Regulating particle morphology during a spray freeze drying ...

REGULATING PARTICLE MORPHOLOGY DURINGA SPRAY FREEZE‐DRYING PROCESS ANDDETECTING PROTEIN STRUCTURE USINGONLINE LIGHT SCATTERINGDer Naturwissenschaftlichen Fakultät derFriedrich‐Alexander‐Universität Erlangen‐Nürnbergzur Erlangung des Doktorgrades Dr. rer. nat.vorgelegt vonGeorg Stralleraus Wassertrüdingen

Als Dissertation genehmigt durch die NaturwissenschaftlicheFakultät der Friedrich‐Alexander‐Universität Erlangen‐NürnbergTag der mündlichen Prüfung:Vorsitzender der Promotionskommission:Erstberichterstatter:Zweitberichterstatter:Prof. Dr. Eberhard BänschProf. Dr. Geoffrey LeeProf. Dr. Wolfgang Frieß

IIFür meinen Großvater

DanksagungDie vorliegende Arbeit wurde am Lehrstuhl für Pharmazeutische Technologie derFriedrich‐Alexander‐Universität Erlangen‐Nürnberg in der Zeit von April 2006 bis Mai2010 angefertigt.Mein Dank gilt an erster Stelle Herrn Professor Dr. Geoffrey Lee, welcher mir dieMöglichkeit gab, die Promotion in seiner Arbeitsgruppe durchzuführen. Für dasangenehme Arbeitsklima, Ihre Unterstützung und Diskussionsbereitschaft beiwissenschaftlichen Fragestellungen danke ich Ihnen sehr.Ferner danke ich der Deutschen Forschungsgemeinschaft für die finanzielleUnterstützung des Projektes (Le 626/9‐2).Des Weiteren gilt mein Dank Prof. Dr. Wolfgang Frieß vom Lehrstuhl fürpharmazeutische Technologie und Biopharmazie der Ludwig‐Maximilian‐UniversitätMüchnen für die Bereitschaft die Aufgabe des Zweitgutachters dieser Arbeit zuübernehmen.Darüberhinaus möchte ich mich bei Professor Dr. Gerhard Winter bedanken,welcher mir die Möglichkeit gegeben hat AF4‐Untersuchungen an seinem Lehrstuhldurchzuführen. Die Zeit in München wurde mir durch die Hilfsbereitschaft undGastfreundschaft der gesamten Arbeitsgruppe erleichtert, ganz besonders möchte ichmich aber für die Unterstützung bei Tim Serno, Andreas Bosch und Elsa Kis bedanken.Großen Dank gebührt Herrn Dr. Roessner und Dr. Scherrers von WyattTechnology Europe für die stete Beantwortung von Fragen und Unterstützung rund umden miniDawn Treos.Ganz besonders gilt mein Dank aber natürlich allen Mitarbeitern des Lehrstuhlsfür Pharmazeutische Technologie der Universität Erlangen‐Nürnberg die mich bei meinerArbeit unterstützt haben.Joseph Hubert danke ich für seine unschätzbare Hilfe bei allen technischenProblemen. Deine gute Laune war stets ansteckend und Du hattest selbst bei größerenProblemen immer eine Lösung parat.

IVPetra Neubarth danke ich für die nette und tatkräftige Unterstützung bei allenVerwaltungsangelegenheiten und die Abnahme zahlreicher organisatorischer Aufgabenbei der Durchführung unterschiedlicher Lehrveranstaltungen.Luise Schedl danke ich für die rasche und geduldige Anfertigung unzähliger SEM‐Aufnahmen von Pulverpartikeln, sowie die kurzweilige Unterstützung bei derStudentenbetreuung im Propädeutikum.Christiane Blaha danke ich für die zuverlässige und schnelle Beschaffung derChemikalien und Materialien sowohl für die wissenschaftliche Arbeit und dieStudentenpraktika.Dr. Stefan Seyferth danke ich für die Unterstützung bei Computerproblemen undRatschläge aller Art. Danke dass Du immer ein offenes Ohr für mich hattest, die lustigenAbende bei der Weiterbildung in Dernbach werden mir immer in Erinnerung bleiben.Dr. Andreas Ziegler danke ich für seine Einführung in die Sprühgefriertrocknung,seine fachliche Unterstützung in der Anfangsphase meiner Arbeit sowie für diezahlreichen interessanten Diskussionen.Desweiteren möchte ich mich bei Dr. Heiko A. Schiffter bedanken. Auch wenn wiruns in der letzten Zeit nur selten in Erlangen gesehen haben, so hast Du mich immerunterstützt, sei es im Studium oder während der Promotion. Ich hoffe wir bleiben inKontakt und ich schaffe es endlich, Dich und Martina in Oxford zu besuchen.Ebenso danke ich meinen Kolleginnen und Kollegen Dr. Henning Gieseler,Dr. Henning Wegner, Dr. Alexander Mauerer, Dr. Joanna Manegold, Dr. Homy Lassner,Dr. Eva Meister, Silja von Graberg, Elke Lorenzen, Felix Wolf, Sabine Ullrich und UlrickePohl für den kollegialen Umgang und das gute Arbeitsklima während unserergemeinsamen Zeit am Lehrstuhl.Einigen meiner Mit‐Doktoranden möchte ich jedoch für die schönen Jahre die wirwährend unserer gemeinsamen Promotionszeit am Lehrstuhl hatten ganz besondersdanken.Dr. Harald Pudritz für die angenehme gemeinsame Zeit im Labor. Du hast mir dieAnfangszeit am Lehrstuhl sehr erleichtert. Deine lustigen Anekdoten und Gesprächehaben mir so manchen mißlungen Versuchstag gerettet.

Desweiteren möchte ich Anne Mundstock danken, mit der ich das letzte Jahrmeiner Assistentenzeit das Labor geteilt habe und oft gelacht habe.Dr. Anke Saß und Dr. Eva Wulsten danke ich für die gute Zusammenarbeit bei derStudentenbetreuung und die vielen schönen Abende in Erlangen/Nürnberg undwährend der gemeinsam besuchten Weiterbildungsseminaren quer durch Deutschland.Mario, ich hoffe wir werden noch viele Club‐Spiele im Stadion anschauen können.Dr. Sebastian Vonhoff danke ich besonders für den gemeinsamen Start alsYoungster am Lehrstuhl. Wir hatten am Anfang der Promotion mit denselbenSchwierigkeiten zu kämpfen ‐ ich wünsche Dir und Deiner Familie alles Gute für Euregemeinsame Zukunft in Kemnath.Dr. Stefan Schneid danke ich nicht nur für das Korrekturlesen dieser Arbeit,sondern für alle Diskussionen und Unterhaltungen außerhalb der Welt derGefriertrocknung.Jakob Beirowski danke ich für seine Freundschaft und den vielen Gesprächenauch außerhalb der Universität. Die spektakuläre Woche in Barcelona werde ich nichtvergessen.Simone Reismann danke ich für die vielen lustigen Unterhaltungen und diezahlreichen Abende an denen wir Kollegen bei Ihr gefeiert haben.Ganz besonders möchte ich mich bei Susanne Rutzinger für Ihre offene undheitere Art bedanken. Es ist schön mit Dir befreundet zu sein, ich hoffe ich werde nochviele Abende mit Dir und Dominik genießen können.Meinen Wahlpflichtfachstudentinnen Julia Schirmer, Conny John und UteGollwitzer danke ich ganz herzlich für Ihr großes Engagement. Auch wenn nicht alleErgebnisse Einzug in diese Arbeit gefunden haben, so ward ihr doch eine unschätzbareHilfe gewesen.Mein größter Dank gilt jedoch meiner Familie, meinen Eltern, Brigitte und Hans,meinen beiden Brüdern Uli und Ludwig sowie meiner lieben Freundin Heike. Ich möchtemich dafür bedanken, dass Ihr mir bei allen wichtigen Entscheidungen immer mit Ratund Tat zur Seite gestanden und mich aufgebaut habt wenn die Maschinen mal wiederanders wollten als ich. Es ist schwer zu beschreiben was und wie viel ich Euch verdanke.

VIParts of this thesis have already been presented or published:I. G. Straller, G. Lee. Determination of Protein Aggregation during Spray‐Freeze‐Drying with online Light Scattering. 6 th World Meeting onPharmaceutics, Biopharmaceutics and Pharmaceutical Technology, Barcelona(SPAIN), April 7‐10, 2008.II.G. Straller, G. Lee. Protein Protection during a Spray‐Freeze DryingProcess with Bovine Carbonic Anhydrase. Jahrestagung der DeutschenPharmazeutischen Gesellschaft, Bonn (Germany), Oktober 8‐11, 2008III.G. Straller, T. Serno, G. Winter, G. Lee. Spray‐Freeze‐Drying of CarbonicAnhydrase with online Light Scattering, Asymmetrical Flow Field FlowFractionation vs. SEC. 7 th World Meeting on Pharmaceutics, Biopharmaceuticsand Pharmaceutical Technology, Valletta, (MALTA), March 8‐11, 2010,

TABLE OF CONTENTS ITable of contents1 INTRODUCTION 12 INTRODUCTION TO SPRAY FREEZE‐DRYING 42.1 Needle‐free injection 42.2 Spray freeze‐drying process 82.2.1 Atomization 102.2.2 Freezing step 112.2.3 Freeze‐drying step 132.2.4 Cryo‐ and lyo‐protection of proteins 152.2.5 Regulation of SFD particle´s appearance 163 LIGHT‐SCATTERING 183.1 Blue sky and red sunsets 183.2 Static light scattering (SLS) 193.2.1 Static light scattering by small particles 193.2.2 The specific refractive index increment, ∂n/∂c 243.2.3 Static light scattering by large particles 243.2.4 Instrumentation and limitation of single angle LS 273.2.5 Dynamic light scattering (DLS) 283.2.6 Instrumentation and limitation of DLS 313.3 Different separation techniques 313.3.1 Size exclusion chromatography (SEC) 323.3.2 Asymmetrical flow field‐flow fractionation (AF4/aF‐FFF) 33

IITABLE OF CONTENTS4 MATERIAL AND METHODS 354.1 Materials 354.1.1 Proteins 354.1.1.1 Bovine Carbonic Anhydrase 364.1.1.2 Bovine Serum Albumin 384.1.1.3 L‐Lactic Dehydrogenase 394.1.2 Excipients and reagents 404.2 Methods 424.2.1 Spray Freeze‐Drying (SFD) 424.2.2 Spray‐drying (SD) 444.2.3 Size Exclusion Chromatography (SEC) 444.2.4 Asymetrical field flow field fractionation (AF4) 464.2.5 Enzymatic Activity Assay of bovine Carbonic Anhydrase 474.2.6 Enzymatic Activity Assay of LDH 484.2.7 Turbidity measurements 494.2.8 Water content, Karl Fischer titration 494.2.9 Differential Scanning Calorimetry (DSC) 504.2.10 Wide‐Angle‐X‐Ray‐Diffraction (WAXD) 504.2.11 Hg Porosimetry 504.2.12 BET measurement 514.2.13 Particle Size Analysis 514.2.14 Native PAGE 514.2.15 Scanning Electron Microscopy (SEM) 524.2.16 Fourier Transformation Infrared Spectroscopy (FTIR) 525 RESULTS AND DISCUSSION 535.1 Working with Light Scattering 535.1.1 LS‐Calibration and Normalization 535.1.2 RI‐Calibration 595.1.3 UV‐calibration 62

TABLE OF CONTENTS III5.1.4 HPLC‐Calibration 635.1.5 Determination of protein extinction coefficients 665.1.6 Determination of dn/dc values of proteins 685.1.7 Protein conjugate analysis 695.1.8 A 2 ‐measurements of proteins 715.1.8.1 Batch‐mode 715.1.8.2 Online‐mode 725.1.9 Solvent contamination 765.1.10 Evaluation of an AF4 separation method 805.2 Bovine Carbonic Anhydrase (bCA) 825.2.1 Different bCA batches 825.2.2 Protein characterization 825.2.3 Denaturation of bCA 885.2.3.1 Denaturation with guanidinium hydrochloride (GuHCl) 885.2.3.2 Thermal denaturation 915.3 SFD/SD of pure bCA 945.3.1 Aggregation of bCA during SFD 945.3.2 Protein aggregation of bCA during SD 1035.3.3 The atomizing step of SFD 1085.3.3.1 Use of different ultrasonic nozzles 1085.3.3.2 Atomization into different media 1155.3.4 SFD of bCA at different protein concentrations 1185.3.5 Stabilization of bCA during SFD with different excipients 1215.3.5.1 SFD powders from LFs with low solid contents (15 % w/w) 1215.3.5.1.1 SFD bCA‐trehalose 15 % (w/w) 1215.3.5.1.2 SFD bCA‐binary mixtures of 15 % (w/w) 1265.3.5.1.3 SFD bCA‐complex mixtures of 15 % (w/w) 1325.3.5.1.4 SFD powders from LFs with high solid contents (30 % w/w) 1375.3.5.2 Comparison of AF4‐ and SEC‐measurements with bCA 142

IVTABLE OF CONTENTS5.4 Lactate Dehydrogenase (LDH) 1475.4.1 Protein characterization 1475.4.2 SFD of pure LDH 1485.4.2.1 Protein aggregation of LDH during SFD 1485.4.2.2 Stabilization of LDH during SFD with trehalose 1515.5 Regulation of SFD‐particle shape 1555.5.1 Excipient‐induced change of particle morphology 1555.5.2 Changes in particle morphology induced by variations in the FD‐cycle 1605.5.2.1 Collapsed bSA particles 1605.5.2.2 Collapsed bCA particles 1716 CONCLUSIONS 1757 ZUSAMMENFASSUNG 1808 REFERENCES 1869 CURRICULUM VITAE 206

ABBREVIATIONS VList of abbreviationsCapital lettersA 2 , A 3C 3D(τ)D pG(τ)I θI 0KMM WN AP θR θTT bT cT gT g´T inT mT outT PVsecond / third virial coefficientpropanetranslational diffusion coefficientpenetration depthautocorrelation functionoverall intensity of scattered lightintensity of incident lightoptical constantmolar massweight averaged molar massAvogadro´s numberangle‐dependent intensity of scattered lightRayleigh ratiotemperature in Kelvinboiling pointcollapse temperatureglass transition temperatureT g of the maximally freeze concentrated solutedrying air inlet temperaturemelting temperatureoutlet temperatureproduct temperaturevolume

VIABBREVIATIONSSmall lettersaacd pi‐C 5ieidimk Bn 0r Hr GrpmrTsctVcVxamino acidconcentrationparticle diameterisopentanintraepidermalintradermalintramuscularBoltzmann´s constantrefractive index of the solventhydrodynamic radiusradius of gyrationrevolutions per minuteretention timesubcutaneoustimechannel‐flow rate (AF4)cross‐flow rate (AF4)Greek letters / numbers1° primary drying2° secondary drying∂n/∂cλλ 0ησν PΓrefractive index increment, also: dn/dcwavelengthwavelength of incident light in vacuumviscositystandard deviationparticle velocitydecay rate

ABBREVIATIONS VIIExpressionsAF4APIbCAbSADLSDSCEPIESCAFDFTIRGPCGuHClHELIRLDHLFLMHLN 2MALSMDRMWCOMWDQELSPAGEPDBPIPNPASDSECasymmetrical flow field‐flow fractionation, also: aF‐FFFactive pharmaceutical ingredientbovine carbonic anhydrasebovine serum albumindynamic light scatteringdifferential scanning calorimetryepidermal powder immunizationelectron spectroscopy for chemical analysisfreeze‐drying / freeze‐driedfourier transformation infrared spectroscopygel permeation chromatographyguanidinium hydrochloridehen egg lysozymeinfraredL‐lactic dehydrogenaseliquid feedlactose monohydrateliquid nitrogenmulti angle light scatteringmonomer‐dimer (peak) ratio obtained from the UV‐signalmolecular weight cut offmolecular weight distributionquasi elastic light scattering, also: DLSpolyacrylamide gel electrophoresisprotein data bank, UniProtKB / Swiss‐Protpowder injectorp‐nitrophenyl acetatspray‐drying / spray‐driedsize‐exclusion chromatography

VIIIABBREVIATIONSSEMSFDSFLSLSSSAUVscanning electron microscopyspray freeze‐drying / spray freeze‐driedspray freezing into liquidstatic light scattering,specific surface areaultra violet

INTRODUCTION 11 IntroductionBiologically manufactured drugs represent a heterogeneous group of proteins andpeptides that resemble or imitate substances produced naturally in the body. Such drugsinclude monoclonal antibodies, fusion proteins, recombinant proteins, growth factors,angiogenetic factors and expression vectors [Boehncke and Radeke 2007]. The recentrisk of a possible pandemic H1N1‐influenza shows that biologicals are in great demand.The European Medicines Agency (EMEA) has been working on a crisis‐management plansince the outbreak of the “swine flu” pandemic in April 2009. A pandemic alert level wascalled by the WHO in the middle of 2009 [EMEA 2009]. It took just 6 months before 3different H1N1‐influenza vaccines received official approval throughout Europe by theEMEA.Usually vaccination is performed using a syringe and needle via injection of liquidinto the muscle tissue (im) or under the skin (sc). Another promising way forimmunization was reported by Chen et al. [2001; 2002; Osorio et al. 2003], the so‐called“epidermal powder immunization” (EPI). Influenza and hepatitis B antigen‐coated goldparticles were tested in mice. The particles were injected needle‐free with a specialdelivery device (PowderJect® device) powered by a small volume of compressed heliumgas. Vaccination under murine skin led to a high antigen concentration in the Langerhanscells with a cellular and humoral immune response.EPI offers some promising advantages over conventional injection with a syringeand needle. During powder injection there is no contact with blood, and the hazard oftransmission of contractible diseases is minimized. Compliance may increase, due to thelack of a cannula. Another advantage of dry powder usage is improved product stabilityduring storage and delivery compared to liquid formulations [Prestrelski and Burkoth1999].For effective vaccine delivery EPI‐powders must exhibit the necessary propertiesto penetrate the skin, i.e. adequate particle size, density and durability [Lahm and Lee2006]. A common method for manufacturing suitable powder particles is spray freeze‐

2 INTRODUCTIONdrying (SFD) [Maa et al. 2004]. Protein solutions are spray‐dried into liquid nitrogen (LN 2 )with a subsequent water removal step in a freeze‐dryer. This procedure leads to highprocess yields of consistent, near‐spherical particles with an adjustable size distribution.Conventionally produced SFD‐powders have a high specific surface area and a lowdensity compared to a spray‐dried product [Maa et al. 1999]. These particles areunsuitable for EPI, although potentially useful for pulmonary application [Edwards et al.1997]. Small spherules with a low aerodynamic size are more likely to reach deeper lungtissue for effective deposition.Sonner et al. [2002] and later Rochelle and Lee [2007] revealed the two mostimportant factors for increasing the density of spray freeze‐dried spherules. These arechanges in freezing conditions, such as an additional annealing step, and excipientcomposition‐induced changes in particle appearance. An anti‐plasticizing effect ofdifferent excipients was discussed. This phenomenon resulted in particle shrinkage withwrinkled surfaces and resulting higher powder tap densities.The first part of the current work focusses on measuring aggregation in proteinloadedformulations with carbinoc anhydrase (bCA) and lactic dehydrogenase (LDH)prepared by SFD. A three‐detector system with multi angle light scattering (MALS) andquasi elastic light scattering (QELS) combined with HPLC‐SEC was established to examinestructural alteration of spray‐freeze‐dried protein particles. Different process stepsduring SFD and their influence on protein integrity were analyzed. Various protein‐ andexcipient‐loaded SFD powders were generated and analyzed. The stability of each batchwas investigated to determine the best drying conditions for the model protein used.Furthermore, advantages of the 3‐detector system over conventional HPLC‐SEC weredemonstrated. With asymmetrical field flow field fractionation (AF4) another separationtechnique was examined. The consistency of bCA‐SFD samples in HPLC‐SEC wascompared to the results from AF4 runs and the differing results were discussed.The subsequent part describes different ways to modify the surface morphologyof protein‐loaded SFD‐powders. Bovine serum albumin (bSA) was used as model protein.At first, different SFD‐batches were created with varying disaccharide‐protein ratios. Theprocess parameters during manufacturing were maintained constant. Differences inparticle morphology were discussed. The next step was to create shrunken particles

INTRODUCTION 3using pure protein solution as liquid feed without any additional excipient. Beside bSAalso bCA was examined. Various adjustments to the freeze drying program wereexecuted, bCA damage during the process was additionally tested.At the end of this project a fairly comprehensive picture of protein aggregationduring SFD is obtained. Additionally shrunken pure‐protein SFD particles had beensuccessfully prepared.

4 INTRODUCTION TO SPRAY FREEZE‐DRYING2 Introduction to spray freeze‐drying2.1 Needle‐free injectionThe concept of needle‐free injection is defined in the ISO definition (InternationalOrganization for Standardization). It describes the injection of any medicinal productthrough the skin by means of pressure without using a needle [ISO/TC 84/WG 4 2004].The first extensively used needle‐free injector was the multiple‐use‐nozzle jet injector(MUNJI). Due to hepatitis B incidences in 1986, where patients were infected with HBVafter receiving injections, the MUNJI´s were resplaced by disposable syringe jet injectors(DSJI´s). Here the API‐containing cartridge and all parts of the injection device in contactwith the skin are exchanged after every actuation, thus avoiding the transmission ofpathogens to subsequent patients [WHO 2009]. With the development of powderinjectors (PI´s) a new form of vaccination method was established.a) b)Figure 1. Different needle‐free injection devices: a) Examples of DSJI´s:1) PharmaJet® (PharmaJet Inc.) and 2) Biojector® 2000 (Bioject) – dispersed vaccinescan be given intradermally [WHO 2009]. The term “jet injection” is a conventionalsynonym for needle‐free application; b) Profile of the rechargeable PowderJect ND®device. The release of 5 mL helium gas from the gas chamber accelerates the dryvaccine particles from the powder cassette into the skin. After recharging the devicewith a new vaccine cassette and filling the gas chamber with He, the PowderJect ND®device is ready to be used again [Chen et al. 2000].EPI now no longer depends on the use of liquids like solutions or suspensions, but on theinjection of dry powders into the skin (see Figure 1).The skin is not only a protective barrier against pathogens; it is also an importantimmune organ. Due to the presence of a mesh of antigen‐presenting cells (APC) the skinis an attractive tissue for powder vaccination [Maa et al. 2003]. Such APC are for

INTRODUCTION TO SPRAY FREEZE‐DRYING 5example the Langerhans cells (LCs) and the dermal dendrocytes, located in the viableepidermis and in the dermis, respectively (Figure 2a). APCs play an important role instimulating both the innate and the adaptive immune responses. Small vaccines that aredelivered into the epidermis act as antigens. The particles in the skin are taken up byAPCs and transported to lymph nodes, where they are presented to CD4+ and CD8+ Thelper cells (Figure 2b). The presentation takes place via the major histocompatibilitycomplex (MHC) class II antigen, and the humoral immune response is initiated[Banchereau and Steinman 1998]. Subsequently, antibodies and B‐lymphocytes aregenerated, and the person is immunized against the same kind of antigen forprospective infections.a) b)Figure 2. a) Top layer of human skin comprising epidermis and dermis. Modifiedgraphic by Benjamin Cummings. b) T cell stimulation due to activated APIs. Modifiedpicture according to RIMD [2009].A comparison of the delivery of vaccine antigens to the dermis or epidermis (intradermaldelivery) to that into muscle or subcutaneous tissue shows a superior protectiveimmune response due to smaller quantities of vaccine antigens via the intradermal (id)route [Henderson et al. 2000; WHO 2009]. As a result of the high APC concentration inthe top skin layer a strong immunization efficiency is deemed possible via an intraepidermalinjection with the use of a powder injector (PI) [Dean et al. 2003].

6 INTRODUCTION TO SPRAY FREEZE‐DRYINGDifferences concerning tissue penetration between needle or needle‐freeinjection are demonstrated in Figure 3a. Unlike needle‐free powder injection,conventional or liquid jet injections practically exclude any intraepidermal delivery (ie).During EPI only a small amount of vaccine reach the dermis – most powder particles aredecelerated during the impact on the skin surface and remain in the viable epidermis.Figure 3b illustrates that id delivery, compared to im or sc delivery may lead toequivalent immune responses with influenza or hepatitis B vaccines although using areduced quantities of antigen for id. Injection [WHO 2009].a.)Figure 3. a) Due to the lack of vasculature in the epidermis, EPI avoids any contactwith blood. Modified picture according to Ziegler [Ziegler 2006]. b) Schematicrepresentation of vaccine dose‐response relationships, according to clinical trialsagainst 11 different diseases such as influenza or hepatitis B [WHO 2009]. Intradermalinjection compared to im or sc injection.b.)To achieve an efficient delivery and therefore optimal immunogenicity during EPI, thesolid vaccines must exhibit specific particle characteristics in order to penetrate the skin(Table 1). Mitchell et al. [2001] demonstrated that the penetration depth D P of smallgold particles depends on the particle density ρ P [kg/m 3 ], its diameter, d P [m] and theparticle velocity, ν P [m/s]:Equation 1.D P ∼ ν P * d P * ρ PWhile working with one PI device, ν P cannot be varied in practice to any substantialdegree, so primarily the particle size, d P , and density, ρ P , are the crucial adjustableparameters for particle penetration. But also here the margin is low: if the particles are

INTRODUCTION TO SPRAY FREEZE‐DRYING 7larger than 75 μm, small unaesthetic bruises are possible; even larger particles couldcause significant skin damage. If the mean mass aerodynamic diameter is less thanabout 20 μm, the vaccines can be stopped or deflected by the skin surface and do notpenetrate into the skin [Burkoth et al. 1999]. The injection velocities can differ from 200– 3,000 m/s, whereas ideal particle densities for EPI lie between 100 – 25,000 kg/m 3 ;ρ P values between 800 and 1,500 kg/m 3 have been suggested as optimal [Sarphie andBurkoth 1997].Physical and mechanical requirementsuniform size (∼30 – 50 μm in diameter)uniform (spherical) shapehigh density (800 – 1500 kg/m 3 )sufficient mechanical strength to withstandimpact with the skinPharmaceutical and biologicalrequirementsappropriate parenterally acceptableexcipientsshort / long term chemical and physicalstabilityappropriate biological properties oncedelivered to the skinviable manufacturabilityTable 1. Powder requirements for needle‐free powder injection according to[Ziegler 2006].Vaccine powders must achieve sufficient mechanical durability to sustain the enormousphysical stress during EPI. Firstly, the particles are entrained by the helium gas whenthey are rapidly accelerated to several 100 m/s. Secondly the strong impact on theepidermis surface forces the particles to decelerate rapidly [Prestrelski et al. 2002].Ziegler et al. [Ziegler et al. 2006] demonstrated that some carbohydrate co‐solutes donot only stabilize proteins during the powder manufacturing process, but also protectthe API from vaccination‐induced stress. Based on the results from catalase SFDparticles, the protein was more damaged by mechanical stress (shear and impact) thanby static pressure. Pure catalase particles reacted more sensitive to any kind of stressthan those formulated with carbohydrates.

8 INTRODUCTION TO SPRAY FREEZE‐DRYING2.2 Spray freeze‐drying processThe typical solid dosage form for biopharmaceutical products is a lyophilized cake in aglass vial. The dried substances are reconstituted in water or buffer to be deliveredparenterally ‐ mostly using a syringe, but also a liquid jet injector is imaginable. Whendry powder is used for vaccination via EPI, conventional freeze‐drying is inappropriatefor particle preparation [Maa et al. 2003]. Certainly, the cake could be milled intosmaller fragments using a second manufacturing step, but the process yield would bereduced and the control of the particles´ properties and bioburden would be poor [Yu etal. 2002; Costantino et al. 2004]. Spray freeze‐drying of proteins is suitable to obtain fineand well‐characterized powders. SFD products are used to achieve sustained drugdelivery or to serve for specialized applications such as dry powder inhalation,intradermal or parenteral delivery [Maa et al. 1999; Sonner et al. 2002] ‐ also dissolutionenhancement for poorly water‐soluble drugs is possible [Leuemberger 2002]. Duringwater‐removal in SFD no substantial shrinkage of the frozen droplets occurs, contrary tothe evaporative water loss during SD [Masters 1991].Generally powder particles produced via SFD are not hollow as frequently observed withSD, despite their more porous character resulting from the freeze‐drying procedure[Maa et al. 1999]. The manufacturing process is composed of two consecutively appliedprocedures. First, the protein is dissolved in water or buffer to the desiredconcentration. The freshly‐prepared solution is atomized typically with an ultrasonicnozzle and directly sprayed into a bowl filled with chilled cryogen (e.g. LN 2 ). Second, theice of the almost instantly frozen, spherical particles is sublimated under vacuum in asubsequent lyophilization step [Maa et al. 2004]. The container filled with the dispersionof frozen spheres is transferred onto the pre‐cooled shelves of a freeze‐dryer, avoidingparticle thawing or melting. Once the cryogenic media is completely vapourized, thedehydration process can be iniciated [Sonner et al. 2002].Different excipients are added to the liquid feed in order to support theparenteral pharmaceutical form or to protect the protein/peptide during themanufacturing process (Table 2). Proteins can be damaged due to different stressesduring atomization, freezing or the subsequent water removal step in the freeze‐dryer[Costantino and Maa 2004] (Table 3).

INTRODUCTION TO SPRAY FREEZE‐DRYING 9Excipients for FD and SFDproductsDesired effect / mechanism• buffering agent control of the pH (different buffers: sodium citrate, sodiumphosphate, Tris(tromethamine) buffer)• stabilizing agento cryprotectant improve protein stability in aqueous solution and upon thefreezing step (PEG, sugars)o lyoprotectant improve protein stability upon lyophilization (water‐removal) andstorage (sucrose, lactose, trehalose)o anti‐oxidant free‐radical scavenger avoiding oxidative damage of the API(ascorbic acid, cystein HCl)o surfactant saturation of hydrophobic interfaces to protect proteins againstdenaturation, aggregation or binding (polysorbat 20 and 80,poloxamer 188)o chelators complexation of heavy metal ions, improving the efficiency of antioxidants(EDTA)• bulking agents structural enhancement of freeze‐dried products / increase ofmass for low dosed potent APIs (mannitol, dextran)• tonicity modifiers parenteral formulations are adjusted to be isotone on redispersal(NaCl, sucrose, mannitol)• antimicrobialpreservative agentsif multiple administrations from the same vial are intended, theformulation must be preserved to avoid microbial contamination(thiomersal, benzyl alcohol)Table 2.2004].Possible e used in FD or SFD products. Abstract list from [Constantinostress/events stress source potential mechanism(s) of damageatomization shear force protein conformation changesatomization air‐water interfaces protein adsorption and conformationchangesfreezing cold denaturation protein activity lossfreezing concentration and pH changes physical or chemical interactions due toenvironmental changesfreezing ice‐liquid interfaces protein adsorption and conformationchangesfreezing mechanical residual stress protein conformation changesdehydrationwater removal to disrupthydrogen bondingprotein conformation changesTable 3. Stress events associated with SFD [Costantino and Maa 2004].

10 INTRODUCTION TO SPRAY FREEZE‐DRYING2.2.1 AtomizationConversion of the LF to fine droplets can either be achieved with an ultrasonic nozzle orwith a two‐fluid nozzle. An ultrasonic nozzle produces ultrasound waves induced by apiezoelectric transducer. High‐frequency electrical energy is translated into mechanicalmotion at a frequency that is capable of atomizing the solution into small droplets. Dueto changing atomization energies, the size distribution of the sprayed droplets can becontrolled by adjusting nozzle frequency or atomization conditions. Higher frequenciesor lower pump velocities result in smaller droplets [Rochelle 2005]. In two‐fluid nozzlesthe LF is nebulized with the help of a pressurized gas‐flow that disintegrates the solutioninto fine droplets. With this method the droplet sizes can be controlled by adjusting theliquid feed rate or the atomizing air flow. It should be noted that droplet size is anessential factor for the subsequent freezing and drying steps. Smaller droplet sizes leadto an increase in specific surface area (SSA) that may be associated with an increase inprotein damage [Costantino et al. 2000].Stress sources during atomization that lead to potential damage of proteinstructure or activity are similar to those in SD [Maa et al. 1999]. Mainly shear forcesduring solute transportation and atomization or protein adsorption at air‐waterinterfaces can result in protein unfolding or aggregation. Surfactants are used asstabilizers in some protein formulations [Wang 1999]. Costantino [2002] demonstratedthat addition of polysorbate 80 near its critical micelle concentration leads to a proteinstabilizing effect with only half of monomer loss for bSA. With electron spectroscopy forchemical analysis (ESCA), Sonner et al. [2002] proved that low concentrations ofpolysorbate 80 are adequate to eliminate proteins from liquid/air interfaces and thusprevent denaturation. The surface concentration of trypsinogen in SFD particlesdecreased, and there was a residual activity increase of 8 % compared to unstabilizedprotein.Heller et al. [1999] suggested that another legitimate concern of SFD could bethe high energy of the sonicating nozzle, that could by itself denature proteins.Vonhoff [2008] showed that (especially for aggressive atomization conditions) atemperature increase or cavitation may lead to high protein damage. The generation offree radicals due to cavitation bubbles produced substantial oxidation of atomized KI‐

INTRODUCTION TO SPRAY FREEZE‐DRYING 11solutions. The residual activity of SFD α‐chymotrypsin was doubled after adding ascorbicacid in a molar ratio of 1:2 [Vonhoff 2010].2.2.2 Freezing stepFreezing is the conversion from a liquid solution into a solid at low temperatures, whereice formation is initiated by nucleation after a certain degree of super‐cooling of thesolution. A fast freezing rate leads to high supercooling and rapid water crystallizationthat creates a large number of small ice crystals. In contrast, fewer but larger ice crystalsare generated if the cooling rate is low. After subsequent water‐removal the small‐sizedice crystals leave behind small pores. The removal of water through tiny pores is nowexpected to be slow during primary drying (1°). In contrast to 1° drying, the waterdesorption rate will increase during secondary drying (2°). Here the drying processdepends on evaporation at the solid‐vapour interface and diffusion of water within thesolid [Pikal et al. 1990]. Both steps will be enhanced by a large specific surface areacaused by numerous, fine channels. After most of the water has been transferred to ice,crystalline and amorphous solutes solidify at the eutectic temperature, T e , or the glasstransition temperature for maximal freeze concentration, T g´, respectively (Figure 4).a) b)Figure 4. Freezing performance of different solutes in water: a) crystalline (NaCl)and b) amorphous solute [Nail and Gatlin 1993].Due to continuous ice formation, proteins and other amorphous excipients remaindissolved in a decreasing volume of water. The mixture is converted from a highlyviscous “syrup” to a rigid amorphous solid glass [Wang 2000]. The freeze‐concentration

12 INTRODUCTION TO SPRAY FREEZE‐DRYINGprocess can damage proteins. Concentration changes, pH‐shifts induced by irregularcrystallization of buffer salts or interactions at ice‐liquid interfaces may cause proteinadsorption or conformational changes [Heller et al. 1999; Costantino 2004].After nebulization the fine droplets descend through the cold vapour phase andbegin to freeze. When striking the surface of the cryogenic liquid the droplets are almostimmediately converted to a solid form [Yu et al. 2006]. Costantino and Maa [2004]calculated the freezing time during SFD for pure water droplets in LN 2 with radii of 10,30 and 100 μm as 0.16 ms, 1.56 ms and 16 ms, respectively. Small ice crystals with largeice‐water interfaces are generated – unlike slow freezing in conventional FD, wherefewer but larger ice crystals are formed [Wang 2000]. The use of other cryogenic agentsapart from LN 2 (i.e. propane: C 3 or isopentane: i‐C 5 ) leads to slight differences in freezingtime and thus in particle morphology [Engstrom et al. 2007b; Gieseler et al. 2009]. Inspite of its low boiling point (Tb LN2 = ‐196°C), liquid nitrogen induces slower cooling ratesthan other cryogens (Tb C3 = ‐190°C, Tb i‐C5 = ‐160°C). Responsible for this is the so‐calledLeidenfrost effect. Nitrogen molecules start to vapourize if an unfrozen droplet is inclose contact with LN 2 . These molecules temporarily enclose the droplet within aninsulating vapour layer and thus decelerate freezing [Engstrom et al. 2007b].This rapid freezing step results in the typical physical properties of SFD particles.Atomized and flash‐frozen droplets maintain their size and spherical shape, even afterwater‐removal, which can explain the high porosity of the dry SFD powder particles andtheir low particle densities [Sonner et al. 2002]. Even size variations within the same SFDexperiment lead to substantial differences in freezing time between the smallest and thelargest droplets (Figure 5). After water‐removal, the smaller particles have a relativelyhigher SSA, due to their finer pore structure [Sonner et al. 2002; Engstrom et al. 2007].Rapid cooling rates reduce crystallization tendencies, which minimizes concentration ofsolutes and pH changes [Yu et al. 2002]. Thus the formation of amorphous structures issupported that mechanically prevent protein relaxation processes and reduce damageby the formation of a glass [Heller et al. 1999]. Because atomized droplets are shockfrozenthere is less time for protein adsorption on water‐ice interfaces [Webb et al.2002; Engstrom et al. 2007b]. Maa and Prestrelski [2000] suggested that auxiliaryprotein damage may be caused by the fast moving freezing front that passes throughthe freezing droplet.

INTRODUCTION TO SPRAY FREEZE‐DRYING 13a) b)c) d)Figure 5. Different freezing behaviour of LN 2 ‐supercooled particles due to varyingdroplet sizes: a + b) dilute/concentrated solution with high supercooling – equitable tosmall atomized droplets. High nucleation rates lead to small domains of water andsolute domains in the unfrozen water; c + d) dilute/concentrated solution with lowsupercooling – equitable to larger droplets; Ice particles are represented as whitedomains and solute precipitate as solid dots or grey regions. Pictures taken fromEngstrom et al. [Engstrom et al. 2007b] are originally presented for SFL, but thetransfer to SFD is legitimate.2.2.3 Freeze‐drying stepDuring the water‐removal step in SFD no substantial shrinkage of the frozen dropletsoccurs, contrary to evaporative water loss during SD [Masters 1991]. Lyophilization andwater sublimation start with the primary drying (1°) phase. The chamber pressure isreduced while the shelf temperature is raised to sustain continuous ice sublimation(Figure 6a). The chamber pressure is well below the vapour pressure of ice, so the frozenwater sublimates and deposits as ice onto the cold plates of the condenser (Figure 6b)[Tang et al. 2006]. During the drying process the product temperature (T P ) should notexceed the maximum allowable temperature, which is determined either by the eutectictemperature, T e , for crystalline, or the T g´ for amorphous substances. Alternatively the

14 INTRODUCTION TO SPRAY FREEZE‐DRYINGcollapse temperature, T c , of the product, which lies some degree above T e / T g´, can beemployed [Tang and Pikal 2004].a) b)Figure 6. a) Phase diagram of water ‐ different steps during lyophilization:1) freezing step, 2a) and 2b) primary drying (1°), 3) secondary drying (2°). Modifiedpicture according to Bauer [Bauer et al. 2006]; b) Schematic construction of a freezedryer[Nail and Gatlin 1993].During the secondary drying step (2°) the water that is either bound at the surface of thecrystals in a crystalline product or embedded in the highly viscous amorphous matrix isremoved [Wang 2000]. This water is also called “unfrozen water”. Due to the smalleramounts of unfrozen water for crystalline products, the secondary drying can be keptshort. The relationship for glassy products is the reverse; here the rate of 2° drying islimited. Much more water is trapped in the glassy matrix, and the mechanism ofmolecular diffusion within the dried particles is slow [Pikal et al. 1990]. Also in the 2°drying step the product temperature must not exceed Tg in order to avoid collapse[Wang 2000]. Nevertheless the shelf temperature during secondary drying should be setas high as possible, since the glass transition temperature rises faster than T P .Additionally, the moisture content decreases rapidly only in the first hours before aplateau is reached that is only dependent on the drying temperature [Pikal et al. 1990].It is not necessary to reduce the chamber pressure for 2° drying step, as it should nothave any measurable influence on the drying rate. As reported in chapter 2.2.2, thedrying rates of the primary and secondary drying steps are strongly influenced by theprevious freezing step and the size of the generated ice crystals [Tang and Pikal 2004].

