Development and Application of Novel ... - Jacobs University

Development and Application of Novel Bioinformatics and
Computational Modeling Tools for Protein Engineering
by
Rajni Verma
A thesis submitted for the degree of
Doctor of Philosophy
in
Computational Chemistry & Bioinformatics
Date of Defense: December 14 th 2012
Supervisor
Prof. Dr. Danilo Roccatano
Jacobs University Bremen, Germany
Co-supervisor
Prof. Dr. Ulrich Schwaneberg
RWTH Aachen University, Germany
External committee member
Dr. Steven Hayward
University of East Anglia, UK
School of Engineering and Science

Acknowledgement
I express my sincere gratitude to my PhD supervisor, Prof. Dr. Danilo Roccatano for
his expert and continuous guidance. Especially, I thank him for his patience and the time he
spent to explain the concepts and ideas that really helped me to accomplish my work. His
constant support, understanding, motivation and valuable discussions provided me a
wonderful learning experience during my PhD.
I am thankful to Prof. Dr. Ulrich Schwaneberg for his constructive comments, fruitful
discussions and his support during this endeavor. It has been a great honor and pleasure to
work with him. I am deeply grateful for his trust on me. I express my respectful gratitude to
Dr. Steven Hayward for being a member of my PhD committee. I am thankful to Dr. Achim
Gelessus for his technical support to utilize efficiently CLAMV facility for scientific
computation throughout this work at Jacobs University Bremen.
I express my whole-hearted thanks to the member of Prof. Roccatano Group, Samira
Hezaveh, Khadga Karki, Susruta Samanta and Edita Sarukhanyan for their wonderful
company and support. I also convey my thanks to the member of Prof. Schwaneberg group
at RWTH Aachen for their cooperation. I express my special thanks to my friends, Kavita,
Amit, Amol, Sagar, Hemanshu, Susruta, Usha and Steffi for providing such a friendly
environment. Especially, I express my whole-souled gratitude to Amol for his continuous
support, encouragement, motivation, patience and care throughout my PhD.
i

Funding
The work described in this PhD thesis was financially supported by European
Union 7 th framework program for the project entitled “Effective redesign of oxidative
enzymes for green chemistry” (Project reference: 212281) in collaboration with Prof. Dr.
Ulrich Schwaneberg from RWTH Aachen University.
ii

List of Publication
1. Verma R, Schwaneberg U, Roccatano D. Conformational dynamics of the FMN-binding
reductase domain of monooxygenase P450BM-3. J Chem Theory and Comput 2012, DOI:
10.1021/ct300723x.
2. Verma R, Schwaneberg U, Roccatano D. Computer-aided protein directed evolution: a
review of web servers, databases and other computational tools for protein
engineering. Computational and Structural Biotechnology Journal 2012, 2 (3),
e201209008.
3. Ruff AJ, Marienhagen J, Verma R, Roccatano D, Genieser HG, Niemann P, Shivange AV,
Schwaneberg U. dRTP and dPTP a complementary nucleotide couple for the Sequence
Saturation Mutagenesis (SeSaM) method. J Mol Catal B-Enzym 2012, 84, 40-47.
4. Verma R, Schwaneberg U, Roccatano D. MAP 2.0 3D: a sequence/structure based server
for protein engineering. ACS Synth Bio. 2012, 1 (4), 139-150.
5. Zhu L, Verma R, Roccatano D, Ni Y, Sun Z, Schwaneberg U. A potential antitumor drug
(arginine deiminase) reengineered for efficient operation under physiological
conditions. ChemBioChem 2010, 11, 2294-2301. [inside cover page]
6. Verma R, Schwaneberg U, Roccatano D. Insight into the redox partner interaction
mechanism in cytochrome P450BM-3 using molecular dynamics simulations.
(manuscript under preparation)
7. Verma R, Schwaneberg U, Roccatano D. A molecular dynamics study of the interactions
between P450BM-3 domains and Coblat(II)Sepulchrate as an electron transfer
mediator. (manuscript under preparation)
iii

Abstract
In the last decades, enzymatic catalysis emerges as a convenient and
environmentally friendly substitute for the traditional chemical processes range from the
synthesis of many pharmaceutical and agrochemical building blocks to fine and bulk
chemicals, and more recently, the components of biofuel. The combination of experimental
and computational methods holds particular promise in the field of enzymatic catalysis to
tailor enzymes for the tasks not yet exploited by natural selection. Therefore, it is
important to develop computational tools that help to exploit this goal. The scope of this
thesis is to propose novel bioinformatics tools and to explore computational methods
aimed to support and guide protein evolution experiments. The thesis is divided into two
parts. First part of the thesis (Part I, Chapter 1 and Chapter 2) is focused on extending the
benchmarking system of random mutagenesis methods (MAP: Mutagenesis Assistant
Program) towards the sequence/structure and structure/function analysis and to evaluate
this approach on commonly used enzymes as biocatalysts. Chapter 1 offers the
comprehensive information about the computational methods used to assist protein
engineering experiments. Chapter 2 describes a completely renewed and improved version
of MAP server, named as MAP 2.0 3D server that correlates the generated amino acid
substitution patterns to the structural information of the target protein. Therefore, the
latter helps to identify in advance the random mutagenesis method that can introduce
mutations having less deleterious effect and to improve protein fitness towards an
expected property, e.g. charged amino acid substitutions to increase solubility of protein in
water. The capability of the server was illustrated by in-silico screening of different
enzymes and the predicted results were in agreement with the experimental findings.
iv

The atomic level understanding of the subtle intertwining among structure,
dynamics and function of enzymes plays an important role to rationally design new or
improved functions. Second part of the thesis (Part II, Chapter 3 – 6) is based on molecular
modeling approach to gain insight into the structural and dynamic properties of P450BM-3
(CYP102) complex in water and in the presence of cobalt(II)sepulchrate (CoSep) as an
electron transfer (ET) mediator. P450BM-3, isolated from Bacillus megaterium is an
attractive target and model system for biochemical (catalyzes the wide variety of
industrially attractive substrates) and biomedical (being a bacterial model for microsomal
P450s system) applications. The comprehensive theoretical aspects of MD simulation are
provided in Chapter 3 with the overview about the system preparation for MD simulation
and the analysis of protein conformation and dynamics in the generated trajectory. In
Chapter 4, the structural and dynamic properties of P450BM-3 FMN (Flavin
mononucleotide) domain as holo-protein, with the cofactor in oxidized and reduced states
and as apo-protein are investigated. The results illustrate the effect of FMN cofactor and its
protonation state on the conformation and dynamics of the FMN domain that can be
related to ET pathway from FMN to HEME cofactor. The study is further extended to garner
insight into the binding modes and the structural determinant of inter-domain ET in
HEME/FMN complex of P450BM-3. MD simulations were performed on both FMN and
HEME domains, isolated and in their crystallographic complex and results are reported in
Chapter 5. HEME/FMN complex undergoes the rearrangement process to decrease the
distance between their redox centers to promote favorable ET rate under physiological
condition. In Chapter 6, MD simulation of P450BM-3 domains (isolated HEME domain and
HEME/FMN complex) were performed in the presence of CoSep, as ET mediator. The
results illustrate the preferential binding modes of CoSep in P450BM-3 domains and the
putative ET pathways from CoSep to the iron center of HEME cofactor and are in agreement
with the experimental findings.
v

Table of Content
Acknowledgement .............................................................................................................................................. i
Funding .................................................................................................................................................................... ii
List of Publication ............................................................................................................................................. iii
Abstract .................................................................................................................................................................. iv
Chapter 1 ................................................................................................................................................................. 1
1.1. Abstract ................................................................................................................................................... 1
1.2. Background ............................................................................................................................................ 1
1.3. Generated diversity and library size ............................................................................................ 5
1.4. Evolutionary conservation based focused library .................................................................. 9
1.5. Structure-based focused library ................................................................................................. 15
1.6. Mutational effects in protein ........................................................................................................ 23
1.7. Summary and outlook..................................................................................................................... 26
1.8. References ........................................................................................................................................... 27
Chapter 2 .............................................................................................................................................................. 38
2.1. Abstract ................................................................................................................................................ 38
2.2. Introduction ........................................................................................................................................ 39
2.3. Methods ................................................................................................................................................ 41
2.3.1. Mutational probability and statistics ............................................................................... 41
2.3.2. MAP indicators .......................................................................................................................... 43
2.3.3. Local chemical diversity and protein structure components ................................. 44
2.3.4. MAP 2.0 3D server description ............................................................................................... 46
2.3.5. MAP 2.0 3D output....................................................................................................................... 48
2.3.6. Model proteins .......................................................................................................................... 48
2.4. Results and discussions ................................................................................................................. 49
2.4.1. D-amino acid oxidase ............................................................................................................. 49
vi

2.4.2. Phytase ......................................................................................................................................... 57
2.4.3. N-acetylneuraminic acid aldolase ..................................................................................... 61
2.5. Conclusions ......................................................................................................................................... 65
2.6. References ........................................................................................................................................... 66
Chapter 3 .............................................................................................................................................................. 71
3.1. Background ......................................................................................................................................... 71
3.2. Setup of the simulated systems ................................................................................................... 75
3.3. Equilibration procedure ................................................................................................................ 76
3.4. Structural and dynamical analysis ............................................................................................. 77
3.5. Cluster analysis ................................................................................................................................. 77
3.6. Principal component analysis ...................................................................................................... 78
3.7. References ........................................................................................................................................... 79
Chapter 4 .............................................................................................................................................................. 81
4.1. Abstract ................................................................................................................................................ 81
4.1. Introduction ........................................................................................................................................ 82
4.2. Methods ................................................................................................................................................ 84
4.2.1. Starting coordinates ............................................................................................................... 84
4.2.2. Molecular dynamics simulation ......................................................................................... 84
4.2.3. FMN binding site analysis ..................................................................................................... 86
4.2.4. Multiple structural alignment of FMN domain ............................................................. 86
4.3. Results ................................................................................................................................................... 87
4.3.1. FMN domain: structural and dynamical properties ................................................... 87
4.3.2. Cluster analysis of FMN domain......................................................................................... 89
4.3.3. FMN binding site ...................................................................................................................... 90
4.3.4. Conservation profile of FMN binding site ...................................................................... 94
4.3.5. Principal component analysis of FMN domain ............................................................. 96
4.3.6. FMN cofactor: structural and dynamical properties .................................................. 99
4.3.7. Cluster analysis of FMN cofactor ..................................................................................... 102
4.3.8. Principal component analysis of FMN cofactor.......................................................... 102
4.4. Discussions and conclusions ...................................................................................................... 103
vii

