympus objectives: the first has
M = 20X, N.A.=0.95 (chosen because of its large working distance W.D. ∼ = 2 mm and
its relatively high numerical aperture. For these reasons this objective lens is considered
one of the most suitable for deep tissue imaging); the second has M=60X, N.A.=1.1 and
W.D. ∼ = 1.5 mm. The galvo mirrors enables to collect image of maximum area of 720 X
720 µm 2 with different scanning modes and speeds.
6.16 The non-descanned detection system
A mode-locked Ti:Sapphire laser (Mai Tai HP, Spectra Physics, CA) that produces pulses
of 120 fs full width at half maximum with 80 MHz repetition rate is coupled to a confocal
scanning system composed of a scanning head (FV300, Olympus, Japan) mounted on
an upright optical microscope (BX51, Olympus, Japan). Non-confocal TPE imaging
can be performed through the FV300 scanning unit because we removed the largest pin
hole from the pin hole wheel. However, as discussed in the previous section, commercial
scanning system modified in order to perform TPE imaging are not efficient enough to be

244 Instruments details
employed, as it is, in in vivo experiments because they suffer of poor collection efficiency
at the excitation power level usually involved in measurements on biological samples.
This problem can be, at least, partially overcome by detecting the signal arising from
the specimen just behind the entrance pupil of the objective by means of a non-descanned
detection unit (ND-unit).
6.16.1 The ND-unit design
The ND-unit was designed to be mounted at right angle respect to the BX51 dichroic
splitter wheel, just above the entrance pupil of the objective lens. The detection unit
had a 2 inch cylindrical connector, that can be easily adapted to different microscopes.
For coupling to the BX51, a hole was made on the microscope structure and one of the
dichroic beam splitter cubes was modified to be mounted, within the filter wheel, at right
angle with respect to the factory position (Figure 6.43).
Figure 6.43: A) Main optical path. The laser beam passes through a λ/2 waveplate and a polarizer and
it is sent to the FV300 scanning unit by means of a beam steering. The non-descanned unit light path is
detailed in the upper right box. The collection lens (L) is indicated. B) The ND-unit mounted on the side
of the BX51 microscope dichroic filters wheel. C) Detail of the modification of the microscope structure
on the side of the filter wheel holder. D) Picture of the ND-unit showing the triangular shaped mounting
that hosts the dichroic filters and the cylindrical mounts for the pass band filters mounted in front of the
PMTs. Fluorescence is entering as indicated by the arrow and the three PMT housings are visible.
The ND-unit avoids the complex optical path back the steers light to the photomultipliers
(PMT) in the FV300. However, in order to maximize the collection efficiency from
highly scattering media, the signal coming from the objective lens must be collected by