INTRODUCTION TO SPRAY FREEZE‐DRYING 15Because residual water plasticizes the amorphous phase and leads to a decreasein glass transition temperature, the generated SFD particles have to be dry enough sothat T g is well above room temperature. This makes transportation and storage easier;also a long shelf‐life is possible. Gieseler and Lee [2009] showed that the rate‐limitingfactor of the primary drying during SFD is the poor heat transfer through the large interparticulatevoids of the packed particle bed at the bottom of the vial. The smaller layerthickness in aluminium bowls (according to 4.2.1) should produce a lower productresistance than that formed in a half filled vial, leading potentially to higher heattransfer.2.2.4 Cryo‐ and lyo‐protection of proteinsAs seen in Table 3, it is important to protect proteins from different kinds of stresssources that may occur during a SFD process. A “cryoprotectant” protects peptides andproteins both in nonfrozen aqueous solutions and during freezing (Figure 7a), whereas“lyoprotection” is effective in the drying step. In the presence of a cryoprotector (i.e.saccharides, polyols, amino acids, etc.) the protein is preferably surrounded by watermolecules (preferential hydration or preferential binding), while the stabilizing agent ispreferentially excluded from the protein´s surface (preferential exclusion, Figure 7b).Unlike in the native state, the unfolded protein has a higher surface area from which thecryoprotector has to be excluded ‐ the chemical potential for the protein would be muchhigher [Carpenter et al. 1997]. Thus the native conformation is preferred over theunfolded state and the protein is stabilized [Pikal 2002]. Other stabilization mechanismsfor cryprotection have been proposed, like reduction of the surface tension or decreaseof diffusion of the reacting molecules by elevating the viscosity of the liquid[Wang 2000].As the hydration shell of proteins is removed during freeze‐drying, the previouslydiscussed theories cannot be used for lyoprotection [Crowe et al. 1993a]. Due to the lackof water, excipients have to substitute its function in generating hydrogen bonds thatare essential to protect the native structure of the protein [Crowe et al. 1993b; Pikal2002]. Well suited stabilizers for the so‐called “water‐replacement hypothesis” aremono‐ and disaccharides. Ambitious sterically structured excipient molecules fail to save

16 INTRODUCTION TO SPRAY FREEZE‐DRYINGthe native character due to steric hindrance and less hydrogen bond formation. The“glassy immobilization” describes another theory for lyoprotection. The addition ofamorphous excipients to the formulation leads to the formation of a highly viscous glassat freezing temperature below the T g´. The viscosity is very high (> 10 10 Pa * s), so thatthe interconversion from one conformational form of the protein to another is sloweddown or totally disrupted [Hagen et al. 1996]. Fragile glasses (i.e. sucrose, trehalose),where the viscosity is more temperature‐dependent than for strong glasses, may have abetter stabilizing efficiency [Hatley 1997]. Furthermore the addition of polymers such asdextran can increase the Tg of the formulation (T g ‐modifiers). Hence the protein is betterprotected during lyophilization and the formulation will be more stable during storage[Wang 2000].a) b)Figure 7. a) Conformational changes during freezing, drying and rehydration. N,native protein; U, unfolded protein [Prestrelski et al. 1993]. b) Schematicrepresentation of preferential exclusion/binding phenomena [Moelbert et al. 2004].2.2.5 Regulation of SFD particle´s appearanceParticles that are designated for vaccination have to be suitable for epidermal delivery(Table 1). The whole SFD process includes numerous parameters that may be modifiedto obtain particles that are suitable regarding size, shape and, most notably, density.As already mentioned, the particle size distribution can induce different freezingand drying rates that may influence the morphology of the SFD powder. Sonner et al.[2002] demonstrated that an additional applied annealing step at the beginning of 1°drying step could lead to partial collapse and shrunken trehalose particles. Annealing

INTRODUCTION TO SPRAY FREEZE‐DRYING 17means the thermal treatment of the frozen product before water‐removal is initiated.Commonly this step is supposed to promote crystallization of some excipients (mannitol,glycine) and growth of ice crystals, to enhance the primary drying rate or to improveproduct homogeneity [Searles et al. 2001]. The trehalose SFD particles produced bySonner et al. [2002] had a cherry‐pick like appearance with a low specific surface areaand high tap density. Noticeable was that mainly small particles were prone to collapseand shrinkage. Great care must be taken to avoid excessive collapse or melting. Rochelleand Lee [2007] observed that the likelihood for such collapse depends on thetemperature dependend viscosity of the amorphous structure. Highly wrinkled andcollapsed SFD‐TMD particles (trehalose/mannitol/dextran in a ratio of 3:3:4) wereobtained, traced back to the fact that dextran (10 kDa) has an anti‐plasticizing effect onthe formulations T g´. During 1° drying the T P exceeded T g´ for nearly the entire time, sothat particles could partially collapse. Another anti‐plasticizer was tested, buthydroxyethyl starch (HES, 200 kDa) gave less shrunken SFD particles with a lower tapdensity. The addition of the protein catalase to TMD (3:3:4) had no effect on themorphology of shrunken SFD particles; it just reduced their relative number in theproduced powder. T P measurements of different catalase TMD ratios during 1° dryingdisclosed that the protein reduced product resistance to cause higher drying rates andhence a lower T P . This fact and the increase in T g´ led to less shrunken particles.The presence of dextran or other anti‐plasticizing co‐solutes appears to beessential for obtaining high‐density (shrunken) SFD particles, although their effects onprotein stability are unfavourable. The question arises if the manufacturing of shrunkenprotein particles is also possible without using any additives. This problem is furtherexamined in this dissertation.

18 LIGHT‐SCATTERING3 Light‐scattering3.1 Blue sky and red sunsetsWithin a homogeneous medium, a light beam is refracted on the boundary surface whenpassing into a new medium with different optical density. This effect is called opticalrefraction. When a medium is inhomogeneous, like a macromolecular dispersion, smokeor aerosol, the incident radiation is scattered in many directions while passing throughthe medium.light of all colours from the sunatmosphere scatters blue light muchmore than red lighta)b)Figure 8. Colour of the sky: sunlight has to cross the atmosphere until sun beamsreach earth´s surface. a) noon: the sun is in zenith, the sunlight is scattered bymolecules or particles of the atmosphere. Light beams with lower wavelength aremainly scattered. b) sunset/sunrise: sunlight has to pass a longer distance throughthe atmosphere. Most of the blue light is scattered away until it reaches thecontemplator – the sky is red coloured, due to less scattered red light.The theory of light‐scattering (LS) for wavelengths between 400 – 700 nm wasdeveloped in the late 19 th century. Lord Rayleigh (John William Strutt, 3 rd Baron Rayleigh1842 – 1919) worked with diluted gases and air when he published his first LS‐theory in1871. Rayleigh was encouraged based on the discoveries by Michael Faraday (1791‐1867) and John Tyndall (1820‐1893) in 1868, who observed a blue light cone

LIGHT‐SCATTERING 19perpendicular to an irradiated white light beam through a gold sol, known as theTyndall‐effect. After J.C. Maxwell (1831‐1879) published his electromagnetic theory tothe Royal Society in 1864, Rayleigh could explain how electromagnetic waves arescattered when they interact with small gas molecules. The intensity of the scatteredlight is proportional to the sixth power of the particle radius and to the inverse value ofthe fourth power of the wavelength of the incident light. These two discoveries couldexplain some well‐known phenomena like the blue colour of the daylight sky and redsunsets (see Figure 8). Debye (1884‐1966) extended Rayleigh´s theory to large particlesand also investigated the effect of intramolecular interactions in light scattering.3.2 Static light scattering (SLS)3.2.1 Static light scattering by small particlesEinstein (1879 – 1955) indicated that LS in liquids only occurs when non‐homogeneousregions are present [Debye 1944]. When light is radiated through a solution the lightbeam will be attenuated due to different phenomena (Figure 9).a) b) c) d)Figure 9. Overview of scattering phenomena that occur when a particle isilluminated with an incident light beam: a) absorption, b) diffraction, c) refraction andd) reflection. Reflection means the abrupt change in direction of a wave front at aninterface between two dissimilar media, so that the wave front returns into themedium from which it originated.

20 LIGHT‐SCATTERINGWhen an electromagnetic field impinges on a medium it leads to polarization of theelectrons in the medium´s molecules. The strength of polarization depends strongly onthe molecules´ chemical structure and quantum mechanical properties [Jackson 1965].Rayleigh considered each scattering centre as an independent dipole oscillator with afixed phase relation between the waves scattered from different points of the sameparticle [Mallamace and Micali 1996]. The oscillating dipoles generated will re‐radiatethe light in all directions within a plane perpendicular to the line of charge acceleration(Figure 10). The intensity of the scattered light depends on the magnitude of themolecule´s polarizability. The more polarization, the greater is the induced dipole andhence the intensity of scattered light.Figure 10. Light scattering of polarized incident light [Malvern Instruments 2009].If the solute concentration is high, each molecule deflects light over this dipolemechanism. The intensity of scattered light is proportional to the concentration of themolecules in solution; twice as many molecules will scatter twice as much light. If theparticle is much smaller than the wavelength of the incident light beam (D < λ/20), theassumption is valid that the particle interacts with only one single photon at a time. Ifthe scattering particle is larger in size, the likelihood of multiple photons simultaneouslyinteracting with the same particle will increase, hence inducing multiple dipoles. Moreconstructive and destructive interferences are produced, making the scattering intensity

LIGHT‐SCATTERING 21of large particles more dependent on the location of the detector, i.e. the scatteringangle (see Figure 11) [Malvern Instruments 2009].a) b)Figure 11. Rayleigh scattering of small and large particles: interaction of a lightbeam with a) a small particle with a diameter D < λ/20 results in an angle‐independentisotropic scattering. b) When the particle´s size increases (D > λ/20), the scatteringintensity will increase upon decreasing scattering angles [Wyatt 2006a].If two or more light beams are scattered from the same particle, they can interact witheach other. If the phase relationship between the two electromagnetic waves is random,then the two light sources are called incoherent with respect to each other. If definitephase relationships between two light beams exist they are said to be “coherent”. Ifthere is no phase difference between E 1 and E 2 (equal to 0°), the two fields“constructively interfere”. The intensities are added, so the scattering intensity is twiceas large as that for two incoherent sources. When the different phases E 1 and E 2 differby 180°, then the observed intensity is reduced to 0 (Equation 2 and Equation 3; Figure12). If two electric fields completely cancel each other out, it comes to a “destructiveinterference” [Wyatt 2006a].a) b)c)Figure 12. Interacting light beams. a) incoherent scattering; b) coherent,constructive interference; c) An interference is called coherent destructiveinterference, when two electric fields cancel each other [Wyatt 2006a].Equation 2. Incoherent sum: I Total ∝ |E 1 | 2 + |E 2 | 2 = I 1 + I 2Equation 3. Coherent sum: I Total ∝ E 1 + E 2 2 ≠ I 1 + I 2

22 LIGHT‐SCATTERINGThe dominant part of the scattered radiation will have the same frequency as theincident electromagnetic field. This effect is known as elastic scattering or static LS.Quasi elastic light scattering (QELS or dynamic LS) occurs when the radiation is scatteredwithout energy loss, but the scattered light has a different frequency than the incidentlight beam. Scattering effects with a loss of energy are named inelastic scattering, as isthe case for Raman‐scattering.The basic equations that describe the scattering of light by infinite dilute dispersions orsolutions of uniform small particles are summarized below. The normalized intensity ofthe scattered light, also known as the Rayleigh ratio (R θ ) is defined as:Equation 4. R θ =(I θ ‐ I θ solvent ) * r 2I 0 * V= A *(I θ ‐ I θ solvent )I 0I 0 stands for the intensity of incident light (per unit scattering volume [cm ‐1 ]), I θ for theoverall intensity of the scattered light and I θ solvent is the intensity of scattered light frompure solvent. The distance between the detector angle and the scattering particle is r[cm]. The Rayleigh ratio depends on the scattering volume of the detector used, V. Theterm r 2 /V is combined in the instrument constant A. Knowledge of R θ at a number ofdifferent angles leads directly to the weight average molar mass and mean square size ofsolute molecules.The Rayleigh ratio is related to the solute concentration, c [g/mL], the molecular weightM of the scattering solute [g/mol], and an optical constant K.Equation 5.R θ I θI 0 r 2 = K * c * M cm As shown in Equation 4, the scattering intensity of the pure solvent is substracted fromthe scattering intensity of the solute to obtain I´θ. When non‐polarized incident light isused, K is given by Equation 6:Equation 6. K =2π 2 * n 2 20 * (∂n∂c)* 1+ cos 2 θλ 04 * NAFor vertical polarizing lasers, only one scattering angle (usually at 90°) is used, andEquation 6 is simplified [Demeester et al. 2005] to:

LIGHT‐SCATTERING 23Equation 7. K =4π 2 * n 2 20 * (∂n∂c)4,λ 0 * NAwhere n 0 is the refractive index of the solvent, ∂n/∂c is the refractive index increment[mL/g] and N A is Avogadro´s number [g/mol]. The wavelength of the incident light invacuum is expressed by λ 0 [nm], and θ stands for the angle between the incident beamand the position of the detector (scattering angle). Equation 5 shows that K and I θdepend on ∂n/∂c . The solute molecules must have a different refractive index as thesolvent to give a good contrast and sufficiently produce scattered light at a reasonablepower of the incident light [Demester et al. 2005]. Equation 6 and Equation 7 are onlyvalid for dilute solutions. If a nondiluted, polydisperse solution, with a constant ∂n/∂c –value is examined with light scattering, the Zimm‐equation is applied (Equation 9)[Zimm 1948]. The Zimm‐equation has its origin in the Rayleigh‐Gans‐Debyeapproximation (RGD), where the correlation between Rayleigh ratio and a particle´smolar mass is given (Equation 8):Equation 8.K * cR θ≈1M W * P θEquation 9.K * cR θ=1M W * P θ+ 2A 2 * c + 3A 3 * c The Rayleigh ratio R θ stands for the function of scattering angle θ and the soluteconcentration c. M W expresses the weight‐averaged solute molar mass and P θ theangular dependence of the scattered light. A 2 is the second and A 3 the third virialcoefficient [mol*mL/g 2 ]. If the protein concentration is set to a very high level, A 2 andeven possibly A 3 can be obtained from LS‐measurements. The second virial coefficientcharacterizes the protein‐solvent interactions. A 2 values can predict proteincrystallization, with a negative algebraic value in the case of attractive and a positivevalue for repulsive interactions [Wilson 2003]. If A 2 is positive, no protein crystallizationoccurs. For A 2 values that are negative, but below ca. ‐8 *10 ‐4 ‐ 0 mol*mL/g ‐2 (crystallizationslot), crystalline solid protein phase separates from the solution. A 2 values above‐8 *10 ‐4 mol*mL/g ‐2 show amorphous precipitation [George and Wilson 1994].

24 LIGHT‐SCATTERING3.2.2 The specific refractive index increment, ∂n/∂cThe deflection of a light beam is proportional to the difference in refractive indexbetween solvent and solution [Brice and Halwer 1951]. The polarizability of the solute isexpressed as the specific refractive index increment. Thus, ∂n/∂c is a solvent‐dependentvariable and can be defined by measuring the difference in refractive index between thesolution (n) and the solvent (n 0 ) using varying solute concentrations:Equation 10. ∆nc=n ‐ n 0c= cThe ∂n/∂c values depend not only on the solvent used, but also on its temperature andthe wavelength of the incident light. The wavelength of the beam in modern LSinstrumentsdiffers from the wavelength in differential refractometers (see 4.2.3, whereλ LS is 658 nm and λ RI is 890 nm). Due to the linear relationship between the specificrefractive index increment and 1/λ 20 in the Cauchy plot, the ∂n/∂c values can becalculated [Huglin 1972]. For protein solutions, values from 0.189 – 0.185 mL/g aresufficient approximations in many applications [Graupner et al. 1999; WTC 2009b]. Inthis thesis, the ∂n/∂c ‐value for proteins was set to 0.185 mL/g.3.2.3 Static light scattering by large particlesIf the scattering particles are larger than about 30 nm (equals ≈ 1/20 of the wavelength),as is the case for large protein aggregates, destructive interferences of the scatteredlight can be generated [Debye 1944]. This phenomenon is angle‐dependent; thereforemore of the scattering light beams are out of phase the larger the scattering angle is(Figure 10). This is in contrast to small particles, where scattered light does not dependon the scattering angle. P θ expresses the angular dependence of the scattered light:Equation 11. P =scattering intensity for a large particlescattered intensity without interferenceThe Zimm equation (Equation 9) can only be applied for large particles when thescattered light is measured at low scattering angles with relatively lower destructiveinterferences. To avoid this limitation, the intensity of the scattered light at larger

LIGHT‐SCATTERING 25scattering angles is extrapolated to the scattering angle zero. As Equation 12 shows, theangle‐dependent intensity of scattered light P θ is influenced by the geometric shape ofthe scattering particle, where r G stands for the radius of gyration:Equation 12. P θ = 1 ‐16π 23λ 2 * r G2 sin 2 * θ 2For solutions of large and non‐uniform particles the dependency of scattered light isexpressed by Equation 13 as a modification of the Zimm‐equation:Equation 13. K * cR θ=1M W+ (1 +16π 23λ 2 * 〈r G 2 〉 * sin 2 θ 2 + … ) + 2A 2 * c + 3A 3 * c 〈r G 2 〉 is the square of the z‐average radius of gyration. r G is defined as the mass averagedistance of each point in a molecule from the molecule’s center of gravity:Equation 14. r G = ∑ m i * r i2∑ m i,where m i stands for the mass point i of the particle and r i for the distance from thecenter of mass of the solute particle to the mass point i. As Equation 12 shows, theradius of gyration can be obtained even without knowledge of the refractive index, therefractive index increment ∂n/∂c , or the molecular weight of the particle M W . To obtainthe molar mass of larger particles from LS, the intensity of scattered light at differentconcentrations and at varying scattering angles is determined (Figure 12). Zimm [1948]developed a graphical model to evaluate a particle´s M W and r G by plotting Kc/R againstsin 2 2+ kc (0). As mentioned before, inter‐ and intramolecular effects can perturb LSmeasurements. To avoid intramolecular effects in the calculation of M W and r G due tointerferences (Figure 12), Equation 13 is extrapolated to a scattering angle of θ0°.This approximation leads to an angle‐independent correction factor, with P(θ) = 1 (seeEquation 12).Equation 15. lim θ → 0 K * cR θ=1M W+ 2A 2 * c + 3A 3 * c 2

26LIGHT‐SCATTERINGa)b)Figure 13. Zimm plot: Light scattering is measured at multiple angles and differentsolute concentrations. After extrapolating the scattering data to an angle of 0° and 0g/mL concentration, the solute´s absolute molar mass and the radius of gyration canbe determined. a) General design of a zimm plot: y‐axis in [mol/g]; the x‐axis isdimensionless [Wyatt 2006a].b) Zimm plot of hydroxyethylcellulose in water.Modified picture from [Demeester et al. 2005]: M W = 1,285,000 g/mol, A 2 = 1.4* *10 ‐4mol*mL/g 2 and r G = 81 nm.The term 3A 3 * c 2 is negligible, so the graph for the scattering angle at 0° is obtained by alinear equation. The second virial coefficient can now be calculated from the slope ofthelinear relationship,the molar mass from the ordinate intercept.Equation 16. lim K * cθ → 0R θ= 2A 2 *c +1M Wy= m * x+ tThe useof higher solute concentrationswill increase the LS‐signal and light beamsareincreasingly scattered multiple times. The extrapolation to an infinite dilution will undointermolecular effects, so light scatteringdue to a single molecule can becalculatedforc = 0. The gradientof the following equation gives the z‐average radius of gyration:Equation 17. lim c → 0K * cR θ= (1 +16π 23λ 2 * 〈r r 2 G 〉 * sin 2 θ2 + … ) +1M Wy =m* x+ t

LIGHT‐SCATTERING 273.2.4 Instrumentation and limitation of single angle LSLS can be used to calculate the molar mass down to values of 2000 g/mol. Themeasurement is absolute and nondestructive, with a very high accuracy and an error ofless than 5 % [Demeester et al. 2005]. The LS‐detector system used for this thesis canmeasure particles down to a gyration radius of 10 nm.When examining protein solutions, the removal of foreign particles can be a seriousproblem. Large particles can contaminate polymer solutions, thereby contributing to thescattered light. Furthermore, the high refractive index of dust particles and their largedimension make scattering measurements liable to large error. Filters with a pore size of20 nm are therefore used to remove impurities from the solvent.a) b)Figure 14. a) Calibration of the detector flow‐cell. θ stands for the actual scatteringand θ´ for apparent (and instrument) scattering angle. b) Varying scattering volumesinfluence scattering of the incident light. Modified pictures from Wyatt [2006a].Modern LS‐instruments allow detection at three or more scattering angles and are thussuited for the characterization of small particles. The quantities measured directly withthe LS‐detector are voltages. A common technique for detector calibration is the use ofa pure organic solvent with a known Rayleigh ratio, like toluene [Wyatt 1993]. Toluenehas one of the highest Rayleigh ratios of all solvents, so it provides a large scatteringsignal. Figure 14 shows a more serious problem. The sample flow cell of the LS‐detectorhas, due to the geometry of the glass cell, different scattering volumes for varyingscattering angles. For LS‐measurements at 90° using a 633 nm laser light the Rayleighratio for pure, filtered toluene equals 1.406 *10 ‐5 cm ‐1 [Kaye and Mc Daniels 1974]. To

28 LIGHT‐SCATTERINGassure correct prediction of particle size and mass, the geometry correction has to becalculated for each sample cell.Equation 18. R θ = R θ, toluene *n 02 * I θn toluene2 * I θ, tolueneThe calibration also includes a reflection correction. Different reflective losses of theintensity of an incident laser beam at each interface, in the sample cell and the glasssolventinterface occure. These effects vary from solvent to solvent. A set ofnormalization coefficients has to be calculated, to relate each detector voltages to the90° detector signal. The coefficients can be determined by analyzing a BSA‐sample asstandard in the same solvent and at the same flow rate as in a SEC measurement. Smallmolecules (with a radius 10 nm) scatter light isotropically, i.e. with equal intensity inall directions (see Figure 11). After normalization, the correction coefficient for the90°‐detector (Det. 2) is set to 1.0, while the other detectors (Det. 1, Det. 3) are adjustedby varying amounts to ensure uniform results [WTC 2009b].3.2.5 Dynamic light scattering (DLS)Dynamic light scattering, also known as quasi elastic light scattering (QELS) or photoncorrelation spectroscopy (PCS), is a noninvasive technique for the determination of thehydrodynamic radius r H , and diffusion coefficient, D , of the molecules. r H is the radius ofa sphere with the same diffusion coefficient or viscosity as the solute macromolecule.Due to their random translational motion (Brownian motion), the solutesgenerate temporary varying fluctuations in the scattered light (Figure 15). Two or morescattered light beams that are random in phase can lead to changing destructive orconstructive interferences, inducing shifted light scattering signals on the detector cells.Figure 16 demonstrates that small and therefore faster moving particles induce fasterbut short‐term intensity variations than larger particles do. Hence the time intervalsbetween destructive and constructive interferences of scattered light beams providemuch information about the velocity of the particle in the solvent and its size. Such timedependentfluctuations in the scattered light are measured by a fast photon counter.

LIGHT‐SCATTERING 29The fluctuations are quantified with an autocorrelation function [G(τ)]:Equation 19. G(τ) =〈 I(t) * I(t + τ) 〉〈 I(t) 〉 2G(τ) is obtained from the fluctuations of the scattered light (I) where the autocorrelationfunction [G(τ)] is the average of the scattered intensity at time t multiplied by thescattered intensity measured at a time interval (t + τ) and later divided by the squaredaverage of the scattered intensity at time t.a) b)Figure 15. a) Changing scattering light intensities due to moving particles in thesolvent. b) The autocorrelation function plots the average overall change in intensitywith time, for a given time interval τ. The autocorrelation function providesinformation on the probability that some time later (t + τ) the same intensity of lightwill be seen by the detector as it was detected before (at time t) [Wyatt 2006b].a) b)Figure 16. Typical intensity fluctuations for large (a) and small (b) particles [Wyatt2006b]. Particles with large radii diffuse more slowly through the solvent than smallones. Hence, fluctuations seen through time are slower but stronger for large particles.

30 LIGHT‐SCATTERINGThe positions of the moving particles at low τ‐values are only slightly different, resultingin high correlating intensities of the scattered light. Later on, i.e. at higher τ‐values, theintensities of scattered light no longer correlate and G(τ) decreases, due to increasingparticles distance to the place of origin.Equation 20 demonstrates that G(τ) decreases exponentially as a function of the timedelay τ when the particles are moving randomly in a solution.Equation 20. G(τ) = A + B exp (‐2Γτ)The variables A and B are instrument factors.The decay rate Γ is associated with the translational diffusion coefficient D(τ of theparticles, q stands for the magnitude of the scattering vector [Demester et al. 2005]:Equation 21. Γ = D τ * q 2Equation 22. q = 4π * nλ 0* sin θ 2The decay rate Γ is strongly dependent of the polydispersity of the sample:polydispersity of 1) the material properties, 2) in size distribution and 3) in shape.The smaller the radius of the solute particle, the larger is its Brownian motion andparticle´s diffusion coefficient D T . Equation 23, known as the Stoke‐Einstein equation,gives a relation between particle size, i.e. the hydrodynamic radius r H [nm], and thediffusion coefficient D τ [m 2 /sec.]:Equation 23. r H = k B * T6 π η D τThe hydrodynamic radius r H is inversely proportional to the particle´s translationaldiffusion and the diffusion coefficient D T . r H additionaly depends on the absolutetemperature T in Kelvin, the Boltzmann´s constant k B (1.3807 * 10 ‐23 J/K) and the solvent´sviscosity η in [g/m*s].

LIGHT‐SCATTERING 313.2.6 Instrumentation and limitation of DLSDLS requires higher solute concentrations than static LS. Particles larger than 5 μmcannot be measured by QELS, due to sedimentation effects [Demeester et al. 2005].When static and dynamic LS‐detectors are combined and r G and r H are measured,derivation of the particles shape and compactness can be made [Kok and Rudin 1981].Figure 17 shows two different types of macromolecule shape.a) b)Figure 17. Molecular conformation: two different molecular shapes: a) solid sphereand b) voluminous molecule with three salient side chains. Modified pictures fromWyatt [2006b]For compact particles, r G is smaller than the hydrodynamic radius. The ratio between thetwo different radii (r G / r H ) therefore is smaller than unity, and lower than the ratio for amore extended particle.3.3 Different separation techniquesWhen a polydisperse sample is analyzed using a separation technique, the sample isdivided into its monodisperse fractions each characterized by its size or molecularweight. Two separation techniques that are commonly used are size exclusionchromatography and asymmetrical field flow field fractionation [Philo et al. 1996].LS‐detectors are always connected to an UV‐ or RI‐detector that are commonlyused as concentration detectors. If the value of dn/dc is known, light scattering canprovide an absolute measurement of the molar mass of macromolecules withoutknowing any other sample information. Thus light scattering can provide a continuousmeasurement of the molar mass if the sample concentrations are high enough toprovide adequate signals.

32LIGHT‐SCATTERING3.3.1Size exclusion chromatography (SEC)a)b)Figure 18. a) Separationprinciple during SEC [Snyder and Kirkland 1979].b) Different movement of small and large molecules in the separation column.Not only the molar mass of the sample is therefore crucial, but also thehydrodynamicradius. After passing the separation column, different molecules are detected by variousconcentration detectors, mainly by UV‐ or RI‐detection. In order to determine unknownBiological macromoleculesand proteins were first separated with “gel filtrationchromatography”by Porath and Flodin [1959]. Theydemonstrated that columns packedwith cross‐linked polydextrangels couldbe used tosize‐separate various water solublesamples. Moore [1964] investigated synthetic macromolecules and usedthe name “gelpermeation chromatography”for his new technique. He separated synthetic polymerssoluble in organic solvents bythe use of cross‐linked polystyrene “gels”. The name “ size‐packingmaterial. The solute particles are transported by the mobile phase throughh theexclusion chromatography (SEC)” is a more accurateterm for this method.The liquid sampleis injectedonto a SEC‐column that is filled with rigid and porouscolumn, where smaller molecules penetrate deeper into pore structures than largermolecules – resulting in lower retentionn time, rT, for larger particles and higher rTforsmallerones (Figure 18).

LIGHT‐SCATTERING33samples correctly, a SEC column hasto be calibrated withstandardmolar mass.samples ofknowna)b)Figure 19. Peril of molecular weight prediction over the rT. LS measurementsinitially could lead to misinterpretation of an unfolded antigen [Philo2006].A problem of the “conventional” calibration is that elutionpositions can change, due tocolumn alteration or if themacromolecule has any tendency to interact with the columnmatrix [Philo et al. 1996] . Electrostatic interactions of positively charged proteins withnegative charges on agarose or Sephacryl gelsmay lead to retardation of theelutedproteins and thus to a wrong molar mass [LeMaire et al. 1987]. Philo et al. [1996]reported that LS could detect an unfolded antigen monomer mixed with a dimer(Figure 19) because of its shorter retention time.3.3.2 Asymmetricalflow field‐flow fractionation (AF4/aF‐FFF)For a sample that contains proteins with a largeaggregate fraction or proteins with widemolar mass ranges, different columns will berequired for SEC separation. Giddings[1966] developed a new technique called field‐flow fractionation (FFF) that is able toseparate particles from the lowest nm‐ up tobinary μm‐range with high resolutioninside a narrowchannel. Nowadays asymmetrical flow FFF probably isthe most versatileapplication [Fraunhofer and Winter 2004]. TheAF4 channel is composed of the non‐permeable upper wall made of Plexiglas, a spacer that regulates the channel height, and

34 LIGHT‐SCATTERINGthe permeable bottom wall comprising a ceramic frit covered by an ultrafiltrationmembrane (Figure 20a).a) b)Figure 20. a) Schematic presentation of an AF4 channel assembly. Modified picturefrom Fraunhofer and Winter [2004]. b) Narrow ribbon‐like separation channel of anAF4 [WTC 2009a].Within the separation channel a parabolic flow profile is generated in the carrier liquidthat is pumped towards the outflow (Figure 20b). The induced laminar Newtonianstream carries the injected particles through the channel, in which solutes move fasterat the center of the channel flow than they do closer to the boundary edges. An externalforce field is adjusted (cross flow) perpendicular to the channel flow that forces thesample components to accumulate at the channel wall and the membrane. Small, fastmoving molecules tend to diffuse to an equilibrium position closer to the middle of theseparation channel than larger molecules do. Hence larger particles are longer retainedin the channel assembly, while smaller particles are quickly transported to the flowoutlet. The prefix “asymmetrical” is used because of the constant loss of axial flow dueto the rectangular channel [Wahlund and Caldwell 2000]. AF4 can be applied for a wideseparation and concentration range. The order of elution of protein samples is thereverse to SEC, where large molecules are first detected [WTC 2009a]. The reducedshearing degradation and also less interactions between particles and stationary phase(than in e.g. SEC separation) makes AF4 a good alternative for protein analytics[Gabrielson et al. 2006].

MATERIAL AND METHODS 354 Material and methods4.1 Materials4.1.1 ProteinsBovine serum albumin (bSA), carbonic anhydrase from bovine erythrocytes (bCA), and L‐lactic dehydrogenase type II from rabbit muscle (LDH) were used as model proteins inSFD experiments. These and further proteins used in this work are listed in Table 4.Product Supplied by Product number LotEnzyme/ProteinAlbumin,Sigma A7906 017K0726, 078K0729from bovine erythrocytesCarbonic anhydrase,from bovine erythrocytesSigma C3934 116K1594, 057K1277,078K1181, 127K1564Carbonic anhydrase,Sigma C3934 058K1031from bovine erythrocytesHuman Immunglobulin G,Sigma I4506 047K7635from human serumL‐Lactic dehydrogenase,Sigma L2500 107K7405from rabbit muscle, type IIMyoglobin,Sigma M0630 028K7002from equine skeletal muscleMolecular weight marker kit Sigma MWGF200 026K6800Albumin,Sigma (A8531) 045K6036from bovine serumAlcohol dehydrogenase, Sigma (A8656) 026K6003from yeastβ‐Amylase,Sigma (A8781) 075K6070from sweet potatoeCarbonic anhydrase,Sigma (C7025) 064K6011from bovine erythrocytesCytochrome C,Sigma (C7150) 016K6016from horse heartRibonuclease A,from bovine pancreasSigma R6513 128K7002Table 4.Proteins used for this work with suppliers, product and lot number.

36 MATERIAL AND METHODS4.1.1.1 Bovine Carbonic AnhydraseCarbonic anhydrase is a globular protein, ellipsoidal in shape with well definedsecondary structures containing Zn 2+ as cofactor. This ubiquitous enzyme is found in allanimals and photosynthesizing organisms, even in some nonphotosynthetic bacteria[Cox and Phillips 2008]. CAs are divided into at least three classes, the α−CA (allmammalian CAs), β−CA and γ−CA class, based on an amino acid homology. Bovine CA ΙΙis structurally related to human CA Ι and ΙΙ (hCA Ι/ΙΙ), with a high degree of sequenceconsistency and active site architecture [Saito R. et al. 2004].a) b)Figure 21. a) Crystal structure of bCA II at 1.95 Å resolution [Saito et al. 2004]. Thestructure is dominated by a central, ten‐stranded twisted β‐sheet, surrounded by 7 α‐helices. b) Congruent overlay of X‐ray structures of hCA I, hCA II and bCA II withhistidine residues in active site highlighted [Whitesides et al. 2008].In 2007 there were more than 270 X‐ray structures of bCA, its isoforms and mutants inthe Protein Data Bank (PDB). The enzyme catalyzes the reversible hydration of carbondioxide to hydrogen carbonate and a proton, thus making CA essential forphotosynthesis and substrate oxidation.CO 2 + H 2 O HCO 3‐ + H+α‐Carbonic anhydrase has various tasks and can be found in different regions of themammal organism. In erythrocytes, CA regulates the transport of carbonic dioxide,

MATERIAL AND METHODS 37thereby producing the aqueous humor in the eyes. The hydrogen carbonate regulation isachieved due to reabsorption in the kidneys – CA also plays a role in the production ofsaliva, pancreatic juices and the secretion of gastric acid.Zink ion as cofactor is located in the active site, surrounded by three histidin residuesand one water molecule in a tetrahedral geometry. The currently accepted reactionmechanism of carbonic anhydrase involves several steps. Primarily carbon dioxideattacks on the Zn 2+ ‐bound hydroxide function to form bicarbonate. Then a watermolecule binds to Zn 2+ , near the bicarbonate ion, initiating leaving of bicarbonate. Thelast step is believed to be the rate‐limiting intermediate step in process. For Zn 2+ ‐reactivation, a proton from the newly bound water is given to the cytosolic buffer,according to Figure 22.Figure 22. Mechanism of catalysis of the hydration of CA [Lindskog and Silverman2000]. CA is an extremely efficient catalyst with one of the highest known turnoverrates. One CA molecule is able to hydrolyze 10 6 molecules of CO 2 per second.In 1982 a stable, partially folded intermediate was identified for the first time, latercharacterized as an inactive molten globule [Henkens et al. 1982]. Ohgushi and Wada

38 MATERIAL AND METHODS[1983] defined the term “molten globule”. “Molten” stands for structural fluctuationsinduced by side chains; “globule” denotes the native‐like compactness of the molecule.Fink [1995] characterized four key features for the molten globule state:1) no or only little tertiary structure,2) significant secondary structure,3) considerably more hydrophobic surface exposed to H 2 O as for native state4) and a higher compactness than the completely unfolded state, closer to thesize of the native molecule.The bovine carbonic anhydrase used in this thesis is a mixture of isoforms with amolecular weight of about 30 kDa [Sigma‐Aldrich 2009a].4.1.1.2 Bovine Serum AlbuminBSA belongs to the class of serum proteins called albumins, which make up about half ofthe proteins in plasma. They are the most stable and soluble proteins in blood and servedifferent purposes. Serum albumin (SA) has carrier functions for small molecules (e.g.cations, lipids, hormones and API´s) and plays an important role in protein‐buffer systemfor pH‐maintenance in the blood. Furthermore SA regulates the colloid‐osmotic pressureand, if required, helps as a fast availableprotein reserve for nutrition [Mutschler et al.2007]. BSA is composed of 584 amino acidsand has a molar composed mass of about 66.5kDa [PDB 2009]. The sequence has 17 disulfidebonds, resulting in 9 loops (see Figure 23).Beside SFD, where BSA was applied as modelprotein, bSA was also used in this work for LSnormalizationand RI‐calibration.Figure 23. X‐ray structure of bSAat 2.25 Å resolution [PDB 2009]

MATERIAL AND METHODS 394.1.1.3 L‐Lactic DehydrogenaseLactic dehydrogenase is a tetrameric enzyme, built up of four subunits with a totalmolecular weight of about 140 kDa. The two different subunit‐types, H‐ (heart) or M‐subunits (muscle) are similar in their molecular weight. Due to variations in their aminoacid composition, the catalytic properties of M‐ and H‐subunits are different [Dawson etal. 1964]. Five isoenzymes are known and can predominantly be found in varying organs:H 4 (LDH‐1) in the heart, MH 3 (LDH‐2) in the RES, M 2 H 2 (LDH‐3) in the lungs, M 3 H (LDH‐4)in the kidneys and M 4 (LDH‐5) especially in the liver and striated muscles.a) b)Figure 24. a) LDH‐homotetramer (LDH from rabbit muscle), monomer subunits(LDH‐M) in different colours [PDB 2009]. b) All dehydrogenases have similarNAD + ‐binding domains: 4 α‐helices and a six‐stranded parallel β‐sheet known asRossmann fold [Berg et al. 2007].LDH is an essential enzyme when oxygen is absent, catalyzing the reduction of pyruvateto lactate as part of glycolysis. In gluconeogenesis, LDH performs the reverse reaction,thus helping to deliver glucose to the blood. NADH/H+ is needed as a coenzyme, whichdelivers the hydrogen for the reaction.Equation 24. pyruvate + NADH + H + lactate + NAD +Figure 25 shows the active site of the protein and the substrate and coenzyme binding.The reaction is stereospecific for the C4‐hydrogens of NADH due to the positioning of

40 MATERIAL AND METHODSthe molecules [Cox and Phillips 2008]. In this thesis, L‐lactic dehydrogenase Type II fromrabbit muscle (LDH) was used as crystalline suspension in 3.2 M (NH 4 ) 2 SO 4 solution, pH6.0.Figure 25. Due to the molecular positioning of NADH and pyruvate it comes to astereospecific reaction occurs, catalyzed by lactic dehydrogenase [Cox and Phillips2008].4.1.2 Excipients and reagentsThe following table gives a summary of all excipients and substances used in this work,ordered based on their intended purpose (Table 5).Product Supplied by ProductnumberLotHPLC/AF4 solventsEthanol 96%, (reagent grade) Roth P075.4 428100648Phosphoric acid Merck 1.59382 K23985682724Sodium chloride Roth 9265.1 49790225Sodium azide Sigma S2002 113HO265Sodium di‐hydrogen phosphate Sigma S0751 054K0143di‐Sodium hydrogen phosphate Sigma S7907 134131832407241Sodium hydroxide, solution 1M Riedel‐de Haën 35256 80360Trizma ® Base Sigma T1503 116K5403Trizma ® HCl, (reagent grade) Sigma T3253 057K5418, 117K5432

MATERIAL AND METHODS 41Product Supplied by ProductnumberLotEnzyme AssaysAcetone Roth CP40.1 9894418β ‐ Nicotinamide adenine dinucleotide,reduced disodium salt hydrateSigma N8129 126K7025097K7016p‐Nitrophenyl acetate Sigma N8130 074K3734Potassium di‐hydrogenRoth P018.1 18887607phosphatePotassium hydroxide, solution 1 M Sigma 35113 20420Sodium pyruvate Sigma P2256 117K0662SFD/SD‐excipientsDextran, Mw 10kDaSigma D9260 045K0672from Leuconostoc mesenteroidesα‐Lactose Sigma L3625 042K0095D‐Mannitol Riedel‐de Haën 15719 70440D‐(+)‐Sucrose (min. 99.5 %) Sigma S9378 096K0026D‐(+)‐Trehalose dihydrate Sigma T9531 127K7350Tween ® 80 Fluka 93781 1324202 30108147Native PAGEAcetic acid Fluka 45731 43510Coomassie ® Serva 17525 29114Towbin ® Buffer 10x Serva 42558 P070143OthersGlycerol Solvay 32887 RBL00091 5BGuanidine hydrochloride Fluka 50940 1209959PEG 6000 Fluka 81255 1153401 31205108Urea Merck 87489 231150186236Table 5.Excipients for this work with suppliers, product and lot number.The water for buffers and reagent solutions was prepared by double distillation of deionizedwater in an all‐glass apparatus (Fi‐streem 4BD, Fisons Scientific Equipment,England) and subsequent filtration using cellulose filter membranes with 0.2 μm poresize (regenerated cellulose, Sartorius AG, Germany).