4.5. References ......................................................................................................................................... 105
Supporting information ............................................................................................................................. 109
Chapter 5 ............................................................................................................................................................ 122
5.1. Abstract .............................................................................................................................................. 122
5.2. Introduction ...................................................................................................................................... 123
5.3. Methods .............................................................................................................................................. 125
5.3.1. Starting coordinates ............................................................................................................. 125
5.3.2. Molecular dynamic simulations ....................................................................................... 125
5.3.2. Electron transfer tunneling ................................................................................................ 127
5.4. Results and discussion.................................................................................................................. 127
5.4.1. Structural properties ............................................................................................................ 127
5.4.2. Cluster analysis ....................................................................................................................... 130
5.4.3. Substrate access channel .................................................................................................... 131
5.4.4. ET tunneling pathways ........................................................................................................ 133
5.4.5. Essential dynamics ..................................................................................................................... 135
5.5. Conclusions ....................................................................................................................................... 139
5.6. References ..................................................................................................................................... 140
Supporting information ............................................................................................................................. 143
Chapter 6 ............................................................................................................................................................ 151
6.1. Abstract .............................................................................................................................................. 151
6.2. Introduction ...................................................................................................................................... 152
6.3. Methods .............................................................................................................................................. 153
6.3.1. Starting coordinates ............................................................................................................. 153
6.3.2. Molecular dynamics simulation and modeling .......................................................... 154
6.4. Results and discussion.................................................................................................................. 156
6.4.1. CoSep binding on P450BM-3 domains .......................................................................... 157
6.4.2. Effect of CoSep binding on substrate access channel .............................................. 159
6.4.3. Effect of CoSep binding on ET tunneling....................................................................... 159
6.4.4. Effect of CoSep binding on P450BM-3 dynamics ...................................................... 162
6.5. Conclusions ....................................................................................................................................... 166
viii

6.6. References ......................................................................................................................................... 167
Supporting information ............................................................................................................................. 170
Summary and outlook................................................................................................................................. 184
Curriculum vitae
ix

PART I: CAPDE
Chapter 1
Computer-Aided Protein Directed Evolution: a Review of
Web Servers, Databases and other Computational Tools
for Protein Engineering
1.1. Abstract
The combination of computational and directed evolution methods has proven a
winning strategy for protein engineering. We refer to this approach as computer-aided
protein directed evolution (CAPDE) and the chapter summarizes the recent developments
in this rapidly growing field. We will restrict ourselves to overview the availability,
usability and limitations of web servers, databases and other computational tools proposed
in the last five years. The goal of this chapter is to provide concise information about
currently available computational resources to assist the design of directed evolution
based protein engineering experiment.
1.2. Background
Protein engineering comprises a large number of techniques applied to evolve or
design protein with desired function.[1] The primary objective in any protein engineering
experiment is to identify specific sequence changes and alter the protein for desired
1

PART I: CAPDE
functional properties.[1,2] Generally, two main approaches are used to design the novel
proteins or enzymes: rational design and directed evolution. The first approach employs
the information of protein structure and focuses mutagenesis to modify protein scaffolds
(e.g. the active site of the biocatalyst). For this approach, the knowledge of the target amino
acid is necessary and can be provided by visual inspection or in-silico prescreening.[3] Both
cases depend on the nature of the problem and show high success rate only for the
prediction of single or double mutations. Indeed, multiple mutations involve cooperative
effects on protein structure and function that are almost inaccessible to the current
computational screening methods as well.
A more challenging de novo design or redesign of synthetic protein or peptide uses
solely structural information and folding rules of the proteins.[4,5] Although the method
offers broadest possibility to design novel fold and function, the success for large proteins
is limited.[6,7] The reasons rely on the limited number of three-dimensional protein
structures (in particular membrane proteins) and the lack of unifying theory for protein
folding mechanisms. Computational approaches based on micro-second to milliseconds
atomistic [8-10] molecular dynamics (MD) simulations of protein folding have recently
given some encouraging success for ab-initio folding of peptides and small proteins. In
addition, the combined approach of quantum mechanics and molecular dynamics methods
have shown the superior capability of physical based method to design new enzymatic
reaction.[11] However, the application of these methods is still limited since they are
considerably computational time demanding.[12] In this chapter, the approaches based on
de novo design, quantum mechanics and molecular dynamics will not be covered. The
reader can refer to different recent papers and reviews on these topics.[13-16]
The second approach is the so-called directed evolution. The method is one of the
most powerful approaches to improve or create new protein function by redesigning the
protein structure.[17] It can, for example, improve activity or stability of biocatalyst under
unnatural conditions (e.g. the presence of organic solvent) by accumulating multiple
mutations.[17,18] Directed evolution involves multiple rounds of random mutagenesis or
gene shuffling followed by screening of the mutant library.[19] The preliminary knowledge
2

PART I: CAPDE
of protein structure is not required in directed protein evolution. However, the structural
information can focus and restrict the approach to specific subsets of amino acids (e.g.
active site residues). A common problem of directed evolution methods is the limited
distribution of generated sequence diversity that reduces the efficient sampling of
functional sequence space.[19,20]
In summary, rational design via site directed or saturation mutagenesis and directed
evolution via random mutagenesis are used as key tools in protein engineering. In both
approaches, the sequence diversity is directly generated as point mutation, insertion or
deletion within a single parental gene. Consequently, the improvement in the quality of
rationally designed libraries and techniques for sequence space exploration and diversity
generation are critical for future advances.
The combination of experimental and computational methods holds particular
promise to tailor the proteins for tasks not yet exploited by natural selection.[21,22] In fact,
most of the computational tools or web servers for directed evolution utilize, when it is
possible, structural data to assist library generation processes. Since it is impossible to test
more than a very small fraction of vast number of possible protein sequences, it urges to
have a directed evolution strategy for generating sequence libraries with the highest
chance to have variants with desired enzymatic properties. Such libraries can be designed
by applying the current knowledge of the protein response towards mutations and
sequence-structure-function relationships.
Thermo stability, solvent stability (pH and salt stability or co-solvents tolerance)
and enzymatic activity (as improvement in both binding affinity and catalytic activity) are
the properties commonly targeted by protein engineering experiments. The first two
effects are subtle to predict due to their distributed effect on protein structure. For the
enzymatic activity, different mutagenesis studies indicate that most of the mutations,
affecting certain enzyme properties, as substrate specificity, enantioselectivity and new
catalytic activities, are located into or near the active site.[21] Rational design approach is
successful in targeting relevant active site residues for site-directed mutagenesis but less
3

PART I: CAPDE
effective for important residues located in the second coordination sphere of the active site.
For these cases, the combination of random mutagenesis and computer-aided protein
directed evolution (CAPDE) approaches can provide a winning strategy. The application of
computational methods in conjunction with directed evolution offers the exciting promise
to generate libraries having high frequency of active and improved variants.[23]
Figure 1.1: Schematic representation of four CAPDE approaches (as the quarters of the circle): (1)
generated diversity and library size (in red), (2) evolutionary conservation based focused library
(in green), (3) structure-based focused library (in purple) and (4) mutational effects in protein (in
cyan). The servers, tools and databases associated with the approaches are shown in boxes.
4

PART I: CAPDE
In this chapter, for the sake of clarity, the CAPDE approaches have been divided in
four major areas, schematically represented in Figure 1.1. The first one comprises tools
used for characterizing the library generated by mutagenesis methods mainly through the
statistical approaches. The second and third areas are represented by tools that consider
the evolutionary and structural information of the target protein to design the focused
library. Multiple sequence or structure alignment (MSA) is the key approach used by these
tools to identify variable or conserved positions in the target protein. The fourth part is
dedicated to the tools for the prediction of mutational effects on protein structure and
function. These tools and/or web servers are based on machine learning, statistical or
empirical approaches and predict mutational effect on protein stability and/or activity by
estimating the relative free energy changes.[24]
This chapter is divided in four parts following the division of CAPDE approaches. It
aims to provide the concise information about currently available CAPDE methods to assist
and design directed evolution experiments with the final goal to enhance the probability
for identifying the mutants with desired properties. In particular, the reader will find a
short overview and classification to novel database, web server and other computational
tools that can provide relevant information for the interpretation of experimental results
and have been developed in the last few years in the field of molecular modeling of protein
structure. Finally and as previously mentioned, we are not going to take in consideration
the methods that involve physical approach based on QM/MM or MD simulations.
1.3. Generated diversity and library size
The unbiased diversity generation followed by the screening of a statistically
meaningful fraction of generated sequence space are fundamental challenges in directed
evolution experiments.[25] The directed evolution strategy comprises two key steps: 1)
generate diverse mutant libraries and 2) screen to identify the improved protein variants.
The success of a directed evolution methods depends upon the quality of the mutant
5

PART I: CAPDE
library. The challenges and advances to generate the functionally diverse libraries have
been reviewed in past year.[20,26] Computational tools can assist directed evolution in
these two steps by in-silico analysis and screening of expected protein sequence space
sampled by generated libraries (summarized in Table 1.1). Publicly available web servers,
MAP (Mutagenesis Assistant Program)[25,27] and PEDAL-AA[28] were developed to
estimate the diversity at protein level in the library generated by random mutagenesis
method.
Table 1.1: Summarizing computational tools to analyze amino acid diversity, size and
completeness of the library generated by mutagenesis methods.
Approach Name Input
Nucleotide
MAP 2.0 3D
sequence or
[25,27]
protein structure.
Statistics of
Nucleotide
generated
sequence,
diversity
PEDEL-AA mutation rate,
[28] library size, indel
rate, nucleotide
mutation matrix.
Library size and
Library size GLUE-IT
randomization
and
[28]
techniques.
completenes
s
Probability
TopLib [30]
required by
Case study
examples
Cytochrome
P450BM-3,[25] D-
amino acid oxidase,
Phytase [27]
α-synuclein,
Phosphoribosylpyro
phosphate
amidotransferase
(purF) [29]
Randomization
scheme: NNK. NDT,
NNB, NAY [28]
Randomization
scheme: NNN, NNB,
URL
http://map.jacob
s-
university.de/su
bmission.html
http://guinevere
.otago.ac.nz/cgi-
bin/aef/pedel-
AA.pl
http://guinevere
.otago.ac.nz/cgi-
bin/aef/glue-
IT.pl
http://stat.haifa.
ac.il/~yuval/topl
6

PART I: CAPDE
library size and
randomization
techniques.
NNK, MAX [30]
ib/
Figure 1.2: a) The MAP 2.0 3D analysis for the amino acid diversity generated by balanced epPCR
(Taq (MnCl 2, G=A=C=T) method. Y-axis shows the original amino acid species and the X-axis shows
the amino acid substitution patterns indicated from red (lowest probability) to blue (highest
probability). The MAP 2.0 3D analysis is restricted to the active site residues (Ala11, Ser47, Thr48,
Tyr137, Ile139, Lys165, Thr167, Gly189, Tyr190). For this analysis, the amino acids are grouped
into four classes according to their chemical nature (charged, neutral, aromatic and aliphatic) with
stop codon ((structure disrupting) and glycine/proline (helix destabilizing) as separate classes. The
7