Chapter 245
a f=100 mm bispherical lens. The signal reaching the ND-unit could be split into three
channels by two dichroic beam splitters, 490DCX and 540DCX (Chroma Technology
Corporation, Rockingham VT, USA) and fed to three Hamamatsu analog output photomultipliers
(HC125-02, Hamammatsu, Japan) whose 21 mm (diameter) photo-cathodes
ensured to collect most of the light during scanning. The fluorescence signal was filtered
by a short-pass 670 nm filter (Chroma Inc., Brattelboro, VT) and by band-pass filters
in order to remove the residual signal due to either first harmonic or to undesired autofluorescence
from the sample. During the optical scanning of the sample (by means of
the FV300 galvo-mirrors) the HC125-02 photocathode was filled for about 80% of the
sensitive area. The microscope objective lens (magnification 20X, N.A. = 0.95, working
distance 2 mm), coupled to the non-descanned unit allowed the imaging of fields of view
460 X 460 µm. The output of the PMTs, called external PMTs, is 0-3 V on 50Ω input
impedance, fully compatible with the Olympus scanning controller. The fluorescence
and SHG images are then processed by means of the Fluoview 3.0 software (Olympus,
Japan) irrespective of the detection path (FV300 or ND-unit).
6.17 Detectors
The choice of the detectors depends on the experimental conditions. Fluorescence Fluctuation
Spectroscopy (FFS) relies on small fluctuations of the fluorescence signal around
an average value and since each photon contributes to the whole statistics, it becomes
very important to detect as many photons as possible. Single Photon Avalanche Diodes
(SPAD) guarantee high single-photon detection efficiency and low dark current. Since
they work close to the breakdown bias, they allow each photon-generated carrier to be
amplified by an avalanche current, resulting in an internal almost infinite gain within the
photo-diode, which increases the effective responsivity of the device making the SPAD
very suitable for photo detection at low count rate (i.e. single photon regime). In Multi-
Photon Microscopy (MPM) the main issue is to collect high resolution 3-dimensional
images limiting the residence time per pixel in order to minimize the photo-damage. In
scanning probe microscopy, where one wants to scan wide samples in few seconds, the
Photo-Multiplier Tubes (PMT) are the most suitable detectors since they have typically
large sensitive areas and allow therefore to collect simultaneously the signal coming from
the whole scanned area. Hereafter we give some additional features of SPAD that has
been used in this PhD report for the analysis of the fluctuations in terms of the correlation
functions. The general characteristics of PMT can be found in several text-book
and technical publications, and will not be discussed here.

246 Instruments details
6.17.1 Single Photon Avalanche Diode (SPAD)
When a photon of enough energy hits the depletion zone of a photo-diode, practically
an inversely biased p-n junction (Figure 6.44), this zone may be the source of an electronhole
couple. The generated charge carriers will move in opposite directions under
the influence of the electric field. Macroscopically the two current flows, generated by
the electrons and the holes, have the same direction and they sum to generate a unique
net current that once amplified gives a measure of the number of detected photons. We
have to underline that a generic Photo-Diode (also known as PD) does not present a
multiplication process: its gain is 1 and therefore it works well in situations where the
flux of the incoming photons on the cathode is high. Unluckily in most applications high
quantum efficiencies (30%), very low dark count levels (≈ 50 Hz) and high temporal resolutions
(FWHM ≈ 1 ns) are required. The Avalanche Photo-Diode (APD) satisfies the
first two requests because the electric field (the inversion polarization voltage is close to
and less, in modulus, than the break-down value) applied to the couple of charge carriers
generated by the photon is very high: during their migration toward the electrodes the
electrons and the holes take a kinetic energy that enable them to create new couple of
carriers just by collision. Since the empty zone in the p-n junction is wide enough also
the two new carriers generated by the collision can create another couple giving rise to
a process called avalanche multiplication. The macroscopic current generated by this
process will be proportional to the flux of detected photons for rates ≤ 3MHz.
Figure 6.44: Schematic representation of a p-n junction.
In the case of experiments where one wants to reach the limit of the single detected
photon, time resolution and single photon sensitivity as high as possible are needed.
For this purpose the most suitable detector is the SPAD (Single Photon Avalanche
Diode). These devices use an inversion polarization voltage larger than the break-down
one. Therefore the current growth after one photon detection is exponential and not
controlled 13 leading to an infinite gain. Due to this feature the SPAD can be considered
similar to a Geiger detector because its output is not proportional to the input; the SPAD
13 This makes the device quite prone to photo-damage.