42MATERIAL AND METHODS4.2Methods4.2.1Spray Freeze‐Drying (SFD)The laboratory scale SFD‐installation was used as illustrated in Figure 26. An ultrasonicnozzle (Sono‐Tek, 60 kHz and120 kHz) was attached 5 cm above a circular aluminumbowl (diameter of 16 cm, height of 6 cm). The bowl was placedon a magnetic stirrerr andfilled with 200 mL liquid nitrogen. After a short pause for temperature equilibration,theprotein solution was transported with a peristaltic pump (Pharmacia, P1, tubing 1.0 mmid.) to the ultrasonic nozzle. The feed rate of the pump was set to 1 mL/min andtheultrasonic nozzle was operated at 5 Watt, unless otherwise stated.a)b)Figure 26. a) Spray freeze‐dryingset‐up: 1)ultrasonic nozzle, 2) aluminiumcontainer filled with LN 2 , 3) peristaltic pump, 4) magnetic stirrer, 5) power generator,6) LF ina Sarstedt tube. Modified picture from Rochelle [2005]. b) 60kHz Sono‐Tekultrasonic nozzle during atomization.The liquid feed was atomizeddirectly into the LN2‐filled 2 bowl, and the frozen dropletsgathered at the bottom. During the process theLN2 was continually stirred with amagnetic stir bar to prevent agglomeration. Typically, batches of 2‐10 mL with a totalsolids content of 15‐30 % (w/w) weree sprayed. After the entire solution has beenatomized, the bowl was topped up with liquid nitrogen and placed on a pre‐cooled shelf(‐45 °C)of the freeze‐dryer. Before starting the primary dryingstep, the cryogenic liquid

MATERIAL AND METHODS 43was allowed to evaporate. For the morphology‐alteration experiments a VirTisAdvantage Plus (SP Industries, US) freeze‐dryer was used to perform subsequent waterremoval.For regular freeze‐drying of bCA and LDH with a standard freeze‐dryingprogram, a Christ freeze‐dryer (Delta 1‐24 KD, Christ, Germany) was utilized (Figure 27).The program was set according to Table 1 and used unless otherwise stated.StepOnset[°C]Endset[°C]Hold [min.]Ramp[°C/min.]Vacuum[mbar]Segment time[min.]Total time[min.]1 ‐45 ‐45 30 equilibrate ‐‐‐‐ 30 302 ‐45 ‐20 ‐‐‐‐ 0.83 0.100 (R) 30 603 ‐20 ‐20 240 equilibrate 0.100 (H) 240 3004 ‐20 +25 ‐‐‐‐ 1.0 0.040 (R) 45 3455 +25 +25 600 equilibrate 0.040 (H) 600 945Table 1. Standard FD‐program for bCA and LDH according to Ziegler et al. [2010].(R) stands for vacuum ramp phase and (H) for vacuum hold phase.After completion of the secondary drying step, the lyophilizer was ventilated withnitrogen gas. The bowls with the SFD‐particles were transferred to a dry‐air purgedglove‐box (0.1 % relative humidity, room temperature), where the powders were filledin Sarstedt‐tubes and sealed with Parafilm. All SFD‐products were stored in a ‐80 °Crefrigerator (Heraeus Instruments, Germany).a) b)Figure 27. The used freeze‐drying units: a Christ Delta 1‐24 KD (a) and a VirTisAdvantage Plus (b).

44 MATERIAL AND METHODS4.2.2 Spray‐drying (SD)For spray drying experiments a Büchi Mini Spray Dryer B‐290 with a standard glasscyclone model was used (Büchi AG, Flawil, Switzerland). The spray dryer was warmed upfor 20 min and the drying temperature was calibrated with de‐ionized water. The inlettemperature was regulated to 130 °C, resulting in an outlet temperature of 66 °C.Sample volumes of 2 mL were atomized with an air‐flow rate of 700 L/h and a pump rateof 2.8 mL liquid feed per minute. The aspirator performance was set to 90 % of themaximum rate of 60 m³/h. The sample was collected directly in a 50 ml Sarstedt tubeand stored in a ‐80 °C refrigerator (Heraeus Instruments, Germany).a) b)Figure 28. a) The mini‐spray‐dryer Büchi B‐290. (b) Sketch of a Büchi Mini SprayDryer B‐290: 1) Two‐fluid‐nozzle, 2) air inlet filter and heating, 3) drying chamber,4) standard glass cyclone with product container, in this work replaced by a Sarstedttube, 5) air outlet filter, 6) aspirator. Photo and sketch taken from Büchi B‐290instruction manual.4.2.3 Size Exclusion Chromatography (SEC)Size exclusion chromatography was used to characterize the proteins, especially fordetermination of soluble aggregates in treated bCA, bSA and LDH‐samples. A Superdex200 (GE Healthcare, Germany) was connected to a Perkin Elmer HPLC‐System (200 LCpumps and ISS 200 autosampler, Perkin Elmer, Germany). The SEC‐column was adjusted

MATERIAL AND METHODS 45to 30 °C in a column heater (CO30, EchoTherm, Germany). The separation systemincludes an UV‐diode array detector (235C, Perkin Elmer) and a refractive index detector(RI‐101, Shodex). Connected between the other 2 detectors was a miniDawn Treos(Wyatt Technology Europe, Germany) multi‐angle light scattering detector (MALS), thatmeasures static (SLS) as well as dynamic light scattering (DLS or QELS). The collectioninterval for SLS was set to 0.5 s and 2.00 s for QELS. The miniDawn was upgraded withthe Comet option (Wyatt Technology Europe, Germany) that facilitates cleaning thedetector flow cell. Before using the LS‐detector for a new experimental series, anormalization was made with bSA in the same solvent according to the recommendedstandard operation [WTC 2009b].Figure 29.[WTC 2009b]Sketch of the used three‐detector system. Modified picture fromFor bCA and bSA‐samples, a tris‐buffer was taken as an isocratic mobile phase (50 mMTrizma HCl/Base and 150 mM NaCl, pH 7.50); for LDH samples, potassium phosphatebuffer (100 mM KH 2 PO 4 and 150 mM NaCl, pH 7.00) was used. The mobile phase wasfreshly prepared daily and filtered through a 0.1 μm PESU‐membrane filter (SartoriusStedim Biotech). As Figure 29 illustrates, particle or gas contamination of the mobilephase was prevented by using a degasser (L‐7614, Merck, Germany) and two inlinefilters(0.1 μm, Whatman), one connected downstream of the pump and the other

46 MATERIAL AND METHODSupstream of the LS‐detector. BCA and bSA were123dissolved in 50 mM Tris‐buffer (50mM TrizmaHCl/Base, pH 7.50) to final concentrations of2 mg/mL – 10 mg/mL. The samples for SEC wereimmediately examined after preparation at a flowrateof 0.5 mL/min with an injection volume of60 μL for high, and 5 μL for low protein loadings.Detection was performed either directly from theliquid feed or by gently dissolving the proteinpowder in a sample of mobile phase. All runs onthe HPLC were evaluated with the Astra Software(Version 5.3.4.14, Wyatt Technology, US) and thenexported to Origin.Figure 30. HPLC‐SEC apparatus: 1) UV‐DAD‐2) LS‐ and 3) RI‐detector4.2.4 Asymetrical field flow field fractionation (AF4)An Eclipse 2 AF4 separation system was equipped with a standard channel (25 cm), a350 μm spacer and a regenerated cellulose membrane (5 kDa cut‐off, Nadir cellulosemembrane, Wyatt Technology, Germany). All samples were analyzed with a threedetectorsystem in the following order: an UV‐MWD detector (λ = 280 nm, AgilentTechnologies, Germany), a Dawn Eos MALS detector (Wyatt Technology Europe,Germany) and a RI‐detector (λ = 690 nm, Agilent Technologies, Germany). Before usingthe AF4, a new normalization with BSA was made. All runs were evaluated with theAstra Software (Version 5.3.4.14, Wyatt Technology) and then exported to Origin. Thechannel flow was set to 0.5 mL/min and the cross flow to 5.0 mL/min. The mobile phasewas identical to the HPLC‐SEC experiments, with only 50 mM NaCl. Microbialcontamination of the mobile phase was avoided by the constant use of an UV‐B lamp(Solaris, Wyatt Technology, Germany). SFD‐bCA‐probes were dissolved in tris‐buffer (50mM) and subsequently analyzed. The results were compared to HPLC‐SECmeasurementsof the same protein solutions.

MMATERIAL AND METHODS47134a)2b)Figure 31.a) AF4 separation system: 1) Agilent HPLC‐system with an UV‐ and RIchannel.b) Different sizes of separation channels [WTCdetector, 2) Wyatt Eclipse2, 3) Wyatt DawnEOS MALS‐detector, 4) AF42009a].4.2.5 Enzymatic Activity Assay of bovine Carbonic AnhydraseTheenzymatic activity is based on the hydrolysis of an ester according to procedures byVerpoorte et al. [1967] and Sigma‐Aldrich [2009b]. The assay was slightly modified byincreasing the reaction temperatureto 25 °C. The principleof the assay is the reaction ofp‐nitrophenyl acetate (PNPA) and water to p‐nitrophenol ( PNP) and acetate catalyzed bybCA:Equation 25.p‐nitrophenyl acetate + H 2 OCA p‐nitrophenol+ acetate + H +All samples were dissolvedd in 50 mMtrizma buffer pH 7.5 (at 25 °C) and stored on ice for15 min. The buffer was composed oftrizma hydrochlorideand base dissolved indoubledistilled water.The pH was adjustedd to 7.5 at 25 °C using hydrochloric acid and sodiumhydroxide. Thefinal protein concentrations of untreated bCA‐solutions activity were inthe range of 100 ‐ 200 units/mL; they were used as references. All spray‐freeze driedsamples were dissolved in Sarstedt tubes and measured three times. A 3 mMPNPA‐up tosolution was freshly prepared daily by dissolving the PNPA in acetone and fillingthe target volume with double‐distilled water. Due to therisk of photolysis, thePNPA‐solution was stored on ice and protected from light. The increasee in absorption at

48 MATERIAL AND METHODS348 nm at 25 °C was measured by UV‐spectroscopy using a Lambda 25 UV/VISspectrometerconnected to a PC (WinLab V 5.0 software, PerkinElmer, Germany). Duringsample measurements the cuvette (10 mm made of quartz) contained 1.90 mL trizmabuffer, 1.00 mL PNPA‐solution and 0.10 mL of the enzyme solution. Blank measurementswere performed using 2.00 mL trizma buffer and 1.00 mL PNPA‐solution. After an initial3 min equilibration time in the cuvette holder, the absorption was recorded for 3 min.The residual activity of the treated sample was calculated by comparing the slopes ofthe treated sample and the untreated reference sample. Residual activity measurementwas expressed in percent deviation from the reference that was defined as 100 %activity.4.2.6 Enzymatic Activity Assay of LDHL‐lactic dehydrogenase (LDH) suspensions were dialyzed prior to use with a specialmembrane (Spectra/Por ® membrane, Spectrum Laboratories, US) with a molecularweight cut off (MWCO) of 12 – 14 kDa. The dialysis tube was placed in a hundredfoldvolume of potassium phosphate buffer on ice for 3 h, and stirred with a magnetic stirrer[Adler 1999]. The phosphate buffer was prepared by dissolving 100 mM potassium dihydrogenephosphate in double‐distilled water. The pH was adjusted to pH 7.0 (at 25 °C)using 1 M potassium hydroxide solution. After replacing the dialysis medium the dialysiswas continued for 14 h. LDH protein solution was concentrated with an Amicon Ultra‐15centrifugal filter device (Millipore Corporation, Billerica, US, MWCO 30 kDa) in acentrifuge (Minifuge RF, Heraeus Sepatech GmbH, Germany). The final concentrationwas determined at 280 nm after equilibration on 25 °C for 1 min in a 10 mm quartzcuvette. The sample was subsequently diluted to the target concentration. Theenzymatic assay is based on the LDH‐catalyzed oxidation of pyruvate and β‐NADH toL‐lactate and β‐NAD+.Equation 26. pyruvate + β‐NADHLDH L‐lactate + β‐NAD+ + H +The procedure is based on instructions by Sigma‐Aldrich [2009b] and Adler and Lee[1999]. Protein samples were dissolved with dialysis‐buffer in a Sarstedt tube and

MATERIAL AND METHODS 49diluted to adequate concentrations with potassium phosphate buffer. Referencesamples of LDH should be adjusted to 1.3 units/mL. The decrease in absorption at340 nm at 25 °C was measured by the decomposition of β‐NADH. The 10 mm quartzcuvette was placed in a Lambda 25 UV/VIS‐spectrometer connected to a computersystem (WinLab V 5.0 software, Perkin Elmer, Germany). 2.80 mL of a 0.13 mM β‐NADHsolution and 0.10 mL of a 69 mM sodium pyruvate solution were mixed together in thequartz cuvette. Both solutions were previously prepared using potassium phosphatebuffer as a solvent. The cuvette was first equilibrated in the cuvette holder at 25 °C forone minute. Then 0.10 mL of the sample solution was added and the absorption at340 nm was measured for 2 min. The slope of both sample and the referencemeasurements were compared ‐ the residual activity was calculated according to 4.2.5.4.2.7 Turbidity measurementsTurbidity values of protein samples of 2 – 5 mg/mL (protein concentration) wereobtained via UV‐spectroscopy (Perkin Elmer Lambda 25 UV/VIS) at 350nm [Banga 1995].All SFD‐powders were dissolved in freshly filtered (0.20 μm, cellulose acetate, FP 30/0.2CA‐S, Whatman) activity buffer (50 mM trizma buffer pH 7.50) and were measured in a10 mm quartz cuvette three times.4.2.8 Water content, Karl Fischer titrationResidual moisture content of SFD‐samples was determined by Karl‐Fischer titration in aMitsubishi Moisture Meter (CA‐06 moisture meter, Abimed) equipped with an oven(Mitsubishi Water Vaporizer VA‐06, Abimed). The VA‐06’s glass sample‐boat was flushedwith nitrogen for 2 min, transferred into the oven and heated up to 140 °C for 2 min.Then the boat was moved back to the sample port and left to cool down for 10 min. Apowder sample of 50 ‐ 150 mg (AT261 DeltaRange, Mettler Toledo) was filled into anequilibrated sample holder made of glass and positioned in a dry‐air purged glove box(0.1 % relative humidity). The glass‐vessel was transferred to the oven´s glass‐boat. Asubsequent reweighing step followed to determine the exact amount of powder in thesample‐vessel. All measurements were taken at a temperature of 140 °C with an initial

50 MATERIAL AND METHODSblank titration rate of ≤ 0.10 μg H2O/min and a N2 gas stream of 200 mL/min. Theresidual moisture content was calculated using the reweighed glass sample holder andexpressed in the percentage of water of the total mass. All measurements wereperformed in triplicate.4.2.9 Differential Scanning Calorimetry (DSC)Thermal transitions of SFD powder were determined using a Mettler Toledo DSC822e.Powder samples of 5 ‐ 15 mg (AT 261 DeltaRange, Mettler Toledo) were sealed in 40 μLAl pans at room temperature at 0.1 % relative humidity within a dry‐air purged glovebox. Nitrogen gas was used for purging and drying of the measuring cell (30 mL/min and150 mL/min, respectively). To determine the Tg‐temperature, each sample wasrepeatedly heated and cooled with a heating / cooling rate of 10 °C/min. The Tg values,given as inflection points of the transitions, were calculated by the Mettler STAReSoftware V 9.01. The second heating run of a sample was taken to ensure its reversibilityand to eliminate interference from enthalpic relaxation [Craig et al. 1999]. To identifythe Tg´ value, 30 μL of the sample solution was sealed in a 40 μL Al pan, cooled down to‐60 °C and then reheated at 10 °C/min to room temperature.4.2.10 Wide‐Angle‐X‐Ray‐Diffraction (WAXD)The physical state of the powders was examined using a Philips model X’pert MPD withCu Kα radiation at 40 kV/40 mA and 25 °C. Powder samples were filled into an Al sampleholder and compressed using a cover glass. All scans were measured in the range2θ = 0.5° ‐ 40° with a step size of 0.02° (time per step = 1s).4.2.11 Hg PorosimetrybSA‐SFD powders were analyzed with Hg porosimetry using a Porosimeter 2000 (CarloErba) in combination with a Pascal 140 (Porotech). Powder samples of about 50 mg wereplaced into a sample container. The maximum intrusion pressure of mercury was set to2000 bar. In contrast to gas absorption methods, mercury is not able to penetrate into

MATERIAL AND METHODS 51very fine pores. Hg porosimetry is therefore an appropriate method for the detection oflarger pores (pore diameters up to 14 μm) in a solid powder [Westermarck et al. 1998].4.2.12 BET measurementDried powders were transferred into the glass BET sample cells inside a glove‐box.Samples were degassed under vacuum for a minimum of 12 h at 30 °C. Specific surfaceareas of spray‐freeze dried powders were measured with an Area‐meter II [StroehleinInstruments, Germany] BET apparatus using liquid nitrogen in a dewar vessel andnitrogen gas. The Brunauer, Emmett and Teller (BET) equation was simplified by using anomogram according to the instrument operation manual [Stroehlein Instruments1988]. The specific surface area (S BET ) was calculated according to Equation 27.Equation 27. S BET = A * ∆h⁄mA is the nomogramm factor, ∆h the manometer pressure difference and m the samplepowder mass. The SFD‐sample weight was 300 ‐ 800 mg (AT 261 DeltaRange, MettlerToledo); all samples were measured in triplicate.4.2.13 Particle Size AnalysisParticle size determinations of the spray‐freeze dried powders were performed based ondigital images from SEM pictures. The SEM‐images were transferred to AxioVision Rel.V4.5 (Carl Zeiss Vision GmbH, Germany). The initial calibration was performed using amicrometer scale, fitted to the scale of SEM‐pictures. The particle diameters of 400‐800representative particles used in each sample were individually determined. The sizedistribution, arithmetic and geometric diameter with the corresponding deviations werecalculated using Origin software, version 8.0G.4.2.14 Native PAGENative PAGE was performed with precast gradient gels (ServaGel TG 4‐12, Serva) in amini‐vertical gel electrophoresis unit (Hoefer SE 260). The runs were controlled by abluepower 3000 power supply (Serva, Germany). Protein samples were dissolved in

52 MATERIAL AND METHODS50 mM trizma buffer pH 7.5 (at 25 °C) with additional 10 % glycerine to a concentrationof 30 ‐ 40 μg/mL. Towbin buffer (Serva, Germany) was used as running buffer (pH 8.50).10 μL of each sample was loaded onto the gel and separated based on their molecularweight. After removal from the apparatus, the gels were fixed overnight andsubsequently silver‐stained according to Blum [1987].4.2.15 Scanning Electron Microscopy (SEM)Solid samples were fixed on aluminium sample stubs (Model G301, Plano) using a selfadhesivefilm and a mellow brush. All samples were then gold sputtered for 1.5 min at20 mA/5kV (Hummer JR Technics) and examined on an Amray 1810T Scanning ElectronMicroscope at 20 kV.4.2.16 Fourier Transformation Infrared Spectroscopy (FTIR)The FT‐IR spectra of both liquid and solid bCA samples were obtained using a NicoletMagna IR 550 FT‐IR spectrometer (Thermo Fisher Scientific Inc., Germany). Theapparatus was constantly purged with dry air (1 bar). Liquid samples containing1 ‐ 2 % (w/w) of protein were examined in a thermostated, specially‐constructed CaF 2window with a fixed sample layer thickness of 5.6 μm. Solid samples were produced bycompressing KBr tablets (2 mg protein Cup to 200 mg KBr) with a ThermoSpectra TechKBr die Model 129 (Thermo Nicolet Cooperation, Madison WI, US) on a Weber hydraulicpress at 5–6 tons pressure. 138 spectra were recorded per sample at a sensitivity of4 cm ‐1 . First, the original spectrum was transformed into absorbance. For the liquidsamples, a water spectrum was first subtracted from the sample spectrum using theNicolet Omnic software. The amide I band range between 1580 ‐ 1720 cm ‐1 was isolatedfrom the complete spectrum and a baseline correction was executed. Individual peakpositions were identified from consideration of both the deconvoluted spectrum and thesecond‐derivative spectrum of this range.

RESULTS AND DISCUSSION 535 Results and Discussion5.1 Working with Light Scattering5.1.1 LS‐Calibration and NormalizationThe LS‐detector had to be calibrated before starting measurements and was recalibratedtwice per year. All PEEK tubings were disconnected from the LS‐detector and filteredsamples (Anotop, Whatman, 20 nm) were directly injected into the flow cell (batchmodus). The detector flow cell was well purged with HPLC‐grade toluene until thevariation in the signal of the 90° detector was 5 mV or less. The calibration‐template(Astra V, Wyatt Technology) was repeatedly executed and a calibration constant K LS wascalculated by averaging the results. All K LS ‐values used in this work were in the range of4.0497*10 ‐5 to 3.9188*10 ‐5 1/V*cm. After calibration with toluene, the flow cell had tobe purged with isopropanol and water before the HPLC‐system was ready for onlineexperiments. Mobile phase (tris‐HCl or KH 2 PO 4 ) was used to equilibrate the SEC systembefore a 5 mg/mL bSA sample was injected onto the connected separation column. Aftera SEC run was completed, all detector baselines had to be determined in order tosubtract the base signal from the collected data. The baseline for each detector wasdetermined by drawing a line from the smooth starting point of each detector signal to aflattened region far enough from the last measured peak.a) b)Figure 32. Sample characterization (5 mg/mL bSA, 120 μL): a) without alignment:the molecular weight is calculated inaccurately. b) After alignment. Dashed line: UVsignal,solid line: LS‐signal, dotted line: RI‐signal.

54 RESULTS AND DISCUSSIONAfter all peaks of the sample were specified in the peak definition option, the raw dataof a run was evaluated and calculated by Astra ® . The volume delays between the 3different detectors result in an inaccurate ratio of LS to RI/UV signals (Equation 30) andan incorrectly calculated molar mass (see Figure 32a) of the monodisperse bSA fractions.The LS and the RI signals were adjusted to the UV signal in the interdetector delayprocedure (alignment) that improved the calculated molar masses due to a morecongruent detector profile (Figure 32b).The next post‐processing step was a correction of peak broadening that increasesfor every detector in series due to sample dilution in the detector flow cells. As the lastinstrument in series, the RI detector has the broadest signal (see Figure 33a) and is set asreference instrument during the band broadening correction procedure.a) b)Figure 33. Interdetector band broadening: detector signals before (a) and after (b)band broadening correction. Dashed line: UV‐signal, solid line: LS‐signal, dotted line:RI‐signal.A monodisperse region is selected as a broadening range marker, such as the BSAmonomer peak. The range marker was selected from a position about halfway up theleading edge of the peak to a point just past the end of the peak where all detectorsignals have returned to the baseline [WTC 2009b]. After band broadening correctionwas completed (Figure 33b), the molar masses of the well separated, monodisperse bSAfractions appear as stair steps (see Figure 34). The values yielded for monomer, dimerand trimer are correct: 67 kDa, 134 kDa and 203 kDa, respectively. The interdetectordelay and the band broadening procedures had to be performed only once, as both

RESULTS AND DISCUSSION 55parameters remain the same unless the tubing length between the instruments or thesequence of the detectors is changed.Figure 34. Molar mass of a bSA (5 mg/mL, 120 μL) sample after normalization;dashed line: UV‐signal, solid line: LS‐signal, dotted line: RI‐signal. Plateaus in molecularweight curve represent monodisperse fractions of bSA mono‐, di‐ and trimer.Figure 35. Hydrodynamic radius of bSA sample (10 mg/mL, 120 μL): QELSmeasurement provides molecular size of protein monomer (3.61 ± 0.24 nm), dimer(4.64 ± 0.43 nm) and trimer (5.94 ± 0.79 nm).

56 RESULTS AND DISCUSSIONAs a small molecule scatters light isotropically with equal intensities in all directions, themonodisperse BSA‐monomer peak was used for the normalization procedure (see 3.2.4).The monomer peak is selected in the peak definition option and assigned to an rms (rootmean square) radius of 3.00 nm. On completion, all LS‐signals from different detectorangles are related to the 90° detector signal and the instrument calibration constant.Figure 34 and Figure 35 show the resulting molecular weight and protein sizedistributions of the bSA‐sample. The molar masses of monomer, dimer and trimer are inthe relative ratio of 1.0 : 2.0 : 3.0, as expected. The rH values show, however, a relativeratio of 1.0 : 1.3 : 1.6 from which can be deduced that bSA does not behave as a hardsphere. Converting the LS results to cumulative or differential weight fractions is analternative representation (Figure 36a and b). The normalization process was performedwith every new batch of mobile phase prepared.a) b)Figure 36. Different plotting variations: a) molar mass of BSA fractions plotted ascumulative weight fraction calculated via bSA concentration (RI‐signal) gives a rapidinsight into sample composition: 79.95 % bSA monomer, 13.51 % dimer and 3.43 %trimer. b) Hydrodynamic radius as differential weight fraction. Peak maximum at3.6 nm for the bSA monomer and 4.6 nm for the dimer.The molar mass can be best obtained by the so‐called “two‐detector approach” [Oliva etal. 2009]. The intensity of the LS‐signal is proportional to the M W , the proteinconcentration, c, the instrumental constant K LS and the square of the refractive indexincrement (Equation 28), whereas RI is proportional to dn/dc and concentration in theequation (Equation 29). The M W can therefore be determined from the ratio of thesignals (as peak height or area) from the LS‐ and RI‐detectors (Equation 30).

RESULTS AND DISCUSSION 57Equation 28. I LS = K LS (c) M W (dn/dc) 2Equation 29. I RI = K RI (c) (dn/dc)Equation 30. M W = K´ * I LS / I RII LS and I RI are the intensities measured with the light scattering and the refractive indexdetector. The constant K´ is calculated via the two instrumental constants K RI and K LS andthe known dn/dc‐value (0.185 mL/g for proteins) [Philo et al. 1996]:Equation 31. K´ =K RIK LS * (dn/dc)To obtain the M W or r G of the solute molecules during separation, a single‐line plot iscreated for each data slice in the chromatogram. Similar to the Zimm plot, Kc/R isplotted against sin 2 θ2, where the intercept of the straight line represents the molarmass, Mw, for that slice, and the initial slope stands for the r G , extrapolated from theangle‐dependent detector signals (see chapter 3.2.3).Figure 37. SEC‐chromatogram of a BSA‐sample (7.5 mg/mL, 60 μL injection volume)in 50 mM tris‐HCl: The three detector signals at time t mpm = monomer‐peak maximumyield the molar mass of 66.4 kDa; the r G is 2.9 ± 0.2 nm.

58 RESULTS AND DISCUSSIONWhen calculating molar masses this way, the second virial coefficient A 2 of the solute isassumed to be negligible [Philo et al. 1996]. Additionally, A 2 cannot be measured in thechromatography mode, since the solute concentration range covered by the peak regionis too small to show any measureable concentration dependence [WTC 2009b]. Thesecond virial coefficient values can only be obtained using a row of concentrated proteinsamples [Wanka and Peukert 2006].a) b)Figure 38. Detector signals: a) 5 mg/mL bSA, b) 3 mg/mL IgG. 60 μL were injectedwith a flow rate of 0.5 mL/min, proteins were dissolved in tris‐HCl buffer (pH 7.50),which was also the mobile phase. The LS‐signal increases with decreasing retentiontime, although the concentration of higher aggregates is relatively low.Figure 38 shows the multiple signals of the three detector‐system for twoproteins. The UV photometer is very sensitive for proteins at 280 nm [Gill and von Hippel1989] due to residual aromatic amino acids. Although RI‐detector signals for proteins at890 nm are lower, a number of different non‐protein co‐solutes can be detected, such assalts and small sugars. Figure 38b demonstrates how the LS signal increases at lowerretention times ‐ equivalent to increasing aggregate size. A decreasing concentrationleads to a lower UV‐ and RI‐signal. Large protein molecules (R G ) lead to higher intensitiesof scattered light that enhance the visualization of large aggregates (Equation 13).

RESULTS AND DISCUSSION 595.1.2 RI‐CalibrationRI‐detector calibration was executed twice per year. Two different methods were usedto obtain the RI calibration constant, i.e. the off‐line batch (without any separationcolumn) and the on‐line chromatography modes.a)b)Figure 39. a) 6 different NaCl concentrations are sufficient for UV‐independentcalibration. b) RI output data were plotted versus varying salt concentration.Off‐line RI calibration was performed with sodium chloride dissolved in de‐ionized water[WTC 2009b]. The Astra RI calibration procedure (Astra V.5.3.4.14) calculates the

60 RESULTS AND DISCUSSIONcalibration constant using the RI detector voltage (plateau) data from several batchinjections. A dilution series from 0.1 mg/mL to 1.2 mg/mL NaCl in water was directlyloaded into the RI detector, and the dn/dc value of 0.174 mL/g used for furtherevaluation (Figure 39).a)b)Figure 40. a) 6 different bSA volumes from one stock solution (10 mg/mL) wereinjected onto the SEC column; equivalent to BSA masses between 48 μg and 430 μg.b) RI output values of peak maxima from each bSA injection were plotted againstprotein concentrations.

RESULTS AND DISCUSSION 61For the on‐line technique, 6 different concentrations of bSA were separated using theSEC column in one run (Figure 40a). The bSA monomer peaks were defined, and thedn/dc values were set to 0.185 mL/g. Each monomer concentration was determinedfrom the UV‐signal (ε bSA = 660 mL/g*cm) before the astra‐DnDc5 template (AstraV.5.3.4.14) was executed to calculate the RI calibration constant (Figure 40b).

62 RESULTS AND DISCUSSION5.1.3 UV‐calibrationThe calibration constant for UV detection can be set to 1.000, provided the UV responsefactor of the instrument is accurate. A high injection volume of bSA (500 μg – 1 mg) wasintroduced onto the column via the autosampler. The molecular weight of the definedmonomer peak was calculated with the RI‐detector calibrated against NaCl. Theconcentration source parameter of the evaluated Astra file was switched to the UVsignal,and the resulting molar mass of the bSA monomer was then compared to thatcalculated based on RI‐signal. If the values were different, the UV‐detector wascalibrated using the molar mass provided by the RI‐signal. The extent of the discrepancyin molar masses gave the deviation factor which was multiplied with the old UVcalibrationconstant. The new UV‐constant now results in a monomer mass equivalent tothe RI‐calculation. After every detector was calibrated, the information gained (Table 6)was saved as an example configuration file that was re‐used for subsequently executedtemplates.LS‐detectorminiDawnTreosUV‐detector 235C RI‐detector RI‐101wavelength 658 nm wavelength 280 nm wavelength 890 nmcalibration const. 4.0497 E‐005 calibration const. 1.000 calibration const. 4.64 E‐004normalizationcoefficients0.8171.0000.790UV response factorCell length2.0001.000fluid connectionLS to RI0.1831 mL fluid connectionLS to RI0.3283 mLband broadeninginstrumental termmixing term1.000 μL50.000 μLband broadeninginstrumental termmixing term1.000 μL132.235 μLband broadeninglast detector,no bandbroadeningnecessaryTable 6.Example for essential detector key data. Bold values are fixed data.

RESULTS AND DISCUSSION 635.1.4 HPLC‐CalibrationThe SEC Superdex 200‐SEC column was calibrated via peak position measurement. Thisconventional calibration procedure was performed to compare the results of theseparation profile with the molar masses of known protein standards. Molar masses ofall standard samples were calculated over the LS‐signal showing a narrow molecularweight distribution (MWD).The rT values of the M W ‐standards used (Table 7, MWGF200, Sigma) wereplotted against the logarithm of the respective M W . The linear extrapolated fit of all datapoints is used as calibration curve (Figure 41). Note that another commonly used (andmore accurate) calibration method is based on data from a viscosity‐detector. Thehydrodynamic volume is obtained via the product of the M W and the intrinsic viscosity[η] of the protein sample [Yau et al. 1979].The molar masses of the 5 proteins (Table 7 and Figure 42) examined with LSmeasurementsare in good agreement to theoretical values [PDB 2009].Approx. M W givenby Sigma [Da]Calculated M W from PDB(Swiss‐Prot * ) [Da]M W , LS measurement [Da]Albumin 66,000 66,430 [1] 66,550Alcohol dehydrogenase 150,000 146,770 [2] 144,850β‐Amylase 200,000 223,790 [3] 218,033Carbonic anhydrase 29,000 28,980 [4] 28,760Cytochrome C 12,400 11,830 [5] 10,610Table 7. Column calibration was performed with 5 proteins of different size andweight. The RI‐signal was taken as concentration detector; dn/dc values were fixed to0.185 mL/g. All proteins were dissolved in tris‐buffer (pH 7.50) to 7.5 mg/mL; 60 μLwere injected in each run. Only the monomer peaks were employed for evaluation.________________________________________________________________________* Protein information from curated protein sequence database Swissprot ExPASy‐Uniprot(http://www.uniprot.org):[1] Serum albumin (ALB, bos taurus)[2] Alcohol dehydrogenase 1 (ADH1, saccharomyces cerevisiae) as homotetramen with Ø 36.69 kDa[3] Carbonic anhydrase 2 (CA2, bos taurus)[4] Beta‐amylase (BMY1, ipomoea batatas) as homotetramen with 55.95 kDa[5] Cytochrome c (CYCS, equus caballus)

64 RESULTS AND DISCUSSIONFigure 41. Relationship between log molecular weight and retention time,determined with a conventional calibration of the SEC with narrow‐MWD standardsinjected in triplicate (see Table 7). Log M W is plotted against the average retentiontime rT (error bars shown); the correlation coefficient R 2 is 0.990.Figure 42. Comparison of conventional calibration and molar masses obtainedfrom LS measurement. 1) β‐Amylase , 2) Alcohol dehydrogenase, 3) Albumin,4) Carbonic anhydrase, 5) Cytochrome C. All proteins were dissolved in tris‐buffer(pH 7.50) to 7.5 mg/mL; 60 μL were injected in each run.

RESULTS AND DISCUSSION 65Two more proteins were investigated using both the calibration curve and the LS datafor calculating their molar masses (Table 8). The molecular weights measured with LSare slightly closer to their literature values, but the conventional calibration alsoprovides acceptable results.Approx. M WSigma [Da]M W PDB *[Da]M W , LSmeasurement [Da]M W , conventionalcalibration [Da]L‐Lactic dehydrogenase 140.000 145.734 [6] 144.850 (‐ 0.6 %) 138.335 (± 5.1 %)Ribonuclease A 13.700 13.690 [7] 13.740 (+ 0.4 %) 16.002 (+ 11.7 %)Table 8. Comparison of LS measurement and molar masses via conventionalcalibration. Both ways to calculate M W´s provide good values, whereas LS results showless deviation from the actual values obtained from the protein sequence [PDB 2009].Both proteins were dissolved in tris‐buffer (pH 7.50) to 10 mg/mL – 60 μL wereinjected. dn/dc was set to 0.185 mL/g for molar mass calculation via RI‐signalAdditionally to molar mass, light scattering measurements also provide information on aprotein´s size as they can give values for rH and rms (Table 9). SLS does not, however,measure particles < 10 nm accurately, so the rms values are imprecise, whereas the rHvalues are accurate and only show small standard deviations.Molar mass, LS measurement [Da] rms [nm] rH [nm]Albumin 66,546 ± 97 2.9 ± 0.2 3.60 ± 0.24Alcohol dehydrogenase 144,850 ± 4,596 3.0 ± 0.9 4.70 ± 0.30β‐Amylase 218,033 ± 153 5.27 ± 0.7 5.45 ± 0.09Carbonic anhydrase 28,776 ± 50 2.9 ± 3.1 2.36 ± 0.21Cytochrome C 10,613 ± 64 4.6 ± 31.8 1.63 ± 0.22L‐Lactic dehydrogenase 144,850 ± 102 5.0 ± 2.4 4.44 ± 0.29Ribonuclease A 13,740 ± 40 n.a. 1.82 ± 0.17Table 9. Hydrodynamic and gyration radii of the analyzed proteins. All data wereobtained with a fixed dn/dc‐value of 0.185 mL/g.________________________________________________________________________* Protein information from curated protein sequence database Swissprot ExPASy‐Uniprot:[6] Lactic dehydrogenase (LDHA, Oryctolagus cuniculus) as homotetramer with 36.43 kDa[7] Ribonuclease A (RNASE1, Bos taurus)

66 RESULTS AND DISCUSSIONAn advantage of LS‐systems is their independence from modifying separation conditionslike changing buffer composition or flow‐rate, and also from column deterioration thatoccurs in the course of time.Even with new columns the pore size distribution of agarose‐ and dextran‐packedcolumns are non‐linear, thus making molar mass definition with conventional calibratedcolumns subject to some variation [Le Maire et al. 1987]. If a solute interacts with thegel‐media, LS‐detectors will still yield correct information about its molecular weight(see chapter 3.3.1).5.1.5 Determination of protein extinction coefficientsIf the primary structure of a protein is known, its theoretical molar extinction coefficientε molar can be calculated from the amino acid composition of the tyrosine, tryptophan andcysteine residues. [Gill and von Hippel 1989; Pace et al. 1995]. For the most proteinsthese values can be obtained with an accuracy of ± 5 %, resulting in an equivalentdiscrepancy for the determination of the molar mass. However, this calculation methodmay be in error since the amino acid composition does not recognize whether cysteine(cys) residues appear alone or as disulfide‐bonds (cystine). The cys residues aretherefore simply defined as half cystines [Gill and von Hippel 1989].A two‐detector system can yield the extinction coefficient of an unknown protein or itsaggregates when calculated from the RI‐signal. The specific extinction coefficient, ε specific,can be obtained from the following equation out of the ε molar and the M W :Equation 32. ε specificε molar * 1000M W[mL/g*cm]A concentrated protein sample (5 ‐ 10 mg/mL for 60 μL injection volume) is put throughthe SEC column and separated into its monodisperse fractions. The baselines and theentire peak of one fraction are defined, and the dn/dc value set to 0.185 mL/g. Afterloading the “UV extinction from RI peak” Astra‐template, the ε specific ‐values arecalculated. The ε specific of bCA determined by the Astra‐software only deviates by 1.1 %from the defined value in the protein data bank (PDB) that was calculated from the

RESULTS AND DISCUSSION 67amino acid composition of the protein (Table 10). The ε specific of ribonuclease A achievedwith the Astra‐software differs from the calculated PDB value because of its highcontent of cys residues.By using the calculated value of ε specific the molar mass of the protein fractions obtainedwith either the UV‐ or the RI‐detector are now identical (Figure 43).Figure 43. Molar masses of bCA were could be calculated with the LS‐ and oneconcentration‐detector. A ample of 7.5 mg/mL was used, 120 μL injected. Dashed line:UV‐signal, solid line: LS‐signal, dotted line: RI‐signal. Molecular weight determinedfrom the RI‐signal (grey line) and UV‐signal (small dashed black line) showed the sameresults. The only abnormality was the peak shoulder between rT = 27 – 28.5 min (seechapter 5.2.2).cys residues inaa sequence [%]ε specific PDB[mL/g*cm]ε specific LS + RI[mL/g*cm]differenceLS/PDB [%]Carbonic anhydrase (# 058K1031) 0 1,740 1,721 1.1Ribonuclease A 6.5 653 ‐ 690 668 3.0 (689*)Table 10. Calculation of ε specific assuming constant dn/dc values of 0.185 mL/g forthe two proteins. ε specific PDB: the first number assumes that all cys residues aredefined as half cystines, the second number stands for pure cystein residues. * Takenfrom Pace et al. [1995]

68 RESULTS AND DISCUSSION5.1.6 Determination of dn/dc values of proteinsThe assumption is made that every protein fraction has a refractive index increment(dn/dc) value of 0.185 mL/g. The inaccuracy of 1.1 % in determining ε specific of bCA(Table 10) may result from this approximation. The dn/dc value is influenced not only bythe protein, but also of the solvent. Ball and Ramsden [1998] demonstrated that dn/dcfrom hen egg white lysozyme (HEL) differed substantially between pure water andvarious buffer substances. Even the addition of different NaCl concentrations(10 ‐ 100 mM NaCl) slightly influenced the resulting dn/dc‐values of 0.186 – 0.188 mL/g.Exchanging buffer salts altered the refractive index increment in the range of0.153 – 0.276 mL/g [Ball and Ramsden 1998].ε specific[mL/g*cm]dn/dc monomer[mL/g]dn/dc dimer[mL/g]dn/dc trimer[mL/g]Carbonic anhydrase (# 058K1031) 1740 0.186 0.183 0.182Ribonuclease A 689 0.194 0.181 ‐‐‐‐‐‐Table 11. dn/dc‐values of different protein fractions are divergent from0.185 mL/g. For both protein samples tris‐buffer (50 mM + 150 mM NaCl, pH 7.50) wasused as mobile phase.If both RI and UV detector are used in combination with MALS, the dn/dc value of apolymer or protein with the respective solvent can be measured on‐line. During HPLCseparation the true masses of the eluted protein fractions can be determined using thedata from the UV‐detector and the known ε specific value. The (UV‐) calculated mass ofeach peak area was set as “injected mass” in the software in order to eliminate theassumption that 100 % of the protein elutes from the SEC column. The Astra template“dndc from peak” uses the RI‐/LS‐signal ratio to calculate dn/dc‐values. Table 11 showsthat the dn/dc values thereby obtained for bCA and ribonuclease A can deviate from0.185 mL/g. Even the dn/dc‐values for monomer and aggregated fractions of the sameprotein differ. The use of an assumed dn/dc of 0.185 mL/g will therefore introduceinaccuracy in the determination of molar mass from the UV‐detector or RI‐detector.