PART I: CAPDE
probabilities of amino acid substitutions were mapped on the protein sequence and structure (PDB
Id: 1NAL) of N-acetylneuraminic acid and represented in b and c, respectively. b) The Jmol [33]
applet is used for the visualization of amino acid substitution patterns using RWB (Red-white-blue)
color gradient scheme and active site residues as sticks. Y-axis shows sequence id, PDB id, amino
acid name and in c) secondary structure elements (T: hydrogen bonded turn and bend, *: loop or
irregular structure), d) normalized Cα b-factor to differentiate flexible (F) and rigid (R) residues,
and e) relative solvent associability to identify exposed (E) or buried (B) residues.
MAP [25] takes nucleotide sequence as input and assists to design better directed
evolution strategy by providing the statistical analysis of random mutagenesis methods on
protein level. The capabilities of MAP was extended in MAP 2.0 3D[27] server that predicts
the residue mutability resulted by the mutational bias of random mutagenesis methods and
correlates the generated amino acid substitution patterns with the structural information
of the target protein. In this way, the server offers the possibility to analyze the
consequences of the limitations of mutational preferences of random mutagenesis methods
on protein level and their effects on protein structure.[25] The capability of the server was
illustrated by the in-silico screening of different enzymes and the predicted results were in
agreement with the experimental results.[27,31,32 ] Figure 1.2 shows an example of the
MAP 2.0 3D output for active site residues of N-acetylneuraminic acid using epPCR
method.[27]
PEDAL-AA returns statistics, at amino acid level and for the libraries generated by
epPCR method, after providing the nucleotide sequence with library size, mutation rate,
indel rate and nucleotide mutation matrix.[28] CodonCalculator and AA-Calculator are two
algorithms developed by Patrik et al. to select an appropriate randomization scheme for
library construction.[28] Two servers GLUE-IT and GLUE estimate amino acid diversity and
completeness in the generated library. Finally, the TopLib [30] web server assists to design
saturation mutagenesis experiment by predicting the size or completeness of the generated
library with the user-defined codon randomization scheme using probabilistic approach.
8

PART I: CAPDE
1.4. Evolutionary conservation based focused library
Multiple sequence or structure alignment (MS) is the most common approach to
identify functionally significant or evolutionary variable regions in protein.[34] In CAPDE,
several servers and databases use MSA with the physical and structural information of
protein or protein superfamilies. Table 1.2 contains a list of the tools considered in this
chapter. ConSurf 2010 [35] server provides the evolutionary conservation profiles of
protein or nucleic acid sequence or structure by first identifying the conserved positions
using MSA and then calculating the evolutionary conservation rate using an empirical
Bayesian inference. ConSurf-DB [36] database make available the evolutionary
conservation profiles of the available protein structures pre-calculated by ConSurf web
server. The 3DM [37] server performs structure based multiple sequence alignments (MSA)
of the members of a protein superfamily and provides the consensus data combined with
other useful information, like interactions and solvent accessibility, about amino acid
positions in protein with published mutation data.
For more focused analysis of protein hotspots or amino acid patches, three
interesting tools are available as standalone programs or web servers. The Joint
Evolutionary Tree (JET) method is more tuned to identify the conserved amino acids
patches on protein interface by taking into account the physical-chemical properties and
evolutionary conservation of the surface residues.[38] The predicted protein interaction
sites or core residues might be used in site-specific mutagenesis experiments. HotSprint
[39] database provides information of the hotspots in protein interfaces using the sequence
conservation score (calculated by Rate4Site algorithm [40]) of the residues and their
solvent accessible surface area. HotSpot Wizard predicts the suitability of the mutagenesis
of the amino acids in or near the active site using their evolutionary conservation
information.[41] The server takes protein structure as input and provides a platform to
experimentalists to select target amino acids for site directed mutagenesis to improve
enzymatic properties like substrate specificities, activity and enantioselectivity.[41]
MAP 2.0 3D [27] (Table 1.1, see previous paragraph) also provides the information of
9

PART I: CAPDE
mutagenic hotspots generated due to the mutational preferences of the random
mutagenesis methods with sequence and structural information of protein. Selecton [42]
web server predicts the selective forces at each amino acid position in protein. The server
performs the codon-based alignment on a set of the homologous nucleotide sequences and
uses the ratio of amino acids altered to silent substitutions (Ka/Ks) to estimate both the
positive (>1) and purifying (

PART I: CAPDE
The web server
performs MSA and
ConSurf 2010
[35]
calculates evolutionary
conservation rate to
identify conserved
positions in protein or
GAL4
transcription
factor [35]
http://consur
f.tau.ac.il/
nucleotide
sequence/structure.
The database provides
ConSurf DB [36]
the predicted results of
ConSurf [35] server for
known protein
Cytochrome c
[36]
http://consur
fdb.tau.ac.il/in
dex.php
structures.
Hotspot
identificatio
n
The Evolutionary trace
based method performs
MSA on a set of
homologous sequences
DNA
polymerase I,
DNA
transferase,
(from PSI-BLAST) after
allophycocya
Gibbs like sampling.
nin, Leucine
JET [38]
The aligned
homologous sequences
are used to construct
distance tree based on
Neighbor Joining
dehydrogena
se, β-trypsin
proteinase,
phosphotrans
ferase,
http://www.i
hes.fr/~carbo
ne/data6/lege
nda.htm
algorithm. The
human
clustering method is
CDC42 gene
parameterized to
regulation
identify protein
protein,
interface or core
oncogene
residues by taking into
protein,
11

PART I: CAPDE
account the physical-
signal
chemical properties and
transduction
evolutionary
protein etc
conservation.
[38]
The database provides
information about
HotSprint
Database [39]
hotspots in protein
interface using
conservation rate and
Numb PTB
domain [39]
http://prism.c
cbb.ku.edu.tr/
hotsprint/
solvent accessibility of
the residues.
Haloalkane
dehalogenase
HotSpot wizard
[41]
The web server predicts
residue mutability of
functionally important
residues and visualizes
it on protein sequence
and structure.
,
Phosphotries
terase, 1,3-
1,4-b-D-
Glucan 4-
glucanohydro
lase, β-
http://loschm
idt.chemi.mun
i.cz/hotspotwi
zard/
Lactamase
[41]
The web server detects
Selecton [42]
selection forces on
biologically significant
sites in the target
protein during
TRIM5α
protein [42]
http://selecto
n.tau.ac.il/ind
ex.html
evolutionary process.
Protein
superfamily
3DM [37]
The database performs
structure based MSA for
α/β
hydrolase
http://3dmcsi
s.systemsbiol
12

PART I: CAPDE
based MSA
a protein superfamily
fold [53]
ogy.nl/
with sequence,
structural, molecular
interaction and
mutational information
from literature.
The Lipase
Engineering
Database
[43,54,55]
Lipases
[43,54,55]
http://www.l
ed.unistuttgart.de/
The database of
Epoxide
epoxide
hydrolases and
haloalkane
dehalogenase
The database performs
hydrolases
and
haloalkane
dehalogenase
http://www.l
ed.unistuttgart.de/
[56]
protein superfamily
[56]
The Laccase
based MSA and
http://www.l
Engineering
annotates functionally
Laccases [45]
cced.uni-
database [45]
relevant amino acid
stuttgart.de/
The Cytochrome
P450
engineering
database [57]
positions with
structural and
mutational information.
Cytochrome
P450s [57]
http://www.c
yped.unistuttgart.de/
The PHA
Depolymerase
Engineering
Database [44]
Polyhydroxya
lkanoates
depolymeras
e [44]
http://www.d
ed.unistuttgart.de/
The Lactamase
Engineering
database [46]
Lactamases
[46]
http://www.l
aced.unistuttgart.de/
13

PART I: CAPDE
SHV Lactamase
Engineering
Database [47]
SHV
lactamases
[47]
http://www.l
aced.unistuttgart.de/cl
assA/SHVED/
PMD [48]
The database provides
literature based protein
mutant information
with structure and
functional annotation.
http://pmd.d
dbj.nig.ac.jp/
~pmd/pmd.ht
ml
The database provides
literature based protein
Literature
based
ProTherm [49-
51]
mutant information
with thermodynamic
parameters and
experimental
conditions integrated
with sequence,
http://gibk26.
bio.kyutech.ac
.jp/jouhou/Pr
otherm/proth
erm.html
protein
structure and function
mutant data
annotation.
The database provides
literature based protein
MuteinDB [52]
mutant information,
kinetic parameters and
experimental
conditions integrated
with user-friendly and
flexible query system to
fetch data using
Cytochrome
P450s [52]
https://mutei
ndb.genome.t
ugraz.at/mute
indb-web-
2.0/faces/init
/index.seam
reaction name or
substrate or inhibitor
14

PART I: CAPDE
name or structure and
mutations.
1.5. Structure-based focused library
The structure based approaches assist rational design and random mutagenesis by
predicting regions in the protein responsible for stability and activity.[2,58] The
computational tools as 3DLigandSite [59], ProBiS [60,61] (Protein Binding Site) and
SiteComp [62] predict ligand binding site in protein [63]. All these tools, in the absence of
crystal structure, use homology model of the target protein and aid the design and tune
ligand binding site by identifying key residues for activity and their molecular interactions
properties. 3DLigandSite [59] performs alignment and clustering of the homologous
structures to predict ligand binding site. ProBiS [60,61] uses MSA to detect structurally
similar binding site in protein and also perform local structural pairwise alignment to
identify functionally relevant binding regions. The pre-calculated results of ProBiS analysis
are available via ProBiS-database [64] as a repository of structurally similar binding sites.
SiteComp [62] characterizes protein binding site using molecular interaction fields based
descriptors. The server evaluates differences in similar binding sites, identification of subsites
and residue contributions in ligand binding. TRITON [65,66] provides the single
platform to protein engineers to model mutants, perform protein-ligand docking and
calculate reaction pathways. In this way, these methods facilitate to study the properties of
protein-ligand complexes.
The knowledge of molecular interactions, contribute to relevant free energy barrier,
and the design of surface charge distribution, can help to understand the molecular basis of
kinetic stability and efficiently modulates the enhancement of protein stability.[58,67] PIC
(Protein Interaction Calculator) server [68] calculates inter or intra protein interactions
using published criteria integrated with solvent accessibility and residue depth
calculations. Recently introduced web server, COCOMAP (bioCOmplexes COntact MAPs)
15

PART I: CAPDE
[69] uses intermolecular interactions to analyze interfaces in biological complexes. The
identification of exposed and buried amino acids also helps to gain insight into protein
stability and to explore the mutational effect on protein. DEPTH [70] employ distance
information between residues and bulk solvent to predict protein stability, conservation or
binding cavity based on information about residue depth and solvent accessibility. SRide
[71] provides residual contribution to protein stability using interactions, evolutionary
conservations and hydrophobicity of their neighboring residues. Patch finder plus [72]
identifies residues that contribute to positively charge patches on protein surface and
might interact with DNA, membrane or the other protein. ConPlex [73] utilizes protein
solvent accessible surface area to identify surface or interface residues and assign residue
specific conservation score on sequence and structure of the protein complex. The server
also provides the pre-calculated ConPlex results of known protein complexes as repository.
Recent studies have suggested that protein flexibility and protein functions are
strongly linked.[24,74,75] Protein flexibility plays an important role in both catalytic
activity and molecular recognition processes. The effect of protein flexibility is particularly
relevant in protein from extremophiles to balance rigidity required for stability and
flexibility necessary for activity [76-78]. In addition, numerous proteins have regions that
adopt different conformation under different conditions, allowing them to take part in
cellular and molecular regulation.[24,79] The residue flexibility in protein has been taken
in account to describe a variety of protein properties including relation with thermal
stability, catalytic activity, ligand binding (induced fit), domain motion, preferential
solvation and molecular recognition in intrinsically disordered protein system. The Debye–
Waller factor, reported in crystallographic atomic resolution structures, provides an rough
estimation of local residue flexibility [80] and different servers provide this information as
an indicator (for example, in MAP 2.0 3D server [27]). If the crystallographic structure is not
available then different tools can be used to estimate flexibility profiles using different
approaches.
The RosettaBackrub [81] server can generate protein backbone structural variability
as consequence of amino acid variations [82] that can be used to design sequence libraries
16