Chapter 247
are just events’ detector. Modern SPAD modules are endowed with an active quenching
circuit that is able to lower the output current after each count (so to lower the device
dead-time 14 ). It can be sketched as a large load resistance ( ∼ = 500 kΩm) in series with
the diode output circuit. The large voltage drop that occurs when the pulse current
drops on this load resistance lowers reverse bias at the leads of the PD, the avalanche
process is stopped and the system is brought back to the initial condition. The Perkin
Helmer SPAD model SPCM-AQR-14 (Figure 6.45) have a circular silicon photo-diode of
180 µm of diameter; because the gap between the conduction and valence bands is 1.12
eV, these SPADs enable to detect photons with λ in a 400- 1100nm range.
Figure 6.45: A Perkin-Helmer SPAD model SPCM-AQR-14.
The two most important characteristics of the SPAD are the photo-sensitiveness, S,
and the quantum efficiency, QE, defined respectively as the ratio between the produced
photo-current, I, and the incident radiant power, P, and the ratio between the number
of photons at the anode, N anode , and the incoming photons, N incoming :
S = I P
(6.3)
QE =
Since the radiant power at the cathode is:
dN anode
dN incoming
(6.4)
P = dN incomingE
dt
(6.5)
14 SPAD as well as other light detectors suffers of two limiting factors: dead time and afterpulsing. In
this thesis, dead time is particularly relevant because it determines the Impulse Response Function that
limits the lifetime resolution. Tipically, the dead time is 50 ns and the time jitter is about 350 ps.

248 Instruments details
where E is the energy of any incident photon and t is the time they need to reach
the detector, and since the photo-current is:
I = dN anodee −
dt
where e − is the electron charge, these two parameters are strictly connected as:
(6.6)
QE = S e − E (6.7)
The QE for the Perkin Helmer SPCM-AQR-14 can be considered constant overall
the sensitive area with a numerical value ≈ 70% (Figure 6.46, Panel A). It depends also
on the detected λ (Figure 6.46, Panel B) and guarantees good performance in the range
400-1060 nm with its best at 700 nm. The typical maximum rate of the SPCM-AQR-14,
certified by the company is ≈ 15-16 MHz (however non-linearity occurs already at 2
MHz) and the dark current is ≈ 50 Hz. The dead-time 3 15 is ≈ 50 ns and the time
resolution is limited by a jitter ≈ 350 ps. Finally, as can be see from Figure 6.46, panel
C, the QE of a SPAD is higher than that of a PMT in the green-red region and this is
the reason that make us to prefer SPADs instead of PMTs.
Figure 6.46: Panel A: detection efficiency as a function of the position in the sensible area. Panel
B; quantum efficiency as a function of λ. Panel C: comparison within the quantum efficiency of SPAD
(indicated as Si-slik), PMT, Ge based detectors and InGaAs based detectors.
6.18 Dynamic Light Scattering:optical system
The dynamic light scattering optical system is composed of four fundamental units: the
laser, the spectrometer, the detector and the correlator unit (figure 6.47).
The output of a He-Ne laser (30 mW, Spectra Physics),vertically polarized at wavelenght
λ ≈ 633 nm,is spatially filtered (by an objective and a pinhole), and slightly collimated
at the centre of the sample cell by a composite spherical plus cylindrical lenses.
The sample housing consists of an outer glass cylinder. Coaxial with this there is a
15 We experimentally evaluated the dead-time for the devices used in our experiments; the values are
very close to the value of 50 ns given by Perkin-Helmer [18].