RESULTS AND DISCUSSION 695.1.7 Protein conjugate analysisThe Astra ® template “protein conjugate analysis” allows the molar mass determinationof two different fractions that elute together in SEC experiments. Two additionaldetectors that have differing sensitivities to the constituent polymers (e.g. RI‐ and UVdetector)in conjunction are needed to calculate the molecular weight out from the LSsignals[WTC 2009a]. A possible application area of that software‐implementation is theidentification of glycosylated or pegylated proteins, as well as membrane proteindetergentcomplexes [WTC‐Appl.N. #5. 1996; WTC‐Appl.N. #18. 2002]. After the peakareas are defined, the dn/dc and UV extinction values for the protein and the additive(modifier) have to be entered in the evaluation window of the template. The Astrasoftwarenow calculates the masses of the single constituents.Some of the generated SFD powders of bCA‐excipient mixtures showpolydisperse elution profiles in SEC (Figure 44). As an example a dextran (10 kDa)‐containing bCA SFD product is analyzed with the 3‐detector system in SEC. RImeasurementsclearly illustrate that the dextran peak elutes with the bCA‐monomer. Allseparated fractions are defined as one peak that consists of bCA (dn/dc: 0.185 mL/g;ε specific : 1.740 mL/g*cm) and dextran 10 kDa as modifier. Due to the fact that dextrandoes not show detectable absorption at 280 nm, the ε specific value was set to 0 mL/g*cm;the dn/dc of 0.150 mL/g was taken from the Wyatt‐dn/dc‐data base [WTC 2009a]. Theelution profile of the reconstituted SFD sample shows that dextran begins to elute withbCA at rT = 30 min and stops at 37 min. The decreased concentration of dextran at lowerrT causes a inaccuracy in molar mass definition. Here M W provided without a denotedmodifier (solid line) differs too much from the calculated protein molar mass of theAstra‐template (dashed line).

70 RESULTS AND DISCUSSIONa)b)Figure 44. a) Molar mass of SFD bCA‐TMD (3:3:4), 30 % w/w at a 1:2 proteinexcipientmixing ratio (old batch: 7.5 mg/mL, 60 μL); Molecular weight distribution ingray colour: molar mass without comprehension of any modifier (solid line); singleprotein fraction (dashed line ‐ dn/dc was set to 0.185 mL/g); molar mass of the singlemodifier (dotted line ‐ dn/dc was set to 0.150 mL/g according to the dn/dc‐data base[WTC 2009a]. Molar masses marked with (M) belong to the modifier (dextran 10 kDa),whereas (P) stands for the molecular weight of the protein fraction.b) Enlargement of the peak fractions from rT 28.5 – 37.0 min.

RESULTS AND DISCUSSION 715.1.8 A 2 ‐measurements of proteinsThe second virial coefficient, A 2 , of the solute cannot be measured in thechromatography mode by a single injection, because the concentration range coveredby the peak region is too low and narrow to show accurate concentration dependence.The Astra‐software does not use a traditional Zimm plot analysis (see 3.2.3). Instead, aglobal fitting algorithm is used to correlate all concentration and angular data[WTC 2009b]. A 2 values are dependent on both the protein and the solvent. Positivevalues indicate predominantly repulsive intermolecular interactions, whereas negativevalues reflect predominantly attractive interactions [Valente et al. 2005]. Negative A 2values are preferable for protein crystallization experiments, whereas a positive A 2 maybe advantageous for stable formulations [Hanlon et al. 2007]. Two different methodscan be used to measure A 2 : the batch and the online‐modes. For both, a dilution seriesof concentrated samples has to be prepared to give an accurate approximation of thesecond virial coefficient.5.1.8.1 Batch‐modeFor A 2 determination in the batch‐mode the samples are introduced into the instrumentwith the help of 1 mL insulin syringes with luer lock adapter. All samples of the dilutionseries are filtered through a 20 nm syringe filter (Anotop, Whatman, US) to avoid noisein the LS‐signal caused by large particles or aggregates. At least 2 mL of each proteinsolution have to be injected each time to completely fill the sample cell of the mini‐Dawn. The backpressure during injection was high, making this type of experimenttechnically difficult. All injections should lead to a stair‐steps‐like diagram (see Figure39a). Unfortunately, manual injections did not induce stable detector signals (data notshown) and A 2 ‐calculation via the Zimm plot failed. A possible error source might havebeen the high pressure fluctuations occurring during non‐uniform sample injection. Theymight be caused by the high viscosity of the concentrated sample and the narrow tubingin the flow channel of the LS‐detector. The experimental setup was therefore changedto the online‐mode, where HPLC pump and autosampler did overcome this technicalproblem.

72 RESULTS AND DISCUSSIONIn addition to complete automation and online dialysis by the separation column, theonline‐method has the added benefit of flushing the light scattering flow cell betweeneach injection [Hanlon et al. 2007].5.1.8.2 Online‐modeSeries of 5‐6 injections with varying bSA or bCA concentrations were injected onto theSEC column [WTC‐Appl.N. #25. 2007]. Only one vial with protein sample (solved in themobile phase) was used for the entire experiment where different sample volumes wereinjected. This means a minor preparative effort compared to batch‐experiments. It isessential to work with concentrated samples, because only one monodisperse region ofthe fraction is taken for A 2 calculation.The baseline was set and the peaks were defined as it is shown in Figure 45 (bSA) andFigure 47 (bCA). A small area enclosing the peak maximum was defined for each peak[Schneider 2008]. To obtain equal analyzed volumes for every protein concentration allpeaks have to be similar in their widths. The concentration of each peak area wascalculated over the RI‐signal. The dn/dc values of the sample monomers were set to0.185 ml/g (bSA) or 0.186 mL/g (bCA). The “online A2” experiment template of thesoftware was now executed on a correctly aligned and normalized SEC experiment(according to 5.1.1).Figure 46 (bSA) and Figure 48 (bCA) illustrate the results of the Zimm plots. Themonomer fraction of bCA (# 058K1031) has an average molecular weight of 28.7 kDaand shows a negative A 2 of ‐2.680 e ‐4 mol*mL/g 2 that stands for attractive interactions.The rms of bCA is calculated more accurately as with a single SEC run (according 5.1.4) to2.8 nm (± 0.3 nm). The bSA monomer weighs 66.8 kDa and has a positive A 2 of1.507 ± 0.035 e ‐4 mol*mL/g 2 – the bSA monomer features therefore repulsiveintermolecular interactions. The rms of the monomer is 2.9 (± 0.4 nm). The RI‐detectorhas been determined as concentration detector for A 2 ‐calculations for bSA and bCA,because the huge injection masses of both proteins resulted in detector signals abovethe detection limit of the UV.

RESULTS AND DISCUSSION73a)b)Figure 45. 6 sample injections inseries from one vial (bSA solved in mobile phaseto 20 mg/mL) ). a) LS‐ and RI‐signals used for evaluation, sample volumes from20 μL ‐ 120 μLL injected. b) Every monomer peak was defined in a time interval of3.100 min.Figure 46.Zimm plot of bSA concentrations according to Figure45. The dn/dc of0.185 mL/g was taken evaluation; the angle and the concentration fit degree were setto “1”. The monodisperse sample has an average molar mass of66.76± 0. .058 e +4 g/mol, an rmsof 2.9 ± 0.4nm and an A 2 of 1.507 ± 0.035 e ‐4 mol*mL/g 2 .

74 RESULTS AND DISCUSSIONa) b)Figure 47. 6 sample injections of bCA (new batch) in series from one vial (solved inmobile phase to 20 mg/mL). a) LS‐signal, sample volumes from 40 μL – 120 μL injected.b) The monodisperse regions of the monomer peaks were defined in a time interval of3.650 min.Figure 48. Zimm plot of bCA according the SEC run of Figure 47. The dn/dc of0.185 mL/g was taken for evaluation; the angle and the concentration fit degree wereset to “1”. The monodisperse sample has an average molar mass of 28.68± 0.001 e +4 g/mol, an rms of 2.8 ± 0.3 nm and an A 2 of ‐2.680 ± 0.050 e ‐4 mol*mL/g 2 .

RESULTS AND DISCUSSION 75A 2 measurements are quite challenging. The separation column has to be suitable forhigh injection masses and the concentration detectors (UV/RI) must have a wide, linearrange. Working with the SEC system used here exposed some error sorces in A 2determination. First, the maximum injection volume of the Perkin Elmer HPLC‐system islimited to 130 μL. Since samples could not be further concentrated than 20 mg/mL inorder to avoid column disintegration, concentration series are limited. Secondly, thatlow flow rates lead to pressure fluctuations caused by the HPLC pump that disturbed theLS‐signal. This problem was worse when A 2 measurements were executed withpolydisperse samples, where the SEC column was removed from the HPLC setup.Schneider [2008] emphasized that a more precise RI‐detector like the Optilab rEX (WyattTechnology, US) would be better suited for A 2 determination because of its increaseddynamic range that is 50 times higher than other RI instruments.

76 RESULTS AND DISCUSSION5.1.9 Solvent contaminationLS‐experiments with unpreserved aqueous solutions can suffer from progressivemicrobial contamination which disturbs the measurement and might lead to blockagesin the tubes of the LS‐detector flow cell (0.005´´ interior diameter). To avoid workingwith toxic excipients like sodium azide, a UV‐B radiation unit (SOLAR, Wyatt Technology,Germany) was used to preserve the mobile phase on‐line. The UV‐lamp was directlyplaced into the solvent vessel (Figure 49a) that was covered with aluminium foil. Tocheck the UV‐lamp´s efficiency, tris‐buffer (50 mM Trizma HCl/Base and 150 mM NaCl,pH 7.50) was examined for a week, either with or without UV‐B exposure. Both mobilephases were investigated for their foreign particle growth.a)b)Figure 49. a) Picture of the Wyatt UV device (SOLAR). The glass vessel was coveredwith aluminium foil when the UV‐B lamp was in use. b) UV‐lamp usage led to anincrease in the solution´s temperature equilibrating to 29 °C and a pH of about 7.60.Figure 50 demonstrates the positive effect of the UV‐B radiation on the LS‐detectornoise. Without the lamp, the impact of microbial contamination is evident ‐ after just 1day the noise of detector 1 (Treos noise template, Astra V.5.3.4.14) separates from theother detector angles, due to an increase in the scattering intensity with decreasingscattering angle. Also noticeable is the lack of particle growth in the two 0.1 μm‐filtersand the security guard column of the SEC‐system. When using buffer without anypreserving agent it must be prepared daily.

RESULTS AND DISCUSSION 77a)b)Figure 50. LS‐detector noise during baseline recording. The HPLC‐SEC setup isaccording to Figure 29. a) With UV‐preservation: after 7 days the detector noise isconstant. b) Already after 40 h the detector noise increased dramatically.To confirm that the high detector noise is caused by microbial contamination, bothmobile phases were investigated with UV/VIS spectroscopy. Figure 51 demonstrates thatstored unpreserved buffer shows a high absorbance increase for the whole UV/Vis range

78 RESULTS AND DISCUSSIONthat could, however, be reduced to the baseline level after passing through a 20 nmfilter. Tris‐buffer exposed to UV‐B radiation only showed a substantial peak maximum inabsorption at 280 nm that could not be reduced by filtration.a)b)Figure 51. Wavelength scan (Perkin Elmer Lambda 25 UV/VIS) of fresh/1‐weekstoredtris‐buffer for the whole UV/Vis range (a) and 200 – 400 nm (b). Several buffersamples were previously filtrated through 20 nm syringe filters (Anotop, Whatman),demineralized water was used as a reference.

RESULTS AND DISCUSSION 79Buffer‐loaded agar plates (nutrient‐rich standard medium, pH 7.50) that were incubatedfor 1 week confirmed the LS and UV results – only one colony was detected with theUV‐B irradiated buffer. In contrast, agar plates inoculated with unpreserved buffer showsubstantial growth of microbial colonies (Figure 52).a) b)Figure 52. Agar plates after 1 week incubation: a.) The effectiveness of UV‐Bradiation was proven, only one microbial colony was detected. b.) More than 200microbial colonies could be counted even after 1 day in the incubator.The potency of the UV‐B lamp has thus been demonstrated. The associated increase insolvent temperature is relatively small (Figure 49b) and may lead to a slight change inthe pH value of the solvent. Additionally note that all experiments were done at 30 °C ina column heater (see chapter 4.2.3), and the pH was regulated with NaOH/HCl to 7.50after reaching 30 °C.UV‐B radiation seems to be a promising alternative to toxic preservatives, but its useshould be reconsidered for UV‐sensitive solvents or buffers.

80 RESULTS AND DISCUSSION5.1.10 Evaluation of an AF4 separation methodThe channel‐ (Vc) and cross‐flow rates (Vx) were varied to ascertain the best separationconditions for bCA in AF4. A high cross‐flow of 5 mL/min is considered to be essential forsharp and monodisperse peaks (program #1, see Figure 53), but might lead to enhancedprotein aggregation. This can be observed, since increasing shear‐stress leads to thecreation of higher aggregates in the LS‐signal. Between Vx = 0 to Vx = 5 mL/min,elevation of Vc produces enhanced peak separation between t = 10 and 15 mins, butpromote aggregate formation at t = 30 min.Figure 53. Variable cross‐ and channel‐flow conditions were tested to separatesamples of 10 mg/mL bCA (LS‐signals shown): Increasing the cross‐flow led to betterpeak separation but caused formation of large aggregates. Vc = channel‐flow rate inmL/min, Vx = cross‐flow rate in mL/min.These larger aggregates have only low concentrations, so they cannot be identified withthe UV‐detector (Figure 54). Increasing the focussing time or reducing the focus‐flow didnot enhance protein separation. The injection mass yielding the best separation wasobserved at 50 μg/injection. Higher injection masses led to polydisperse peaks merginginto each other. Lower masses showed weak light scattering signals. Recoveries in AF4are high, over 90 % for unstressed bCA‐samples and over 85 % for SFD‐samples.

RESULTS AND DISCUSSION 81Figure 54. Variable cross‐ and channel‐flow conditions were tested to separate bCAsamples of 10 mg/mL (UV‐signals shown): Pure solvent showed an increase in UVsignalat rT = 20 min, similar to bCA‐samples, clearly not induced by protein. Higheraggregated fractions as seen in LS could not be visualized with the UV‐detector.Vc = channel‐flow rate in mL/min, Vx = cross‐flow rate in mL/min.Step No.Time[min]Δ Time[min]EventVx start[mL/min]Vx end[mL/min]Focus‐flow[mL/min]1 3 3 elution 5 5 ‐‐‐‐‐‐2 1 4 focusing ‐‐‐‐‐‐ ‐‐‐‐‐‐ 33 2 6 focusing + injection ‐‐‐‐‐‐ ‐‐‐‐‐‐ 34 2 8 focusing ‐‐‐‐‐‐ ‐‐‐‐‐‐ 35 10 18 elution 5 5 ‐‐‐‐‐‐6 10 28 elution 5 0.1 ‐‐‐‐‐‐7 5 33 elution 0.1 0.1 ‐‐‐‐‐‐8 5 38 stop 0 0 ‐‐‐‐‐‐#1: channel‐flow at 0.50 mL/min; Injection flow at 0.20 mL/min;Table 12.Separation conditions with AF4 to analyze small and large aggregates.

82 RESULTS AND DISCUSSION5.2 Bovine Carbonic Anhydrase (bCA)5.2.1 Different bCA batchesBCA was obtained from Sigma as a dialyzed, lyophilisated powder (C3934). Differentbatches of protein powder had similar concentrations (units/mg powder). After a longtime working with C3934, however, the composition of the Sigma bCA powder hadobviously changed. Sigma would not comment on this issue, but either the proteinextraction method changed or Sigma has several suppliers for bCA using differentchemical treatment steps for preparation of their protein powder. The new batch(058K1031) was a white crystalline‐looking powder, whereas the old batches had beenyellow/white‐coloured fluffy powders. The new batch appeared to dissolve faster, had ahigher enzymatic activity and exhibited lower concentrations of the aggregated fractionsin the LF after SEC. Additionally, SFD led to decreasing inactivation compared to previousexperiments. This will be discussed in detail in chapter 5.3.1.5.2.2 Protein characterizationSEC measurements clearly identify higher aggregates that are invisible to the RI or UVdetectors(Figure 55). Molar mass in an online‐measurement of a pre‐dialyzed7.5 mg/mL bCA‐solution exhibits mono, di‐, tri‐ and tetramer structures of about28.80 kDa, 56.35 kDa, 86.74 kDa and 121.7 kDa, respectively. Also a fragment orimpurity at 9 kDa is visible (Figure 55). According to Philo [2004] the aggregated states ofbCA may be allocated to their multimeric fractions by dividing the molar masses by themonomer weight (28.8 kDa). Figure 56 demonstrates that the molar masses ofmonomer, dimer, trimer and also tetramer are indeed in the relative ratio of 1 : 2 : 3 : 4.Hydrodynamic radii obtained from QELS could only be measured for the monomer forthe concentration of 7.5 mg/mL. An increase in the bCA concentration to 50 mg/mLresults gives an rH for the bCA monomer and dimer of 2.5 nm (± 0.2nm) and 3.4 nm(± 0.3nm), respectively (Figure 57). This result for the native monomer is equal to thedetermination of Wong and Tanford [1973]. The rH‐ratio of bCA monomer to dimer is1 : 1.4, thus bCA does not behave as a hard sphere.

RESULTS AND DISCUSSION 83Figure 55. Molar mass of bCA (old batch: 7.5 mg/mL, 120 μL); dashed line: UVsignal,solid line: LS‐signal, dotted line: RI‐signal. Plateaus in molecular weight curverepresent monodisperse fractions of bCA. Monomer: 28.80 kDa ± 0.002 kDa; dimer:56.35 kDa ± 0.012 kDa; trimer: 86.74 ± 0.053 kDa, tetramer: 121.7 kDa ± 0.017 kDa andfragment/impurity: 9.0 kDa ± 0.256 kDa.Figure 56. Molar mass of bCA (old batch: 7.5 mg/mL, 120 μL) divided by themonomer weight results in the relative masses of multiple aggregated monomers;dashed line: UV‐signal, solid line: LS‐signal, dotted line: RI‐signal. Aggregates up to thetetramer peak could be clearly assigned. Molar mass calculated from the RI‐signal.

84 RESULTS AND DISCUSSIONFigure 57. Hydrodynamic radii of a concentrated bCA sample (old batch, 50 mg/mL,120 μL): QELS measurement calculated the molecular size of the protein monomer as2.5 ± 0.2 nm and of the dimer as 3.4 ± 0.3 nm. 10 mg/mL bCA at an injection volumeled to an rH monomer of 2.24 ± 0.3 nm, but rH dimer could not be evaluated precisely (1.92 ±0.5 nm) due to the low concentration.The hydrodynamic radius of the trimer could not be detected. A comparison of Figure 57and Figure 56 shows a decreased dimer fraction of the concentrated bCA (50 mg/mL)compared to the lower protein concentrations.A shoulder is evident in Figure 55 ‐ Figure 57 between the bCA‐monomer‐ and dimerpeakthat is visible in the LS‐ and RI‐ but not in the UV‐detector. The shoulder has aslightly increased rH and molar mass compared to the monomer fraction of 2.66 ±0.20 nm (Figure 57) and 33 kDa (Figure 55), respectively. Since there is no UV‐signal at280 nm, a proteinous origin such as a partly unfolded molten globule state is unlikely.Uversky [1993] found, however, that native bCA‐II had a Stokes diameter of 5.0 nm,compared to 5.2 nm for its molten globule.Experiments with the Astra‐software “protein conjugate analysis” template show thatthe peak shoulder could also come from an additive whose peak merges together withthe bCA monomer fraction. The dn/dc‐value of the so‐called modifier was set to0.140 mL/g in order to show how a protein‐foreign substance might influence the molar

RESULTS AND DISCUSSION 85mass determination. The value 0.140 mL/g was taken as a dn/dc‐average from PEG anddextran (dn/dc PEG : 0.135 mL/g and dn/dc dextran : 0.150 mL/g [WTC 2009a]).Figure 58 demonstrates how a protein‐foreign substance may complicate the molarmass determination of bCA in SEC. It is assumed that an additive elutes together withthe whole peak composition from rT = 20 min – 34 min. Two conclusions can now bedrawn from the result. First, the influence on the average molar mass of the assumedsample modifier peak ends at 28 min elution time and thus does not change the molarmass of the bCA monomer at its peak maximum. Secondly, the modifier does not seemto elute with the dimer and trimer fraction due to the wrongly calculated protein molarmass of 52 kDa compared to 56 kDa (not shown in Figure 58).Figure 58. Molar mass of a bCA (old batch: 7.5 mg/mL, 60 μL – the bCA LF waspreviously atomized into air with a 60 kHz ultrasonic nozzle at 5 W); LS‐signal takenfrom detector 2 (90°), the molecular weight was calculated over the RI‐signal. Solidline: molar mass without comprehension of any modifier – sample dn/dc was set to0.185 mL/g. Dashed line: molar mass of the single protein fraction (0.185 mL/g),dotted line: molar mass of the single modifier (0.140 mL/g)The new batch (058K1031) of bCA shows not just a lower concentration of aggregatedprotein fraction, but also lacks the undefined peak shoulder between monomer anddimer (Figure 59). The decreased aggregate fractions may, however, come from a lowerconcentration/or no molten globule.

486 RESULTS AND DISCUSSIONFigure 59. LF of a new bCA batch (058K1031) at 10 mg/mL. 60 μL were injected.The aggregated fraction was decreased compared to older batch. No shoulderbetween monomer‐ and dimer‐peak was detectable. Monomer: molar mass of 28.68± 0.003 kDa and rH of 2.34 ± 0.2 nm; dimer: 63.42 ± 0.051 kDa; trimer and tetramermasses could not be calculated accurately due to low concentration; Molar masscalculated from the RI‐signal.Figure 60. FTIR difference spectra of amide I region of bCA from different batches.In each case, the second derivative spectrum, as well as the original Fourierdeconvoluted spectrum are shown. Both LFs had a concentration of 50 mg/mL.

RESULTS AND DISCUSSION 87FTIR measurements (Figure 60) demonstrate that the two batches of bCA also differslightly in their absorbance in the range of the amide I bands and its nearby regions at1800 cm ‐1 – 1580 cm ‐1 . Differences appear in the composition of intramolecular β–sheet(1623 cm ‐1 – 1641 cm ‐1 ), α‐helix (1648 cm ‐1 – 1657 cm ‐1 ) and intermolecular β–sheet(∼ 1616 cm ‐1 ). The intermolecular β‐sheet position are taken from Dong et al. [1995],while other positions were taken from Barth and Zscherp [2002]. The new batch of bCApowder has an increased band intensity of intramolecular β–sheet and α‐helix, with adecreased intermolecular β–sheet fraction.Examination of bCA‐samples with AF4 (Figure 61 and Figure 62) leads to similar molarmasses of monomer, dimer and trimer fraction as determined using SEC. The order ofpeak separation is reverse compared to SEC, here a small protein elutes earlier than alarger one. Notable is the missing shoulder on the monomer peak, and also the missingfragment previously seen in SEC. The time of injection is visualized by the void peak inthe LS‐signal. Measurement of protein size with dynamic light scattering cannot beobtained due to the very low injected masses using AF4 compared to SEC separation.Exceeding 50 μg bCA for one injection leads to lower monodispersity with a resultingdecrease in peak‐sharpness (see chapter 5.1.10).Figure 61. Molar mass of bCA (old batch: 10 mg/mL, 5 μL) determined using AF4separation. Dashed line: UV‐signal, solid line: LS‐signal, dotted line: RI‐signal. The largepeaks > 20 min indicate aggregates. Vc = 0.5 mL/min, Vx = 5.0 mL/min.

88RESULTS AND DISCUSSIONFigure 62. Magnificationof molar mass of bCA (old batch: 10 mg/mL, 5 μL)determined by AF4. Dashed line: UV‐signal, solid line: LS‐signal, dottedd line: RI‐signal.Plateaus in the molecular weight curve represent monodisperse fractions of bCA.Monomer: 29.72 kDa ± 0.033 kDa; dimer: 58.29 kDa ± 0.096 kDa;trimer: 86.81± 0.478kDa.5.2.3Denaturation of bCAProteins in their molten globule stateare considered to be particularly pronetoaggregation [Pittsyn 1995]. bCA was stressed using either chemical or physicalstandardized procedures to determine alterations inaggregation and rH.5.2.3.1Denaturation with guanidinium hydrochloride (GuHCl)bCA was dissolvedd in tris‐HCl (50 mM, pH 7.50) toa concentration of 10 mg/mL. Theprotein solution was mixed with 10 M GuHCl (also dissolved intris‐HCl) ina ratio of1:1,resulting in a 0.5 % (w/w) LF with 5 M denaturant. The sample was put on ice andanalyzed via SEC after various timepoints (Figure 63). Turbidity measurementsareprovided in Table 13. Directly after mixing bCA with GuHCl, sample turbidity increases.Also the SEC measurementsshow a decrease in the LS‐signal for higher aggregates atrT = 15 min. The residual activity measured after 2 min decreases to 32. 2 %. After 18 hthe bCA has unfolded and aggregatedto a large extent, and the activity further

RESULTS AND DISCUSSION 89decreased to 28 %, although the turbidity of the corresponding bCA‐solutions did notincrease. Turbidity remains more‐or‐less constant at all times after addition of GuHCl.Table 13 summarizes the SEC results and shows how monomer and dimer are lost, asboth fragments and higher aggregates increase greatly.a) b)Figure 63. bCA unfolds after a short time. The monomer peak decreases and higheraggregates are generated. The high LS‐signal at rT = 15 min cannot be seen with theUV, which indicates that large aggregates exist in a low concentration.Figure 64. Molar mass of aggregated bCA fractions after 6 h GuHCl‐exposure.Molecular weight calculated via RI‐signal.

90 RESULTS AND DISCUSSIONFigure 65.Hydrodynamic radii of bCA‐multimers after 8 h GuHCl‐exposure.Calculating the molar mass from the RI‐signal demonstrates the generation of largeaggregates up to several GDa (Figure 64) with an only marginal increase in size. Thehydrodynamic radii of the aggregates reach values up to 6 nm (Figure 65). Afterdenaturation of bCA with GuHCl it is known that the protein can renature spontaneouslyafter removal of the denaturant [Yazgan and Henkens 1972; Mc Coy and Wong 1981;Henkens et al. 1982]. However, Mc Coy and Wong also demonstrated that renaturedbCA could not regain its initial activity. Even high recoveries for refolding of bCA do notimplicate a return to an active enzyme. No renaturation studies were performed in thiswork.monomer dimer trimer fragment aggregate > turbidity λ 350nm600 kDa [5 mg/mL]t = 0 h 72.35 19.70 2.99 1.16 1.09 0.146t = 1 h 68.61 12.64 3.78 3.98 3.78 0.094t = 3 h 62.18 13.26 4.73 5.74 5.08 0.092t = 6 h 49.28 13.96 6.97 9.22 7.41 0.094t = 18 h 37.38 12.16 6.58 17.92 13.15 0.097Table 13. Alterations in peak areas and turbidity values of unfolded bCA‐samples.Peak areas were taken from the UV‐signal.

RESULTS AND DISCUSSION 915.2.3.2 Thermal denaturationSmall changes in temperature are known to be able to cause large changes in theconformation of a protein leading to denaturation or aggregation [Tsai et al. 1998]. bCAsamplesof 0.5 % (w/w) are heated to 50 °C in a water bath to investigate the effect ofincreased temperature on the enzyme. After 2 min the protein solution already starts toget turbid (Figure 66), and 10 min later large precipitating particles become visible.There is, however, an activity loss of just 5.6 % after 42 min heating at 50 °C.Temperature‐stressed bCA‐samples are also analyzed via SEC after previouscentrifugation at 10,000 rpm to remove the higher aggregates that would otherwisedamage the SEC column. The relative concentration of bCA‐monomer increases whilethe di‐ and trimer fractions decrease (Figure 67).Figure 66. Turbidity measurements for bCA at 5 mg/mL heated to 50 °C. Theprotein solution started to precipitate after 12 minutes.

92 RESULTS AND DISCUSSIONa) b)Figure 67. SEC chromatogram of 5 mg/mL bCA‐samples heated to 50 °C. Themonomer fraction increases, the longer the temperature treatment step was, and themultimer fractions (tetramer, trimer and dimer) decrease in turn. UV‐ (a) and LSsignalsshown (b).Figure 68. FTIR difference spectra of amide I region of untreated and precipitatedbCA. In each case, the second derivative spectrum is shown, as well as the originalFourier deconvoluted spectrum. Spectra taken out of KBr tablets.

RESULTS AND DISCUSSION 93Mc Coy and Wong [1981] showed that bCA precipitates after thermal treatment in trisbuffer at the protein´s melting temperature, T M , of 65 °C. This should be a result oftemperature‐induced unfolding. Although bCA precipitated after a few minutes(Figure 66), the negligible decrease in residual activity indicates refolding after a fastdilution step with the activity buffer. The temperature of 50 °C used here is 15 °C underthe protein´s T m . The SEC result in Figure 67 indicates that small bCA‐aggregatesdissociate to monomers on heat dissipation during ice treatment. Some form largeraggregated fractions that precipitate (Figure 66). Lavecchia and Zugaro [1991]discovered that the temperature at which the catalytic activity of bCA was first lost isequal to its melting temperature, which explains the very small loss seen here at 50 °C.After collecting and drying the precipitate in a desiccator, FTIR measurements wereobtained using KBr tablets (Figure 68). The decrease in peak intensity of intramolecularβ–sheet (1623 cm ‐1 – 1641 cm ‐1 ) is accompanied by an increase in intermolecularβ–sheet (∼ 1616 cm ‐1 ). This indicates alteration in bCA secondary structure with achange to aggregated forms in the precipitate.

94 RESULTS AND DISCUSSION5.3 SFD/SD of pure bCA5.3.1 Aggregation of bCA during SFDPure bCA in a concentration of 15 % (w/w) suffers substantially during SFD. Spraying,freezing and drying results in a loss in residual activity of the redispersed particles of 30% (Table 14). The concentrated enzyme solutions show suitable sprayability with onlyrudimental foaming during the spraying step owing to the protein´s surface activity[Prins et al. 1998]. BCA solutions were prepared and exposed individually to eachindividual process during SFD: 1) atomizing of the LF into air, 2) into liquid nitrogen, 3)freeze‐drying and 4) the entire spray freeze‐drying process.SDS‐PAGE and native PAGE are not suitable to clarify the changes occurring in proteinstructure (Figure 69). No clear difference can be seen between a SFD bCA powder andthe LF; both gel electrophoresis techniques actually indicate large protein aggregateseven in the untreated protein solution. SEC and residual activity tests can provide moreinformation about these individual intermediate steps during SFD.Carbonic anhydrase[mg/g]Solid content[% w/w]Residual activity± σ [%]bCA standard 0.6 0.06 100.00Liquid feed 150.0 15.0 90.96 ± 1.05LF atomized in air (5 W) 150.0 15.0 84.23 ± 0.68LF atomized in LN 2 (5 W) 150.0 15.0 87.43 ± 1.70FD process 150.0 15.0 87.20 ± 2.40SFD process 150.0 15.0 70.13 ± 5.36Table 14. Residual activity after different process steps during SFD of 15 % (w/w)bCA in tris‐HCl buffer (pH 7.50). All activity‐ and SEC‐measurements were obtainedimmediately after treatment. Experiments were performed in triplicate; 120 μL wereinjected in each SEC run, after diluting to 1 % (w/w) with activity buffer. The referencevalue for 100 % residual activity was obtained with bCA standard at a concentration of0.6 mg/mL.The preparation of a high‐concentrated solution of bCA in tris‐buffer alreadyleads to an inactivation of 9 % (Table 14). The protein was placed on ice and took nearly30 min for complete dissolving – sufficient for aggregation of bCA monomers, resultingin a decrease in monomer and an increase in dimer peaks with SEC (Figure 70). There is

RESULTS AND DISCUSSION95alsoa reduced concentration of higher aggregates in the diluted liquidfeed compared tothe undiluted standard of 10 mg/mL.Theliquid feedwas nebulized into air and collected in a beaker (process parametersequal to 4.2.1) ). A fractionof bCA aggregates after this atomization into air (Figure 71).Theprotein activity is reduced by additional 6 % (Table 14) , but no increase in multimersis detected in SEC (Figure71). Since the protein shows anadditionalloss in multimericfractions, the di‐ and trimers may have been merged to insoluble aggregates. It isknownthatt the generation of large air‐water interfaces during atomizing may lead to proteinaccumulation at the droplet surface where molecule unfolding can beinduced, showingfirstorder kinetics [Serrien et al. 1992]. When using ultrasonic energy for atomization,cavitation mayalso play an important role for bCA inactivation [Vonhoff 2010].Figure 69. Native PAGE (silver‐stained): 0.035 μg of15 % (w/w) bCA‐samples,0.015 μg of bSA (line 4), albumin egg (line 5) and 0.025 μg of myoglobin (line 6) eachband: bCA‐SFDat line 1, 7, 10; SFDbCA‐trehalose 1:2 atline 2, 8, 11; Atomized bCAat line 3, 9, 12. Probablydue to theexistencee of higher aggregates in all bCA‐samplesandinexact results ingel staining (silver‐/coomassie‐stained), no significantdifferences in protein structure can be derived from SDS‐step. A 15 % (w/w) bCA solution wasand native PAGE.A surprising result is obtained for thespray‐freezingatomized into a bowl filled with LN 2 and then thawed at room temperature. A further

96 RESULTS AND DISCUSSIONactivity loss of 3.5 % (Table 14) can be detected compared to the LF. The residual activitywas, however, 3 % higher as with single nebulizing into air. Figure 71 demonstrates thisparadoxical result. bCA evidently refolded in the cryogen on thawing, thus resulting in aMDR of 12.32 (Table 15). There is therefore a monomer increase (+ 3.8 %) and a dimerdecrease (‐ 3.8 %).Figure 70. Process‐induced stress on bCA when dissolving a high concentration ofprotein: 150 mg/mL LF diluted to 10 mg/mL (Liquid feed) and bCA of 10 mg/mL(bCA standard); 120 μL injected.Similar results for 100 mg/mL trypsinogen solutions were reported by Sonner et al[2002]. Atomization of trypsinogen into air with a 120 kHz ultrasonic nozzle led to aninactivation of 14.5 ± 2.0 % that was attributed to irreversible partial unfolding of theprotein. The previously nebulized and quench‐cooled protein solution showed nomeasurable loss in enzyme activity after thawing. Sonner et al. postulated that also hereunfolding occurred during atomization, but the change in protein structure was reversedduring quench‐freezing in LN 2 and subsequent thawing. Yu et al [2006] give anotherexplanation for this observation. The atomized droplets start to freeze as they fallthrough the vapor phase and immediately freeze upon contact with the cryogen. Thus

RESULTS AND DISCUSSION 97the surface‐active protein has a shorter time frame to adsorb to the air/liquid interfaceand suffers less inactivation.Figure 71. Process‐induced stress on bCA during atomization: treated anduntreated 150 mg/mL LF samples, subsequently diluted to 10 mg/mL; 120 μL injected.The complete SFD process using the concentrated bCA solution leads to an activitydecrease of ≈ 30 % (Table 14) and a conspicuous increase in the dimer (+ 10.5 %) andtrimer (+ 1.2 %) fraction compared to a standard solution (Figure 72 and Table 15).This indicates that the drying step is responsible for the largest part of inactivation,almost 17 %, although direct freeze‐drying of the non‐atomized LF only leads to aninactivation of 3.80 %, almost identical to that caused by spraying into LN 2 (Table 14).Addition of the separate activity decreases induced by the individual process steps doesnot result in the high protein damage measured during the entire SFD procedure. Onemay conclude that the large interfaces present during freezing and lyophilization areresponsible for the substantial differences in residual activity observed during SFD.Further experiments with surfactants may probably clarify this assumption. The proteinintegrity of freeze‐ and spray freeze‐dried rhMAb gave similar results

98 RESULTS AND DISCUSSION[Maa and Prestrelski 2000]. Atomization into air with a two‐fluid nozzle or bulk freezingwith LN 2 had only little influence an rhMAb. Dehydration during SFD, however, led to anaggregate increase of 21.6 % compared to 4.3 % for individual lyophilization. Maa et alconjectured that freezing stress generated at the freezing front (ice‐water interface) wasresponsible for protein denaturation, which ultimately led to aggregation during dryingand reconstitution.Figure 72. Process‐induced stress on bCA during FD and SFD: 150 mg/mL LF dilutedto 10 mg/mL and 10 mg/mL of bCA‐FD and bCA‐SFD powder reconstituted with tris‐HCl buffer (pH 7.50); 120 μL injected.UV‐measurment Monomer Dimer Trimer Recovery MDRfraction [%] fraction [%] fraction [%] [%]Standard of 10 mg/mL 83.0 11.0 2.4 100.0 7.55Liquid feed 83.7 12.7 1.1 91.8 6.60LF atomized in air 86.1 11.1 0.5 77.5 7.76LF atomized in LN 2 89.9 7.3 0.5 82.6 12.32FD process 77.7 16.2 2.5 85.9 4.80SFD process 70.6 21.5 3.6 72.8 3.28Table 15. Peak distribution of HPLC‐SEC experiments calculated from UV‐signal.Standard recovery was set to 100 %.

RESULTS AND DISCUSSION99SFDwith the new bCA batch (#058K1031) doesnot yield results comparable to the SFDpowder of theolder batch. As already reported in chapter 5.2.1the two proteinpowders are different in their composition. The new batch includesfewer aggregatedfractions and an increasedenzymaticactivity. Additionally,SFD leads to less inactivationof bCA for thenew batch. Figure 73shows only a slight increase inaggregates and asmall decreasee in monomer after SFD. There isno substantial changein the MDR afterSFDand only a marginal loss in residual activity of 4.5% (Table 16). The turbiditymeasurementss of redispersed SFD powders of the two different batches also indicate alarge difference in aggregated fraction.Figure 73. Process‐induced stress on new batch of bCA (#058K1031) during SFD:10 mg/mL LF and 10 mg/ /mL SFD powder reconstituted with tris‐HCl buffer (pH 7.50);60 μL injected.UV‐measurementLiquid feed old batchSFD process old batchLiquid feed newbatchSFD process newbatchMonomerfraction [%]83.770.696.395.5Dimerfraction [%]12.721.51.82.2MDR6.603.2853.2343.02Residual activity± σ [%]90.96 ± 1.0570.13 ± 5.3698.79 ± 0.8993.91 ± 0.76Turbidity[2 mg/mL]0.0460.3170.0210.049Table 16. Comparison of SFD bCA of the old and the new batch: Peak distributionof SEC experiments (UV‐signals), residual activity and turbidity measurement.