PART I: CAPDE
for experimental screening and to predict protein or peptide interaction specificity. The
server generates Rosetta scored modeled structures for variant with single or multiple
point mutations in monomeric proteins. It also generates near-native structural ensembles
of protein backbone conformations and sequences consistent with those ensembles.
Finally, it can predict sequences tolerated by proteins or protein interfaces using flexible
backbone design methods. The tCONCOORD [83] method generates conformational
ensembles to gain insight in the conformational flexibility and conformational space of the
protein.
FlexPred [84] specially predicts residue flexibility using pattern recognition
approach to identify residue positions in conformations switches integrated with their
evolutionary conservation and normalized solvent accessibility (if structure is available) as
the Support Vector Machine (SVM) predictors.
Different simplified methods have been proposed to identify local flexibility or large
scale motions in protein at coarse-grained level [85-87] Many of these methods are based
on Gaussian network model (GNM) [88] or its extension, the anisotropic network model
(ANM) [89] to study protein dynamics using Normal Mode Analysis (NMA) (see the review
[90] for a general overview about these topics). Table 1.3 shows the tools available to
analyze conformational flexibility on protein structure (for more details see [91]). ElNemo
[92] and WEBnb@ [93] servers are reported here to complete the information about NMA
based tools. Both the servers perform NMA using coarse grain model to analyze the
conformational changes in protein. FlexServ [94] server estimates protein flexibility using
three different coarse-grained approaches: 1) discrete molecular dynamics (DMD), 2)
normal mode analysis (NMA) and 3) Brownian dynamics (BD). The server characterizes
protein flexibility by analyzing different structural and dynamic properties of the protein
such as structural variations, essential modes, stiffness between the interacting residues
and dynamic domains and hinge points. Different tools are available to identify hinge
bending residues on large-scale protein motions. HINGEprot [95] server predicts hinge
motion in protein using coarse grained GNM and ANM model. DynDom [96] use a rigorous
approach to describe domain motion. The method determines hinge axes and hinge
17

PART I: CAPDE
bending residues using two conformations of the protein. A recent addition to DynDom is
the ligand-induced domain movements in enzymes database.[97] Furthermore, the
Dyndom3D [98] server provides a more advanced and generic tool that can be used to
study any kind of polymer.
The reader should be noticed that the connection between protein flexibility and
function has been investigated theoretically and experimentally only in the last few years.
[87,99-101] The methods based on this approach provide a qualitative estimation of
protein dynamical properties but they do not take in account many effects (such as direct
solvent effects) that are important for protein functionality. Till now, the atomistic
simulation (MD or QM/MD) is the best approach to quantitatively study protein flexibility
and dynamics.[8,87,99] Nevertheless, even to this level of accuracy, the connection
between flexibility and functionality is still puzzling. In addition, the simulation approaches
are still time consuming and unpractical for high-throughput modeling and analysis of
protein structural dynamics.
Table 1.3: Summarizing the computational tools for structure-based focused library
generation.
Approach Name Description
The web server
identifies ligand
3DLigandSite
binding site via MSA
[59]
Ligand
and clustering
binding site
algorithm.
The web server
ProBiS [60,61] detects binding site
using MSA and
Case study
examples
Target
T0483 in
CASP8
Biotin
carboxylase,
TATA
URL
http://www.sbg.bio.i
c.ac.uk/~3dligandsit
e/
http://probis.cmm.ki
.si/
18

PART I: CAPDE
characterizes it
binding
using local
protein [60],
structural pairwise
D-alanine–
alignment.
D-alanine
ligase,
Protein
kinases C
[61]
The database
provides
ProBiS-
structurally similar
Cytochrome
http://probis.cmm.ki
database [64]
protein binding site
c [64]
.si/?what=database
using ProBiS
algorithm.
The web server
SiteComp [62]
characterizes ligand
binding site using
molecular
interaction
Cyclooxygen
ase,
adenylate
kinase [62]
http://scbx.mssm.ed
u/sitecomp/sitecom
p-web/Input.html
descriptors.
The method
facilitates to model
mutant, dock ligand
TRITON
[65,66]
in the protein and
calculates reaction
pathways for the
characterization of
PA-IIL lectin
and its
mutants
[65]
http://www.ncbr.mu
ni.cz/triton/descripti
on.html
protein-ligand
interactions using
Semi-empirical
19

PART I: CAPDE
quantum-mechanics
approach.
The web server
PIC [68]
calculates the
molecular
interactions using
-
http://pic.mbu.iisc.er
net.in/job.html
published criteria.
The web server
Protein
interaction
COCOMAPS
[69]
analyzes and
visualizes interfaces
in biological
complexes using
intermolecular
contact maps based
Hen egg
lysozyme
interaction
with two
antibodies
https://www.molnac
.unisa.it/BioTools/co
comaps/
on distance or
[69]
physicochemical
properties.
The web server
predicts binding
cavity and
West Nile
mutational effect on
Virus
http://mspc.bii.a-
DEPTH [70]
protein stability
NS2B/NS3
star.edu.sg/tankp/int
Residue
using residue depth
protease
ro.html
depth and
and solvent
[70]
stability
accessible surface
area.
SRIde [71]
The web serve
predicts the
contribution of
residues in protein
TIM-barrel
proteins
[102]
http://sride.enzim.h
u/
20

PART I: CAPDE
stability using
interactions with its
spatial neighbors
and their
evolutionary
conservation.
The web server
identifies large
positively charged
DNA binding
Patch finder
plus [72]
electrostatic patches
on protein surface
using Poisson
domain of
TATA
binding
http://pfp.technion.a
c.il/
Protein
Boltzmann
protein [72]
surface and
electrostatic
interface
potential.
The web server
performs
Rho–
ConPlex [73]
evolutionary
RhoGAP
http://sbi.postech.ac.
conservation
complex
kr/ConPlex/
analysis of the
[73]
protein complex.
The web server
performs flexible
https://kortemmelab
Protein
flexibility
RosettaBackru
b [81]
backbone modeling
using Backrub [103]
method to design
hGH-hGHr
interface
[104]
.ucsf.edu/backrub/cg
i-
bin/rosettaweb.py?q
tolerated protein
uery=index
sequences.
tCONCOORD
The method
Osmoprotec
http://wwwuser.gw
[83]
generates
tion protein
dg.de/~dseelig/tcon
21

PART I: CAPDE
FlexPred [84]
ElNemo [92]
WEBnm@
[93]
FlexServ [94]
HINGEprot
[95]
conformation
ensemble and
transitions using
geometrical
constrains based
prediction of
protein
conformational
flexibility
The web server
predicts residue
flexibility in the
protein using SVM
approach.
The web server
predicts large
amplitude motions
in the protein using
NMA.
The web server
determines and
analyzes protein
flexibility using
coarse-grained
modeling approach.
The web server
detects hinge region
[83] coord.html
Human PrP http://flexpred.rit.al
[105] bany.edu/
HIV-1
protease, E. http://igs-
coli server.cnrs-
membrane
mrs.fr/elnemo/index
channel .html
protein TolC
Calcium http://apps.cbu.uib.n
ATPase [93] o/webnma/home
http://mmb.pcb.ub.e
-
s/FlexServ/input.ph
p
Calmodulin http://www.prc.bou
protein, n.edu.tr/appserv/prc
22

PART I: CAPDE
in the protein using
hemoglobin
/hingeprot/
both GNM and ANM.
[95]
The web server
predicts domain
Hemoglobin,
DynDom3D
motions using
70S
http://fizz.cmp.uea.a
[98]
conformational
ribosome
c.uk/dyndom/3D/
changes in the
[98]
protein.
1.6. Mutational effects in protein
For biotechnological applications, the enhancement of protein thermal stability or
tolerance is a common requested task in protein engineering.[106] Highly stable structure
correlates with well-packed highly compact structure and has increased tolerance to
mutation because mostly the mutations are deleterious i.e. related to instability of
protein.[107] Generally the effect of the mutation on protein has been calculated by the
free energy differences between two states of protein like thermodynamic stability as
change in free energy in folded and unfolded state (ΔΔG). The mutational effect has been
predicted by using different machine learning and selection methods (as SVM, Decision
Tree (DT) or Random Forest (RE) [108]) for classification or regression of data or by using
statistical or empirical methods taking into account the atomic interactions or structural
properties like solvent accessibility. Most of the servers based on these approaches use
available information of mutational effects (fetched from databases like PMD [48],
ProTherm [51]) to predict the effect of new substitutions. Table 1.4 summarizes the
available tools to predict mutational effects on protein stability and activity using different
methods. I-Mutant2.0 [109] and MUpro [110] are SVM based methods to predict stabilizing
or destabilizing amino acid substitutions based on free energy change (ΔΔG). iPTREE-STAB
[111] server employ a DT approach to predict the effect of single mutation on protein
stability considering physicochemical properties and contact information of the substituted
23

PART I: CAPDE
amino acid with their neighboring amino acids. WET-STAB [112] server performs a similar
prediction with an additional feature to predict protein stability changes upon double
mutations from amino acid sequence. ProMAYA [113] uses RF machine learning algorithm
to predict protein stability based on free energy difference. MuD (Mutation detector) uses
the same algorithm for the classification of amino acid substitutions as neutral or
deleterious by taking into account structure- and sequence-based features as solvent
accessibility, binding site, sequence identity.[114] SDM (Site Directed Mutator) [115] and
PopMuSic2.1 [116] are statistical derived force field potential based methods for protein
stability prediction using relative free energy differences. In PopMuSic2.1 [116], however,
the parameters of statistical derived force field potential depend on protein solvent
accessibility. FoldX plugin [117] and PEAT-SA [118] program suite utilize empirical force
field to calculate, from three-dimensional protein or peptides structures, the relative free
energy difference determined by the changes of interactions in the mutated structures.
CUPSAT [119] estimates the effect of mutations on the protein stability using protein
environment specific mean force potentials. The potentials are derived from statistical
analysis of protein structure data sets. AUTO-MUTE [120,121] provides either energy based
or machine learning methods for the prediction of protein stability by providing protein
structure, mutation and experimental condition. SIFT (Sorts Intolerant From Tolerant)
[122] server helps to explore the effect of mutation on protein function using sequence
homology approach. The multiple alignment information is used to identify tolerated and
deleterious substitutions in the query sequence.
A quantitative in-silico screening of the virtual libraries based on the cooperative
effect of multiple mutations to the stability and functionality is still out of reach. However,
the current methods allow a qualitative indication of possible mutation sites that can
increase the chances to get higher population of stable and functionally active variants in
the library. The available knowledge of mutational effects on protein provided by all these
CAPDE approaches help to limit library size and focus to generate unpredictable
substitutions that may lead to large effects. These libraries based on in-silico screening
generally show a higher success rate when the starting protein has sufficient stability.
24