Chapter 249
smaller quartz cell (Hellma, 5 cm height, 10 and 8 mm of internal amd external diameter
respectively). The space between the cell and the outer cylinder is filled with water
which acts as index matching liquid in order to lower light flares at the walls of the cell
where the laser beam enters or exists. The index matching water is filtered with millipore
0.22 µm filters. The cell is thermostated by a chamber filled with water flowing from a
thermostat.
In order to fill and empty the cell two teflon needles passing through the sylicon cell cap
were used. A short fixed needle is used to fill the cell: this touches the cell walls in order
to avoid bubbles when introducing the sample. To empty the cell a long movable needle
is used. During the measurements the needle is raised in order not to intercept the laser
beam.
The scattered radiation is detected by a photomultiplier tube (9863KB,EMI,UK,
photoncounting mode), mounted on a (0-360) ◦ goniometer, coaxial with the cell. A
diaphragmed spherical lens is placed between the cell and the photomultiplier tube in
order to focalize the scattered light at the centre of a double slit (vertical and horizontal)
inside the PMT housing. The slits are adjusted in order to select a convenient coherence
area A coh =λ 2 /Ω, where λ is the wavelenght of the radiation and Ω is the solid angle
with which the scattering volume is subtended at the detector 16 ; it is essential that the
photomultiplier samples an area (A s ) not very different from the coherence area (A coh ) in
order not to average out the light intensity fluctuations. The polarization of the detected
light can be selected by a polarizer mounted in front of the collecting lens.
The discriminated signals are then fed to the correlator board (ISS, Urbana Champaign,
IL) to compute the auto-correlation functions (ACF) g (2) (CAP.). Each ACF was tipically
collected for ≈ 30s with a sample time variable from 10 kHz to 750 kHz depending
on the sample characteristics.
The alignment has been checked both in the polarized and in the depolarized configuration
measuring the light scattered by double distilled and filtered water (in order to
test the cell cleaning) and by polystyrene spheres ≈ 200 nm nominal diameter dissolved
in distilled water. The test of the setup in depolarized light was performed on a sample of
intrinsically optically anisotropic particles of PTFE (polytetrafluorethylene), which are
approximately prolate ellipsoids with length ≈ 0.33 µm and diameter ≈ 0.15 µm. The
normalized g (2) (t) ACFs, collected in depolarized configuration, have been fitted with a
single exponential decay finding a relaxation rate Γ = DK 2 + 6R ⊥ versus the scattering
vector K 2 =(4(π/λ)nsin(θ/2)) 2 , where λ is the incident wavelenght, θ the scattering angle
and n the refraction index (see chapter ...).
16 divergence from the beam waist is tipically 1/10 rad

250 Instruments details
CAP...
The Dynamic Light Scattering theory and the analysis methods are described in
Figure 6.47: Schematic representation of the optical setup.
6.18.1 Softwares
In this section the softwares employed to acquire the fluorescence trace, perform FCS and
burst analysis are presented. In particular, ALV5000 correlator board was used for the
calculation of G(τ) in classic FCS; the fluorescence emission signal was fed to a digital
TimeHarp 200 (Picoquant, Berlin, D) board and its analysis was performed with the
SymPhoTime program by Picoquant in order to compute the rate trace at the desired
sampling time and the lifetime histograms.
ALV5000 correlator board
The ALV5000 correlator architecture (ALV-Laser, Vertriebsgesellschaft m.b.H., Langen,
Germany) is conceived on the idea proposed by Schatzel 17 based on multiple sampling
and delay times (Figure 6.48 ). ALV5000 has a quasi-logarithmic time scale where each
channel has an individual sampling time (the bin width) and delay time (the delay from
the measurement at time 0). With this quasi-logarithmic time-scale structure the sampling
time increases with the delay time. This approach allows to explore a wide range
of delay times with a limited number of correlation channels. For example, delay times
between 0.2 µs and 50 ms can be obtained using only 128 channels instead of the 250000
17 Schatzel K. 1985. New concepts in correlator design. Inst. Phys. Conf. Ser. 77: 175-184; Schatzel
K. 1991. Photon Correlation Spectroscopy Multi-component System. SPIE. 1430: 109-115.