100 RESULTS AND DISCUSSIONTo avoid inaccurate quantitation of protein by SEC due to bCA aggregation onto theseparation column, SFD samples are additionally analyzed with AF4 as an orthogonalmethod. Compared with SEC measurements, the results for stressed bCA samples in AF4show an increase in aggregated bCA fractions, the recoveries are higher than with SEC(Figure 74 + Table 17). Discrepancies between the two separation methods arediscussed in chapter 5.3.5.2.Figure 74. AF4 measurements: Process induced stress on bCA during FD and SFD.All samples prepared at 10 mg/mL ‐ 5 μL injected. Molar mass taken from theUV‐signal (bCA‐pure, SFD 15 % w/w).UV‐measurement Monomer Dimer Trimer Recovery MDRfraction [%] fraction [%] fraction [%] [%]LF 10 mg/mL 73.4 22.1 2.7 96.4 3.32FD process 65.6 26.8 4.6 97.8 2.44SFD process 61.5 32.0 4.8 95.0 1.92LF 10 mg/mL (new batch) 97.7 2.3 0.3 99.7 42.48SFD process (new batch) 96.6 2.4 0.8 99.2 40.25Table 17.Peak distribution of AF4 experiments calculated from the UV‐signal.

RESULTS AND DISCUSSION101a)b)200 μm20 μmc)d)100 μm20 μme)f)50 μm2000 μmFigure 75.SEM of SFDpure bCA:a) 100x, b)1000x; SFD particles were ground in aLN 2 ‐filled mortar to reveal particle´sinner structure: c) 400x, d) 1500x; The poresize ofan FD productis larger: FD‐producthas poreswith a diameter of about 40 – 60 μmcompared to 5 μm for SFD‐particles. e)500x; f) 200x, bCA FD according to [Gieseler2004, standardd cycle P‐29].SFDparticles have a diameter between 20 and 60 μm (Figure 76), witha nearly perfectlyspherical shape and an even but porous surface. After crushing LN 2 ‐frozen powderparticles in a mortar, the inner structure of small channels is clearly visible (Figure 75candd).

102 RESULTS AND DISCUSSIONIn contrast to the round SFD‐particles, FD‐bCA consists of unformed, flaked pieces.Freeze‐drying of bCA with a more conservative program in a glass vial according toGieseler [2004, standard cycle P‐29] results in an residual activity of 86.5 %. Thedifference in protein residual activity of the fast (SFD) and the slower but moremoderate (FD) freeze drying cycle is therefore negligible. The slower freezing step duringFD leads, however, to an increase in pore size of the dried cake compared to the highlyporous SFD particles (Figure 75d + f).Figure 76. SFD particle size: The cumulative particle fraction indicates a d 50 ‐value ofapproximately 31.5 μm (old batch) and 28.3 μm (new batch). 540 (old batch) and 523particles (new batch) were selected for optical computation. The produced powderparticles lie in the size‐range suitable for EPI (20 – 75 μm, see chapter 2.1).Figure 77 shows the DSC thermograms of untreated and SFD carbonic anhydrase (oldbatch). Untreated protein powder has a detectable T g at 124.4 °C. Melting ordecomposition at T m beyond 142 °C is also detectable, similar to the SFD powder. Sprayfreeze‐dried bCA also shows, however, a weak T g with an inflection point of 127.9 °C.

RESULTS AND DISCUSSION 103a)Figure 77.b)DSC of untreated (a) and SFD bCA, old batch (b) at 1 st scan.5.3.2 Protein aggregation of bCA during SDSpray‐drying of bCA at a concentration of 15% w/w leads to an activity loss of 27 %(Table 18), similar to SFD where the residual activity dropped to 70.6 %. SEC experimentsare consistent with the measured activity values (Figure 78). Compared to the bCAstandard, bCA looses 7 % of its enzymatic activity after atomization with a 2‐fluid nozzleinto a beaker at 2 bar. Due to the high‐pressure liquid stream, some small droplets were,however, blown out of the beaker and lost. This leads to a weak increase in bCA in theliquid feed, demonstrated in an unexpectedly high recovery of 107 % (Table 19). Thehigh shear forces evidently generate large aggregates; the atomized bCA showsincreased turbidity ‐ three‐times higher than the LF. Additionally, the dimer fraction ofbCA rises up to 17.6 %, whereas the MDR decreases from 6.61 to 4.45 (Table 19).Carbonicanhydrase[mg/g]Solidcontent[% w/w]Residual activity± σ [%]bCA standard 0.6 0.06 100.00Liquid feed 150.0 15.0 98.96 ± 0.64LF atomized into air (2‐fluid nozzle, 2 bar) 150.0 15.0 93.05 ± 0.92SD process 150.0 15.0 72.12 ± 0.95Table 18. Process steps during SD of 15 % (w/w) bCA in tris‐HCl buffer (pH 7.50).All activity‐ and SEC‐measurements were obtained immediately after respectivetreatment. Experiments were performed in triplicate; 120 μL were injected for eachSEC run after previously diluting to 7.5 mg/mL with activity buffer. The reference valuefor 100 % residual activity was obtained with bCA at a concentration of 0.6 mg/mL.

104RESULTS AND DISCUSSIONFigure 78. Impact of isolated process steps during SD: 150 mg/mL LF were dilutedto 10 mg/mL, and10 mg/mLof FD and SFD bCA powder were reconstituted with tris‐HCl buffer (pH 7.50); 120 μL for each sample were injected.UV‐measurementsMonomerfractionDimerfractionTrimerfractionMDRRecovery[%]Turbidity[2 mg/ /mL][%][%] [%]Standard of 10 mg/mL Liquid feedLF atomized into air(2‐fluid nozzle, 2 bar)SD process83.0 83.7 78.4 72.2 11.0 12.7 17.622.01 2.4 1.1 1.1 2.3 7.55 6.61 4.45 3.28100.098.8107.177.50.0560.0950.2510.363Table 19. Process stepss during SD of bCA: Peak distribution of HPLC‐SECexperiments calculated from UV‐signal, residual activity assay and turbiditymeasurement.The subsequent water removal step lowers the residual activity to 72 %, and theturbidity value of reconstituted SD‐powder is further increasedto 0.36. The morphologyof SD‐particles is different to the approximately ten‐times larger SFD‐particles. Theirsizerange between 1 – 8 μm (Figure 82) is suitable for inhalation, but not for EPI (seechapter2.1). Mostof the smooth particles are spherical (Figure 79a), but few of themare donut‐shapedwith a holein the middle (Figure 79b + c). This phenomenon has been

RESULTS AND DISCUSSION 105observed for spray‐dried proteins by other authors. Maa et al [1997] indicated that it is aresult of high drying rates. A decreasing drying rate leads to a change in particle shapefrom irregular (e.g. "donut") to spherical. AF4 measurements of SD bCA show a peakdistribution of 70.12 % for the monomer‐, 26.81 % for the dimer‐ and 2.74 % for thetrimer‐peaks (MDR of 2.62, Figure 80).a) b)20 μm 10 μmc)Figure 79.5 μmSEM of SD pure bCA particles: a) 1000x, b) and c) at 3000x magnification;Table 20. Peak distribution of HPLC‐SEC experiments with modified mobile phase(50 mM tris‐HCl + 150 mM NaCl + 25 % (v/v) glycerol). Values calculated from UVsignal.UVmeasurmentMonomer Dimer Trimer Recovery [%] MDRfraction [%] fraction [%] fraction [%]Liquid feed 81.7 (83.7) 12.6 (12.7) 1.8 (1.7) 98.8 (100) 6.49 (6.59)FD process 78.1 (77.7) 17.1 (16.2) 1.9 (2.5) 96.6 (85.9) 4.58 (4.80)SFD process 70.0 (70.6) 23.7 (21.5) 2.8 (3.6) 94.1 (72.8) 2.83 (3.28)SD process 70.3 (72.2) 22.9 (21.5) 3.0 (3.6) 93.6 (77.5) 3.28 (3.08)

106 RESULTS AND DISCUSSIONFigure 80. AF4 measurements of SD bCA at 10 mg/mL ‐ 5 μL injected. Solid line: LSsignal,dashed line: UV‐signal and dotted line: RI‐signal. Molar mass taken from theUV‐signal.Figure 81. Process induced stress on bCA. Samples of 10 mg/mL reconstituted withtris‐HCl buffer (50 mM + 150 mM NaCl, pH 7.50), 120 μL injected. 25 % (v/v) glycerolwas added to the mobile phase to increase the recovery (according chapter 5.3.5.2).Molar mass taken from the RI‐signal of the SFD sample.

RESULTS AND DISCUSSION 107A satisfactory recovery of the highly stressed samples in SEC could be obtained by theaddition of 25 % (v/v) glycerol to the mobile phase (Figure 81 and Table 20). Theaddition of 5 – 20 % of a nondenaturating solvent such as glycerol can eliminatehydrophobic interactions between unfolded bCA and the column matrix [Sigma‐Aldrich1997] that could reduce recovery.Figure 82. SFD particle size: The cumulative particle fraction indicates a d 50 ‐value ofapproximately 2.5 μm. 467 particles were selected for optical computation.The spray‐dried protein powder does not show a T g in its DSC thermogram in Figure 83.Only melting or decomposition beyond T m = 145 °C is detectable.a) b)Figure 83.DSC of untreated (a) and SD bCA (b) at 1 st scan.

108 RESULTS AND DISCUSSIONDuring all three drying methods, SFD, FD or SD, similar alteration in protein activity isfound. Compared to the results of chapter 5.3.1, no big difference is detectable betweenmethods that involve an atomization step (SFD and SD) and that which don´t, i.e.conventional FD. Due to the fact that the protein´s native conformation is partlystabilized by hydrogen bonds of surrounding water molecules [Izutsu et al. 1993], itseems evident that the removal of water from the protein’s immediate vicinity isresponsible for the decrease in enzymatic integrity during the manufacturing process ofdry powders.5.3.3 The atomizing step of SFD5.3.3.1 Use of different ultrasonic nozzlesThe effect of atomization into air was further investigated with both pure bCA solutionsand with bCA‐trehalose mixtures containing 2 % (w/w) protein. A 60 kHz and 120 kHzultrasonic nozzle was used for nebulization, the energy was set to 5 W and the pumprate was 1 mL/min. Furthermore, the effect of a surfactant was investigated which wasadded to the LF for SFD‐experiments. For this purpose, 0.005 and 0.01 % (w/w) ofTween 80 ® was employed to stabilize bCA.Table 21 shows that the integrity of non‐stabilized bCA suffers during atomization. Asubstantial decrease in residual activity is seen that is combined with an increase inturbidity. The higher frequency of the 120 kHz ultrasonic nozzle leads to a largerdecrease in protein activity than that observed with the 60 kHz nozzle. The mediandroplet size is known to be inversely proportional to frequency to the 2/3 power[Sono‐Tek 2010], so the surface area of the droplets will be lower with 60 kHz. Dropletsize measurements were not executed, but an increasing damage to bCA with higherfrequencies can be expected. Vonhoff [2010] showed that TMD‐SFD particles (3:3:4,200 mg/mL) had higher diameters if the 60 kHz ultrasonic nozzle was used foratomization instead of the 120 kHz‐model. A reduced specific surface area of atomizeddroplets at lower frequencies is equivalent to a decrease in stress affecting the protein.

RESULTS AND DISCUSSION 109Therefore it is no surprise that atomization with the 120 kHz nozzle leads to morebCA‐inactivation and higher turbidity values than atomization using 60 kHz (Table 21).The complete SFD‐process with the high nozzle frequency leads to a decrease in residualactivity of 4 %, an increase in turbidity, and also a larger dimer peak area of 0.6 %compared to the powder that was produced with a 60 kHz‐nozzle (Table 21 and Figure84). The porous SFD particles generated with the 60‐ and the 120 kHz nozzles are similarin their spherical shape (Figure 86a ‐ d). The d 50 ‐value for the 120 kHz‐powder lies at21.5 μm and is therefore 8 μm smaller than the SFD‐powder generated with the 60 kHznozzle(Figure 87).Ultrasonic nozzlefrequency [kHz]Residual activity [%]± SD [%]Turbidity[2mg/mL]bCA, LF ‐‐‐‐‐‐‐‐ 98.96 ± 0.64 0.048bCA atomized into air 60 90.63 ± 4.29 0.088120 88.84 ± 4.64 0.107bCA‐trehalose 1:1, LF ‐‐‐‐‐‐‐‐ 106.69 ± 0.07 0.038bCA‐trehalose 1:1, atomized into air 60 95.29 ± 1.06 0.077120 92.05 ± 0.79 0.079bCA‐trehalose 1:2, LF ‐‐‐‐‐‐‐‐ 98.22 ± 0.78 0.044bCA‐trehalose 1:2, atomized into air 60 98.49 ± 1.73 0.075120 97.06 ± 0.99 0.071bCA‐SFD 15 % (w/w) (old batch) 60 70.13 ± 5.36 0.317bCA‐SFD 15 % (w/w) (old batch) 120 66.15 ± 4.06 0.340bCA‐SFD + 0.005 % (w/w) Tween 80 ® ,15 % (w/w) (old batch)bCA‐SFD + 0.010 % (w/w) Tween 80 ® ,15 % (w/w) (old batch)60 70.20 ± 2.93 0.18960 73.89 ± 2.38 0.128Table 21. Stabilizing effect of trehalose and Tween 80 ® : liquid bCA‐samples of20 mg/mL with and without trehalose addition of trehalose were examined on theirresidual activity and turbidity after proper dilution. Analyzed SFD‐samples of15 % (w/w) were previously redispersed in tris‐buffer (pH 7.50).The addition of trehalose to the LF has a positive effect on the protein activity. Themonomer fraction increases by 2.8 %, whereas the dimer is 2.5 % reduced compared tounstabilized LF (Table 21). Trehalose is expected to facilitate bCA‐refolding, indicated bya large increase in the MDR in SEC (Table 22) and slightly decreased turbidity values

110 RESULTS AND DISCUSSION(Table 21). CA‐trehalose mixtures in a weight ratio of 1:2 are suitable for stabilizationand cause lowering of turbidity during atomization. This LF shows no difference whetherit is atomized with 60 or 120 kHz (Table 21), the residual activity of bCA almost remainsunchanged.Trehalose is known to be a superior stabilizer of biological materials due to its high rateof preferential hydration of proteins. Preferential exclusion stabilizes the folded (native)structure of proteins in solution [Timasheff 2002]. Former observations of Timasheff[Lin and Timasheff 1996] regarding RNase A showed that the addition of trehaloseincreases the surface tension of the medium, which also led to preferential exclusionfrom the protein.Figure 84. SFD samples produced with a 60‐ and a 120 kHz‐ultrasonic nozzle. SFDpowder was reconstituted to 7.5 mg/mL with tris‐HCl buffer (pH 7.50); 60 μL wereinjected each run. UV‐ and LS‐signals are shown, molar mass calculated from the RIsignal.

RESULTS AND DISCUSSION 111UV‐measurementsMonomerfraction [%]Dimerfraction [%]Trimerfraction [%]MDRSECbCA, LF 74.98 16.79 3.78 4.47bCA atomized into air 73.57 16.54 3.52 4.45bCA‐trehalose 1:2, LF 77.81 14.31 3.33 5.44bCA‐trehalose 1:2, atomized into air 77.89 14.48 3.37 5.38SFD bCA 15 % (w/w) with 60 kHz,old batchSFD bCA 15 % (w/w) with 120 kHz,old batch70.61 21.55 3.38 3.2869.66 22.12 3.95 3.15SFD bCA + 0.005 % (w/w) Tween 80 ® 69.47 21.71 3.88 3.20SFD bCA + 0.010 % (w/w) Tween 80 ® 70.40 21.29 3.82 3.31Table 22. SEC‐results of the stabilizing effect of trehalose and Tween 80 ® :Analyzed SFD‐samples of 15 % (w/w) and liquid samples were redispersed/diluted intris‐buffer (pH 7.50) to 7.5 mg/mL. 60 μL injection volume.Figure 85. SFD particle size distribution produced with a 60 kHz‐ and a 120 kHzultrasonicnozzle: The cumulative particle fraction of 120 kHz‐powder indicates ad 50 ‐value of 21.5μm, whereas the 60 kHz‐powder shows a d 50 of 31.5 μm. 540 (60 kHznozzle)and 432 (120 kHz‐nozzle) particles were applied for optical computation.

112 RESULTS AND DISCUSSIONa) b)500 μm 50 μmc) d)500 μm50 μme) f)500 μmFigure 86. SEM of SFD bCA after atomization with a 60 or 120 kHz ultrasonic nozzle.Pictures at 50 fold (a, b, c, e) and 500 fold magnification (d, f). a, b) SFD pure bCA,atomized with 60 kHz; c, d) SFD pure bCA, atomized with 120 kHz; e, f) SFD bCA +0.01 % (w/w) Tween 80 ® .50 μmTween 80 ® added to the LF in concentrations of 0.005 or 0.01 % w/w shows only littleimpact on protein stability after SFD. The protein solution was easier to handle, due toless bubble formation during atomization. Addition of surfactant at a concentration of0.01 % (w/w) increases the residual activity by 3.7 % (Table 21); however, no change inthe MDR can be observed in SEC‐measurements (Table 22 and Figure 87). Nevertheless,

RESULTS AND DISCUSSION 113the addition of surfactant has a positive effect on protein aggregation. The reducedturbidity values (Table 21) correlate well with the decreasing peak area of largeaggregated bCA‐multimers that can be seen at rT = 15 min in Figure 87. Due to the factthat both the enzyme and the tenside competitively inhibit each other from reaching thedroplet surface during atomization, increasing the surfactant concentration enhancesthe residual activity [Maa et al. 1998].SFD particles containing Tween 80® are smaller compared to pure bCA‐SFD powder,with a narrower particle size distribution (Figure 88). This observation can be explainedby the decrease in surface tension which leads to smaller droplet sizes duringatomization. Adler and Lee [1999] showed that polysorbat 80 inhibits LDH accumulationin the surface of spray‐dried particles. ESCA measurements demonstrated that thesurfactant displaced the protein from the particle surface. The results correlated with areduction in LDH inactivation during SD.Figure 87. SFD product with and without addition of surfactant (Tween 80 ® ). SFDpowder was reconstituted to 7.5 mg/mL with tris‐HCl buffer (pH 7.50); 60 μL injected.UV‐ and LS‐signals shown, molar mass calculated from the RI‐signal.

114 RESULTS AND DISCUSSIONFigure 88. SFD particle size produced with a 60 kHz‐ultrasonic nozzle withTween 80 ® : The cumulative particle fraction (CPF) of the powders indicates d 50 ‐valuesof 31.5 μm for pure bCA‐SFD (540 particles), 25.8 μm for 0.005 % Tween 80 ®(249 particles) and 21 μm for 0.01 % Tween 80 ® (368 particles).

RESULTS AND DISCUSSION 1155.3.3.2 Atomization into different mediabCA at 15 % w/w sprayed into air shows a decrease in residual activity to 84.2 %,whereas atomization into LN 2 results in a regain in enzymatic activity of 3 % (see chapter5.3.1). The atomization process with an ultrasonic nozzle (60 kHz) into air and LN 2 wastherefore further investigated with bCA at 1.5 % (w/w) with varying ultrasonic energy.All samples were atomized into an aluminium bowl and the frozen bCA wassubsequently thawed in a glove‐box and immediately measured by SEC without previousdilution. Residual activity and turbidity measurements were performed after dilution toproper concentrations.Table 23 and the SEC chromatograms (Figure 89) confirm the results of the concentratedsamples (15 % w/w). Quench cooling and subsequent thawing does not lead to bCAinactivation, in fact the residual activity of LF at 1.5 % (w/w) shows a light increase afterthe treatment with LN 2 . bCA sprayed into air shows a lower MDR compared toatomization into LN 2 when using energies of 0 – 5 Watt. At higher atomization energiesthis stabilizing effect of LN 2 appears to be lost. As shown in Figure 90, these SEC resultsare consistent with turbidity‐ and enzyme activity determinations.air [MDR SEC ] LN 2 [MDR SEC ]LF atomized at 0 Watt in: 4.88 6.25LF atomized at 2 Watt in: 5.98 7.02LF atomized at 5 Watt in: 3.69 5.33LF atomized at 8 Watt in: 5.38 4.42Table 23. Different MDR of atomized bCA‐samples (7.5 mg/mL) in SEC afteratomization into air/LN 2 .Figure 90 illustrates a linear decrease in residual activity and a linear increase in theturbidity for air‐atomized bCA when the atomization energy is elevated to 8 W. Turbidityvalues of flash‐frozen and thawed bCA are therefore always lower or at least equal tosamples that were atomized into air. Residual activity almost remains unaltered whenspraying into LN 2 , whereas atomization into air reduces the enzyme activity to 90 %.Experiments also reveal that damage inflicted on bCA during atomization into air isreversed during quench‐freezing in LN 2 (Figure 90). This results agree with the research

116 RESULTS AND DISCUSSIONof Sonner et al [2002], who measured a regain in trypsinogen residual activity of 14 %after atomization into LN 2 compared to nebulization into air.MacRitchie [MacRitchie 1998] stated that protein aggregation induced upon atomizationmust not necessarily result in irreversibly aggregated structure. Based on bCA´s facilityof refolding [Wong and Tanford 1973; Chen et al. 2003], a spontaneously recovery to theprotein´s native conformation is imaginable ‐ although it is not clear how this could takeplace during quench cooling in LN 2 .Figure 89. 15 mg/mL bCA‐samples atomized into air and LN 2 . SEC runs wereexecuted without any dilution step with 120 μL injection volume.The results from the experiments obtained at 8 W are questionable, as they showinconsistent turbidity‐ and activity results with high standard deviations (Figure 90). Areason for this behavior might be the observed turbulent and non‐uniform spray cloudthat was generated with the ultrasonic nozzle after the atomization energy wasincreased from 5 W to 8 W.

RESULTS AND DISCUSSION 117Figure 90. Turbidity and residual activity values of bCA‐samples at 15 mg/mL afteratomization. Atomization was conducted into air or LN 2 at varying spraying energies.Remarkable are the SEC results from atomized samples compared with the LF. Smalldifferences in peak area distribution occur over time, the monomer fraction slightlyincreases and the dimer fraction decreases at the same time (Table 24). Stressed bCAseems to be able to refold to a minor degree after atomization. The aggregatesgenerated are partially reversible, whereas reconstituted SFD‐ and SD‐powders do notshow a shift in the MDR. Table 24 indicates that the complete SFD‐ and SD‐processeslead to irreversible damage to protein structure giving higher MDR values.MDR, SEC‐run 1 MDR, SEC‐run 2 MDR, SEC‐run 3LF, 15mgmL 4.87 4.88 4.88LF atomized at 5 Watt into air 3.54 3.67 3.86LF atomized at 5 Watt into LN 2 5.14 5.33 5.52SDF bCA 15% w/w 3.27 3.29 3.28SD bCA 15% w/w 3.32 3.26 3.26Table 24. Different peak composition of SFD‐bCA and treated LFs (old batch,7.5 mg/mL) in UV‐signal. bCA‐samples that were previously sprayed into air/LN 2 (5 W)show a change in MDR over time, whereas redispersed SD‐/SFD‐samples areunaltered.

118 RESULTS AND DISCUSSION5.3.4 SFD of bCA at different protein concentrationsNon‐stabilized bCA solutions suffer substantially during spray freeze‐drying. In order toexamine the influence of enzyme concentration on bCA‐integrity, pure protein solutionswith solid content from 0.25 – 15 % (w/w) were used for SFD‐experiments. SEM pictures(Figure 91) show that a concentration of at least 5 % (w/w) is required to obtainspherical particles. These results agree with results from Sonner [2002] who suggestedthat a minimum solid content is crucial to form spherical particles during SFD. Lowersolid contents lead to undefined, sponge‐like powders with high porosity and very fragilestructure (Figure 91a ‐ c). If the SFD powders are intended for needle‐free ballisticinjection superior particle stability and density should be targeted, enabling particles topenetrate through the skin. A high protein load is essential for this purpose [Kendall etal. 2004; Ziegler 2006]. An increased solids content due to elevated bCA concentration inthe LF leads to distinct spheres with an improved stability; fixing SFD samples for SEMsputtering with a brush (see 4.2.15) does not seem to cause any damage to them. Thepowder particles exhibit less chipped scratches on their surfaces (Figure 91d ‐ f). Theparticles show a less porous structure (Figure 91f) with most likely a higher density.Figure 92 shows that the results of the MDRs (AF4 and SEC) and the enzyme activity lossof bCA samples behave inverse. Changing the protein content from 10 % to 15 % (w/w)leads to a higher activity loss of 18 %. The difference between MDR SEC and MDR AF4shows that the dimer concentration in AF4 is higher compared with the SEC results.Further explanations of this are given in chapter 5.3.5.2.Turbidity values are consistent with bCA‐inactivation, showing a direct proportionality tothe residual activity (Figure 93). Also turbidity and SEC measurements show majoralterations when the bCA‐concentration is raised from 10 % to 15 % (w/w).Costantino et al [2002] showed that the loss of monomer of atomized bSA was notaffected by concentration over a range from 0.2 – 20 mg/mL protein. At higher proteinconcentrations there was a marked increase in particle size, along with a minorimprovement in stability.It can be concluded that an increase in the protein content affects various stresses aslisted in Table 3 (chapter 2.2), leading to an escalating loss in monomer fraction andtherefore a reduction in residual activity.

RESULTS AND DISCUSSION119a)50 μmb)50 μmc)50 μmd)20 μme)f)20 μm20 μmFigure 91.SEM pictures of SFD bCA particles with different solid content: a ‐ c) 500fold, d ‐ f) 1000 fold magnification. SFD‐particles from solutions at: a) 0.25 % (w/w)bCA, b) 0.5 % (w/w) bCA, c) 1.0 % (w/w) bCA, d) 5.0 % (w/w) bCA, e) 10.0 % (w/w) bCAandf) 15.0 % ( w/w) bCA content.

120 RESULTS AND DISCUSSIONFigure 92. Residual activity loss and MDR of pure SFD bCA‐powders with varyingsolid contents. MDR values of SEC and AF4 measurements obtained from the UVsignal.The difference in MDR SEC and MDR AF4 are further discussed in chapter 5.3.5.2.Figure 93. Turbidity and activity loss for pure SFD bCA‐powders of varying solidcontents. Turbidity measurements were obtained with samples of 2 mg/mL.

RESULTS AND DISCUSSION 1215.3.5 Stabilization of bCA during SFD with different excipientsIt has been shown that bCA suffers substantial damage during SFD. By addition ofdifferent excipients to the LF, it should be possible to eliminate the protein alterations.Stabilizing substances were selected based on their cryo‐ and lyoprotectant properties(according to 2.2.4), based on the preferential exclusion and the water replacementtheories [Wang 1999; Wang 2000]. First, some low concentrated (15 % w/w) proteinstabilizermixtures are reviewed. Protein integrity of simple and complex mixtures isanalyzed with SEC and AF4. Secondly, 30 % (w/w) SFD‐batches are examined.5.3.5.1 SFD powders from LFs with low solid contents (15 % w/w)5.3.5.1.1 SFD bCA‐trehalose 15 % (w/w)Costantino et al [2004] reported that a protein‐sugar mass ratio of at least 1:1 isnecessary for complete protein stabilization. SEC measurements indicate that even lesstrehalose is needed to prevent the aggregation of bCA monomers (Figure 94).Figure 95a illustrates that a bCA‐trehalose ratio of just 5:1 is sufficient for almostcomplete stabilization. The residual activity of the SFD‐powder is 24 % above thatwithout any stabilizing excipients. A plateau level is reached, and further increase intrehalose concentration does not lead to any change in the MDR or enzyme activity,except regarding the turbidity measurements (Figure 95). The results obtained from the3 methods are consistent. Although the LS‐signal (Figure 94a) indicates that there arestill high aggregated fractions present in trehalose‐stabilized bCA samples yet, themonomer peak area from the UV‐signal (Figure 94b) for samples with 5:1 protein/sugarratio increases to 13.2 %. The dimer‐ and trimer peak areas decrease to 8.7 and 2.6 %,respectively.SEM pictures (Figure 96a‐f) of the SFD powders show that the addition of trehalosemodifies the structure of the 15 % (w/w) particles. Trehalose mixed to the LF changesthe porous surface of pure bCA‐SFD particles into smooth spheres (Figure 100a ‐ c).Higher trehalose contents lead to a rough surface (bCA‐trehalose 1:1) and later on with

122 RESULTS AND DISCUSSIONincreasing sugar contents (bCA‐trehalose 1:2) to wrinkled particle surfaces(Figure 100d ‐ f).Figure 94. Reconstituted SFD‐bCA‐trehalose samples analyzed with SEC. 7.5 mg/mLsamples, 60 μL injection volume. Molar mass was taken from the bCA‐trehalose 5:1‐run over the RI‐signal. LS‐signals (a) and UV‐signals (b) shown.The structural alteration of the SFD‐powders may be related to the decrease in theT g´‐value when the trehalose concentration in the mixtures is increased (Table 25 andFigure 97). The lyophilization program was not modified at higher trehaloseconcentrations, so it cannot be ruled out that the drying temperature exceeded T g´during the ramping step into secondary drying and led to particle shrinkage.T g´ values could not be detected either with pure bCA or with bCA‐trehalose in a ratio of5:1. Nevertheless no negative consequences could be detected, as the protein isstabilized further with increasing trehalose addition.The SFD powders give acceptable T g ‐values (Table 25). All T g´s are more than than 20 °Cabove room temperature, promising an acceptable shelf live for the trehalosecontainingbCA powders. Table 25 and Figure 98 demonstrate that high protein contentsincrease the T g ‐value – with a bCA‐trehalose ratio larger than 2:1 no T g can be detected(Figure 98).

RESULTS AND DISCUSSION 123a)b)Figure 95. Protein integrity of bCA‐trehalose SFD powders with varyingcomposition. a) Residual activity loss and compared to MDR of SEC and AF4measurements obtained from the UV‐signal. The difference in MDR SEC and MDR AF4 arefurther discussed in chapter 5.3.5.2. b) Turbidity and residual activity measurementsshow consistent results. A high activity loss is accompanied with an increase inturbidity.

124 RESULTS AND DISCUSSIONa) b)20 μm20 μmc) d)50 μm50 μme)50 μm 10 μmFigure 96. SEM of SFD bCA‐trehalose particles: a) pure bCA SFD particles (1000x),b) bCA‐trehalose 5:1 (1000x), c) bCA‐trehalose 2:1 (500x), d) bCA‐trehalose 1:1 (500x),e) bCA‐trehalose 1:2 (500x), f) bCA‐trehalose 1:5 (1000x).f)

RESULTS AND DISCUSSION 125a) b)Figure 97.T g´ measurements with DSC of varying bCA‐trehalose solutions.a) b)Figure 98.DSC of bCA‐trehalose SFD‐powders, (a) 1 st scan and (b) 2 nd scan.The addition of trehalose to bCA has therefore a positive effect on enzyme stability afterSFD. Ziegler [2006] reported that higher contents of trehalose added to catalase also ledto an increase in protein stability after actuation of the SFD‐powder with an injectiondevice. A trehalose/catalase mass ratio of 3:1 showed almost complete preservation ofthe initial enzymatic activity, also at high actuation pressures.

126 RESULTS AND DISCUSSIONLF of differentbCA‐trehaloseratiosT g´‐values[°C](20°C/min)bCA‐SFD powders of varyingtrehalose addition1st heatingT g[°C](10°C/min)T g2nd heating[°C](10°C/min)bCA, old batch n.d. bCA‐pure, SFD 15 % (w/w), old batch n.d. n.d.bCA‐trehalose 5:1 n.d. bCA‐trehalose 5:1, SFD 15 % (w/w) n.d. n.d.bCA‐trehalose 2:1 ‐ 24.32 bCA‐trehalose 2:1, SFD 15 % (w/w) 102.11 n.d.bCA‐trehalose 1:1 ‐ 25.19 bCA‐trehalose 1:1, SFD 15 % (w/w) 67.07 63.74bCA‐trehalose 1:2 ‐ 25.21 bCA‐trehalose 1:2, SFD 15 % (w/w) 49.61 59.65bCA‐trehalose 1:5 ‐ 27.03 bCA‐trehalose 1:5, SFD 15 % (w/w) 47.96 56.30trehalose ‐ 31.37 Trehalose pure, SFD 15 % (w/w) 88.01 89.08Table 25. Results from DSC‐measurements: T g´‐values were obtained from 30 μLprotein/trehalose solutions. T g ‐values of SFD‐powders were determined with twoheating steps.5.3.5.1.2 SFD bCA‐binary mixtures of 15 % (w/w)The protein solution containing trehalose exhibited superior stabilizing capabilities,enzyme stability could be completely preserved for at least short term storage. Twohypotheses have been proposed to explain the stabilization mechanism of sugars onproteins during freeze‐drying and storage, the “water replacement” and the “glassimmobilization” hypotheses. However, it must be recognized that sometimes boththeories simultaneously play an important role in protein stabilization [Pikal 2004].An influence of sucrose, lactose monohydrate and mannitol on SFD‐bCA was alsoobserved. These three additives are known for their potential stabilizing features withproteins [Costantino et al. 1998a; Costantino et al. 1998b; Imamura et al. 2003; Wang etal. 2008]. Protein‐excipient mixtures in the ratio of 1:1 and 1:3 were prepared from15 % (w/w) LFs under the same conditions given in chapter 4.2.1.All SFD‐powders containing trehalose are spherical and less porous than pure bCAparticles(Figure 99 and Figure 100). bCA‐sucrose powders are different in their particlemorphology. The low sucrose ratio leads to smooth particle surfaces. Spheres obtainedfrom higher sucrose contents are sometimes wrinkled, whereas pure sucrose SFDparticlesof 15 % w/w are shrunken. Compared to trehalose the T g´ for sucrose isdecreased from ‐31 °C to ‐27 °C [Meister and Gieseler 2008], so the producttemperature may exceed T g´ during the fast ramping to 2° drying in the FD‐cycle wherethe particles start to collapse as they partially collapsed.

RESULTS AND DISCUSSION 127a)b)20 μm 50 μmc) d)20 μm 20 μmFigure 99. SEM pictures of stabilized bCA‐SFD particles for binary mixtures: a) bCAsucrose1:1 (1000x), b) bCA‐sucrose 1:3 (500x), c) bCA‐sucrose 1:3 (1000x), d) puresucrose, 15 % (w/w) (1000x).Mannitol‐containing bCA‐SFD particles show very smooth surfaces that are covered withsmall folds (Figure 100c + d). LMH particles indicate an increased tendency to breakwhen they were collected out of the aluminium bowl after FD. The broken spheres showa very porous inner particle structure, indicating this powder´s low mechanical stability(Figure 100b).Figure 101 demonstrates a slight difference between the stabilizing effects of trehalosecompared to sucrose, mannitol and LMH. This result agrees with the conclusion ofCostantino et al [2004] who suggested that a protein‐sugar mass ratio of at least 1:1 isrequired for complete protein stabilization. Nevertheless, protein‐excipient ratios of 1:1reveal less effect than seen with the trehalose mixtures. Sucrose, mannitol and LMHwith bCA exhibit nearly 10 % higher protein inactivation than the trehalose formulation.

128 RESULTS AND DISCUSSIONa) b)200 μm50 μmc)b) d)50 μm 20 μmFigure 100. SEM of stabilized bCA‐SFD particles of binary mixtures: a) bCA‐LMH 1:1(100x), b) bCA‐LMH 1:3 (500x), c) bCA‐mannitol 1:1 (500x), d) bCA‐mannitol 1:1 (1000x)Turbidity values of bCA‐mannitol formulations are equal to results of bCA‐trehalosesamples, whereas the results for reconstituted LMH‐ and sucrose‐containing powdersare doubled. Also here, the MDR and the turbidity measurements illustrate consistency(Figure 101b). Noticeable is the fact that MDR values obtained with AF4 are slightlyshifted towards the aggregated fractions, whereas the SEC results for the MDR are muchhigher. Further increase of excipient content to a ratio of 1:3 leads to excellent bCAstabilizationproperties for all three additives respectively. Higher excipientconcentrations lead to an increase in residual activity and MDR‐values after SECmeasurements(a + b), while the sample turbidity is further decreased.Chang et al [2005] showed that an IgG 1 antibody formulation could be stabilized in aweight ratio of 1:1 with sucrose as well as with trehalose. Other investigations provedthat the physical stability of most lyophilized proteins seems to improve monotonicallywith increasing sucrose ratios, until a certain level of sugar is reached that leads to aplateau of maximum stability [Wang et al. 2008].

RESULTS AND DISCUSSION 129a)b)Figure 101. Protein integrity of SFD bCA ‐ binary mixtures with varying composition.a) Activity loss plotted against sample turbidity ‐ both methods show consistentresults. b) Sample turbidity plotted against MDR measurements (SEC and AF4)obtained from the UV‐signal.

130 RESULTS AND DISCUSSIONFigure 102. Reconstituted SFD‐bCA‐mannitol samples analyzed with SEC. 7.5 mg/mLsamples, 60 μL injection volume. Molar mass was taken from the bCA‐mannitol 1:3‐run over the RI‐signal.Ziegler [2006] showed that lactose could mechanically stabilize catalase SFD particlesnearly as well as trehalose. LMH can be taken as a substitute for trehalose withoutcompromising long‐term storage stability. Lactose effectively inhibits protein unfoldingduring lyophilization [Carpenter et al. 1997] but as a reducing sugar it is also known toreact with amino groups in proteins via the Maillard reaction [Carpenter et al. 1997].After storage the lactose‐containing formulations are therefore expected to shownegative deviation in their stabilizing efficiancy. Costantino et al. [1998b] examined theshelf life of lyophilized recombinant human growth hormone (rhGH) stabilized withlactose. After 4 weeks of incubation under elevated stability conditions, up to 20 % ofthe rhGH was found in an altered form likely a glycosylated monomer [Costantino et al.1998b].Mannitol in the amorphous state may be suitable for protein stabilization, although thestabilization efficiency is limited by its tendency towards crystallization over time[Costantino et al. 1998a]. Ziegler et al. [2006] reported that catalase‐mannitol loadedSFD particles with high mannitol contents showed reduced protein stabilization when

RESULTS AND DISCUSSION 131actuated with the PowderJect ® device. The authors implied that the amorphous state isnot only crucial for protein stabilization during SFD, but also for the actuation process.Figure 103.WAXD of different mannitol containing SFD formulations.a) b)Figure 104. DSC of bCA‐mannitol SFD‐powders, (a) 1 st scan and (b) 2 nd scan. NoT g ‐values are detectable.WAXD measurements demonstrate that SFD powders with mannitol are at least partiallycrystalline (Figure 103); no T g ‐values for the SFD‐powders can be detected (Figure 104).Nevertheless, SEC‐ and residual activity measurements show that the native structure ofbCA could be preserved during the SFD‐process. T g ‐values for the bCA‐LMH mixtures are

132 RESULTS AND DISCUSSIONmore than 20 °C higher than those of sucrose‐containing bCA‐particles (Table 26). Thelow T g of SFD sucrose‐blends is due to the lower T g ‐value of sucrose compared to otherdisaccharides, and the higher water content in the hygroscopic powder.bCA‐SFD powders of binary mixtures1st heatingT gT g2nd heating[°C][°C]bCA‐sucrose 1:1, SFD 15 % (w/w) 41.42 55.94bCA‐sucrose 1:3, SFD 15 % (w/w) 41.80 51.87bCA‐LMH 1:1, SFD 15 % (w/w) 65.79 74.28bCA‐LMH 1:3, SFD 15 % (w/w) 67.29 78.56bCA‐mannitol 1:1, SFD 15 % (w/w) n.d. n.d.bCA‐mannitol 1:3, SFD 15 % (w/w) n.d. n.d.Table 26.mixtures.Results from DSC‐measurements of bCA‐SFD powders of binaryThe results of the SFD experiments for bCA mixed with only one excipient are inagreement with observations of Kawai and Suzuki [2007] who examined the stabilizationefficiency of different sugars on LDH during FD. All sugars and mannitol are suitable toobtain stable bCA powders, whereas trehalose‐protein mixtures show increased residualactivity even in small concentrations of trehalose. The sugar does not tend to react withthe protein and the SFD‐particles exhibits a high T g value.5.3.5.1.3 SFD bCA‐complex mixtures of 15 % (w/w)The last two chapters have shown that single mixtures with sugar or mannitol canstabilize bCA during SFD. In the literature some complex formulations with more thanone excipient are mentioned [Maa et al. 2004; Rochelle and Lee 2007; Ziegler et al.2010]. The benefits of some excipients are supposed to compensate typicaldisadvantages of other individual components.Sonner [2002] reported that amorphous substances like trehalose show an increase intheir hygroscopicity after SFD. High water uptake rates and particle rigidity of the drypowders can, however, be reduced with the addition of crystalline substances such asmannitol [Maa et al. 2004]. Dextran is known to increase a formulation´s elasticity inorder to reduce particle brittleness and yield more dense particles due to a highertendency for partial shrinkage of the frozen particles in 1° and 2° drying

RESULTS AND DISCUSSION 133[Rochelle and Lee 2007]. Furthermore dextran was found to increase T g of the driedpowder which might lead to more stable powder particles in long time storage tests.Maa et al [2004] showed that an influenza vaccine powder formulation that wassuccessfully stabilized with a TMD formulation in the ratio of 3:3:4 still possessed goodphysical and biochemical stability after 1 year storage at room temperature. Based onobservations by Gloger et al [2003] who found that dextran with low molecular weightshowed an increase stabilization of the protein aviscumine, the polymer taken here forfurther experiments had a molar mass of 10 kDa. bCA blends with trehalose‐mannitol,trehalose‐dextran and TMD 3:3:4 were taken for SFD experiments (Table 27).SFD bCA‐complex mixtures of 15 % (w/w) Residual activity [%]± SD [%]Turbidity[2mg/mL]MDRSECMDRAF4bCA‐SFD (old batch) 70.13 ± 5.36 0.317 3.31 1.92bCA‐mannitol‐trehalose 1:1:1, SFD 91.40 ± 3.73 0.064 7.18 7.38bCA‐mannitol‐trehalose 1:2:2, SFD 94.37 ± 0.32 0.067 7.30 7.42bCA –trehalose‐dextran (10 kDa) 1:1:1, SFD 73.83 ± 3.22 0.124 4.19 4.71bCA –trehalose‐dextran (10 kDa)1:2:2, SFD 89.14 ± 2.24 0.088 4.82 5.58bCA –trehalose‐dextran (10 kDa)1:3:3, SFD 89.45 ± 2.68 0.096 4.88 5.81bCA‐TMD (10 kDa) 1:3:3:4, SFD 97.04 ± 2.79 0.038 6.37 ‐‐‐‐Table 27. Complex SFD‐bCA mixtures, 15 % (w/w). Residual activity, turbiditymeasurements and peak composition of reconstituted SFD powder.The SEM picture of the SFD bCA‐trehalose‐mannitol mixture in the ratio of 1:1:1(Figure 105a) shows a wrinkled structure that even collapsed with a further increase inexcipient content to the 1:2:2‐blend (Figure 105b). Due to the extremely low T g ‐values ofthese two powder compositions (Table 28) a further structural degradation of theparticles during SEM preparation cannot be ruled out.bCA‐SFD powders of complex mixtures1st heatingTg2nd heatingTg[°C][°C]bCA‐mannitol‐trehalose 1:1:1, SFD 27.96 47.80bCA‐mannitol‐trehalose 1:2:2, SFD 31.78 50.72bCA –trehalose‐dextran (10 kDa) 1:1:1, SFD 77.60 73.44bCA –trehalose‐dextran (10 kDa) 1:2:2, SFD 67.71 77.55bCA –trehalose‐dextran (10 kDa) 1:3:3, SFD 88.41 89.08bCA‐TMD (10 kDa) 1:3:3:4, SFD 56.55 55.65Table 28. Results from DSC‐measurements of bCA‐SFD powders of complexmixtures of 15 % (w/w).