PART I: CAPDE
Table 1.4: Summarizing the computational tools to analyze the mutational effect on
protein stability and activity.
Approach Name Description URL
SVM
I-Mutant2.0
[109]
MUpro [110]
The web server predicts
protein stability change
upon point mutation.
http://folding.uib.es/imutant/i-mutant2.0.html
http://mupro.proteomics.ic
s.uci.edu/
The web server predicts
iPTREE-STAB
protein stability change
http://210.60.98.19/IPTRE
[111]
with residues
Er/iptree.htm
Decision tree
information.
(DT)
The web server predicts
WET-STAB [112]
protein stability change
upon double mutation
with residue
http://210.60.98.19/WETr
/wet.htm
information.
Random
forests (RF)
ProMAYA [113]
MuD [114]
The web server predicts
mutational effect on
protein function.
http://bental.tau.ac.il/Pro
Maya/
http://mud.tau.ac.il/
Statistical
potential
SDM [115]
The web server predicts
mutational effect on
protein stability.
http://mordred.bioc.cam.ac
.uk/sdm/sdm.php
based
method
PopMuSic2.1
[116]
The web server predicts
thermodynamic stability
change upon mutation.
http://babylone.ulb.ac.be/
popmusic/
Empirical
force field
FoldX [117]
The plugin predicts
mutational effect on
http://foldx.crg.es/
25

PART I: CAPDE
protein and facilitates in-
silico alanine screening,
mutant homology
modeling and
interaction energy
calculation.
The program suite
PEAT-SA [118]
predict mutational effect
on protein stability,
ligand affinity and pKa
http://enzyme.ucd.ie/PEA
TSA/Pages/FrontPage.php
values.
The web server predicts
CUPSAT [119]
mutational effect on
http://cupsat.tu-bs.de/
protein stability.
The web server predicts
RF, SVM, Tree
and SVM
regression
AUTO-MUTE
[121]
mutational effect on
protein stability and
activity (up to 19
http://proteins.gmu.edu/a
utomute/
mutations).
Evolutionary
conservation
SIFT [122]
The web server predicts
mutational effect on
protein function.
http://sift.jcvi.org/
1.7. Summary and outlook
In this chapter, the recent additions to the CAPDE arsenal of computational tools,
servers and databases have been briefly reviewed. The rapid accumulation of the
knowledge on protein structures and sequence-structure-function relationships foresee
the continuous amelioration of these methods. In particular, machine-learning approaches,
26

PART I: CAPDE
in which the volume of data is the heuristic key to access the hidden knowledge, statistical
based force fields for coarse-grained approaches will surely benefit this trend. These
approaches are not only the convenient aids to support lab experiments but also the
workbench for heuristically blueprinting novel molecules. In addition, the availability of the
low cost and high performance computers will soon transform currently expensive
physically based simulations to the convenient and very accurate high throughput
computational tools. This will make possible to predict structural stability and folds of
small or medium sized proteins and will open a new working style paradigm in protein
engineering. In addition, the physical based approach has already shown promising results
to understand enzyme activity.[123,124]
1.8. References
1. Bornscheuer UT, Huisman GW, Kazlauskas RJ, Lutz S, Moore JC, et al. (2012)
Engineering the third wave of biocatalysis. Nature 485: 185-194.
2. Lutz S (2010) Beyond directed evolution-semi-rational protein engineering and
design. Curr Opin Biotech 21: 734-743.
3. Gerlt JA, Babbitt PC (2009) Enzyme (re)design: lessons from natural evolution and
computation. Curr Opin Chem Biol 13: 10-18.
4. Jackel C, Kast P, Hilvert D (2008) Protein design by directed evolution. Annu Rev
Biophys 37: 153-173.
5. Damborsky J, Brezovsky J (2009) Computational tools for designing and engineering
biocatalysts. Curr Opin Chem Biol 13: 26-34.
6. Suarez M, Jaramillo A (2009) Challenges in the computational design of proteins. J R
Soc Interface 6 (Suppl 4): S477–S491.
7. Pantazes RJ, Grisewood MJ, Maranas CD (2011) Recent advances in computational
protein design. Curr Opin Struct Biol 21: 467-472.
8. Dror RO, Dirks RM, Grossman JP, Xu H, Shaw DE (2012) Biomolecular simulation: a
computational microscope for molecular biology. Annu Rev Biophys 41: 429-452.
27

PART I: CAPDE
9. Lee EH, Hsin J, Sotomayor M, Comellas G, Schulten K (2009) Discovery through the
computational microscope. Structure 17: 1295-1306.
10. Schlick T, Collepardo-Guevara R, Halvorsen LA, Jung S, Xiao X (2011)
Biomolecularmodeling and simulation: a field coming of age. Q Rev Biophys 44: 191-
228.
11. McGeagh JD, Ranaghan KE, Mulholland AJ (2011) Protein dynamics and enzyme
catalysis: Insights from simulations. BBA-Proteins Proteom 1814: 1077-1092.
12. Klepeis JL, Lindorff-Larsen K, Dror RO, Shaw DE (2009) Long-timescale molecular
dynamics simulations of protein structure and function. Curr Opin Struct Biol 19:
120-127.
13. Barrozo A, Borstnar R, Marloie Gl, Kamerlin SCL (2012) Computational protein
engineering: bridging the gap between rational design and laboratory evolution. Int
J Mol Sci 13: 12428-12460.
14. Frushicheva MP, Cao J, Warshel A (2011) Challenges and advances in validating
enzyme design proposals: the case of kemp eliminase catalysis. Biochemistry 50:
3849-3858.
15. Frushicheva MP, Warshel A (2012) Towards quantitative computer-aided studies of
enzymatic enantioselectivity: the case of Candida antarctica lipase A. Chembiochem
13: 215-223.
16. van der Kamp MW, Mulholland AJ (2008) Computational enzymology: insight into
biological catalysts from modelling. Nat Prod Rep 25: 1001-1014.
17. Turner NJ (2009) Directed evolution drives the next generation of biocatalysts. Nat
Chem Biol 5: 567-573.
18. Arnold FH, Moore JC (1997) Optimizing industrial enzymes by directed evolution.
Adv Biochem Eng Biotechnol 58: 1-14.
19. Tracewell CA, Arnold FH (2009) Directed enzyme evolution: climbing fitness peaks
one amino acid at a time. Curr Opin Chem Biol 13: 3-9.
20. Wong TS, Roccatano D, Schwaneberg U (2007) Steering directed protein evolution:
strategies to manage combinatorial complexity of mutant libraries. Environ
Microbiol 9: 2645-2659.
28

PART I: CAPDE
21. Chica RA, Doucet N, Pelletier JN (2005) Semi-rational approaches to engineering
enzyme activity: combining the benefits of directed evolution and rational design.
Curr Opin Biotech 16: 378-384.
22. Kazlauskas RJ, Bornscheuer UT (2009) Finding better protein engineering
strategies. Nat Chem Biol 5: 526-529.
23. Romero PA, Arnold FH (2009) Exploring protein fitness landscapes by directed
evolution. Nat Rev Mol Cell Biol 10: 866-876.
24. Tokuriki N, Tawfik DS (2009) Stability effects of mutations and protein evolvability.
Curr Opin Struct Biol 19: 596-604.
25. Wong TS, Roccatano D, Zacharias M, Schwaneberg U (2006) A statistical analysis of
random mutagenesis methods used for directed protein evolution. J Mol Biol 355:
858-871.
26. Shivange AV, Marienhagen J, Mundhada H, Schenk A, Schwaneberg U (2009)
Advances in generating functional diversity for directed protein evolution. Curr
Opin Chem Biol 13: 19-25.
27. Verma R, Schwaneberg U, Roccatano D (2012) MAP2.03D: a sequence/structure
based server for protein engineering. ACS Synth Biol 1: 139-150.
28. Firth AE, Patrick WM (2008) GLUE-IT and PEDEL-AA: new programmes for
analyzing protein diversity in randomized libraries. Nucleic Acids Res 36: W281-
W285.
29. Patrick WM, Matsumura I (2008) A study in molecular contingency: glutamine
phosphoribosylpyrophosphate amidotransferase is a promiscuous and evolvable
phosphoribosylanthranilate isomerase. J Mol Biol 377: 323-336.
30. Nov Y (2011) When second best is good enough: another probabilistic look at
saturation mutagenesis. Appl Environ Microbiol 78: 258-262.
31. Rasila TS, Pajunen MI, Savilahti H (2009) Critical evaluation of random mutagenesis
by error-prone polymerase chain reaction protocols, Escherichia coli mutator strain,
and hydroxylamine treatment. Anal Biochem 388: 71-80.
32. Ruff AJ, Marienhagen J, Verma R, Roccatano D, Genieser H-G, et al. (2012) dRTP and
dPTP a complementary nucleotide couple for the Sequence Saturation Mutagenesis
(SeSaM) method. J Mol Catal B-Enzym 84: 40-47.
29

PART I: CAPDE
33. Jmol: an open-source Java viewer for chemical structures in 3D.
http://www.jmol.org/
34. Pei J (2008) Multiple protein sequence alignment. Curr Opin Struct Biol 18: 382-386.
35. Ashkenazy H, Erez E, Martz E, Pupko T, Ben-Tal N (2010) ConSurf 2010: calculating
evolutionary conservation in sequence and structure of proteins and nucleic acids.
Nucleic Acids Res 38: W529-533.
36. Goldenberg O, Erez E, Nimrod G, Ben-Tal N (2009) The ConSurf-DB: pre-calculated
evolutionary conservation profiles of protein structures. Nucleic Acids Res 37:
D323-D327.
37. Kuipers RK, Joosten H-J, van Berkel WJH, Leferink NGH, Rooijen E, et al. (2010) 3DM:
Systematic analysis of heterogeneous superfamily data to discover protein
functionalities. Proteins 78: 2101-2113.
38. Engelen S, Trojan LA, Sacquin-Mora S, Lavery R, Carbone A (2009) Joint
Evolutionary Trees: a large-scale method to predict protein interfaces based on
sequence sampling. PLoS Comput Biol 5: e1000267.
39. Guney E, Tuncbag N, Keskin O, Gursoy A (2008) HotSprint: database of
computational hot spots in protein interfaces. Nucleic Acids Res 36: D662-D666.
40. Pupko T, Bell RE, Mayrose I, Glaser F, Ben-Tal N (2002) Rate4Site: an algorithmic
tool for the identification of functional regions in proteins by surface mapping of
evolutionary determinants within their homologues. Bioinformatics 18: S71-S77.
41. Pavelka A, Chovancova E, Damborsky J (2009) HotSpot Wizard: a web server for
identification of hot spots in protein engineering. Nucleic Acids Res 37: W376-
W383.
42. Stern A, Doron-Faigenboim A, Erez E, Martz E, Bacharach E, et al. (2007) Selecton
2007: advanced models for detecting positive and purifying selection using a
Bayesian inference approach. Nucleic Acids Res 35: W506-W511.
43. Pleiss Jr, Fischer M, Peiker M, Thiele C, Rolf D (2000) Lipase Engineering Database:
understanding and exploiting sequence-structure-function relationships. J Mol Catal
B-Enzym 10: 491-508.
30