Chapter 251
needed to a linear correlator 18 .
Channels 1-16 have a sampling time ∆τ i=1...16 = 0.2 µs. Each following group of eight
channels has an individual sampling time twice as large as the preceding group (channels
17-24, ∆τ i=17...24 = 0.4 µs ; channels 25-32, ∆τ i=25...32 = 0.8 µs; up to channels 121-128,
∆τ i=121...128 = 3.2768 ms). The delay time of each channel is the sum of the sampling
time of all the preceding channels (channel 1, 0 µs; channel 2, 0.2 µs; ...channel 17, 3.2
µs; channel 18, 3.6 µs; up to channel 128, 49.152 ms). For each channel there is a delayed
monitor M del that accumulates all counts sampled in that channel. For every group of
channels with equal sampling time there is a direct monitor M dir that accumulates all
counts without delay time at a particular sampling time.
The calculation of the G(τ) is performed as follows. First it must be considered that the
counting event occurs always with a sampling time of 0.2 µs. The countings on multiples
of this sampling time are obtained by summing up successive 0.2 µs countings. Then,
every channel is multiplied according to its delay time by a channel at 0-delay-time, that
possesses the same sampling time. For example, channels m = 1,2,..16 (delay times from
0 to 3.0 µs and sampling time ∆τ= 0.2 µs) are multiplied by the intensity signal that is
presently measured during 0.2 µs (delay time 0, sampling time 0.2 µs). Channels 17-24
(delay times from 3.2 to 6.0 µs and sampling time ∆τ= 0.4 µs) are multiplied by the
intensity signal that is presently measured during two successive 0.2 µs sampling times
(delay time 0, sampling time 0.4 µs). The results of the multiplication are summed up
over time for the calculation of the G(τ). The counts of each channel are accumulated
in its delayed monitor M del . The counts of the sample at delay time 0 with sampling
times of 0.2 µs, 0.4 µs, etc. are summed up in the direct monitor M dir of each group of
channels with equal sampling time.
The G(τ) is then calculated by:
with
G(m∆τ i ) =
∑
1 k=M−m
k=1
n(k∆τ i )n(k∆τ i + m∆τ i )
(6.8)
M − m
M del,i M dir,i
and
M del,i =
k=M
1 ∑
n(k∆τ i ) (6.9)
M − m
k=m
18 Wohland T. Rigler R. and Vogel H. 2001. The Standard Deviation in Fluorescence Correlation
Spectroscopy. Biophys. J. 80: 2987-2999.

252 Instruments details
M dir,i =
k=M−m
1 ∑
n(k∆τ i ) (6.10)
M − m
Here m is an integer that span a group of channels, ∆τ i is the sampling time (channel
width) of the i-th group of channels and m∆τ i is the delay time of the m-th channel; M
is the number of measurements with sampling time ∆τ i , and is given by M = T/∆τ i ,
where T is the total measurement time; n(k∆τ i ) is the number of photons at time k∆τ i ,
sampled with a channel width of ∆τ i , and n(k∆τ i +m∆τ i ) the number of photons at
time m∆τ i later; M-m is the number of possible products n(k∆τ i )n(k∆τ i +m∆τ i ) over
which the summations extend in eqs. 6.8-6.10. The G(τ) is symmetrically normalized 19
with the direct and delayed monitor channels M dir,i and M del,i of corresponding channel
i, respectively.
k=1
MANCANO TIMEHARP E SYMPHOTIME
6.19 TimeHarp and Symphotime softwares
Fluorescence lifetime measurements in the time domain are commonly performed by
means of Time-Correlated Single Photon Counting (TCSPC). This is a histogramming
technique based on precise timing and time binned counting of single photons emitted
on pulsed laser excitation [1]. However, in many fluorescence applications it is of
great interest not only to obtain the fluorescence lifetime(s) of the fluorophore(s) but to
record and use more information on the fluorescence dynamics. This is most often the
case when very few or even single molecules are observed.For instance,single molecules
flushed through capillaries (e.g in DNA analysis applications) will emit short bursts of
fluorescence, that are of interest for further analysis. The resulting fluorescence intensity
dynamics on a time scale of milliseconds can be used to identify single molecule transits
and to discriminate these events against background noise.
The desired capturing of the complete fluorescence dynamics can be achieved by recording
the arrival times of all photons relative to the beginning of the experiment (time
tag), in addition to the picosecond TCSPC timing relative to the excitation pulses. This
is called Time-Tagged Time-Resolved (TTTR) mode [2]. Figure 6.49 shows the relationship
of the time figures involved. As in conventional TCSPC, a picosecond timing
between laser pulse and fluorescence photon is obtained. In addition to that, a coarser
timing is performed on each photon with respect to the start of the experiment. This is
done with a digital counter running at typically 50 or 100 ns resolution. Even though
19 Schatzel K. Drewel M. and Stimac S. 1988. Photon correlation measurements at large lag times:
improving statistical accuracy. J. Mod. Opt. 35: 711-718.