134 RESULTS AND DISCUSSIONNevertheless, reconstituted bCA exhibits almost unaltered residual activity (Table 27).bCA samples show only low concentrations of aggregates in SEC and turbidity values arenegligible (Figure 106).a) b)20 μm20 μmc) d)200 μm20 μme) f)200 μmFigure 105. SEM pictures of bCA‐SFD particles of complex mixtures: a) bCAtrehalose‐mannitol1:1:1 (1000x), b) bCA‐trehalose‐mannitol 1:2:2 (1000x), c + d) bCAtrehalose‐dextran10 kDa 1:1:1 (100x, 1000x), e + f) bCA‐TMD 1:3:3:4 (100x, 1500x)20 μmTrehalose‐dextran (10 kDa) particles have a very smooth surface with only some brokenpieces from the shell (Figure 105c + d). Dextran in combination with trehalose at an

RESULTS AND DISCUSSION 135excipient‐protein ratio of 1:1 leads to more bCA inactivation than a single SFD 1:1 bCAtrehalosemixture (Table 27). This corresponds well with the results of Rochelle [2005]who observed that pure dextran 10 kDa was unable to protect catalase fromdenaturation during SFD. Addition of 4 parts of polymer that was mixed to the proteinfailed to stabilize catalase, and led to a poor residual activity value of only 59.6 %. Laterinvestigations proved that dextran weakened the stabilizing effects of trehalose andmannitol on catalase. The high turbidity value of reconstituted SFD bCA‐trehalosedextran1:1:1 powder (Table 27) is consistent with the very large aggregates at the rT of14 min in SEC (Figure 106). An increase in the bCA‐excipient ratio enhances bCA stabilityafter SFD. The sample turbidity decreases, although reconstituted bCA still exhibits ahigh concentration of di‐ and trimers in the SEC experiment (Table 27). Further increasein the trehalose‐dextran content did not result in any further stabilizing effect, and theresidual activity remained at 89 %.Figure 106. SEC‐measurements of reconstituted complex SFD‐bCA mixtures,15 % (w/w). UV‐signal (solid line) and LS‐signal (dashed line) shown.

136 RESULTS AND DISCUSSIONMolar mass determination of dextran‐containing bCA samples in SEC is morecomplicated than for samples without dextran (10 kDa). Most peaks are notmonodisperse because the dextran peak converges with bCA (as already seen inFigure 44). The “protein conjugate analysis” template of the Astra‐software was used tosolve this problem (see 5.1.7). Noticeable are the high T g ‐values for dextran‐containingpowders, lying all in the range from 70 – 90 °C (Table 28 and Figure 107). Since dextran isunfavourable for protein stability (Table 27), the polymer´s concentration should be highenough to produce the required effects on particle morphology or density, but still lowenough to maximize protein process stability [Rochelle and Lee 2007].a) b)Figure 107. DSC of bCA‐trehalose‐dextran SFD‐powders, (a) 1 st scan and (b) 2 nd scan.The SEM pictures of bCA‐TMD 1:3:3:4 particles (Figure 105e + f) show an extremelywrinkled structure with no visible pores on the surface. Rochelle and Lee [2007]reported that the presence of dextran 10 kDa produced high‐density, shrunken SFDprotein particles from catalase‐TMD (3:3:4) mixtures. Due to similar freeze‐drying cyclesalso here the collapsed, shriveled structure might be a result of elevated producttemperature during 1° drying that exceeded T g´. Nevertheless, residual activity andturbidity measurements illustrate the suitability of the complex excipient mixture forbCA stabilization (Table 27). The SFD protein remains unaltered in its enzymaticreactivity and the SEC runs show a low aggregate content.

RESULTS AND DISCUSSION 1375.3.5.1.4 SFD powders from LFs with high solid contents (30 % w/w)Chapter 5.3.4 showed that an increase in protein solid content had noticeable effects onthe morphology and mechanical stability of pure bCA SFD particles. Ziegler [2005]demonstrated that a low LF concentration of TMD 3:3:4 led to very fragile SFD powderscompared to increased solid contents. Powders produced from 10 % (w/w) solutionssuffered more mechanical disruption than particles from 30 % w/w solutions with thePowderJect® and Venturi® devices. The higher solid content created denser particlesthat were able to compensate high attrition effects.To evaluate the influence of increased solid contents on protein integrity during SFD, LFsof 30 % w/w were taken for SFD experiments. Trehalose, sucrose, mannitol and thecomplex mixture of TMD 3:3:4 were used to stabilize 10 % w/w bCA samples in aprotein‐excipient ratio of 1:2.The 30 % w/w SFD powders generated show a very smooth and dense structure withless chipped powder fragments (Figure 108 and Figure 109).Residual activity [%]± SD [%]Turbidity[2mg/mL]MDRSECbCA‐SFD 10 % (w/w), old batch 88.44 ± 4.31 0.233 5.15bCA‐trehalose 1:2, SFD 30 % (w/w) 96.25 ± 2.68 0.214 5.76bCA‐sucrose 1:2, SFD 30 % (w/w) 97.88 ± 2.79 0.168 6.57bCA‐mannitol 1:2, SFD 30 % (w/w) 93.42 ± 3.10 0.261 6.29bCA‐TMD (3:3:4) 1:2, SFD 30 % (w/w) 91.87 ± 2.42 0.463 4.51Table 29. SFD‐bCA mixtures of 30 % (w/w) at a protein‐excipient ratio of 1:2.Residual activity, turbidity measurements and peak composition of reconstituted SFDpowder.Sucrose‐ and mannitol‐containing particles showed a partially shrunken morphology,whereas TMD‐ and trehalose‐blends resulted in even particle surfaces (Figure 110). Theincrease in protein content compared to the wrinkled 15 % (w/w SFD bCA‐TMD 1:3:3:4particles (Figure 105e + f) resulted in the expected anti‐plasticising effect of the proteinon the TMD formulation [Sonner et al. 2002, Rochelle and Lee 2007]. Compared to pureSFD bCA from the LF of 10 % (w/w), all 30 % (w/w)‐bCA formulations showed an increasein protein stability, whereas turbidity measurements suggested evidence of someaggregates (Table 29).

138 RESULTS AND DISCUSSIONa)b)500 μm 20 μmc) d)100 μm 20 μmFigure 108. SEM of SFD bCA‐excipient 1:2, 30% (w/w): a + b) bCA‐trehalose 1:2 (50x,1000x), c + d) bCA‐sucrose 1:2 (200x, 1000x).Trehalose and sucrose added in a twofold amount to the produces good stabilization atthe high solid contents. The residual activity was nearly 100 %, and the disaccharidecontainingbCA powders showed low turbidity values and a decrease in aggregatesdetectable with SEC (Figure 111).Mannitol showed slightly lower stabilization potential compared to trehalose andsucrose. The residual activity of the mannitol‐containing SFD powder is 5.4 % higherthan the non‐stabilized SFD protein. The monomer fraction of the mannitol‐stabilizedpowder is increased, although the sample turbidity does not show any difference topure SFD bCA 10 % (w/w).

RESULTS AND DISCUSSION 139a) b)100 μm20 μmc)d)200 μm 10 μmFigure 109. SEM photographs of SFD bCA‐excipient 1:2, 30% (w/w): a + b) bCAmannitol1:2 (200x, 1000x), c + d) bCA‐TMD (334) 1:3 (50x, 1000x).The insufficient protein protection with TMD (3:3:4) lead to unsatisfactory results. Whilea 15 % bCA‐TMD (3:3:4) formulation suffers no noteworthy alteration on SFD, thecomplex mixture fails to avoid bCA‐denaturation for a 30 % (w/w) LF. TMD in a 2:1 massratio can only elevate bCA residual activity by 3.4 % compared to the pure SFD protein.Turbidity measurements and the increase in dimer and trimer peak fraction in SECexperiments suggest additional creation of larger aggregates (Figure 111). The highdextran content in the TMD mixture leads to a higher T g ‐value of the manufacturedpowder compared to pure trehalose (Table 30). The sucrose‐formulation exhibits a verylow glass transition temperature, which probably would lead to problems during longtimestability tests.The insufficient protein protection with TMD (3:3:4) lead to unsatisfactory results. Whilea 15 % (w/w) bCA‐TMD (3:3:4) formulation suffers no noteworthy alteration on SFD, thecomplex mixture fails to avoid bCA‐denaturation for a 30 % (w/w) LF.

140 RESULTS AND DISCUSSIONFigure 110. SFD particle size for powders prepared from 30 % (w/w) solutions: Thecumulative particle fraction (CPF) of the powders indicate d 50 ‐value of: a) bCAmannitol1:2 at 23.8 μm (407 particles), b) bCA‐sucrose 1:2 at 23 μm (450 particles),c) bCA‐trehalose 1:2 at 40 μm (649 particles), d) bCA‐TMD 1:2 at 41 μm (543 particles)and e) bCA‐pure 15 % (w/w) at 31.5 μm (540 particles).Figure 111. SEC runs of reconstituted SFD samples of 30 % (w/w). 7.5 mg/mLsamples, 60 μL injection volume. Molar mass was taken from the bCA‐trehalose 1:2‐run over the RI‐signal.

RESULTS AND DISCUSSION 141No T g ‐value could be detected for the bCA‐mannitol powder, although a partiallyamorphous state of the additive is likely.Figure 112.DSC of 30 % (w/w) bCA‐SFD‐powders, (a) 1 st scan and (b) 2 nd scan.T1st heatingT gbCA‐mannitol 1:2 , SFD 30 % (w/w) n.d. 74.45bCA‐sucrose 1:2, SFD 30 % (w/w) 41.42 34.94bCA‐trehalose 1:2, SFD 30 % (w/w) 70.12 64.79bCA‐TMD 1:2, SFD 30 % (w/w) 41.11 77.43Table 30. Results from DSC‐measurements of bCA‐SFD powders of complexmixtures.g2nd heating

142 RESULTS AND DISCUSSION5.3.5.2 Comparison of AF4‐ and SEC‐measurements with bCASEC experiments have shown their suitability in the determination of protein alterationsafter an SFD process, although the recoveries can be low (see 5.3). The highly stressedbCA samples did not elute completely from the separation column. Denatured proteincan bind to the guard column and to the surface of the cross‐linked agarose/dextranmatrix of the Superdex 200 column via charge‐charge or hydrophobic interactions.These high inner surfaces are not present in the separation channel of an AF4 system,leading to higher recovery with an elevated ratio of detectable bCA aggregatescompared to SEC runs (Figure 115). Preliminary tests with modified compositions of themobile phase with an increase in salt concentration did not lead to improved recoveries.Herold [1993] has shown that NaCl addition has strong effects on the elution of proteinsin SEC analysis. The addition of high concentrations of NaCl to the mobile phase doesnot, however, lead to an increase in protein recovery (Figure 113). However,hydrophobic interactions between sample analytes and the column packing materialincrease with rising ionic strength. Thus a ballance must be achieved between the needto reduce ionic interactions and the need to limit hydrophobic interactions. Yumioka etal. [2010] suggested that especially new columns show a stronger tendency to bindproteins by adsorption. He found out that a better recovery was achieved in thepresence of arginine. Arginine solubilizes aromatic and weakly hydrophobic compounds,but to a lesser extent than organic solvents [Arakawa et al. 2010]. Arginine appears,however, to have little impact on the structure and stability of native bCA (Figure 113).Experiments with a modified mobile phase (50 mM tris‐HCl + 200 mM and 400 mMarginine) indeed lead to a slightly higher recovery (Figure 113), but also to a change inthe peak distribution (Figure 114). Arginine is known to suppress aggregation of bCA, soit is possible that it may break down losley associated aggregates and lead tounderestimation of aggregate content [Baynes et al. 2005]. Finally the addition of25 % (v/v) glycerol to the mobile phase of 50 mM tris‐HCl and 150 mM NaCl lead toadequate recovery values of about 95 % for unstabilized SFD‐ and SD‐samples (see5.3.2).

RESULTS AND DISCUSSION 143Figure 113. Changes in the recoveries of a SFD bCA sample with different mobilephase compositions of SFD‐bCA (60 μL injected volume, concentration of 10 mg/mL).Figure 114. Reconstituted bCA FD 15 % (w/w) sample to 5 mg/mL with tris‐HClbuffer (pH 7.50) and 150 mM NaCl or 200 mM arginine, respectively. 60 μL wereinjected for each sample. UV‐ and LS‐signals shown, molar mass calculated from theRI‐signal (FD sample run in 50 mM tris‐HCl + 150 mM NaCl).

144 RESULTS AND DISCUSSIONAF4 and SEC measurements clearly illustrate consistent results for stabilized proteinsolutions (Figure 95/Figure 101, Table 28). Well‐stabilized bCA‐samples have anincreased residual activity and show a high monomer fraction with both separationtechniques (Figure 116). Two points should be noted. First, more stressed samples havea higher increase in large aggregates detectable in AF4 than in SEC; and secondly, theshoulder between the bCA‐monomer‐ and dimer‐peak seen in SEC is missing in AF4‐runs(Figure 115).Denatured and unfolded bCA may bind to the column‐matrix, leading to a decrease inprotein recovery with SEC. AF4 evaluations with non‐centrifugated SFD samples did notshow any increase in higher, soluble aggregates (results not shown). The poorly solubleor insoluble larger aggregates could not be detected with this method, although theseruns suffered much from particle‐induced blockages in the LS‐detector flow cell.Figure 115. Comparison of SEC and AF4 measurements with a highly‐stressed and astabilized bCA sample.It can be concluded that the different analytical methods for bCA‐separation, SEC andAF4, lead to consistent results for undamaged protein solutions, but an increasingdiscrepancy for stressed bCA. The protein recovery is different between AF4 and SEC. Amore stressed sample shows a strong tendency to be retained in the SEC column. AF4

RESULTS AND DISCUSSION 145shows therefore a high recovery, but it should be mentioned that an increase in thecross‐flow could induce high shear forces that might generate further aggregates. Thedifferences in MDR for both methods is not easy to explain. It might be possible that theincreased salt concentration in the mobile phase of the SEC compared to the AF4method could lead to the diffusion of reversible aggregates. Even a refolding of bCAduring a SEC‐process that is discussed in the literature is imagineable [Gua et al. 2003].However, SEC analysis that was repeated with the same mobile phase used in AF4experiments showed a slight reduction in the MDR discrepancy, although the recoveryin SEC was still low (Table 31).Samples MDR AF4 MDR SEC recovery AF4 / SEC residual activitySFD bCA 15 % (w/w), pure 1.86 2.42 95.0 / 75.9 70.13SFD bCA‐lactose 1:1 4.58 4.39 93.4 / 74.8 91.19SFD bCA‐trehalose 2:1 7.07 6.62 98.8 / 75.1 95.39SFD bCA‐trehalose 1:1 7.49 7.05 96.0 / 72.2 98.00Table 31. Comparison of bCA‐samples (10mg/mL, 5 μL) in AF4 and SEC in the samemobile phase (50 mM tris‐HCl + 50 mM NaCl, pH 7.50).

146 RESULTS AND DISCUSSIONa)b)Figure 116. Reconstituted SFD‐bCA‐trehalose samples analyzed with AF4. 10 mg/mLsamples, 5 μL injection volume. Molar mass was taken from the bCA‐trehalose 5:1‐run.LS‐signals (a) and UV‐signals (b) shown.

RESULTS AND DISCUSSION 1475.4 Lactate Dehydrogenase (LDH)5.4.1 Protein characterizationSEC measurements of LDH show a clear LS‐signal for the protein monomer and only asmall signal for the aggregated fraction – almost invisible in the RI‐ or UV‐detector signal(Figure 117). The molar mass of a pre‐dialyzed 10 mg/mL LDH‐solution (see chapter4.2.6) exhibits three peaks at about 146 kDa, 292 kDa and 576 kDa. The molar masses ofmonomer, dimer and the third peak are in the relative ratio of 1 : 2 : 4, indicating thatthe third peak is not a trimer but a tetramer. Nevertheless, the signals for the third peakfor both concentration detectors are too low to allow a reliable molar mass calculation.Hydrodynamic radii obtained from QELS (Figure 118) could only be measured for themonomer at a concentration of 10 mg/mL. The protein monomer, actually made up of 4subunits (4.1.1.3), has an rH of 4.40 ± 0.29 nm according to Engstrom et al [2007a].Figure 117. Molar mass of a pre‐dialyzed LDH sample (7.5 mg/mL, 60 μL); dashedline: UV‐signal, solid line: LS‐signal, dotted line: RI‐signal. Plateaus in molecular weightcurve represent monodisperse fractions of LDH, calculated via the UV‐signal(ε specific = 1370 ml/g*cm). Monomer: 146.2 kDa ± 0.004 kDa; dimer: 292.0 kDa± 0.108 kDa; trimer: 576 ± 366 kDa.

148 RESULTS AND DISCUSSIONFigure 118. Hydrodynamic radius of a pre‐dialyzed LDH sample (10 mg/mL, injectionvolume of 60 μL): QELS measurement calculated the molecular size of the proteinmonomer (rH monomer ) to be 4.40 ± 0.29 nm.5.4.2 SFD of pure LDH5.4.2.1 Protein aggregation of LDH during SFDPure LDH in a concentration of 1 % (w/w) suffers substantially during SFD, even morethan bCA. The entire SFD‐process results in an activity loss of 59 % for reconstitutedpowder (Table 32). LDH solutions were prepared and exposed to each individual processduring SFD: 1) atomization of the LF with a 60 kHz nozzle at 5 W into air and 2) intoliquid nitrogen. 3) FD of the LF with 30 min freezing on ‐45 °C pre‐cooled shelves; and4) the complete spray freeze‐drying process with previous quench freezing in LN 2(according to chapter 5.3.1). Residual activity‐, turbidity‐ and SEC‐measurements wereexecuted after every single process step in triplicate. Activity measurements for LDHshow an increased standard deviation compared to bCA measurements, probablycaused by the longer sample preparation step of the LF (see 4.2.6). The protein wasdialyzed overnight and finally diluted to a concentration of 10 mg/mL with 100 mM

RESULTS AND DISCUSSION 149potassium di‐hydrogene phosphate buffer (pH 7.00). The liquid feed was set as the100 % standard for activity measurements.Atomization of the LDH solution into air leads to inactivation of 37.4 % with a substantialincrease in sample turbidity at 350 nm from 0.040 to 0.370. Nevertheless there is nodifference detectable in SEC‐measurements except for a small decrease in recovery(Table 32 and Table 33). Atomization with subsequent quench‐freezing shows similarresults as reported for bCA. The residual activity gains 3.2 % compared to nebulizationwithout freezing, and the turbidity decreases. Water removal during FD reduces LDHresidual activity to almost 44 %, whereas the entire SFD‐process leads to 59 %inactivation. SEC measurements do not show significant changes in peak composition ofFD and SFD samples (Figure 119) that would explain the large alteration in LDHfunctionality compared to the LF. The slight increase in higher aggregates in SEC isconsistent with higher turbidity values. Noticeable is an observed decrease in SECrecovery, which might be an indication for protein denaturation and precipitation(Table 33).The results of the residual activity measurements show that protein denaturation duringSFD is comparable to that in the atomization (activity loss of 34.2 %) and the waterremoving step (activity loss of 24.8 %).FD‐powder consists of unformed, flaked pieces, with clean edges on the particle surface(Figure 120a + b). SFD particles where more smooth (Figure 120c + d).LDH [mg/g] Solid content[% w/w]Residual activity± σ [%]Liquid feed 10.0 1.0 100 ± 1.05LF atomized in air (5 W) 10.0 1.0 62.58 ± 7.74LF atomized in LN 2 (5 W) 10.0 1.0 65.80 ± 5.30FD process 10.0 1.0 43.90 ± 7.08SFD process 10.0 1.0 41.04 ± 4.56Table 32. Process steps during SFD of 1 % (w/w) LDH in KH 2 PO 4 buffer (pH 7.00).All activity‐ and SEC‐measurements were obtained immediately after particulartreatment. Experiments were obtained in triplicate; 60 μL were injected for each SECrun, each sample in a concentration of 10 mg/mL. FD and SFD powders werereconstituted in LDH‐activity buffer. The activity value of the LF was set to 100 %residual activity.

150 RESULTS AND DISCUSSIONUV‐measurementsMonomerfraction [%]Dimerfraction [%]MDR Recovery[%]Turbidity[10 mg/mL]Liquid feed 97.05 1.65 58.82 101.60 0.040LF atomized into air (5 W) 96.14 1.59 60.47 98.49 0.370LF atomized into LN 2 (5 W) 97.40 1.55 62.84 99.17 0.274FD process 95.63 2.62 36.50 96.48 0.280SFD process 95.24 2.89 32.96 103.16 0.540Table 33. Process steps during SFD of LDH: Peak distribution of HPLC‐SECexperiments calculated from UV‐signal, turbidity measurement of undiluted samples.Figure 119. Process‐induced stress on LDH during single steps of a SFD‐process.Relative LS‐signals of stressed LDH during SFD. Monomer peaks were set to “1” asrelative scale; aggregated fractions are seen in their relative ratios for each sample.The LF and reconstituted powders were prepared in potassium phosphate buffer(100 mM KH 2 PO 4 , pH 7.00) at a concentration of 10 mg/mL; 60 μL injection volumeeach sample.

RESULTS AND DISCUSSION 151a) b)50 μm 20 μmc) d)50 μmFigure 120. SEM pictures of FD and SFD LDH, 1% (w/w): a + b) FD LDH (500x, 1000x),c + d) SFD LDH (200x, 1000x).20 μm5.4.2.2 Stabilization of LDH during SFD with trehaloseChapter 5.3.5.1.1 showed the high stabilization efficiency of trehalose as a cryo‐ andlyoprotectant for bCA. The disaccharide was now added to LDH solutions of 10 mg/mL indifferent concentrations to evaluate LDH conservation during SFD.Trehalose addition increases the solid content of the LF; thus adding 60 mg/mL sugar toLDH results in spherical particles (Figure 121c, d). Less concentrated sugar‐LDH mixturesdo not form a spherical structure. The surface of the powder seems to have thawedduring FD. The high buffer salt content (100 mM KH 2 PO 4 ) compared to the sugar/proteinfraction in the formulation might be a reason for this, because it leads to a lowerT g´ value and enables the frozen particles to collapse. Elevated trehalose concentration(LDH‐trehalose 1:6, 7 % w/w) leads to spheres with partially thawed surfaces. Nochannels or other surface texture can be seen, and sometimes the powder particles arefused (Figure 121d). With further addition of trehalose, the powder particles become

152 RESULTS AND DISCUSSIONstable and not aggregated (Figure 121e, f). A 1:1 mass ratio of LDH with trehalosedecreases the sample turbidity and improves the residual activity up to 24 % comparedto SFD of pure protein (Figure 122). The enzyme activity increases with higher trehalosecontent up to almost 93 %.a) b)50 μm 20 μmc) d)200 μm 50 μmf)20 μmFigure 121. SEM pictures of SFD bCA‐trehalose particles: a + b) SFD LDH‐trehalose1:1, 2 % (w/w) (500x, 1000x), c + d) SFD LDH‐trehalose 1:6, 7 % (w/w) (100x, 500x),e) SFD LDH‐trehalose 1:20, 21 % (w/w) (1000x).

RESULTS AND DISCUSSION 153The sample turbidity of a reconstituted SFD 1:20 LDH‐trehalose powder (21 % w/w) isonly half of the value of unstabilized 1 % (w/w) SFD LDH (Figure 122).Figure 122. Turbidity and activity loss of pure SFD LDH‐powders. The LDHconcentration was set to 10 mg/mL. The trehalose addition was varied, so the solidcontents were different. Turbidity measurements were obtained with samples of aLDH concentration of 2 mg/mL.SEC‐measurements of SFD LDH‐trehalose products show a small loss in dimer peak area.Reconstituted pure 1 % (w/w) SFD LDH sample consists of 2.64 % dimer, whereas theSFD LDH‐trehalose 1:20 powder shows just 1.43 % dimer content for the overall peakfraction.Trehalose alone is therefore not able to fully protect LDH during SFD. Sonner [2002]showed that the combination of trehalose and Tween 80 to a LF of LDH could lead toalmost undamaged SFD products. The surfactant additionally prevents proteininactivation during the atomizing step [Adler 2000] and freezing step [Chang et al. 1996]due to the protein exclusion from generated interfaces by the tensid.

154 RESULTS AND DISCUSSIONFigure 123. SFD LDH – protected and unprotected. The LF and the two reconstitutedpowders were prepared in dialysis buffer (100mM KH 2 PO 4 , pH 7.00) at a concentrationof 10 mg/mL. 60 μL injected. The trehalose can also be detected with the LS‐detector.

RESULTS AND DISCUSSION 1555.5 Regulation of SFD‐particle shapeSFD‐powders show benefits compared to FD‐cakes such as their highly porous characterwhich makes them suitable for pulmonary delivery or the improved dissolution of poorlywater‐soluble drugs [Costantino and Maa 2004]. For other applications, such as EPI,these powders suffer from their low‐density that will limit the skin penetration ofvaccine particles. This chapter describes modification of the surface morphology ofprotein‐loaded SFD‐powders by addition of different excipients or changes in processparameters during the FD step.5.5.1 Excipient‐induced change of particle morphologyMaa et al [2004] showed that the excipients used for SFD influence the resulting particlemorphology. Working with influenza vaccines, Maa and later Rochelle and Lee [2007]showed that mainly protein‐TMD mixtures tend to shrink during lyophilization of thefrozen droplets due to the plasticizing effect of the additives – most notably caused bydextran.For further investigations on this issue, bSA was used as model protein. SFD powders of15 % (w/w) LF were generated with a more aggressive drying cycle (Table 34). Trehaloseand sucrose were employed as stabilizers in varying mass ratios with the protein;6 mL of each LF were used for SFD experiments.StepOnset[°C]Endset[°C]Hold [min.]Ramp[°C/min.]Vacuum[mTorr]Segment time[min.]Total time[min.]1 ‐45 ‐45 30 equilibrate ‐‐‐‐ 30 302 ‐45 ‐15 ‐‐‐‐ 2.0 75 (R) 15 453 ‐15 ‐15 240 equilibrate 75 (H) 240 2854 ‐15 +25 ‐‐‐‐ 0.88 40 (R) 45 3305 +25 +25 600 equilibrate 40 (H) 600 930Table 34.FD‐programm #2 with the VirTis Advantage Plus freeze‐dryer.SFD pure bSA particles are round spheres with myriad small channels on their surfaces(Figure 125a). With the addition of trehalose to the protein in equal mass, the particlesurface becomes smoother. A further increase in sugar (bSA‐trehalose 1:3) leads to

156 RESULTS AND DISCUSSIONpartially shrunken particles (Figure 125c + d). A difference of approximately 1.5 °C in theformulation´s T g´ compared to bSA‐trehalose (1:1) might cause this change in particlemorphology (Table 35).Different protein/excipient ratiosLF/SFD powdersT g´ [°C](10°C/min)T g´ [°C](3°C/min)T g [°C](1 st scan)T g [°C](2 nd scan)Watercontent [%]bSA‐pure, SFD 15 % (w/w) n.d. n.d. n.d. n.d. 1.62bSA‐trehalose 1:1, SFD 15 % (w/w) ‐ 25.09 ‐ 26.51 93.29 87.36 1.88bSA‐trehalose 1:3, SFD 15 % (w/w) ‐ 26.68 ‐ 28.09 84.00 83.53 1.69bSA‐trehalose 1:6, SFD 15 % (w/w) ‐ 27.21 ‐ 27.71 84.44 82.12 1.77trehalose‐pure, SFD 15 % (w/w) ‐ 30.22 ‐ 30.98 71.50 71.51 2.51Table 35. DSC measurements of the LFs (T g´) and the generated SFD powders (T g )of bSA‐trehalose mixtures. Water contents were analyzed by KF titration.The product temperature went above T g´ during primary drying, which promoted viscousflow of the solid phase (Figure 124). SFD bSA‐trehalose powder at a ratio of 1:6 alsoshowed a shrunken structure, but the particles were molten together to some extent(Figure 125e).Figure 124. FD‐cycle #2: Thermocouples were positioned such that they contactedthe bowl´s base through the powder layer.

RESULTS AND DISCUSSION 157a) b)20 μm 20 μmc)d)20 μm 20 μme)20 μmFigure 125. SEM of SFD bSA‐trehalose particles of 15 % (w/w): a) pure bSA SFDparticles (1000x), b) bSA‐trehalose 1:1 (1000x), c, d) bSA‐trehalose 1:3 (1000x, 1000x),e) bSA‐trehalose 1:6 (1000x), f) pure trehalose SFD particles (500x).f)50 μmPure trehalose particles (Figure 125f) are completely collapsed with loss of their particlecharacteristics. The anti‐plasticizing effect of the protein is lost while T p probablyexceeded the formulation´s decreased T g´ / T c during the drying step.Although the drying cycle is very short, the water content of all powders is below 2 %.The dry product can be attributed to a low mass transfer resistance because of a thin

158 RESULTS AND DISCUSSIONdrying cake of the frozen spheres [Maa et al. 2004]. Only SFD pure trehalose gives ahigher residual moisture value of 2.5 %. The collapse probably caused the sealing ofcapillaries that led to a reduced dehydration during FD.The repetition of SFD experiments with sucrose instead of trehalose led to similarresults. A 1:1 protein‐sucrose ratio yielded round particles that already showed partiallythawed surfaces with large channels and scratches (Figure 126b). Increase in sugarcontent to a ratio of 1:3 led to a decrease in T g´ (Table 36) and causesd particleshrinkage. The “spheres” are edged and finely wrinkled. At a six‐fold sucrose mass, theouter shell is completely collapsed, no surface texture is visible and the particles arepartially molten together (Figure 126e). Pure SFD sucrose particles are even moreaggregated and consist of several fused individual spheres.T g´ [°C](10°C/min)T g´ [°C](3°C/min)T g [°C](1 st scan)T g [°C](2 nd scan)Watercontent [%]bSA‐pure, SFD 15 % (w/w) n.d. n.d. n.d. n.d. 1.62bSA‐sucrose 1:1, SFD 15 % (w/w) ‐ 26.63 ‐ 28.13 56.17 56.67 3.32bSA‐sucrose 1:3, SFD 15 % (w/w) ‐ 28.37 ‐ 29.63 48.99 44.17 2.58bSA‐sucrose 1:6, SFD 15 % (w/w) ‐ 28.27 ‐ 29.81 49.80 28.67 2.05sucrose‐pure, SFD 15 % (w/w) ‐32.85 ‐ 35.10 48.49 61.82 1.50Table 36. DSC measurements of the LFs (T g´) and the generated SFD powders (T g )of bSA‐sucrose mixtures. Water contents were analyzed by KF titration.Because of the fixed FD‐cycle the stronger decrease in T g´ of sucrose‐bSA mixtures leadsto a higher tendency to collapse than for the trehalose formulations. The use of SEMprovides only a qualitative impression of particle properties. More precise, quantitativecorrelations such as specific surface area measurements or powder tap densityexaminations were not investigated in this work. The water contents of sucrose‐basedSFD powders are somewhat higher than for trehalose mixtures. This might be caused byinferior dehydration or increased hygroscopicity.

RESULTS AND DISCUSSION 159a)20 μmb)20 μmc)500 μmd)20 μme)20 μmFigure 126. SEM pictures of SFD bSA‐sucrose particles of 15 % (w/w) LF:a) pure bSA SFD particles (1000x). b) bSA‐sucrose 1:1 (1000x). c) bSA‐sucrose 1:3(200x). d) bSA‐sucrose 1:3 (1000x). e) bSA‐sucrose 1:6 (1000x). f) pure sucrose SFDparticles (1000x).f)20 μm

160 RESULTS AND DISCUSSION5.5.2 Changes in particle morphology induced by variations in theFD‐cycle5.5.2.1 Collapsed bSA particlesSonner [2002] demonstrated that a modified FD cycle that included an annealing stepcould decrease SFD powder porosity by partial collapse and subsequent shrinkage of theparticles. Hg porosimetry and a particle embedding technique with a sectioned resinshowed that the annealing step reduced the number of pores and led to more compactparticles. Sonner tested various excipients and trypsinogen‐mixtures, but not singleprotein solutions. Annealing generally facilitates the crystallization process and Ostwaldripening, and is mainly used for formulations including bulking agent such as glycine ormannitol. This thermal treatment is not only used to provide elegant FD cakes or toavoid vial breakage, but also to optimize the FD process. Due to ice‐crystall ripening,larger pore structures occur in the crystalline matrix that accelerates the drying time[Overcashier et al. 1999]. The induced crystallization of stabilizing agents can eliminateprotein‐excipient hydrogen bond interactions and thus reduce protein integrity.Sonner [2002] reported that a two‐hour thermal treatment at 0 °C with a trypsinogen/trehalose/mannitol (15:15:70) formulation led to an activity loss of 16.5 %, whereas aformulation without mannitol (trypsinogen/trehalose 15:85) remained unharmed.For collapsed protein SFD particles bSA was used as model protein. The protein wasdissolved in 50 mM tris‐buffer (pH 7.50) to a concentration of 15 % (w/w). 10 mL of the LFwas used for each SFD experiment in order to sustain the same particle filling level in thealuminium bowl. The first variations in the FD‐cycle were done on the basis of theannealing experiments used by Sonner et al [2002], setting the annealing equilibrationtime to 1h, 2h or 6h (#3, #4 and #5; Table 37).Experiments with FD‐cycle #3 ‐ #5 did not lead to satisfactory results. With 120 minannealing at +5 °C the particles are spherical and have a smooth surface (Figure 127).Even with decreased (#5) and increased (#6) annealing times no shrunken particles areobtained after SFD of the 15 % (w/w) pure protein solution.

RESULTS AND DISCUSSION 161Step Onset [°C] Endset [°C] Hold[min.]Ramp [°C/min.]Vacuum[mbar]Segment time[min.]1 ‐45 ‐45 30 equilibrate ‐‐‐‐ 302 ‐45 +5 (+10, #6) ‐‐‐‐ 1.11 0.100 (R) 45 (50, #6)3 +5 (+10, #6) +5 (+10, #6) 120 (60, #4)(360, #5)(120, #6)equilibrate 0.100 (H) 120 (60, #4)(360, #5)(120, #6)4 +5 (+10, #6) ‐20 ‐‐‐‐ 0.55 0.100 (H) 45 (55, #6)5 ‐20 ‐20 1560 equilibrate 0.100 (H) 15606 ‐20 +25 ‐‐‐‐ 1.00 0.040 (R) 457 +25 +25 1200 equilibrate 0.040 (H) 1200Table 37. FD‐program #3 executed with the Christ Delta 1‐24 KD freeze‐dryer.Variations of #3 (#4 ‐ #6) shown in brackets.a) b)100 μm 20 μmc) d)50 μm 20 μmFigure 127. SEM of pure SFD bSA 15 % w/w: a+b) FD‐program #3: 300x and 1000x;c+d) FD‐program #6: 500x and 1000x.Since annealing at +5 °C does not have strong effects on particle shrinkage, thetemperature was further increased from 5 °C to 10 °C (#6). SEM pictures of the resulting

162 RESULTS AND DISCUSSIONSFD particles are again disappointing. Only a few spheres have a slightly waved surface,but no particle collapse is induced with higher annealing temperature.The annealing concept of Sonner et al [2002] seems therefore to fail for the generationof shrunken pure protein particles. The protein shows an excessive anti‐plasticizingeffect that may not allow assessable viscous flow at annealing temperatures duringlyophilization. The next set of experiments was executed according to 5.5.1 (Table 38).Primary drying temperature was set to ‐15 °C and secondary drying to +25 °C. SEMpictures of SFD powders generated with FD‐cycles #7 ‐ #11 do not show any collapse ofthe particle surface. Figure 128 demonstrates that at a 1° drying temperature of at least+5 °C the outer shells of the spheres are smooth and pervaded with fewer but largerpores. No evidence of particle alteration is seen in the particle inside. SFD powderswhich were ground in a LN 2 filled mortar did not reveal a collapsed inner structure, but asponge‐like texture with pore diameters of several μm (Figure 128).Step Onset [°C] Endset [°C] Hold[min.]Ramp[°C/min.]Vacuum[mbar]Segmenttime [min.]Total time[min.]1 ‐45 ‐45 30 equilibrate ‐‐‐‐ 30 302 ‐45 ‐15 (‐10, #8)3 ‐15 (‐10, #8)(‐5, #9)(+5, #10)(+8, #11)4 ‐15 (‐10, #8)(‐5, #9)(+5, #10)(+8, #11)(‐5, #9)(+5, #10)(+8, #11)‐15 (‐10, #8)(‐5, #9)(+5, #10)(+8, #11)‐‐‐‐ 1.0 (1.17, #8)(1.33, #9)(1.67, #10)(1.77, #11)0.100 (R) 30 60240 equilibrate 0.100 (H) 240 300+25 ‐‐‐‐ 0.89 (0.77, #8)(0.67, #9)(0.44, #10)(0.38, #11)0.040 (R) 45 3455 +25 +25 600 equilibrate 0.040 (H) 600 945Table 38. FD‐program #7. Experiments performed with the Christ Delta 1‐24 KDfreeze‐dryer. The 1° drying temperature was varied to ‐10 °C (#8), ‐5 °C (#9), 5 °C (#10)and 8 °C (#11), all other adjustments have been retained unchanged.The SFD experiments of 15 % (w/w) bSA particles show that the particle morphologycannot be changed by an increase in primary drying temperature during lyophilization.While pure sugars or protein‐excipients mixtures tend to collapse (see 5.5.1), frozen

RESULTS AND DISCUSSION 163particles with pure protein are stable and do not exhibit shrinkage caused by viscousflow during FD. An explanation is that the product temperature of the frozen proteinparticles does not exceed the “critical temperature” during 1° drying. Sublimation at lowtemperature apparently cools the product too fast, so that no particle breakdown ispossible. Generally the shrinkage phenomenon is thought to be a result of the producttemperature at least approaching the glass transition temperature at some point duringthe drying process [Rambhatla et al. 2005].a) b)20 μm 20 μmc) d)20 μm 20 μmFigure 128. SFD bSA 15 % (w/w) particles were ground in a LN 2 ‐filled mortar toreveal particle´s inner structure: a) FD‐program #8 (1000x); b) FD‐program #9 (1000x);c) FD‐program #10 (1000x); d) FD‐program #11 (1000x).The literature T g´ value of ‐11 °C [Chang and Randall 1992] for bSA was measured with asalt‐free dialyzed solution. The determination of an accurate T g´ value of pure protein inthis work was unsuccessful, as no clear glass transition was detectable with the MettlerToledo DSC (Figure 129). So the T g´ of bSA in tris‐buffer is assumed to be decreased by its

164 RESULTS AND DISCUSSIONsalt components. The collapse temperature, T c of bSA was measured by Meister [2009]at ≈ ‐5 °C.The vacuum pump was started with increased temperature in the 1° drying step (Table39). Vacuum turn‐on at ‐15 °C leads to very smooth particles with less channels on theirsurfaces (Figure 130a). With a starting temperature of +5 °C the outer shells of thefrozen spheres flake off in irregular chips (Figure 130b + c).a)b)Figure 129.DSC‐measurement of 15 % w/w bSA (a) and bCA (b) in 50mM tris‐buffer.An additional equilibration step of 15 min at +5 °C in FD‐program #8 leads to completeproduct collapse with a complete loss of particle characteristics (Figure 130d). Settingthe 1° drying temperature to +7 °C and accelerating the vacuum ramp step causes thegeneration of few craters on the particle surface (Figure 131).