PART I: CAPDE
44. Knoll M, Hamm TM, Wagner F, Martinez V, Pleiss J (2009) The PHA Depolymerase
Engineering Database: a systematic analysis tool for the diverse family of
polyhydroxyalkanoate (PHA) depolymerases. BMC Bioinformatics 10: 89.
45. Sirim D, Wagner F, Wang L, Schmid RD, Pleiss J (2010) The Laccase Engineering
Database: a classification and analysis system for laccases and related multicopper
oxidases. Database 2011: bar006.
46. Thai QK, Bos F, Pleiss J (2009) The Lactamase Engineering Database: a critical
survey of TEM sequences in public databases. BMC Genomics 10: 390.
47. Thai QK, Pleiss J (2010) SHV Lactamase Engineering Database: a reconciliation tool
for SHV beta-lactamases in public databases. BMC Genomics 11: 563.
48. Kawabata T, Ota M, Nishikawa K (1999) The Protein Mutant Database. Nucleic Acids
Res 27: 355-357.
49. Gromiha MM, Uedaira H, An J, Selvaraj S, Prabakaran P, et al. (2002) ProTherm,
thermodynamic database for proteins and mutants: developments in version 3.0.
Nucleic Acids Res 30: 301-302.
50. Gromiha MM, An J, Kono H, Oobatake M, Uedaira H, et al. (2000) ProTherm, version
2.0: thermodynamic database for proteins and mutants. Nucleic Acids Res 28: 283-
285.
51. Bava KA, Gromiha MM, Uedaira H, Kitajima K, Sarai A (2004) ProTherm, version 4.0:
thermodynamic database for proteins and mutants. Nucleic Acids Res 32: D120-121.
52. Braun A, Halwachs B, Geier M, Weinhandl K, Guggemos M, et al. (2012) MuteinDB:
the mutein database linking substrates, products and enzymatic reactions directly
with genetic variants of enzymes. Database 2012.
53. Kourist R, Jochens H, Bartsch S, Kuipers R, Padhi SK, et al. (2010) The alpha/betahydrolase
fold 3DM database (ABHDB) as a tool for protein engineering.
Chembiochem 11: 1635-1643.
54. Fischer M, Pleiss J (2003) The Lipase Engineering Database: a navigation and
analysis tool for protein families. Nucleic Acids Res 31: 319-321.
55. Widmann M, Juhl PB, Pleiss J (2010) Structural classification by the Lipase
Engineering Database: a case study of Candida antarctica lipase A. BMC Genomics
11: 123.
31

PART I: CAPDE
56. Barth S, Fischer M, Schmid RD, Pleiss J (2004) The database of epoxide hydrolases
and haloalkane dehalogenases: one structure, many functions. Bioinformatics 20:
2845-2847.
57. Sirim D, Wagner F, Lisitsa A, Pleiss J (2009) The cytochrome P450 engineering
database: Integration of biochemical properties. BMC Biochem 10: 27.
58. Gong S, Worth CL, Bickerton GR, Lee S, Tanramluk D, et al. (2009) Structural and
functional restraints in the evolution of protein families and superfamilies. Biochem
Soc Trans 37: 727-733.
59. Wass MN, Kelley LA, Sternberg MJ (2010) 3DLigandSite: predicting ligand-binding
sites using similar structures. Nucleic Acids Res 38: W469-473.
60. Konc J, Janezic D (2010) ProBiS algorithm for detection of structurally similar
protein binding sites by local structural alignment. Bioinformatics 26: 1160-1168.
61. Konc J, Janezic D (2012) ProBiS-2012: web server and web services for detection of
structurally similar binding sites in proteins. Nucleic Acids Res 40: W214-221.
62. Lin Y, Yoo S, Sanchez R (2012) SiteComp: a server for ligand binding site analysis in
protein structures. Bioinformatics.
63. Liang J, Tseng YY, Dundas J, Binkowski TA, Joachimiak A, et al. (2008) Predicting and
characterizing protein functions through matching geometric and evolutionary
patterns of binding surfaces. Adv Protein Chem Struct Biol 75: 107-141.
64. Konc J, Cesnik T, Konc JT, Penca M, Janezic D (2012) ProBiS-database: precalculated
binding site similarities and local pairwise alignments of PDB structures. J Chem Inf
Model 52: 604-612.
65. Prokop M, Damborsky J, Koca J (2000) TRITON: in silico construction of protein
mutants and prediction of their activities. Bioinformatics 16: 845-846.
66. Prokop M, Adam J, Kriz Z, Wimmerova M, Koca J (2008) TRITON: a graphical tool for
ligand-binding protein engineering. Bioinformatics 24: 1955-1956.
67. Sanchez-Ruiz JM (2010) Protein kinetic stability. Biophys Chem 148: 1-15.
68. Tina KG, Bhadra R, Srinivasan N (2007) PIC: Protein Interactions Calculator. Nucleic
Acids Res 35: W473-476.
32

PART I: CAPDE
69. Vangone A, Spinelli R, Scarano V, Cavallo L, Oliva R (2011) COCOMAPS: a web
application to analyse and visualize contacts at the interface of biomolecular
complexes. Bioinformatics.
70. Tan KP, Varadarajan R, Madhusudhan MS (2011) DEPTH: a web server to compute
depth and predict small-molecule binding cavities in proteins. Nucleic Acids Res.
71. Magyar C, Gromiha MM, Pujadas G, Tusnady GE, Simon I (2005) SRide: a server for
identifying stabilizing residues in proteins. Nucleic Acids Res 33: W303-305.
72. Shazman S, Celniker G, Haber O, Glaser F, Mandel-Gutfreund Y (2007) Patch Finder
Plus (PFplus): a web server for extracting and displaying positive electrostatic
patches on protein surfaces. Nucleic Acids Res 35: W526-W530.
73. Choi YS, Han SK, Kim J, Yang J-S, Jeon J, et al. (2010) ConPlex: a server for the
evolutionary conservation analysis of protein complex structures. Nucleic Acids Res
38: W450-W456.
74. Teilum K, Olsen JG, Kragelund BB (2011) Protein stability, flexibility and function.
Biochim Biophys Acta 1814: 969-976.
75. Teilum K, Olsen JG, Kragelund BB (2009) Functional aspects of protein flexibility.
Cell Mol Life Sci 66: 2231-2247.
76. Henzler-Wildman K, Kern D (2007) Dynamic personalities of proteins. Nature 450:
964-972.
77. Mittermaier AK, Kay LE (2009) Observing biological dynamics at atomic resolution
using NMR. Trends Biochem Sci 34: 601-611.
78. Martinez R, Schwaneberg U, Roccatano D (2011) Temperature effects on structure
and dynamics of the psychrophilic protease subtilisin S41 and its thermostable
mutants in solution. Protein Eng Des Sel 24: 533-544.
79. Ma B, Nussinov R (2010) Enzyme dynamics point to stepwise conformational
selection in catalysis. Curr Opin Chem Biol 14: 652-659.
80. Zhang H, Zhang T, Chen K, Shen SY, Ruan JS, et al. (2009) On the relation between
residue flexibility and local solvent accessibility in proteins. Proteins 76: 617-636.
81. Lauck F, Smith CA, Friedland GF, Humphris EL, Kortemme T (2010) RosettaBackruba
web server for flexible backbone protein structure modeling and design. Nucleic
Acids Res 38: W569-W575.
33

PART I: CAPDE
82. Mandell DJ, Kortemme T (2009) Backbone flexibility in computational protein
design. Curr Opin Biotech 20: 420-428.
83. Seeliger D, Haas Jr, de Groot BL (2007) Geometry-based sampling of conformational
transitions in proteins. Structure 15: 1482-1492.
84. Kuznetsov IB, McDuffie M (2008) FlexPred: a web-server for predicting residue
positions involved in conformational switches in proteins. Bioinformation 3: 134-
136.
85. Bahar I, Lezon TR, Yang L-W, Eyal E (2010) Global dynamics of proteins: bridging
between structure and function. Ann Rev Biophys 39: 23-42.
86. Bahar I, Rader AJ (2005) Coarse-grained normal mode analysis in structural biology.
Curr Opin Struct Biol 15: 586-592.
87. Kamerlin SCL, Vicatos S, Dryga A, Warshel A (2011) Coarse-grained (multiscale)
simulations in studies of biophysical and chemical systems. Annu Rev Phys Chem
62: 41-64.
88. Bahar I, Atilgan AR, Erman B (1997) Direct evaluation of thermal fluctuations in
proteins using a single-parameter harmonic potential. Fold Des 2: 173-181.
89. Atilgan AR, Durell SR, Jernigan RL, Demirel MC, Keskin O, et al. (2001) Anisotropy of
fluctuation dynamics of proteins with an elastic network model. Biophys J 80: 505-
515.
90. Skjaerven L, Hollup SM, Reuter N (2009) Normal mode analysis for proteins. J Mol
Struc- Theochem 898: 42-48.
91. Liu X, Karimi HA (2007) High-throughput modeling and analysis of protein
structural dynamics. Brief Bioinform 8: 432-445.
92. Suhre K, Sanejouand Y-H (2004) ElNemo: a normal mode web server for protein
movement analysis and the generation of templates for molecular replacement.
Nucleic Acids Res 32: W610-W614.
93. Hollup S, Salensminde G, Reuter N (2005) WEBnm@: a web application for normal
mode analyses of proteins. BMC Bioinformatics 6: 52.
94. Camps J, Carrillo O, Emperador A, Orellana L, Hospital A, et al. (2009) FlexServ: an
integrated tool for the analysis of protein flexibility. Bioinformatics 25: 1709-1710.
34