Chapter 253
Figure 6.48: Channel architecture of the correlator. (A) The fluorescence intensity from the confocal
volume is registered with a channel width of 0.2 µs. After every measurement, three tasks are executed:
(1) all channels are shifted to the right, channel (n - 1) to channel n, channel (n - 2) to channel (n - 1),
... , channel 1 to channel 2. The new measurement is stored in channel 1. (2) The products (depicted as
X) of channel 1 and 2, of channel 1 and 3, etc., are calculated and added (S) to the correlation function
G(∆τ i), G(2∆τ i), G(3∆τ i), etc., respectively. The value of G(0) can be calculated by multiplying channel
1 with itself. (3) The delayed monitor is a register for every channel that sums up all counts that pass
through a channel. Therefore, after each shift of channels the content of every channel is added to its
delayed monitor. (B) Channels 15 and 16 are summed up and shifted to channel 17 which now acquires a
width of 0.4 µs. This happens at the end of every channel group. The last two channels with 0.4 µs width
(channel 23 and 24) are added to yield channel 25 with 0.8 µs width and so on. (C) Those channels that
have a width larger than 0.2 µs (all channels after channel 16) are now correlated with a channel at 0
delay time of equal length. To achieve this, several channels can be summed up. Channels 1+2 act as the
0-delay-time channel for channels 17-24 (0.4 µs). The sum of channels 1-4 act as 0-delay-time channel
for channels 25-32 (0.8 µs), and so on. Note that, for channels 1-16, the correlation is performed after
every 0.2 µs measurement, but, for channels with a width of 0.4 µs, the correlation will be done only
every 0.4 µs, i.e. after 2 measurements of 0.2 µs and so on. (D) The 0-delay-time channel with 0.4 µs
width is shown. It is used for the correlations of channels with a width of 0.4 µs. The direct monitor
(not shown) is a register for every group of channels with equal length. In this register, are stored the
counts that pass through the channel at 0-delay-time. Therefore, the 0-delay-time channel will be added
to the direct monitors after the correlation for a group of channels with equal width. For example, the
0-delay-time channel of 0.2 µs will be added to the direct monitor for channels 1-16 every 0.2 µs. The
0-delay-time channel of 0.4 µs will be added to the direct monitor for channels 17-24 every 0.4 µs, etc.

254 Instruments details
Figure 6.49: Timing figures in TTTR data acquisition.
this is much higher than what most applications mentioned above would require, modern
hardware provides this resolution at no extra cost. Since the TCSPC timing typically
covers the time scale just below 100 ns, it is indeed sensible to chose a time tag resolution
just above that range, thereby covering the whole time range for ultimate flexibility in
further data analysis. The two timing figures (TCSPC time and Time Tag) are stored as
one photon record. In order to work efficiently with current host computers, the photon
record is typically chosen as a 32 bit structure. A hardware First In First Out (FIFO)
buffer for 128 k events is used to average out bursts and deliver a moderate constant
data rate to the host interface. This way sufficient continuous sustained transfer rates
are possible in real-time.
Each single photon is recorded with its global arrival time and the ’microscopic’ delay
time with respect to the corresponding laser pulse. While the microscopic delay time
is evaluated in lifetime related analyses, the global arrival time can be used to form a
fluorescence intensity time trace, making all related analyses possible, like FCS, on / off
analysis etc. In addition this global arrival time can be synchronized with external trigger
pulses, for example the line or frame clock of LSMs, which is usedto extract imaging
information.
Intensity traces over time, as traditionally obtained from Multi-Channel-Scalers (MCS),
are obtained from TTTR data by evaluating only the time tags of the photon records.
Sequentially stepping through the arrival times, all photons within the chosen time bins
(typically milliseconds) are counted. This gives access to e.g. single molecule bursts (in
flow) or to blinking dynamics. The bursts can be further analyzed e.g. by histogram-