RESULTS AND DISCUSSION165a)50 μmb)200 μmc)d)20 μm50 μmFigure 130.SFD bSA 15% w/w particles: a) FD‐program#8 (1000x); b) FD‐program#7 (200x); c) FD‐program#7 (1000x); d) FD‐program #7 with 10 min equilibration at+5 °C (1000x).Step12Onset [°C]‐45‐45Endset [°C]‐45+5 (‐20, #13)( +7, #14)3 +5 (‐20, #13)( +7, #14)+5( +7, #14)456+5 (+7, #14)+5 (+7, #14)+25+5 (+7, #14)+25+25Hold[min.]30‐‐‐‐‐‐‐‐240‐‐‐‐600Ramp [°C/min.]Vacuum[mbar]Segment time[min.]equilibrate 1.11 (0.83, #13)1.73 (0.83, #14)0 (0.83, #13)equilibrate 0.44equilibrate ‐‐‐‐‐‐‐‐ 0.1000 (R)0.1000 (H)0.0400 (R)0.0400 (H)3045 (30, #13)(30, #14)15 ( 30, #3)24045600Table 39. FD‐program#12. #12 was modified toperformed with the Christ Delta 1‐24KD freeze‐dryer.#13 and#14. ExperimentsThedrying conditions seem to be too rough for selectivee particle shrinkage; probablycycle #9 gives not enough time to develop the structural alterations. The idea of an

166 RESULTS AND DISCUSSIONadditional cooling step to ‐20 °C and equilibration for 2 h between step 2 and 3 in FDprogram#12 does not, however, lead to any morphology improvements (Figure 131b).In fact there are rather more disruptions and craters in the particles´ surface (Figure131).a) b)20 μm 20 μmFigure 131. SEM pictures of particles produced with FD‐cycle: a) #14 (1000x);b) Modified #12 with an additionally cooling step to ‐20 °C and subsequentequilibration of 2h (attached between step 2 and 3, 1000x).Step Onset [°C] Endset [°C] Hold[°min.]Ramp[°C/min.]Vacuum[mbar]Segmenttime [min.]1 ‐45 ‐45 30 equilibrate ‐‐‐‐ 302 ‐45 ‐15 (‐10, #16)3 ‐15 (‐10, #16)(‐8, #17)(‐5, #18)4 ‐15 (‐10, #16)(‐8, #17)(‐5, #18)5 ‐15 (‐10, #16)(‐8, #17)(‐5, #18)6 ‐15 (‐10, #16)(‐8, #17)(‐5, #18)(‐8, #17)(‐5, #18)‐15 (‐10, #16)(‐8, #17)(‐5, #18)‐15 (‐10, #16)(‐8, #17)(‐5, #18)‐15 (‐10, #16)(‐8, #17)(‐5, #18)‐‐‐‐ 1.50 (1.75, #16)(1.85, #17)(2, #18)‐‐‐‐ 201800 equilibrate ‐‐‐‐ 180015 equilibrate 0.100 (R) 15900 equilibrate 0.100 (H) 900+25 ‐‐‐‐ 0.88 (0.77, #16)(0.73, #17)(0.66, #18)0.040 (R) 457 +25 +25 600 equilibrate 0.040 (H) 600Table 40. FD‐program #15 and variations. Experiments performed with the ChristDelta 1‐24 KD freeze‐dryer. The 1° drying temperature was varied to ‐10 °C (#16),‐8 °C (#17) and ‐5 °C (#18).

RESULTS AND DISCUSSION 167Levi and Karel [1995] who observed volumetric shrinkage problems with FD excipientsprocessed above their T g´ values, concluded that the collapse temperature, T c , stronglydepends on the duration of the experiment. While T g´ is independent of time, collapsecan only be induced when a critical viscosity in the range of 10 5 – 10 8 Pa*s is reached[King 1975]. Due to the high anti‐plasticizing effect of proteins [Rochelle and Lee 2007]and the failed shrinkage experiments, a decrease in product viscosity is assumed here totake much more time than a few hours.The lyophilization program of the next experimental series is a modification ofFD‐cycle #2 (Table 40). The second equilibration time and the primary dryingtemperature are now considerably increased. Since the formulation´s T g´ is not known,the 1° drying temperature is elevated from ‐15 °C to ‐5 °C. The temperature ramp timesof the modified FD‐cycles are taken from #15, so every increase in 1° drying temperatureis followed by an increase in temperature ramp time in 1° and a decrease in 2° drying.a) b)50 μm 20 μmc) d)50 μmFigure 132. SEM pictures of SFD bSA 15 % (w/w) with: a, b) FD‐cycle #15: 1° at‐15 °C (500x, 1000x); c, d) FD‐cycle #16: 1° at ‐10 °C20 μm

168 RESULTS AND DISCUSSIONThe additional time for the FD‐process leads to more structural alterations than shownbefore. Small SFD particles obtained with FD‐cycle #15 are partially shrunken, whereaslarger ones still remain as smooth spheres (Figure 132 a+b). Changing the 1° dryingtemperature to ‐10 °C causes an inconsistently collapsed product. Also large particleshave a rugged and wrinkled surface (Figure 132 c+d). Another temperature increase to‐8 °C shows completely shrunken raisin‐like particles (Figure 133a + b). The free‐flowingwhite SFD powder was easy to empty out from the aluminium pan. Elevating thetemperature to ‐5 °C increased the collapse of the sphere structure (Figure 133c + d).a) b)100 μm 20 μmc)200 μmFigure 133. SEM of SFD bSA 15 % (w/w) with: a, b) FD‐cycle #17: 1° at ‐8 °C(300x, 1000x); c, d) FD‐cycle #18: 1° at ‐5 °C (100x, 500°C).d)50 μmDue to superficial thawing, the formerly wrinkled surface has now become smooth andthe particles are molten together as large aggregates. The brown‐white colouredpowder stickes to the aluminium bowl and could only be transferred in a Saarsted‐tubeby scraping with a spatula. Hg porosimetry measurements of unaltered and shrunkenbSA SFD particles confirm the results of the SEM interpretations. Undamaged roundspheres obtained with the standard FD‐cycle show two main peaks in their pore size

RESULTS AND DISCUSSION 169distribution, namely the intra‐ and the interparticular spaces (Figure 134). Theinterparticular spaces are those between different compressed powder particles,whereas intraparticular spacings can be seen as facile pores of the non‐collapsedspheres. Using a model calculation for cylindrical pores, an average pore diameter of0.84 μm is measured (Figure 134, Table 41) for sponge‐like powder particles.a) b)c) d)Figure 134. a) Hg porosimetry of SFD bSA generated with FD‐cycle #2 (Table 40)given as pore frequency (histogram) and cumulative pore frequency (solid line);b) Intramolecular pore volume in detail; c) Partially shrunken SFD powder generatedwith #16; d) Shrunken SFD powder generated with #17.The average pore diameter of particles manufactured from FD‐cycle #16 and #17 arehigher, indicating that there are fewer pores after particle collapse. With incipientcollapse at #16 the SSA obtained by BET measurements (Table 41) decreased by up toone third compared to the non‐shrunken bSA powder of #2. BET results are consistent

170 RESULTS AND DISCUSSIONwith Hg porosimetry analysis, where the absences of small pores for particles of FD‐cycle#17 show the lowest value for cumulative volume and the SSA. Thus, the interpretationsof the SEM pictures are confirmed. Collapsed and shrunken SFD particles show ameasureable decrease in SSA.It is concluded that modification of the lyophilization procedure leads to uniform particleshrinkage of SFD pure bSA particles. Because of the strong anti‐plasticising effect of theprotein the addition of a standard annealing step does not cause particle collapse in anycase. Even a remarkable increase in 1° drying temperature did not result in particlecollapse as was seen in chapter 5.5.1. A longer equilibration and 1° drying timecombined with a temperature near the formulation´s T g´ can, however, improve the lowdensity matrix of the spray frozen particles during lyophilization.SFD bSA obtained withFD‐cycle no.:S BET [m 2 /g] Ø pore diameter [μm] Calculated cumulativevolume [cm 3 /g]#2 26.41 ± 0.29 0.84 6.36#16 9.46 ± 0.21 8.58 2.27#17 8.27 ± 0.19 41.32 1.17Table 41.BET and Hg porosimetry results of different SFD bSA powders.

RESULTS AND DISCUSSION 1715.5.2.2 Collapsed bCA particlesThe FD procedure used for the shrunken bSA particles was now used for shrinkageexperiments of pure bCA particles (Table 42). Due to missing T g´ values in the literatureand no reliable DSC results at either high or low heating rates (data not shown), 4different 1° drying temperatures were chosen by trial and error. Instead of generatingcollapsed powder particles the first two FD‐cycle variations yielded a completely thawedand molten product (Figure 135 a+b). The powder particles did not consist of individualspherical particles but of large plates without any surface texture.Step Onset [°C] Endset [°C] Hold [min.] Ramp [°C/min.] Vacuum[mbar]Segment time[min.]1 ‐45 ‐45 30 equilibrate ‐‐‐‐ 302 ‐45 ‐5 (‐8, #20)3 ‐5 (‐8, #20)(‐12, #21)(‐14, #22)4 ‐5 (‐8, #20)(‐12, #21)(‐14, #22)5 ‐5 (‐8, #20)(‐12, #21)(‐14, #22)6 ‐5 (‐8, #20)(‐12, #21)(‐14, #22)(‐12, #21)(‐14, #22)‐5 (‐8, #20)(‐12, #21)(‐14, #22)‐5 (‐8, #20)(‐12, #21)(‐14, #22)‐5 (‐8, #20)(‐12, #21)(‐14, #22)‐‐‐‐ 2 (1.85, #20)(1.65, #21)(1.55, #22)‐‐‐‐ 201800 equilibrate ‐‐‐‐ 180015 equilibrate 0.100 (R) 15900 equilibrate 0.100 (H) 900+25 ‐‐‐‐ 0.66 (0.73, #20)(0.82, #21)(0.87, #22)0.040 (R) 457 +25 +25 600 equilibrate 0.040 (H) 600Table 42. FD‐program #19 and variations. Experiments performed with the ChristDelta 1‐24 KD freeze‐dryer. The 1° drying temperature was varied to ‐5 °C (#19),‐8 °C (#20), ‐12 °C (#21) and ‐14 °C (#22).SEM pictures show that small particles that were lyophilized with FD‐cycle #22 arealready shrunken, whereas larger ones still exhibit a spherical structure with large pores

172 RESULTS AND DISCUSSIONon their surfaces (Figure 135 c+d). An increase in 1° drying temperature by 2 °C led toformation of a free‐flowing powder with raisin‐like structure even for larger particles(Figure 135 e+f).a) b)200 μm 200 μmc) d)100 μm 20 μme)100 μmFigure 135. SEM pictures of SFD bCA 15 % w/: a) 1° at ‐5 °C 100x; b) 1° at ‐8 °C 100x;c, d) 1° at ‐14 °C (#22: 250x, 1000x); e, f) 1° at ‐12 °C 250x, 1000x.Contrary to annealing experiments with different proteins reported in the literature[Sonner C. 2002; Webb et al. 2003], where identical contents or even an increase innative‐like secondary structures were obtained, bCA integrity suffers during thef)20 μm

RESULTS AND DISCUSSION 173generation of collapsed spheres. SEC evaluation (Figure 136) illustrates that a bCAformulation obtained with recipe #21 had an increased dimer peak fraction compared toSFD protein prepared under moderate conditions (#1), with an enzymatic residualactivity of 65.8 % instead of 70.1 %.a)b)Figure 136. Detector signals of reconstituted shrunken SFD‐bCA samples analyzedwith SEC. 7.5 mg/mL samples, 60 μL injection volume. Molar mass was taken from theSFD bCA‐#9‐run from the RI‐signal. a) LS‐signals; b) UV‐signals.

174 RESULTS AND DISCUSSIONA change in 1° drying temperature up to ‐5 °C in a short FD‐cycle resulted in an activityloss of 48.2 % and a further increase in aggregates in SEC. Here the dimer peak is nowlarger than the monomer peak. Increased molecular mobility caused by increase indrying temperature near the T g´ may cause more problems of protein instability due topH shifts or phase separation. Mentionable is the high robustness of bCA againstcomplete denaturation. Even a completely collapsed and thawed protein formulationstill obtains 47 % of its unaltered residual activity.

CONCLUSIONS 1756 ConclusionsThis thesis deals with the preparation of protein‐loaded microparticles by spray freezedryingand the investigation of process‐induced structural alterations in proteinstructure using online light scattering. Furthermore the regulation of powder particlemorphology and the generation of dense SFD particles was a focus of this work.Spray freeze‐drying as a relatively new method for powder preparation is stillincompletely investigated in detail. The manufacturing process consists of twoconsecutively applied procedures. First the liquid feed is atomizated with an ultrasonicor two‐fluid nozzle above a cryogen (e.g. LN 2 ). The aluminium bowl filled with the “flashfrozen”protein particles has to be transferred onto pre‐cooled shelves of a freeze‐dryerto avoid particle thawing or melting after the LN 2 has boiled off. In the second processstep the water in the frozen powder particles is removed by sublimation in a freezedryingcycle at low temperature and pressure.The first part of this thesis introduces a three‐detector system including UV‐, RI and LSinstrumentsthat were connected to a SEC‐HPLC system. Today such detector systemsare well‐established in quality control divisions of pharmaceutical companies, due totheir versatility. The larger molecular weight of the analyzed substance, the higher is thesensitivity of the LS‐signal. Large aggregates of protein formulations can be monitoredwith LS even when there is no signal in the UV‐ or RI‐detector. The calibration andnormalization procedure of the LS‐detector was presented. A large advantage of LSdetectorsis the calculation of absolute molar masses of analyzed samples. It could beshown that molecular masses of proteins could be calculated more accurately with LSthan with a UV calibration curve. Furthermore this method is independent of the flowrate and the composition of the mobile phase. A further feature of the combination of aLS‐unit with concentration‐detectors is the simplified determination of refractive indexincrements, extinction and second virial coefficients compared to other applications. A 2determination with an SEC column proved its suitability and rapidness compared withthe batch‐method, although more precise results could have been obtained with a moresensitive RI‐detector. Auxilliary to SEC, AF4 was investigated as a further analyticalmethod for protein characterization. Appropriate separation conditions were developed,

176 CONCLUSIONSand an UV‐B lamp was used to replace toxic preservative agents. The germ‐killingefficiency was measured with UV‐ and LS‐experiments and finally confirmed withincubated agar plates.The second part of this thesis describes protein damage induced by a spray freeze dryingprocess; bovine carbonic anhydrase (bCA) was taken as model protein. Despite anunchanged product number for bCA, different qualities of protein batches weredelivered. Whereas a fresh prepared solution of an old batch showed a large content ofaggregates, the new batch (058K1031) was characterized by its improved integrity andincreased encymatic activity. Moreover, old bCA batches were, contrary to the newbatch, impurified with an unknown substance that formed a shoulder between themonomer and the dimer peak in the LS‐ and RI‐signal. The new batch must have beenmanufactured or purified with a different SOP. Denaturation experiments of bCA withGuHCl rapidly led to a reduced encyme activity of 30 %, whereas large aggregates andsmall protein fragments could be detected with SEC. After the heating of a bCA solutionto 50 °C it only took a few minutes until the protein precipitated, but no increase inaggregates could be detected with SEC. While storing on ice, the protein refolded to thenative state which resulted in a residual activity of 94 %. Experiments with bCA solutionsof 15 % (w/w) gave a comprehensive picture of protein alteration during the differentprocess steps of SFD. The atomization of the LF induced only low protein damage(‐6.5 %), whereas the preparation of high bCA concentrations (‐9 %) and most notablythe drying step led to a large activity loss (‐27 %). The addition of Tween 80 ® to the LFshowed a moderate increase in the enzymatical performance (+4 %), whereas a doubledatomization frequency (120 kHz) had negligible negative influence (‐4 %). Noticeable wasthe recovery in enzymatic activity after atomization into LN 2 (+3.2 %), that also has beenreported in the literature for other proteins [Sonner 2002]. The atomization tests wererepeated with diluted bCA LFs at different Watt adjustments. Contrary to bCAnebulization into air, atomization into LN 2 showed an increase in the monomer peak anda decrease in the dimer concentration. Previously generated bCA aggregates evidentlycould at least partly divide into monomers and further refold back into their native state.SFD experiments with varying protein contents have shown that a minimum solidscontent of 5 % (w/w) is needed to obtain spherical powder particles. Noticeable was a

CONCLUSIONS 177distinct increase in protein damage between 10 – 15 % (w/w), which resulted in adoubled enzyme activity loss.SD of bCA showed similar results compared to SFD experiments. There was a largedecrease in protein activity observed during the drying process (‐21 %), whereas theatomization of the LF with a two‐fluid nozzle induced only marginal damage (‐7 %). Bothdrying methods, SFD and SD, showed comparable alteration in bCA structure. Thepowder samples obtained showed consistently increased dimer and trimer fractions andhad a residual activity of 70 %. Repeated SFD experiments wit the new batch (058K1031)led to altered results. The redispersed SFD powder showed only little change in its peakcomposition or turbidity compared to the LF, whereas the enzyme activity was still 94 %.Variable results for different batches of the same product numbers are not exceptionalwhile working with proteins, but such strong deviations are notable.The attempt to stabilize the proteins during SFD with excipients was successful. First,simple blends with bCA and sugars/sugar alcohols were examined. Trehalose proved itsoutstanding suitability, a protein‐excipient mixture of only 5:1 was sufficient to increasethe residual activity from 70 to 95 %. Higher mixing ratios of 1:1 or 1:3 were needed toobtain comparable results with LMH, sucrose or mannitol. SFD powders of complexexcipient mixtures with dextran (10 kDa) showed lower enzyme activities than withoutdextran, but the formulations obtained higher T g ‐values due to the polymer addition.SFD experiments with bCA‐excipient ratios of 1:2 at solid contents of 30 % (w/w) led togood results, most notably for trehalose and sucrose blends (96.3 and 97.9 %,respectively). However, bCA mixtures with trehalose, mannitol and dextran (10 kDa)could only be stabilized to 92 %. Residual activity measurements were consistent withturbidity‐ and SEC/AF4 measurements, high turbidity values and low monomer contentswere an indication for increased protein damage.Because of the low recovery of 70 – 80 % for highly stressed samples in SEC compared tostabilized SFD samples, the mobile phase had to be further modified. The addition of25 % (v/v) glycerol could solve this problem and increased the values up to ~ 95 %. Adirect comparison of the SEC‐ and the AF4‐method was difficult, since non‐stabilized SFDsamples exhibited increased aggregate concentrations in AF4 compared to SEC.

178 CONCLUSIONSLactic dehydrogenase (LDH) was taken as second model protein for observation of theimpact of different process steps of SFD. LFs of 1 % (w/w) were prepared for theseexperiments. An increase in protein degradation during the drying process was expectedbecause of the more complex molecule structure of LDH (quartary structure) comparedwith bCA. The high sensitivity of LDH to denaturation during atomization mentioned inthe literature [Adler and Lee 1999] was confirmed, with a decrease of almost 38 % in theprotein activity. LDH also showed a marginal recovery in residual activity after thesubsequent quench‐cooling in LN 2 (+3.3 %). The following drying step in the FD led tofurther protein damage of 24.8 %; the produced LDH powder finally exhibited only 41 %of its original activity. Although the sample turbidity was higher after the process steps,only a small increase in dimer fraction could be measured compared to the LF (+ 1.25 %).However, the high activity loss is not explicable due to higher protein aggregation orfragmentation. Stabilization experiments were performed with single trehalosemixtures. SFD of a protein‐sugar mixture in a 1:20 ratio showed an activity improvementof +51.5 % (92.5 %). Additionally mixed surfactant in an adequate concentration mighthave further stabilizing effects, but experiments of that kind were not executed in thiswork.Powders that are suitable for EPI have to exhibit a specific particle shape andsize, and also have to possess a high density. This last demand is quite challenging. Goldparticles with adsorbed antigens were recently described for the needle‐free injectionapplication in the literature [Chen et al. 2001]. But the adoption of powder particlesmade of organic, and thus biodegradable materials (such as sugars or polymers), isdifficult due to their low density. Sonner et al. [2002] primarily had success in thegeneration of densified, shrunken powder particles with an additional annealing step.The last part of this thesis therefore discusses controlled modification of the particlemorphology during SFD. BSA and bCA were taken as model proteins. At first it could beshown that the particle structure could be changed by selective excipients addition. SFDexperiments were performed with different bSA‐trehalose and bSA‐sucrose mixtures of15 % w/w. After atomization of the LFs, all formulations were dried with the sameaggressive FD‐cycle. Alterations in the particle morphology were induced by thedifferent additives and the varying protein‐excipient ratios. Despite the short dryingtime pure protein powder could be dried to a residual water content of 1.6 %, the

CONCLUSIONS 179powder particles were spherical and had small pores on their surface. As the excipientaddition leads to a decrease in the formulations´ T g´‐ values, already smalltrehalose/sucrose additions resulted in changed particle morphologies. At firstsuperficial thawing induced the loss of pore structure, whereas higher excipientconcentrations were responsible for the generation of shrunken protein particles. Dueto the missing protein fraction, pure trehalose/sucrose particles completely collapsedunder these rough drying conditions and melted together.The emerging consideration was to manufacture such collapsed powder particles out ofpure protein solutions. Experimental series with bSA and bCA were dissolved in 50 mMtris‐buffer (pH 7.50) to 15 % (w/w). Annealing experiments with bSA solutions accordingto Sonner et al. [2002] failed, because no visual structural alterations of the powderparticles were detectable. Even high annealing temperatures (+5 °C/+10 °C) andelevated annealing duration (6h) did not lead to shrunken particles. A further increase ofthe primary drying temperature was ineffective, just as the short‐term thawing andrefreezing of the protein particles. Only an extended drying cycle at a primary dryingtemperature above the T g´ value could finally manufacture collapsed, shrunken powderparticles. Hg porosimetry and BET‐measurements proved the impressions of the SEMpictures. Compared to SFD bSA powders from previous short FD‐cycles, the SFD particlesof the modified FD‐programm showed no small pores on their surface and evenexhibited only 1 / 3 of their original specific surface area. For the preparation of shrunkenbCA particles, the 1° drying temperature had to be decreased by 4 °C. Residual activitymeasurements of redispersed bCA particles indicated that the modification of the FDcyclecaused additional protein damage that led to a further activity loss of 4.3 %.A new three‐detector system could be established for this work, and its versatileapplications could be presented. Prospective A 2 ‐experiments might be easier andprobably would lead to more accurate results if a modern HPLC system was connectedwith the high end detectors. SFD and SD experiments with bCA and LDH could give abetter insight in process‐induced alterations in protein structure. Furthermore, differentexcipients were tested for their protein‐stabilizing efficiency. All analytical methods thatwere used to characterize the SFD particles showed deviating results of SEC and AF4runs.

180 ZUSAMMENFASSUNGConducted modifications of the FD‐cycle could be used to manufacture collapsed,densified protein powder particles. Possible applications and the suitability of suchpowders have to be shown in prospective studies.7 ZusammenfassungDiese Dissertation behandelt die Herstellung von proteinhaltigen Mikropartikeln mittelsSprühgefriertrocknung und die Untersuchung von prozeßbedingten Schädigungen derProteinstruktur mit Hilfe der Lichtstreuung. Ein zweiter Schwerpunkt dieser Arbeit ist dieSteuerung der Partikelmorphologie solcher Pulver, um Pulverteilchen mit möglichsthohen Dichten zu erzielen. Die Sprühgefriertrocknung ist ein relativ neuesHerstellungsverfahren für Pulver, das bis heute noch nicht umfassend untersuchtworden ist. Die SFD ist in zwei aufeinander folgende Teilschritte untergliedert. Zuerstwird eine Proteinlösung mit einer Ultraschall‐ oder Zweistoffdüse über einem flüssigenKühlmittel zerstäubt (z.B. flüssiger Stickstoff). Um das Antauen oder Schmelzen derPartikel nach dem Abdampfen des flüssigen Stickstoffes zu vermeiden, muss dieAluminiumschale mit den „schockgefrorenen“ runden Proteinpartikeln auf vorgekühlteStellflächen eines Gefriertrockners überführt werden. Im zweiten Prozessschritt, derGefriertrocknung, wird durch niedrige Temperaturen und Druck das Eis der gefrorenenPulverpartikel durch Sublimation getrocknet.Im ersten Teil dieser Arbeit wird das verwendete Drei‐Detektoren‐System vorgestelltwelches mit einer HPLC‐SEC Einheit verbunden ist und aus einem UV‐, einemBrechungsindex‐ und einem Lichtstreudetektor besteht. Heutzutage sind solcheDetektorsysteme aufgrund ihrer Vielseitigkeit in der Qualitätskontrolle weit verbreitet.Die Sensitivität des LS‐Signals ist umso höher, je größer die zu analysierenden Substanzist. Somit können Aggregate aus Proteinformulierungen detektiert werden, die miteinem UV‐ oder RI‐Detektor aufgrund ihrer geringen Konzentration nicht erkennbar sind.Die Kalibrierung und Normalisierung des LS‐Detektors wurden näher vorgestellt. Eingroßer Vorteil von Lichtstreudetektoren ist die Berechnung der absoluten Masse der zuuntersuchenden Proben. Es konnte gezeigt werden, dass Molekulargewichte vonProteinen durch LS genauer errechnet werden können als es mit einer Kalibriergerademöglich wäre. Darüberhinaus ist das Verfahren unabhängig von der Flussrate und

ZUSAMMENFASSUNG 181Zusammensetzung der mobilen Phase. Eine weitere Besonderheit des LS‐Detektors inKombination mit Konzentrationsdetektoren ist die im Gegensatz zu anderen Methodenvergleichsweise einfache Bestimmung von Brechungsindexinkrementen, Absorptionskoeffizientenund sekundären Virialkoeffizienten. Die A 2 ‐Messung mittels SEC‐Säuleerwies sich im Gegensatz zur Batch‐Methode als äußerst praktikabel und zeitsparend,obwohl offensichtlich wurde, dass genauere Ergebnisse mit einem sensitiveren RI‐Detektor erzielt werden könnten. Neben der SEC wurde die AF4 als weiteresanalytisches Verfahren zur Proteincharakterisierung etabliert. Es wurde eine geeigneteTrennmethode entwickelt wobei eine UV‐B Lampe zum Einsatz kam, die den Gebrauchvon toxischen Konservierungsmitteln überflüssig machte. Die keimabtötende Wirkungder UV‐B Strahlung wurde mit Hilfe von LS‐ und UV‐Versuchen untersucht und mitbebrüteten Nährmedien bestätigt.Im zweiten Teil der Arbeit wurde näher auf die Proteinschädigung bei einem SFD Prozeßeingegangen, dabei dienten bCA als erstes Modelproteine. Trotz unveränderterBestellnummer hatten sich die Enzymchargen bei bCA stark unterschieden. Währendfrisch hergestellte Lösungen alter Chargen große Mengen an Aggregaten aufgewiesenhaben, zeichnete sich die neue Lieferung (058K1031) durch seine Unversehrtheit undhoher Aktivität aus. Alte Proteinchargen waren darüberhinaus mit einer fremdenSubstanz verunreingt. Diese erzeugte eine Peakschulter zwischen dem Monomer‐ unddem Dimerpeak im LS‐ und RI‐Signals, die bei bCA Proben der neuen Lieferung fehlte.Aufgrund dieser Unterschiede kann angenommen werden, dass die verschiedenen bCACharge auf eine andere Art und Weise hergestellt oder aufgereinigt worden sind.Denaturierungsversuche von bCA mit GuHCl führten schon nach kurzer Zeit zu einerSenkung der Aktivität auf 30 %, wobei große Aggregate und Teilfragmente mittels SECnachgewiesen werden konnten. Bei der Erwärmung einer bCA Lösung auf 50 °C kam esnach wenigen Minuten zur Präzipitation, dennoch konnten keine Aggregate in der SECnachgewiesen werden. Das Enzym konnte sich bei der Lagerung auf Eis in die nativeForm zurückfalten, was eine Restaktivität von 94 % ergab.Versuche mit 15 %igen bCA Lösungen ergaben ein umfassendes Bild über dieProteinschädigung bei den unterschiedlichen Prozessschritten der SFD. Während dieZerstäubung der Lösung nur geringfügigen Enzymschädigung hervorrief (‐6.5 %), kommt

182 ZUSAMMENFASSUNGes beim Lösen hoher Konzentrationen (‐9 %) und v.a. der Trocknung zu beträchtlichenAktivitätsverlusten (‐27 %). Die Zugabe des Tensids Tween 80 ® zur LF ergab eine mäßigeVerbesserung der Enzymleistung (+4 %), während eine Verdoppelung der Zerstäubungsfrequenz(120 kHz) einen ebenfalls vernachlässigbaren negativen Einfluss hatte (‐4 %).Auffallend nach der Vernebelung der Proteinlösung in flüssigem Stickstoff war einemessbare Rückgewinnung der Enzymaktivität (+3.2 %), die in der Literatur auch fürandere Proteine beschrieben ist [Sonner 2002]. Der Versuch wurde beiunterschiedlichen Wattzahlen mit verdünnten bCA Lösungen wiederholt. Im Vergleichzur Vernebelung von bCA in Luft war beim Zerstäuben in LN 2 ein deutlicher Zuwachs desMonomerpeaks in der SEC erkennbar, während der Dimergehalt sank. Aus unbekanntenGründen könnten sich zumindest teilweise zuvor gebildete Aggregate der bCA beimschnellen Einfrieren in Monomere teilen und in die native Form zurückfalten.SFD Versuche mit unterschiedlichem Proteingehalt haben gezeigt, dass einFeststoffgehalt von mindestens 5 % (m/m) nötig ist um runde Pulverpartikelherzustellen. Auffallend war ein deutlicher Sprung in der Proteinschädigung zwischen 10‐ 15 % (m/m), die eine Verdopplung der Enzymschädigung zur Folge hatte.Bei der SD von bCA kam es zu ähnlichen Ergebnissen wie bei der SFD. Auch hier führtedie Trocknung des Proteins zu hohen Aktivitätseinbußen (‐21 %), wobei die Zerstäubungder Lösung mit einer Zweistoffdüse geringeren Schaden verursachte (‐7 %). Die beidenunterschiedlichen Trocknungsmethoden, SFD und SD, zeigten vergleichbareVeränderungen der Proteinstruktur der bCA. Die Pulver wiesen einen ähnlichen Zuwachsan Di‐ und Trimeren auf und hatten Restaktivitäten von etwa 70 %. Zu erwähnen ist dieTatsache, dass die Wiederholung der SFD Versuche mit der neuen bCA Charge(058K1031) vollkommen andere Resultate lieferte. Das redispergierte SFD Pulver wies imVergleich zur LF nur geringe Änderungen in der Peakzusammensetzung und der Trübheitauf, während die Restaktivität noch bei 94 % lag. Schwankende Ergebnisse fürunterschiedliche Chargen desselben Produkts sind beim Arbeiten mit Proteinen nichtsUngewöhnliches, diese starken Abweichungen sind jedoch bemerkenswert.Versuche die Proteinschädigung bei der SFD durch die Zugabe von Hilfsstoffen zuverbesseren waren erfolgreich. Zuerst wurden Einzelmischungen vonZucker/Zuckeralkoholen mit bCA untersucht. Trehalose eignete sich besonders gut zurStabilisierung, ein Protein‐Hilfsstoff Verhältnis von 5:1 war ausreichend um die

ZUSAMMENFASSUNG 183Restaktivität von 70 auf 95 % zu erhöhen. Mischungsverhältnisse von 1:1 bzw 1:3 warennötig um ähnliche Ergebnisse mit LMH, Saccharose und Mannitol zu erzielen. SFD Pulveraus komplexe Hilfsstoffmischungen mit Dextran (10 kDa) zeigten zwar eine geringereEnzymaktivität als ohne Dextran, dafür hatten die Formulierungen aufgrund derPolymerzugabe einen erhöhten T g Wert. SFD‐Versuche mit bCA und der doppeltenMenge an Hilfsstoff bei einem Feststoffgehalt von 30 % (m/m) lieferten v.a. mitTrehalose und Saccharose gute Ergebnisse (96.3 und 97.9 %). Eine Mischung desProteins mit Trehalose, Mannitol und Dextran (10 kDa) im Verhältnis 3:3:4 konntehingegen nur auf 92 % stabilisiert werden. Alle Aktivitätsuntersuchungen standen inEinklang mit Trübheits‐ und SEC/AF4 Messungen, wobei hohe Trübheitswerte und einniedrigerer Monomergehalt ein Indiz für die Enzymschädigung darstellte. Der Vergleichder beiden Auftrennungsmethoden lieferte für stabilisierte Proben ähnliche Ergebnisse.Da die Wiederfindung für denaturierte Proben in der SEC nur etwa 70 – 80 % betrug,musste die mobile Phase verändert werden. Durch den Zusatz von 25 % (v/v) Glycerolkonnte dieses Problem erfolgreich gelöst werden und der Wert verbesserte sich auf~ 95 %. Bei gestressten (unstabilisierten SFD‐) Proben lieferte die AF4‐Methode imGegensatz zur SEC höhere Signale für Aggregate, was einen direkten Vergleich schwierigmachte.LDH diente als zweites Modellprotein für die Betrachtung der unterschiedlichenProzessschritte bei der SFD. Anhand 1 % iger Lösungen wurden die verschiedenenZwischenschritte betrachtet. Durch den komplexeren Molekülaufbau von LDH(Quartärstruktur) im Vergleich zur bCA wurde eine stärkere Veränderung derProteinstruktur nach der Trocknung erwartet. Die in der Literatur erwähnteDenaturierungsanfälligkeit während der Zerstäubung [Adler and Lee 1999] wurde durcheine fast 38 % ige Senkung der Proteinaktivität bestätigt. Auch bei LDH kam es beimanschließenden Einfrieren in LN 2 zu einer minimalen Aktivitätszuwachs (+3.3 %). Derdarauffolgende Trocknungsschritt im Gefriertrockner hatte weitere Einbußen von 24.8 %zur Folge, wobei das fertige Pulver nur noch 41 % der Ursprungsaktivität besaß. Obwohlsich die Trübheit der LDH Lösungen bei den jeweiligen Zwischenschritten unterschiedenhaben, konnte mittels SEC nur ein geringer Zuwachs im Dimergehalt im Vergleich zur LFgemessen werden (+ 1.25 %). Der Aktivitätsverlust ist somit nicht über eine mögliche

184 ZUSAMMENFASSUNGAggregation oder Fragmentierung des Enzyms erklärbar. Stabilisierungsversuche vonLDH wurde mit Trehalose durchgeführt. Die SFD einer Protein‐Zucker Mischung imVerhältnis von 1:20 führte zu einer Verbesserung der Restaktivität um +51.5 % (92.5 %).Durch den Zusatz eines Tensides in geeigneter Konzentration wäre eine weitereStabilisierung von LDH denkbar, Versuche dieser Art wurden jedoch nicht für dieseArbeit vorgenommen.Pulver die für die EPI eingesetzt werden müssen eine bestimmte Partikelform und‐größe, aber auch eine hohe Dichte aufweisen. Gerade die letztgenannte Anforderung istbis heute eine große Herausforderung. Antigenbehaftete Goldpartikel sind bereits fürdie nadelfreien Injektion in der Literatur beschrieben [Chen et al. 2001]. Doch derEinsatz von Pulverpartikel aus organischen und somit abbaubaren Materialien(z.B. Zuckern, Polymeren) ist aufgrund ihrer geringen Dichte und Festigkeit für diesenZweck immer noch fraglich. Sonner et al. [2002] gelang es, SFD Pulver durch einenzusätzlichen Annealingschritt zum Schrumpfen zu bringen und dadurch zu verdichten.Der letzte Abschnitt dieser Abeit befasst sich näher mit der Veränderung derPartikelform bei der SFD. Als Modellproteine dienten bSA und bCA. Zuerst konntegezeigt werden, dass die Struktur der Pulverteilchen durch gezielte Hilfsstoffzugabeverändert werden konnte. SFD Versuche wurden mit 15 %igen bSA‐Trehalose und bSA‐Saccharose Mischungen durchgeführt. Nach dem Vernebeln der Proteinlösungenwurden sämtliche Formulierungen mit demselben aggressiven Gefriertrocknungsprogrammgetrocknet. Änderungen der Partikelmorphologie lagen somit nur an denunterschiedlichen Hilfsstoffen und dem variablen Protein‐Hilfsstoff Verhältnis. ReineProteinpulver konnten trotz der kurzen Trocknungszeit bis auf einen Restwassergehaltvon 1.6 % getrocknet werden, die Pulverteilchen waren rund und wiesen kleine Porenauf der Oberfläche auf. Da die Zugabe der verwendeten Hilfsstoffe eine Senkung desT g´‐Wertes der Formulierungen zur Folge hatte, kam es schon bei niedrigenTrehalose/Saccharose‐Zugaben zu Veränderungen der Partikelmorphologie. Zuerst kames nur zu einem oberflächlichen Tauen und Verlust der kleinen Poren, bei erhöhtenHilfsstoffzugaben entstanden geschrumpfte Proteinpartikel. Reine Hilfsstoffpartikelschmolzen unter diesen Trocknungsbedingungen zusammen und kollabierten aufgrunddes fehlenden Proteinanteils vollständig.