PART I: CAPDE
95. Emekli U, Schneidman-Duhovny D, Wolfson HJ, Nussinov R, Haliloglu T (2008)
HingeProt: automated prediction of hinges in protein structures. Proteins 70: 1219-
1227.
96. Hayward S, Berendsen HJC (1998) Systematic analysis of domain motions in
proteins from conformational change: New results on citrate synthase and T4
lysozyme. Proteins 30: 144-154.
97. Qi GY, Hayward S (2009) Database of ligand-induced domain movements in
enzymes. BMC Struct Biol 9.
98. Poornam GP, Matsumoto A, Ishida H, Hayward S (2009) A method for the analysis of
domain movements in large biomolecular complexes. Proteins 76: 201-212.
99. Glowacki DR, Harvey JN, Mulholland AJ (2012) Taking Ockham's razor to enzyme
dynamics and catalysis. Nature Chemistry 4: 169-176.
100. Pisliakov AV, Cao J, Kamerlin SCL, Warshel A (2009) Enzyme millisecond
conformational dynamics do not catalyze the chemical step. Proc Natl Acad Sci USA
106: 17359-17364.
101. Roca M, Vardi-Kilshtain A, Warshel A (2009) Toward accurate screening in
computer-aided enzyme design. Biochemistry 48: 3046-3056.
102. Gromiha MM, Pujadas G, Magyar C, Selvaraj S, Simon I (2004) Locating the
stabilizing residues in (α/β)8 barrel proteins based on hydrophobicity, long-range
interactions, and sequence conservation. Proteins 55: 316-329.
103. Davis IW, Arendall WB, 3rd, Richardson DC, Richardson JS (2006) The backrub
motion: how protein backbone shrugs when a sidechain dances. Structure 14: 265-
274.
104. Humphris EL, Kortemme T (2008) Prediction of Protein-Protein Interface Sequence
Diversity Using Flexible Backbone Computational Protein Design. Structure 16:
1777-1788.
105. Kuznetsov IB (2008) Ordered conformational change in the protein backbone:
prediction of conformationally variable positions from sequence and low-resolution
structural data. Proteins 72: 74-87.
106. Bloom JD, Arnold FH (2009) In the light of directed evolution: pathways of adaptive
protein evolution. Proc Natl Acad Sci USA 106 Suppl 1: 9995-10000.
35

PART I: CAPDE
107. Tokuriki N, Stricher F, Serrano L, Tawfik DS (2008) How protein stability and new
functions trade off. PLoS Comput Biol 4: e1000002.
108. Saeys Y, Inza I, Larranaga P (2007) A review of feature selection techniques in
bioinformatics. Bioinformatics 23: 2507-2517.
109. Capriotti E, Fariselli P, Casadio R (2005) I-Mutant2.0: predicting stability changes
upon mutation from the protein sequence or structure. Nucleic Acids Res 33: W306-
310.
110. Cheng J, Randall A, Baldi P (2006) Prediction of protein stability changes for singlesite
mutations using support vector machines. Proteins 62: 1125-1132.
111. Huang LT, Gromiha MM, Ho SY (2007) iPTREE-STAB: interpretable decision tree
based method for predicting protein stability changes upon mutations.
Bioinformatics 23: 1292-1293.
112. Huang LT, Gromiha MM (2009) Reliable prediction of protein thermostability
change upon double mutation from amino acid sequence. Bioinformatics 25: 2181-
2187.
113. Wainreb G, Wolf L, Ashkenazy H, Dehouck Y, Ben-Tal N (2011) Protein stability: a
single recorded mutation aids in predicting the effects of other mutations in the
same amino acid site. Bioinformatics 27: 3286-3292.
114. Wainreb G, Ashkenazy H, Bromberg Y, Starovolsky-Shitrit A, Haliloglu T, et al.
(2010) MuD: an interactive web server for the prediction of non-neutral
substitutions using protein structural data. Nucleic Acids Res 38: W523-W528.
115. Worth CL, Preissner R, Blundell TL (2011) SDM:a server for predicting effects of
mutations on protein stability and malfunction. Nucleic Acids Res 39: W215-W222.
116. Dehouck Y, Kwasigroch J, Gilis D, Rooman M (2011) PoPMuSiC 2.1: a web server for
the estimation of protein stability changes upon mutation and sequence optimality.
BMC Bioinformatics 12: 151.
117. Van Durme J, Delgado J, Stricher F, Serrano L, Schymkowitz J, et al. (2011) A
graphical interface for the FoldX forcefield. Bioinformatics 27: 1711-1712.
118. Johnston MA, Søndergaard CR, Nielsen JE (2011) Integrated prediction of the effect
of mutations on multiple protein characteristics. Proteins 79: 165-178.
36

PART I: CAPDE
119. Parthiban V, Gromiha MM, Schomburg D (2006) CUPSAT: prediction of protein
stability upon point mutations. Nucleic Acids Res 34: W239-242.
120. Masso M, Vaisman II (2008) Accurate prediction of stability changes in protein
mutants by combining machine learning with structure based computational
mutagenesis. Bioinformatics 24: 2002-2009.
121. Masso M, Vaisman II (2010) AUTO-MUTE: web-based tools for predicting stability
changes in proteins due to single amino acid replacements. Protein Eng Des Sel 23:
683-687.
122. Kumar P, Henikoff S, Ng PC (2009) Predicting the effects of coding non-synonymous
variants on protein function using the SIFT algorithm. Nat Protoc 4: 1073-1081.
123. Adamczyk AJ, Cao J, Kamerlin SC, Warshel A (2011) Catalysis by dihydrofolate
reductase and other enzymes arises from electrostatic preorganization, not
conformational motions. Proc Natl Acad Sci USA 108: 14115-14120.
124. Ishikita H, Warshel A (2008) Predicting drug-resistant mutations of HIV protease.
Angew Chem Int Edit 47: 697-700.
Part of this chapter is adapted with permission from ‘Verma R, Schwaneberg U,
Roccatano D. Computational and Structural Biotechnology Journal 2012, 2 (3),
e201209008.’
37

PART I: MAP 2.0 3D
Chapter 2
MAP 2.0 3D: A Sequence/Structure Based Server for
Protein Engineering
2.1. Abstract
The Mutagenesis Assistant Program (MAP) is a web-based tool to provide statistical
analyses of the mutational biases of directed evolution experiments on amino acid
substitution patterns. MAP analysis assists protein engineering in the benchmarking of
random mutagenesis methods that generate single nucleotide mutation in a codon. Herein,
we describe a completely renewed and improved version of the MAP server, named as
MAP 2.0 3D server that correlates the generated amino acid substitution patterns to the
structural information of the target protein. This correlation helps to select more suitable
random mutagenesis method with specific biases on amino acid substitution patterns. In
particular, the new server represents MAP indicators on secondary and tertiary structure,
and correlates them to specific structural components like hydrogen bonds, hydrophobic
contacts, salt bridges, solvent accessibility and crystallographic B-factors. Three model
proteins (D-amino oxidase, phytase and N-acetylneuraminic acid aldolase) are used to
illustrate the novel capability of the server. MAP 2.0 3D server is available publicly at
http://map.jacobs-university.de/map3d.html.
38

PART I: MAP 2.0 3D
2.2. Introduction
Over the past two decades directed protein evolution has been proven to be a
powerful algorithm to tailor protein properties through iterative rounds of random
mutagenesis and screening for improved protein variants.[1,2] Directed evolution methods
are especially useful for improving properties difficult to rationalize and, hence, to identify
amino acids and protein regions that can guide to further enhancements using site directed
and saturation mutagenesis methods.[3,4] The success of a directed evolution campaign
depends highly on the quality of the mutant library and on the employed random
mutagenesis method. Random mutagenesis methods are based on specific error prone
polymerase (enzymatic methods), DNA modifying chemicals (e.g. nitrous acid) or mutator
strains (e.g. Escherichia coli mutA).[5] The quality of a mutant library is determined by the
generated genetic diversity and corresponding protein sequence space.[6] Since the
number of muteins boost with the increasing number of amino acid exchanged in the
protein, protein engineers are challenged with an astronomically vast sequence space[7].
Despite advances in high-throughput screening, it is very difficult to screen the
theoretically generated diversity even in the case of a small protein sequence.[8,9]
Therefore, generating high quality mutant libraries enriched with functional trait is of high
importance. To deal with the challenge to access and screen such a large sequence space,
protein engineers usually adopt two strategic approaches.[10-12] The first approach
consists in the random mutagenesis of the target protein and the subsequent identification
of ‘mutagenic hot spots’. Random mutagenesis can be followed by recombination of the
best variants by site directed mutagenesis or saturation mutagenesis.[13] The second
approach involves the identification of a subset of specific residues using rational or semirational
design with the help of computational tools.[14] Up to five amino acid positions
can be efficiently targeted with focused mutagenesis methods allowing the generation of
focused mutant libraries of a number of variants that can be screened with the state of the
art in flow cytometry methods.[13] Focused mutagenesis is normally employed to improve
the properties of target protein such as activity or selectivity, by mutating residues in close
39

PART I: MAP 2.0 3D
proximity to a specific protein region like the active site. In this case, random mutagenesis
methods are complementary to the rational design since they can identify important amino
acid positions, especially in the second and third coordination sphere, which would have
been overlooked rationally. Nevertheless, random mutagenesis methods are biased toward
certain nucleotide exchanges (e.g. many epPCR methods prefer transition mutations). The
mutagenic preferences resulted by biased random mutagenesis methods affect the
generated diversity. The analysis of the effects of mutational bias on the amino acid
diversity provides a useful indicator in the selection of the mutagenesis method with
diverse and complementary amino acid substitution patterns. The generated
complementary mutant libraries extend the sampling of the vast protein space and
enhance the chance to obtain improved variants.[15,16]
Recently, we have introduced a freely available web-based statistical analysis tool
(MAP[17]). The server statistically analyzes the effects of mutational bias of 19 different
random mutagenesis methods on the level of amino acid substitutions for a given
nucleotide sequence of the target protein. The analysis is returned in terms of MAP
indicators that allow a rapid comparison of different random mutagenesis methods on the
protein level. It has been shown that this approach can be used to predict the type, extent,
and chemical nature of the genetic diversity generated by different mutagenesis
methods.[17,18] Recently, Rasila and co-workers[19] reported a comparative evolution of
commonly used random mutagenesis methods. They found the experimentally induced
substitution patterns very similar to those obtained by MAP server and suggested the use
of combination of mutagenesis methods to generate high diversity.[19]
One of the limitations of the original MAP server is the absence of the analysis tools
relating the MAP indicators to the structural properties of the target protein. The nature of
the amino acid change in different region of the protein can affects its global and local
structural and thermodynamics properties.[14,20,21] Therefore, the possibility to
correlate the generated diversity with structural properties help to identify in advance the
random mutagenesis method that has the least number of “deleterious” mutations on the
protein stability and the higher probability to introduce amino acid substitutions that may
40

PART I: MAP 2.0 3D
improve the fitness toward an expected property, e.g. substitutions to charged amino acid
residues to increase solubility in water. For this reason, we have expanded the capability of
the server by introducing these new features. The new server (MAP 2.0 3D) can correlate the
mutational propensity at amino acid level of a gene for 19 random mutagenesis methods
(and now also for a user customized random mutagenesis method) with the
crystallographic or homology modeled structure (if available in Protein Data Bank[22]
format) of the target protein. MAP 2.0 3D analyses the three-dimensional structure of the
target proteins by calculating secondary structure elements, important local interactions
(such as hydrogen bonds, hydrophobic contacts, salt bridges, disulphide bridges, solvent
accessibility), and amino acid motilities from the crystallographic B-factors. These
combined information help to identify biased amino acid substitutions that may improve
stability and function of the protein.[23-25]
To correlate the sequence-based analysis to the structural data analysis, a new
indicator, the residue mutability indicator ‘µ’ (amino acid substitution probability leading
to amino acid change at specific position), has been introduced (see Methods). The
mutability indicator allows a rapid identification of mutagenic hot spots and, more easy
comparison of experimental data to the predicted ones.
This chapter illustrates the new features of MAP 2.0 3D server in detail, performing
the analysis on three model proteins. The results of the MAP 2.0 3D analysis are compared
with the results of protein engineering experiments reported in the literature. The three
examples show possible uses of the server for computational pre-screening of the target
protein to evaluate and select mutagenesis method for directed evolution.
2.3. Methods
2.3.1. Mutational probability and statistics
41