Chapter 255
ming for burst height and frequency analysis. Fluorescence lifetimes can be obtained by
histogramming the TCSPC (start-stop) times and fitting of the resulting histogram, as
in the conventional approach. In single molecule applications with very few counts per
histogram, faster algorithms based on maximum likelihood criteria may be used.
Lifetime fitting can either be done using iterative reconvolution, taking into account
the influence of the instrument response function (IRF), or as a tailfit, neglecting this
influence. Decay models up to four exponential components can be applied. A Maximum
Likelihood Estimator (MLE) method can be used to account for regions with low signal
intensity.
The strength of the TTTR format is used when both time figures are used together. For
instance, one can first evaluate the MCS trace to identify single molecule bursts, and
then use the TCSPC times within those bursts, to evaluate fluorescence lifetimes for
individual bursts.

Collaborations and Manuscripts
List of the collaborations that have make possible this PhD work:
• General Chemistry Department, University of Pavia, Pavia, Italy In particular the
group of Prof. P.Pallavicini
• Environment and Terrytory Department, University of Milano-Bicocca, Milan,
Italy
In particular Dr. P.Mantecca, Dr. G.Soncini and Dr. Maurizio Gualtieri (Polaris
Project)
• Department of Biotechnology and Bioscience, University of Milano-Bicocca, Milan,
Italy.
In particular Prof. Francesca Granucci, Dr. Ivan Zanoni, Dr. Tatiana Gorletta
and Dr. Marco Di Gioia (ENCITE project)
List of the manuscripts related to this PhD work:
• M. Caccia, L. Sironi, M. Collini, G. Chirico, I. Zanoni, and F. Granucci. ”Image filtering
for two-photon deep imaging of lymphonodes”, Eur. Biophys. J., 37:979987,
Jul 2008.
• S. Freddi, L. DAlfonso, M. Collini, M. Caccia, L. Sironi, G. Tallarida, S. Caprioli,
and G. Chirico. ”Excited-state lifetime assay for protein detection on gold colloidsfluorophore
complexes.”, J. Phys. Chem. C, 113:27222730, Jan 2009.
• L.Sironi, S.Freddi, L.D’Alfonso, M. Collini, T. Gorletta, S.Soddu, G.Chirico. ”P53
detection by fluorescence lifetime on a hybrid fluorescein-isothiocyanate gold nanosensor”,
J.Biomedical nanotechnology, 5: 683-691, 2009.
• L. Sironi, S. Freddi, L. DAlfonso, M. Collini, T. Gorletta, S. Soddu, G. Chirico.
”In-vitro and in-vivo detection of p53 by fluorescence lifetime on a hybrid FITCgold
nanosensor”, Proceedings of the SPIE, 7574:757403, 2010
256

Chapter 257
• M. Caccia, T. Gorletta, L. Sironi, I. Zanoni, C. Salvetti, M. Collini, F. Granucci., G.
Chirico. ”Two Photon Microscopy Intravital Study of DC-Mediated Anti-Tumor
Response of NK Cells”, Proceedings of the SPIE, 7565:75650Q, 2010
• P. Pallavicini, G. Chirico, M. Collini, G. Dacarro, A. Don, L. DAlfonso, A. Falqui,
Y. A. Diaz-Fernandez, S. Freddi, B. Garofalo, A. Genovese, L. Sironi and A. Taglietti.
”Synthesis of branched gold nanoparticles with tunable Near-InfraRed Localized
Surface Plasmon Resonance using a zwitterionic surfactant for the seed-growth
method”, Chem. Commun., 2011, DOI: 10.1039/C0CC02682D