ZUSAMMENFASSUNG 185Die Frage war nun, ob es möglich ist, auch aus reinen Proteinlösungen kollabiertePulverpartikel herzustellen. Für diese Versuchsreihe wurden 15 %ige (m/m) bSA‐ bzw.bCA‐Lösungen verwendet, die mit 50 mM Tris‐Puffer hergestellt wurden.Annealingversuche mit bSA, ähnlich wie sie bei Sonner et al. [2002] vorgestellt wurden,zeigten keine sichtbare Veränderung der Pulverpartikel. Selbst bei hohen Temperaturen(+5 °C/+10 °C) und langer Dauer (6h) des Annealings kam es zu keiner Schrumpfung derPulverteilchen. Eine weitere Veränderung der Primärtrocknungstemperatur war ebensoerfolglos wie das kurzzeitige Antauen und Wiedereinfrieren der Proteinpartikel. Erst einsehr langer Trocknungszyklus bei einer Primärtrocknungstemperatur über derangenommenem T g´ erzeugte kollabierte Partikel. Hg Porosimetrie und BET‐Messungenbestätigten die Ergebnisse der REM‐Aufnahmen. Im Vergleich zu bSA Pulvern ausvorhergehenden Versuchen mit kurzem Gefriertrocknungsprogramm hatten die bSAPartikel nach dem modifizierten FD‐Zyklus weniger/keine Poren und nur etwa ein Drittelder ursprünglichen Gesamtoberfläche. Für die Herstellung von kollabierten bCAPartikeln musste die Trocknungstemperatur im Gegensatz zu bSA gesenkt werden.Aktivitätsmessungen der redispergierten, geschrumpften bCA Partikel zeigten, dass esdurch die Veränderung des Trocknungszykluses zu weiterer Proteinschädigung kommt,die sich in einem Aktivitätsverlust von weiteren ‐4.3 % ausdrückt.Zusammenfassend war es möglich, ein neues Dreidetektorsystem am Lehrstuhl zuetablieren und in seiner Vielseitigkeit vorzustellen. In Kombination mit einermoderneren HPLC‐Anlage könnten jedoch zukünftig (A 2 ‐) Messung einfacher undgenauer durchgeführt werden. SFD und SD Versuche mit bCA und LDH konnten einenbesseren Einblick in die prozeßinduzierten Schädigungen von Proteinen geben.Darüberhinaus wurden verschiedene Hilfsstoffe auf ihre proteinschützende Wirkung hingetestet. Alle analytischen Methoden, die für die Charakterisierung der SFD Pulververwendet wurden haben sich als aussagekräftig erwiesen, abweichende Ergebnisse vonSEC‐ und AF4‐Läufen wurden ausführlich diskutiert. Durch Veränderungen desTrocknungsprogramms in der Lyophilisation konnten kollabierte, verdichteteProteinpartikel hergestellt werden. Die Einsatzmöglichkeiten und Alltagstauglichkeitdieser Pulver werden sich jedoch erst mit zukünftigen Untersuchungen zeigen.

186 REFERENCES8 References[Adler 1999][Adler and Lee 1999][Arakawa et al. 2010]Adler, M. (1999). Sprüheinbettung von Proteinen inGerüstbildner: Stabilität und Oberflächenanalyse.Division of Pharmaceutics. Friedrich‐AlexanderUniversity, Erlangen Germany. PhD. ThesisAdler, M. and Lee, G. (1999). Stability and surfaceactivity of lactate dehydrogenase in spray driedtrehalose. J. of Pharm Sci 88, (2): p. 199‐208.Arakawa, T., Ejima, D. et al. (2010). The critical roleof mobile phase composition in size exclusionchromatography of protein pharmaceuticals. J. ofPharm Sci 99, (4): p. 1674‐1692.[Ball and Ramsden 1998] Ball, V. and Ramsden, J. J. (1998). Bufferdependence of refractive index increments ofprotein solutions. Biopolymers 46: p. 489‐492.[Banchereau and Steinman 1998] Banchereau, J. and Steinman, R. M. (1998).Dendritic cells and the control of immunity. Nature392, (6673): p. 245‐252.[Banga 1995][Barth and Zscherp 2002][Bauer et al. 2006].Banga, A. K. (1995). Therapeutic peptides andproteins: formulation, processing and deliverysystems. Technomic Publication Company,Lancaster, US, p. 42‐ 43.Barth, A. and Zscherp, C. (2002). What vibrationstell us about proteins. Q Rev Biophys 35, (4): p. 369‐430.Bauer, K. H., Frömming K.‐H. et al. (2006). Lehrbuchder Pharmazeutischen Technologie ‐ Verfahren undGrundoperationen. Stuttgart, WVG, p. 129.

REFERENCES 187[Baynes et al. 2005][Berg et al. 2007]Baynes, B. M., Wang, D. I. C. et al. (2005). Role ofarginine in the stabilization of proteins againstaggregation. Biochemistry 44, (12): p. 4919‐4925.Berg, J. M., Tymoczko, J. L. et al. (2007]. StryerBiochemie. Elsevier GmbH, Spektrum AkademischerVerlag, Heidelberg, München, Germany[Bhatnagar et al. 2008][Bhugra et al. 2007][Bhugra et al. 2006][Blum 1987][Boehncke and Radeke 2007].[Brice and Halwer 1951]Bhatnagar, B. S., Pikal, M. J. et al. (2008). Study ofthe individual contributions of ice formation andfreeze‐concentration on isothermal stability oflactate dehydrogenase during freezing. J. of PharmSci. 97, (2): p. 798‐814.Bhugra, C., Rambhatla, S. et al. (2007). Prediction ofthe onset of crystallization of amorphous sucrosebelow the calorimetric glass transitiontemperature from correlations with mobility. J. ofPharm Sci 96, (5): p. 1258‐1269.Bhugra, C., Shmeis, R. et al. (2006). Predictions ofonset of crystallization from experimentalrelaxation times I‐correlation of molecular mobilityfrom temperatures above the glass transition totemperatures below the glass transition. PharmRes 23, (10): p. 2277‐2310.Blum, H. (1987). Improved silver staining of plantproteins, RNA and DNA in polyacrylamide gels.Electrophoresis 8: p. 93‐99.Boehncke, W.‐H. and Radeke, H. H. (2007). Biologicsin General Medicine, Chapter 1: Introduction;Definition and Classification of Biologics, p. 1‐2,Springer, Berlin, Heidelberg.Brice, B. A. and Halwer, M. A. (1951). A differentialrefractometer. J. of Opt Soc Am 41: p. 1033‐1037.[Burkoth et al. 1999] Burkoth, T. L., Bellhouse, B. J. et al. (1999).Transdermal and transmucosal powdered drugdelivery. Critical reviews in therapeutic drugcarrier systems 16, (4): p. 331‐384.

188 REFERENCES[Carpenter et al. 1997][Chang et al. 1996][Chang et al. 2005][Chen et al. 2000][Chen D. et al. 2001][Chen et al. 2002][Chen et al. 2003]Carpenter, J. F., Pikal, M. J. et al. (1997). Rationaldesign of stable lyophilized protein formulations:some practical advice. Pharm Res 14, (8): p. 969‐975.Chang, B. S., B. S. Kendrick, et al. (1996). Surfaceinduceddenaturation of proteins during freezingand its inhibition by surfactants. J. of Pharm Sci85, (12): 1325‐1330.Chang, L. L., Shepherd, D. et al. (2005). Mechanismof protein stabilization by sugars during freezedryingand storage: native structure preservation,specific interaction, and/or immobilization in aglassy matrix? J. of Pharm Sci 94, (7): p. 1427‐1444.Chen, D., Endres, R. L. et al. (2000). Epidermalimmunization by a needle‐free powder deliverytechnology: Immunogenicity of influenza vaccineand protection in mice. Nature Medicine 6: p. 1187‐1190.Chen D., E. C., Weis, K. F. et al. (2001). Epidermalpowder immunization induces both cytotoxic T‐lymphocyte and antibody responses to proteinantigens of influenza and hepatitis B viruses.Journal of Virology 75, (23): p. 11630‐11640.Chen D., Maa., Y‐F. et al. (2002). Needle‐freeepidermal powder immunization. Expert Rev.Vaccines 1: p. 265‐276.Chen, Y.‐J., Huang, L.‐W. et al. (2003). Temperatureresponsivepolymer‐assisted protein refolding.Enzyme and Microbial Technology 32: p. 120‐130.[Constantino 2004] Constantino, H. R. (2004). Lyophilisation ofbiopharmaceuticals. Excipients for use inlyophilized pharmaceutical peptide, protein andother bioproducts. Arlington, AAPS Press, p. 139‐228.[Costantino et al. 1998a]Costantino, H. R., Andya, J. D. et al. (1998). Effect ofmannitol crystallization on the stability and aerosolperformance of a spray‐dried pharmaceuticalprotein, recombinant humanized anti‐IgE

REFERENCES 189monoclonal antibody. J. of Pharm Sci 87, (11): p.1406‐1411.[Costantino et al. 1998b] Costantino, H. R., Carrasquillo, K. G. et al. (1998).Effect of excipients on the stability and structure oflyophilized recombinant human growth hormone.J. of Pharm Sci 87, (11): p. 1412‐1420.[Costantino et al. 2000] Costantino, H. R., Firouzabadian, L. et al. (2000).Protein spray‐freeze drying. Effect of atomizationconditions on particle size and stability. Pharm Res17, (11): p. 1374‐1383.[Costantino et al. 2002] Costantino, H. R., Firouzabadian, L. et al. (2002).Protein spray freeze drying. 2. Effect of formulationvariables on particle size and stability. J. of PharmSci 91, (2): p. 388‐395.[Costantino and Maa 2004] Costantino, H. R. and Maa, Y. F. (2004).Lyophilisation of biopharmaceuticals. Sprayfreeze‐drying of biopharmaceuticals: Applicationsand stability considerations. Arlington, VA, AAPSPress, p. 519‐561.[Cox and Phillips 2008][Craig et al. 1999][Crowe et al. 1993a][Crowe et al. 1993b]Cox, M. M. and Phillips, G. N. J. (2008). Handbook ofProteins: Structure, Functions and Methods.John Wiley & Sons, Chichester, England.Craig, D. Q., Royall, P. G. et al. (1999). The relevanceof the amorphous state to pharmaceutical dosageforms: glassy drugs and freeze dried systems.International journal of pharmaceutics 179, (2):p. 179‐207.Crowe, J. H., Crowe, L. M. et al. (1993). Preservingdry biomaterials: the water replacementhypothesis. Part 1. Biopharm., April 1993: p. 28‐37.Crowe, J. H., Crowe, et al. L. M. (1993). Preservingdry biomaterials: the water replacementhypothesis. Part 2. Biopharm., May 1993: p. 40‐43.

190 REFERENCES[Dawson et al. 1964] Dawson, M. D., Goodfriend, L. T. et al. (1964).Lactic dehydrogenases: Functions of the two types.Science, Vol. 143, p. 929‐933.[Dean et al. 2003][Debye 1944][Demeester et al. 2005][Dong et al. 1995][Edwards et al. 1997][EMEA. 2009][Engstrom et al. 2007a]Dean, H. J., Fuller, D. et al. (2003). Powder andparticle‐mediated approaches for delivery of DNAand protein vaccines into the epidermis.Comparative Immunology, Microbiology andInfectious Diseases 26: p. 373‐388.Debye, P. (1944). Light scattering in solutions.Journal of Applied Physics 15, p.338‐342.Demeester, J., De Smedt, S. et al. (2005). Methodsfor structural analysis of protein pharmaceuticals.Chapter 7: Light Scattering. Arlington, AAPS Press,VA, p.245‐275.Dong, A., Prestrelski, S. J. et al. (1995). Infraredspectroscopic studies of lyophilization‐ andtemperature‐induced protein aggregation. J PharmSci 84, (4): p. 415‐424.Edwards, D. A., Hanes, J. et al. (1997). Large porousparticles for pulmonary drug delivery. Science 276:p. 1868‐1871.EMEA (2009). Pandemic influenza (H1N1) website,http://www.emea.europa.euEngstrom, J. D., Simpson, D. T. et al. (2007). Stablehigh surface area lactate dehydrogenase particlesproduced by spray freezing into liquid nitrogen. EurJ. of Pharm Biopharm 65, (2): p. 163‐174.[Engstrom et al. 2007b] Engstrom, J. D., Simpson, D. T. et al. (2007).Morphology of protein particles produced by sprayfreezing of concentrated solutions. Eur J. of PharmBiopharm 65, (2): p. 149‐162.[Fink 1995]Fink, A. L. (1995). Molten globules. Methods MolBiol. 40: p. 343‐360.

REFERENCES 191[Fraunhofer and Winter 2004]Fraunhofer, W. and Winter, G. (2004). The use ofasymmetrical flow field‐flow fractionation inpharmaceutics and biopharmaceutics. European J.of Pharmaceutics and Biopharmaceutics 58:p. 369‐ 383.[Gabrielson et al. 2006] Gabrielson, J. P., Brader, M. L. et al. (2006).Quantitation of aggregate levels in a recombinanthumanized monoclonal antibody formulation bysize‐exclusion chromatography, asymmetrical fieldflow fractionation and sedimantation velocity. J. ofPharm Sci 96, (2): p.268‐279.[George and Wilson 1994][Giddings 1966][Gieseler 2004][Gieseler et al. 2007][Gieseler and Lee 2009][Gill and von Hippel 1989][Gloger et al. 2003]George, A. and Wilson, W. W. (1994). Predictingprotein crystallization from a dilute solutionproperty. Acta Crystallographica Section D ‐Biological Crystallography 50: p. 361‐365.Giddings, J. C. (1966). A new separation conceptbases on a coupling of concentration and flownonuniformities. Sep. Sci. 1: p. 123‐125.Gieseler, H. (2004). Product morphology and dryingbehavior delineated by a new freeze‐dryingmicrobalance. Division of Pharmaceutics. Friedrich‐Alexander University, Erlangen Germany. PhD.ThesisGieseler, H., Kessler, W. J. et al. (2007). Evaluationof tunable diode laser absorption spectroscopy forin‐process water vapor mass flux measurementsduring freeze drying. J. of Pharm Sci 96, (7): p. 1776‐1793.Gieseler, H. and Lee, G. (2009). Gravimetricmeasurement of momentary drying rate of sprayfreeze‐dried powders in vials. J. of Pharm Sci 98,(9): p. 3447‐3455.Gill, S. C. and von Hippel, P. H. (1989). Calculation ofprotein extinction coefficients from amino acidsequence data. Analytical Biochem. 182: p. 319‐326.Gloger, O., Witthohn, K. et al. (2003). Lyoprotectionof aviscumine with low molecular weight dextrans.

192 REFERENCESInternational Journals of Pharmaceutics 260: p. 59‐68.[Graupner et al. 1999][Gua et al. 2003][Hagen et al. 1996]Graupner, H., Haalck, L. et al. (1999). Moleculardynamics of microbial lipases as determined fromtheir intrinsic tryptophan fluorescence. Biophys. J.77: p. 493‐504.Gua, Z., Zhub, X. et al. (2003). Inhibition ofaggregation by media selection, sample loadingand elution in size exclusion chromatographicrefolding of denatured bovine carbonic anhydraseB. J. Biochem. Biophys. Methods 56, p. 65‐175.Hagen, S. J., Hofrichter, J. et al. (1996). Germinaterebinding and conformational dynamics ofmyoglobin embedded in a glass at roomtemperature. J. Phys. Chem. 100: p. 12008‐12021.[Hanlon et al. 2007] Hanlon, A., Trainoff, S. et al. (2007). Online A2:Rapid characterization of the second virialcoefficient, Poster presentation, Wyatt TechnologyCorp., Santa Barbara, CA.[Hatley 1997][Heller et al. 1999]Hatley, R. H. (1997). Glass fragility and the stabilityof pharmaceutical preparations. Pharm DevTechnol 2, (3): p. 257‐264.Heller, M. C., Carpenter, J. F. et al. (1999). Proteinformulation and lyophilization cycle design:prevention of damage due to freeze‐concentrationinduced phase separation. Biotechnol Bioeng 63,(2): p. 166‐174.[Henderson et al. 2000] Henderson, E. A., Louie, T. J. et al. (2000).Comparison of higher dose intradermal hepatitis Bvaccination to standard intramuscular vaccinationof healthcare workers. Infect Control HospEpidemiol 21, (4): p. 264‐269.[Henkens et al. 1982]Henkens, R. W., Kitchel, B. B. et al. (1982). Detectionand characterization using circular dichroism andfluorescence spectroscopy of a stable intermediate

REFERENCES 193conformation formed in the denaturation ofbovine carbonic anhydrase with guanidiniumchlorid. Biochemistry 21, (23): p. 5918‐5923.[Herold 1993]Herold, M. (1993). Size exclusion chromatography:Influence of salt concentration in the mobilephase. Am Lab 25: p. 35‐38.[Huglin 1972] Hugglin, M. (1972). Specific refractive indexincrements. London, Academic Press, p. 165‐331.[Imamura et al. 2003][ISO/TC 84/WG 4 2004][Izutsu et al. 1993][Jackson 1965][Kawai and Suzuki 2007][Kaye and Mc Daniel 1974][Kendall 2004][King 1975]Imamura, K., Ogawa, T. et al. (2003). Effects ofTypes of Sugar on the Stabilization of Protein in theDried State. J. of Pharm Sci 92, (2), p. 266‐274.Business plan ISO/TC 84/WG 4, InternationalOrganization for Standardization, TechnicalCommitee 84 Workgroup 4; 2004/11/08.Izutsu, K., Yoshioka, S. et al. (1993). Decreasedprotein‐stabilizing effects of cryoprotectants dueto crystallization. Pharm Res 10, (8): p. 1232‐1237.Jackson, J. D. (1965). Classical electrodynamics.Wiley, New York.Kawai, K. and Suzuki, T. (2007). Stabilizing effect offor types of disaccharide on the enzymatic activityof freeze‐dried lactate dehydrogenase: Step bystep evaluation from freezing to storage. PharmResearch 24, (10): p.1883‐1890.Kaye, W. and Mc Daniel, J. B. (1974). Low‐anglelight scattering: rayleigh factors and depolarizationratios. Appl Opt 13: p. 1934‐1937.Kendall, M., Mitchel, T. et al. (2004). Intradermalballistic delivery of micro‐particles into excisedhuman skin for drug and vaccine applications. J.Biomechanics 37, (11), p. 1733‐1741.King, C. J. (1975). Freeze‐drying and advanced foodtechnology. Chapter 21: Application of freezedryingin food products. Academic Press, London.

194 REFERENCES[Kok and Rudin 1981][Lahm and Lee 2006][Lavecchia and Zugaro 1991][Le Maire et al. 1987][Leuemberger 2002][Levi and Karel 1995][Lin and Timasheff 1996][Lindskog and Silverman 2000][Luthra et al. 2007]Kok, C. M. and Rudin, A. (1981). Relationshipbetween the hydrodynamic radius and the radiusof gyration of a polymer in solution. Makromol.Chem Rapid Commun. 2: p. 655‐659.Lahm, K. and Lee, G. (2006). Penetration ofcrystalline powder particles into excised humanskin membranes and model gels from a supersonicpowder injector. J. of Pharm Sci. 95: p. 1511‐1526.Lavecchia, R. and Zugaro, M. (1991). Thermaldenaturation of erythrocyte carbonic anhydrase.FEBS Lett. 292, p. 162‐164.Le Maire, M., Ghazi, A. et al. (1987). The use of gelchromatography for the determination of size andrelative molecular masses of proteins. Biochem. J.243: p. 399‐404.Leuemberger, H. (2002). Spray freeze‐drying ‐ Theprocess of choice for low water soluble drugs? J. ofNanopart Res 4: p. 111‐119.Levi, G. and Karel, M. (1995). Volumetric shrinkage(collapse) in freeze‐dried carbohydrates abovetheir glass transition temperature. Food ResearchInternational 28, (2): p. 145‐151.Lin, T.‐Y. and Timasheff, S. N. (1996). On the role ofsurface tension in the stabilization of globularproteins. Protein Sci. 5: p. 372‐381.Lindskog, S. and Silverman, D. N. (2000). TheCarbonic Anhydrases: New Horizons. Basel,Switzerland, Birkhäuser Verlag.Luthra, S., Obert, J. P. et al. (2007). Investigation ofdrying stresses on proteins during lyophilization:differentiation between primary and secondarydryingstresses on lactate dehydrogenase using ahumidity controlled mini freeze‐dryer. J Pharm Sci96, (1): p. 61‐70.

REFERENCES 195[Maa et al. 1997][Maa et al. 1998][Maa et al. 1999][Maa et al. 2003][Maa et al. 2004]Maa Y. F., Costantino, H. R., et al. (1997). The effectof operating and formulation variables on themorphology of spray‐dried protein particles.Pharmaceutical development and technology 2, (3)p. 213‐223.Maa, Y. F., Nguyen, P. A. et al. (1998). Spray‐dryingof air‐liquid interface sensitive recombinant humangrowth hormone. J Pharm Sci 87, (2): p. 152‐159.Maa, Y. F., Nguyen, P. A. et al. (1999). Proteininhalation powders: spray drying vs spray freezedrying. Pharm Res 16, (2): p. 249‐254.Maa, Y. F., Shu, C. et al. (2003). Optimization of analum‐adsorbed vaccine powder formulation forepidermal powder immunization. Pharm Res 20,(7): p. 969‐977.Maa, Y. F., Ameri, M. et al. (2004). Influenza vaccinepowder formulation development: spray‐freezedryingand stability evaluation. J Pharm Sci 93, (7):p. 1912‐1923.[Maa and Prestrelski 2000] Maa, Y. F. and Prestrelski, S. J. (2000).Biopharmaceutical powders: particle formationand formulation considerations. Curr PharmBiotechnol 1, (3): p. 283‐302.[MacRitchie 1998][Mallamace and Micali 1996][Malvern Instruments. 2009]MacRitchie, F. (1998). Reversibility of proteinadsorption, in: Proteins at liquid interfaces,Möbius, D. and Miller, R. (Eds). Amsterdam,Elsevier Science.Mallamace, F. and Micali, N. (1996). Low angle lightscattering and its applications, in Brown, W, (Eds),Light scattering: Principles and development.Oxford, Clarendon Press, p. 381‐435Malvern Instruments. (2009). Static light scattering‐ overview. Frequently asked questions.http://www.malvern.com

196 REFERENCES[Masters 1991]Masters, K. (1991). Spray‐Drying Handbook. NewYork, Wiley & Sons.[Mc Coy and Wong 1981] Mc Coy, L. F. J. and Wong, K.‐P. (1981).Renaturation of bovine erythrocyte carbonicanhydrase B denatured by acid, heat anddetergent. Biochemistry 20, (11): p. 3062‐3067.[Meister 2009] Meister,E (2009). Methodology, data interpretationand practical transfer of Freeze drymicroscopy. DivisionofPharmaceutics.Friedrich‐Alexander University, Erlangen, Germany. PhD.Thesis[Meister and Gieseler 2008][Mitchell et al. 2001]Meister, E. and Gieseler, H. (2008). Freeze‐drymicroscopy of protein/sugar mixtures: Dryingbehavior, interpretation of collapse temperaturesand a comparison to corresponding glass transitionData. J. of Pharm Sci 98, (9), p. 3072‐3087Mitchell, T. J., Kendall, M. A. F. et al. (2001). Microparticlepenetration to the oral mucosa.Bioengineering Conference ASME 2001, BED‐Vol.50: p. 763‐764.[Moelbert et al. 2004] Moelbert, S., Normand, B. et al. (2004).Kosmotropes and chaotropes: modellingpreferential exclusion, binding and aggregatestability. Biophys Chem 112, (1): p. 45‐57.[Moore 1964] Moore, J. C. (1964). Gel permeationchromatography. A new method for molecularweight distribution of high polymers. J. Polym. Sci.,Part A, Vol 2, p. 835‐843.[Mutschler et al. 2007] Mutschler, E., Schaible, H. G. et al. (2007).Anatomie, Physiologie, Pathophysiologie desMenschen. WVG Wissenschaftliche VerlagsgesellschaftmbH Stuttgart.[Nail and Gatlin 1993]Nail, S. and L. Gatlin (1993). Freeze Drying:Principles and practice. p. 163‐333. New York,Marcel Dekker.

REFERENCES 197[Ohgushi and Wada 1983][Oliva et al. 2009][Osorio et al. 2003]Ohgushi, M. and Wada, A. (1983). Molten‐globulestate: a compact form of globular proteins withmobile side‐chains. FEBS Lett 164, (1), p. 21‐24.Oliva, A., Llabres, M. et al. (2009). Estimation ofuncertainty in size‐exclusion chromatography witha double detection system (light‐scattering andrefractive index). Talanta 78: p. 781‐789.Osorio, J., Chen, D. et al. (2003). Immune responseto hepatitis B surface antigen following epidermalpowder immunization. Journal of Immunology andCell Biology 81: p. 52‐58.[Overcashier et al. 1999] Overcashier, D. E., Patapoff, T. W. et al. (1999)Lyophilization of protein formulations in vials:investigation of the relationship betweenresistance to vapor flow during primary drying andsmall‐scale product collapse. J. of Pharm Sci 88, (7):p. 688‐695.[Pace et al. 1995]Pace, C. N., Vajdos, F. et al. (1995). How to measureand predict the molar absorption coefficient of aprotein. Protein Science 4: p. 2411‐2423.[PDB. 2009] PDB (2009). RCSB Protein Data Bank.http://www.pdb.org.[Philo 2004][Philo 2006][Philo et al. 1996]Philo, J. S. (2004). Methods for structural analysisof protein pharmaceuticals. Chapter 7: LightScattering. Demeester, J., De Smedt, S. et al.(2005). Arlington, AAPS Press, VA, p.265.Philo, J. S. (2006). Is any measurement methodoptimal for all aggregate sizes and types? AAPSJournal 2006, 8, (3), Article 65, p. 564‐571.Philo, J. S., Wen, J. et al. (1996). Size exclusionchromatography with online light scattering.Analytical Biochem. 240, (2): p. 155‐166.

198 REFERENCES[Pikal et al. 1990][Pikal et al. 2007][Pikal 2002]Pikal, M. J., Shah, S. et al. (1990). The secondarydrying stage of freeze drying : drying kinetics as afunction of temperature and chamber pressure.International j. of pharmaceutics Vol. 60, (3): p. 203‐217.Pikal, M. J., Rigsbee, D. R. et al. (2007). Solid statechemistry of proteins: I. glass transition behavior infreeze dried disaccharide formulations of humangrowth hormone (hGH). J. of Pharm Sci 96, (10):p. 2765‐2776.Pikal, M. J. (2002). Freeze Drying, in: Encyclopediaof Pharmaceutical Technology. Marcel Dekker, NewYork, p. 1299‐1326.[Pikal 2004] Pikal, M. J. (2004). Mechanisms of proteinstabilization during freeze‐drying and storage: Therelative importance of thermodynamicstabilization and glassy state relaxation dynamics.Drug Pharm Science 137: p. 63‐107.[Pittsyn 1995]Pittsyn, O. B. (1995). Molten globule and proteinfolding. Adv Protein Chem.: p. 83‐229.[Porath and Flodin 1959] Porath, J. and P. Flodin. (1959). Gel filtration: Amethod for desalting and group separation.Nature 183, (4676): p. 1657‐1659.[Prestrelski et al. 1993] Prestrelski, S. J., Arakawa, T. et al. (1993).Separation of freezing‐ and drying‐induceddenaturation of lyophilized proteins using stressspecificstabilization: II. Structural studies usinginfrared spectroscopy. Archives of Biochemistryand Biophysics 303, (2): p. 465‐473.[Prestrelski et al. 2002] Prestrelski, S. J., Burkoth, T. L. et al. (2002).Production of hard, dense particles. WO02100380.PCT Int. Apll.[Prestrelski and Burkoth 1999] Prestrelski, S. J. and Burkoth, T. L. (1999).Supersonic powder injection for transdermaland transmucosal delivery of drugs and vaccines.Book of Abstracts, 217 th ACS National Meeting,Anaheim, Calif., March 21‐25: POLY‐197.

REFERENCES 199[Prins et al. 1998][Rambhatla et al. 2005][RIMD 2009][Rochelle 2005][Rochelle and Lee 2007][Saito et al. 2004][Sarphie and Burkoth 1997][Schneider 2008][Searles et al. 2001]Prins, A., Bos, M. A. et al. (1998). Relation betweensurface rheology and foaming behaviour ofaqueous protein solutions. Amsterdam, ElsevierScience B.V.Rambhatla, S., Obert, J. P. et al. (2005). Cakeshrinkage during freeze drying: a combinedexperimental and theoretical study. Pharm DevTechnol 10, (1): p. 33‐40.RIMD, R. I. f. M. D., Osaka University (2009). T‐cellactivation and migration into the skin. ResearchProjects. http://www.biken.osaka‐u.ac.jp.Rochelle, C. (2005). Spray‐freeze‐dried proteinpowders for needle‐free injection. Division ofPharmaceutics. Friedrich‐Alexander University,Erlangen Germany. PhD. ThesisRochelle, C. and Lee, G. (2007). Dextran orhydroxyethyl starch in spray‐freeze‐driedtrehalose/mannitol microparticles intended asballistic particulate carriers for proteins. J. ofPharm Sci 96, (9): p. 2296‐2309.Saito, R., Sato, T. et al. (2004). Structure of bovinecarbonic anhydrase II at 1.95 Å resolution. ActaCrystallographica D60, (2): p. 792‐795.Sarphie, D. F. and Burkoth, T. L. (1997). Method forproviding dense particle compositions for use intransdermal particle delivery. PCT Int. Apll.: 31 pp.Schneider, M. (2008). Different ways to determinethe online A2. A little overview. Posterpresentation, WTC Europe, Germany.Searles, J. A., Carpenter, J. F. et al. (2001). Annealingto optimize the primary drying rate, reducefreezing‐induced drying rate heterogeneity, anddetermine T(g)' in pharmaceutical lyophilization. J.of Pharm Sci 90, (7): p. 872‐887.

200 REFERENCES[Serrien et al. 1992][Shamblin et al. 2006][Sigma‐Aldrich 1997][Sigma‐Aldrich 2009a][Sigma Aldrich 2009b][Snyder and Kirkland 1979][Sonner et al. 2002][Sonner 2002][Sono‐Tek. 2010][Stroehlein Instruments 1988][Tang and Pikal 2004]Serrien, G., Geeraerts, G. et al. (1992). Dynamicsurface properties of adsorbed protein solutions:BSA, casein and buttermilk. Colloids and surfaces68, (4): p. 219‐233.Shamblin, S. L., Hancock, B. C. et al. (2006). Couplingbetween chemical reactivity and structuralrelaxation in pharmaceutical glasses. Pharm Res 23,(10): p. 2254‐2268.Sigma‐Aldrich (1997).Gel filtration chromatographywith biocompatible SigmaChrom GFC columns.Bulletin 891A.Sigma‐Aldrich (2009). Encymatic assay of bovinecarbonic anhydrase. http://www.sigmaaldrich.com.Sigma‐Aldrich (2009). Encymatic assay of L‐lacticdehydrogenase. http://www.sigmaaldrich.com.Snyder, L. and Kirkland J. J. (1979). Introduction toModern Liquid Chromatography, John Wiley &Sons, Inc., New Jersey, USSonner, C., Maa, Y. F. et al. (2002). Spray‐freezedryingfor protein powder preparation: particlecharacterization and a case study with trypsinogenstability. J Pharm Sci. 91: p. 2122‐2139.Sonner, C. (2002). Protein‐loaded powders byspray freeze drying. Division of Pharmaceutics.Friedrich ‐Alexander University, Erlangen Germany.PhD. ThesisSono‐Tek. (2010). Drop size and distribution.http://www.sono‐tek.comStroehlein Instruments (1988). Operation manualArea‐meter II. JUWE Laborgeräte GmbH, Viersen,Germany.Tang, X. and Pikal M. J. (2004). Design of freezedryingprocesses for pharmaceuticals: practicaladvice. Pharm Res 21, (2): p. 191‐200.

REFERENCES 201[Tang et al. 2006]Tang, X. C., Nail, S. L. et al. (2006). Evaluation ofmanometric temperature measurement, a processanalytical technology tool for freeze‐drying: part IImeasurement of dry‐layer resistance. AAPS PharmSci Tech 7, (4): p. 93.[Timasheff 2002] Timasheff, S. N. (2002). Protein hydration,thermodynamic binding and preferentialhydration. Biochemistry 41, (46): p. 13473‐13482.[Tsai et al 1998][Uversky 1993][Valente et al. 2005][Verpoorte et al. 1967][Vonhoff 2008][Vonhoff 2010]Tsai, A. M., van Zanten et al. (1998). Study ofprotein aggregation due to heat denaturation.Biotechnology and Bioengineering 59, (3), p. 273‐280Uversky, V. N. (1993). Use of fast protein sizeexclusionliquid chromatography to study theunfolding of proteins which denature through themolten globule. Biochemistry 32: p. 13288‐13298.Valente, J. J., Verma, K. S. et al. (2005). Secondvirial coefficient studies of cosolvent‐inducedprotein self‐interaction. Biophysical Journal 89:p. 4211‐4218.Verpoorte J.A., Mehta S., et al. (1967). Esteraseactivities of human carbonic anhydrase B and C.Journal of Biological Chemistry 242, (18): p. 4221‐4229.Vonhoff, S., Schiffter, H. A. (2008). Influence ofnozzle‐type and atomization conditions on proteinseconddary and tertiary structure during sprayfreeze‐drying.Poster presentation, AAPS AnnualMeeting, Atlanta.Vonhoff, S. (2010). The influence of atomizationconditions on protein secondary and tertiarystructure during microparticle formulation byspray‐freeze‐drying. Division of Pharmaceutics.Erlangen, Friedrich‐Alexander University. PhD.

202 REFERENCES[Wahlund and Caldwell 2000][Wang 1999][Wang 2000][Wang et al. 2008][Wanka and Peukert 2006][Webb et al. 2003][Webb et al. 2002][Westermarck et al. 1998]Wahlund, K. G., and Caldwell, K. D. (2000). Methodsfor structural analysis of protein pharmaceuticals.Chapter 11: Asymmetrical flow field‐flowfractionation. Arlington, AAPS Press, VA, p. 245‐275.Wang, W. (1999). Instability, stabilization, andformulation of liquid protein pharmaceuticals. IntJ. of Pharm 185, (2): p. 129‐188.Wang, W. (2000). Lyophilization and developmentof solid protein pharmaceuticals. Int J. of Pharm203, (1‐2): p. 1‐60.Wang, B., Tchessalov, S. et al. (2008). Impact ofsucrose level on storage stability of proteins infreeze‐dried solids: II. Correlation of aggregationrate with protein structure and molecular mobility.J Pharm Sci 98, (9): p. 3131‐3144.Wanka, J. and Peukert, W. (2006). Die Bedeutungdes zweiten osmotischen Virialkoeffizienten für dieProteinkristallisation. Chemie Ingenieur Technik 78,(3): p. 273‐278.Webb, S. D., Cleland, J. L. et al. (2003). Effects ofannealing lyophilized and spray‐lyophilizedformulations of recombinant human interferongamma.J Pharm Sci 92, (4): p. 715‐729.Webb, S. D., S. L. Golledge, et al. (2002). Surfaceadsorption of recombinant human interferongammain lyophilized and spray‐lyophilizedformulations. J Pharm Sci 91, (6): p. 1474‐1487.Westermarck, S., A. M. Juppo, et al. (1998). Porestructure and surface area of mannitol powder,granules and tablets determined with mercuryporosimetry and nitrogen adsorption. EuropeanJournal of Pharmaceutics and Biopharmaceutics 46,(1): p. 61‐68.

REFERENCES 203[Whitesides et al. 2008][WHO 2009][Wilson 2003][Wong and Tanford 1973][WTC. 2009a][WTC. 2009b]Whitesides, G. M., Krishnamurthy, V. M. et al.(2008). Carbonic anhydrase as a model forbiophysical and physical‐organic studies of proteinand protein‐ligand binding. Chemical Review,Author Manuscript 108, (3): p. 946‐1051.WHO (2009). Intradermal delivery of vaccines. Areview of the literature and the potential fordevelopment for use in low‐ and middleincomecountries. http://www.who.int.Wilson, W. W. (2003). Light scattering as adiagnostic for protein crystal growth ‐ a practicalapproach. J Struct Biol 142: p. 56‐65.Wong, K.‐P. and Tanford, C. (1973). Denaturation ofbovine carbonic anhydrase B by guanidiniumhydrochloride. J Biol Chem 248, (24): p. 8518‐8523.Homepage of Wyatt Technology Europe GmbH,Germany (2009). http://www.wyatt.euWyatt Technology Cooperation, US. (2009). Astra Vonline help. Astra V, 5.3.4.14.[WTC‐Appl.N. #5 1996] Wyatt Technology Cooperation, US. (1996).Protein Application Note #5: Protein PEGylation.WTC Europe, Germany. http://www.wyatt.eu[WTC‐Appl.N. #18 2002][WTC‐Appl.N. #25 2007][Wyatt 1993]Wyatt Technology Cooperation, US. (2002). ProteinApplication Note #18: Membrane proteins. WTCEurope, Germany. http://www.wyatt.euWyatt Technology Cooperation, US. (2007). ProteinApplication Note #25: Automated, online secondvirial coefficient (A 2 ) measurements. WTC Europe,Germany. http://www.wyatt.euWyatt, P. J. (1993). Light scattering and theabsolute characterization of macromolecules. AnalChim Acta 272: p. 1‐40.

204 REFERENCES[Wyatt 2006a] Wyatt, G. (2006). General MALS theory.Introduction to light scattering for absolutemacromolecular characterization. WTC, US, ppt.‐presentation.[Wyatt 2006b][Yau et al. 1979][Yazgan and Henkens 1972][Yu et al. 2002][Yu et al. 2006][Yumiokaet al. 2010][Ziegler 2006]Wyatt, G. (2006). Measuring molecular size andshape. Introduction to light scattering for absolutemacromolecular characterization, WTC, US, ppt.‐presentation.Yau, W. W., Kirkland, J. J. et al. (1979). Modern sizeexclusionliquid chromatography, Chapter 9:Calibration. John Wiley & Sons, Inc., New York.Yazgan, A. and Henkens, R. W. (1972). Role of zinc(II) in the refolding of guanidine hydrochloridedenatured bovine carbonic anhydrase.Biochemistry 11, (7): p. 1314‐1318.Yu, Z., Rogers, T. L. et al. (2002). Preparation andcharacterization of microparticles containingpeptide produced by a novel process: sprayfreezing into liquid. Eur J Pharm Biopharm 54, (2):p. 221‐228.Yu, Z., Johnston, K. P. et al. (2006). Spray freezinginto liquid versus spray‐freeze drying: influence ofatomization on protein aggregation and biologicalactivity. Eur J Pharm Sci 27, (1): p. 9‐18.Yumioka, R., Sato, H. et al. (2010). Mobile phasecontaining arginine provides more reliable SECcondition for aggregation Analysis. J. of PharmSci 99, (2): p. 618‐620.Ziegler, A. S. (2006). Needle‐free injection. Attritionforces acting on spray‐dried and spray‐freeze driedprotein loaded particles with needle‐free injectors.Division of Pharmaceutics, Friedrich‐AlexanderUniversity, Erlangen, PhD.[Ziegler et al. 2006] Ziegler, A. S., Schluecker, E. et al. (2006).Inactivation effects on protein in a needle‐freevaccine injector. Eng. Life Sci. 6, (4): p. 384‐393.

REFERENCES 205[Ziegler et al. 2010][Zimm 1948]Ziegler, A. S., Lee, G. et al. (2010). Comminution ofspray‐freeze‐dried microparticles on actuation of aballistic powder injector. J. of Pharm Sci (in press).Zimm, B. H. (1948). The scattering of light and theradial distribution function of high polymersolutions. J Chem Phys 16: p. 1093‐1099.

206 CURRICULUM VITAE9 Curriculum VitaePersonal DataNameDate of BirthMarital StatusNationalityGeorg StrallerFebruary 13 th , Wassertrüdingen (Germany)SingleGermanSchool educationSep 1986 – Jul 1990Sep 1990 – Jul 1999Jul 1999Elementary school in Wassertrüdingen, GermanyGymnasium Dinkelsbühl, GermanyDiploma from German secondary school qualifying foruniversity administration and matriculationCommunity ServiceSeptember 1999 – September 2000Alternative civilian serviceSonderschule “Zum Guten Hirten”, Wassertrüdingen, GermanyProfessional and academic trainingOct 2000 – Oct 2004Jul 2002Sept 2004Friedrich‐Alexander‐University, Erlangen, GermanyStudies of Pharmacy1 st part of state examination in pharmacy2 nd part of state examination in pharmacyOct 2004 – Mar 2005 Institution of Forensic Medicine, Ludwig‐MaximilianUniversity, Division of Forensic Toxicology, Munich, GermanyInternship at universityApr 2005 – Oct 2005Dec 2005Dec 2005Jan 2006 – Mar 2006Wittelsbacher Apotheke, Munich, Germany Internship atcommunity pharmacy3 rd part of state examination in pharmacyLicensed pharmacistPharmacist, Stadtapotheke Wassertrüdingen, GermanyApr 2006 – Present Friedrich‐Alexander‐University, Erlangen, GermanyPhD. studies in Pharmaceutics under the supervision of Prof.Dr. G. Lee, assistant teacher for undergraduate students(solid and liquid dosage forms)Further education / Additional businessApr 2006 – Apr 2009 Advanced training “Fachapotheker für PharmazeutischeTechnologie”Mar 2006 – PresentPharmacist, Altstadt Apotheke Nürnberg, Germany