PART I: MAP 2.0 3D
The MAP 2.0 3D server performs statistical analysis on a given nucleotide sequence
based on the mutational spectra of different random mutagenesis methods that were
slightly elaborated as follow to be used in the analysis.[17] First, insertions and deletions
with an occurrence frequency between 0.80 % and 13.9 % were neglected and remaining
nucleotide substitution frequencies were scaled proportionally to 100 %. Second,
mutations in upper and lower DNA strand were considered to occur with equal frequency.
The scaled mutational frequencies are used in the analysis to calculate the probability of
amino acid substitutions resulting from one nucleotide exchange in one codon of the gene.
The analysis is performed as follows. Consider a gene coding for a protein of L amino acids.
For each nucleotide of a codon (named as X,Y,Z) in the gene sequence, the corresponding
single nucleotide substitutions X´,Y´,Z´ (with {X, Y, Z, X´, Y´, Z’ ∈ {A, T, G, C} | X´ ≠ X, Y’ ≠ Y, Z’ ≠
Z}) are considered. For each one of the 19 random mutagenesis methods, matrix P (in
equation 2.1) gives the 16 mutational probability values for the given nucleotide
substitution into another three (e.g. X → X´). The values of matrix P have been already
reported in Table 1 of our previous publication.[17]
⎛
⎜
P= ⎜
⎜
⎜
⎝
A → A A → T A → G A → C
T → A T → T T → G T → C
G → A G → T G → G G → C
C → A C → T C → G C → C
⎞
⎟
⎟
⎟
⎟
⎠
(2.1)
In equation 2.2, the binary vector U and V are then used to select a given probability
(f) from the matrix P. The four elements of U and V correspond to the nucleotide (A, T, G, C)
that can be selected by assigning a value of one or zero. U selects the original nucleotide
and V the mutated one. In the equation 2.2, an example for the epPCR method with the Taq-
Polymerase (unbalanced dNTPs) is given as matrix P. In this example the mutational
probability for the transformation of nucleotide A → T gives a value of f = 9.7.
42

PART I: MAP 2.0 3D
⎛ 0.0 9.70 19.34 16.14⎞⎛0⎞
⎜
⎟⎜
⎟
T
⎜9.70
0.0 16.14 19.34⎟⎜1⎟
f = UPV =
⎜
=
4.82 0.0 0.0 0.0 ⎟⎜0⎟
⎜
⎟⎜
⎟
0.0 4.82 0.0 0.0
0
⎝
⎠⎝
⎠
( 1 0 0 0) 9. 7
(2.2)
By applying this procedure to each single nucleotide substitution in the codon, nine
probability values (three for each nucleotide) are obtained. Each of these values gives the
k th f ( i
mutational probability (
k
) α → β
) that change the i th amino acid (α) expressed by the
native codon into the one (β, which also comprises the stop codon) expressed by the
k
f( i) mutated codon (e.g. X,Y,Z → X´,Y,Z). Therefore, the 9 probabilities ( α→
get the normalization factor (Ni) for the i th residue of the protein sequence
β
) are summed to
N
i
=
9
∑
k = 1
f
( i)
k
α →β
(2.3)
hence, the normalized probability for the substitution of amino acid α → β is given by
φ(
i)
k
α
→
= f ( i)
β
k
α → β
N
i
(2.4)
2.3.2. MAP indicators
Three indicators protein structure indicator, amino acid diversity indicator and
chemical diversity indicator are used to summarize the characteristics of random
mutagenesis method for the target gene on amino acid level. The amino acid diversity for
the substitution of amino acid α → β in the protein sequence (L) is calculated by
43

PART I: MAP 2.0 3D
∆
α→ β
=
L
1
∑φ(
i)
L 1
i=
α→β
(2.5)
The amino acid diversities are summed together to calculate the values for MAP indicators.
I
α →S
=
r '
∑
r = 1
∆
r
α →β
( r )
(2.6)
where S indicates different subset of amino acids or stop codons and r´ represents the
elements in these subsets. The chemical diversity indicator quantifies the generated
chemical diversity by the random mutagenesis method. For this indicator, the S consists of
one of the subset of amino acids: charged (D, E, H, K, R; S = ch), neutral (C, M, S, P, T, N, Q; S
= ne), aromatic (F, Y, W; S = ar) and aliphatic (G, A, V, L, I; S = al). For example, Iα
ch indicate
the total probability of a given amino acid α to substitute into charged amino acids (ch) is
calculated by Δα
β(r), where the substituted residue β(r) can be one of the charged residues
(E, D, R, K and H) i.e. r´ = 5. The protein structure indicator signifies the fraction of single
nucleotide substitution resulting in protein structure/function-disrupting (stop codons; S =
st) and likely destabilizing (glycine or proline; S = gp) amino acid substitutions. Finally, the
amino acid diversity indicator measures the fraction of variants with preserved amino acid
substitutions (S = pr) and average amino acid substitutions per residue. This is
complemented by codon diversity coefficient that measures the distribution of random
mutations among the codons of the gene.
2.3.3. Local chemical diversity and protein structure components
Two new sequence based indicators are introduced with the MAP 2.0 3D server to
complement the single amino acid structural analysis. The substitution probability of the i th
44

PART I: MAP 2.0 3D
amino acid (α) that leads to change in the amino acid (β) with the side chain of same
chemical nature is calculated by
δ(i) α→S
=
r'
∑
r=1
r
φ(i) α→β
(2.7)
where, x and r´ represents the amino acid group and its members, respectively (as
described for equation 2.6). The amino acid mutability of the i th amino acid (a special case
of the equation 2.7 with r´ = 1) is given by
µ(i) =1−φ(i) α→α (2.8)
where
φ(i)
α →α
is the normalized probability for the substitution does not lead to amino
acid change (α → α) at the i th residue. The local structure environment of the amino acid
residue influences the acceptance of the amino acid substitutions.[23,24] Local structural
environment of the protein comprises secondary structure element, residue flexibility and
solvent accessibility. Intra protein interactions contribute to define secondary structure
elements and residue flexibility in a target protein and help to understand molecular basis
of the stability and activity of the protein.[26] To illustrate the effect of generated chemical
diversity on protein structural environment, these factors are mapped with amino acid
substitution patterns.
The secondary structure elements are derived using DSSP[27] while Relative
Solvent Accessibility (RSA) has been calculated by the number of water molecules in
contact of residue[27] divided by total surface area of the residue.[28] A threshold value of
0.16 is used to differentiate between exposed (RSA >= 0.16) or buried residues (RSA <
0.16). Crystallographic B-factors are used as indicators of the residue flexibility.[29] The B-
factors of Cα atoms are normalized by the
45

B´=
( B − B )
σ
PART I: MAP 2.0 3D
(2.9)
where ‹B› is the average value for Cα atom (after omitting first and last 3 residues) and σ
the standard deviation.[30] The relative B-factor values after normalization is employed to
differentiate flexibility and rigidity of the residue.[31]
Finally, the new server calculates from the crystallographic protein structure, using
criteria reported in literature, the following intra-protein interactions: disulphide
bonds,[27] salt bridge,[32] hydrophobic interaction,[33] aromatic interaction[34] and side
chain hydrogen bond.[35] The default parameters are taken from the widely accepted
primary literature for the calculation of molecular interactions and can me modified by the
user.
2.3.4. MAP 2.0 3D server description
MAP 2.0 3D analysis was performed on gene sequence along with the 3D coordinates
of target protein for a random mutagenesis method at a time. Figure 2.1 shows the query
interface of the server available at http://map.jacobs-university.de/map3d.html.
The server is flexible to accept the gene sequence in commonly used sequence
format (fasta, GenBank, GCG) or as the raw sequence. The 3D coordinates is accepted in
PDB file format[36]. The protein sequence, after translation from gene sequence, is aligned
with protein sequence, extracted from protein coordinates, by using Smith Waterman
algorithm[37] for local sequence alignment. For the complete analysis, the sequences
should have appropriate identity (default >= 70 %). In case of multi-protein chain files, the
analysis performs on first chain or can be defined by the user. The analysis is performed on
a user selected mutagenesis method (chosen among the MAP library of commonly used
methods or, as a feature of the server by directly introducing the values of the probability
46

PART I: MAP 2.0 3D
of transformation matrix P). By default the results include the analysis of all the residues
that can be changed by selecting predefined group of amino (charged, neutral, aromatic
and aliphatic or, accordingly to their relative solvent accessibility, exposed or buried) or by
providing a set of amino acid residues, which can be extended to residues within a given
range (in Å) from the given set of amino acids. Finally, the advanced user interface section
allows the change of the parameters used for the calculation of molecular interactions.
47

PART I: MAP 2.0 3D
Figure 2.1: Query interface for MAP 2.0 3D. Black boxes show two ways to query the sever: (1)
sequence based analysis that take nucleotide sequence as an input (red box) and (2) structure
based analysis, which takes protein coordinates (crystallographic structure or homology model),
nucleotide sequence and a random mutagenesis method as input (red boxes). The options given in
the green boxes can be used to customize the query like (1) 19 commonly used mutagenesis
methods are included in the server as default, new method can be included by defining its
mutational spectra, (2) selection of chain in case of multi chain protein, (3) restrict the search for a
group of amino acids either selecting the predefined groups based on (a) the chemical property of
their side chain like charged, neutral, aromatic, and aliphatic, (b) the solvent accessible area like
buried or exposed and (c) the given set of amino acids and define cutoff (in Å) to include residues in
the defined diameter of given residues in the analysis, and (4) altering the threshold used for the
calculation of molecular interactions.
2.3.5. MAP 2.0 3D output
Along with the sequence based MAP analysis indicators, the implemented indicators
in MAP 2.0 3D correlate the generated amino acid substitution patterns of random
mutagenesis methods to the protein structure (by using the Jmol applet,
http://www.jmol.org/) and includes a residue mutability indicator and taking secondary
structure elements, residue flexibility, relative solvent accessibility and intra protein
interactions into account (see above). Generated results are also available to download for
further use in text format. The modified coordinate files (with amino acid substitution
probabilities) in pdb format are also available as downloads.
2.3.6. Model proteins
The enzyme selected for the analysis by MAP 2.0 3D are: 1) D-amino acid oxidase from
Rhodotorula gracilis (EC: 1.4.3.1; EMBL-Bank: AAB93974.1[36]; PDB Id: 1C0I[37]), 2)
Phytase from Escherichia coli (EC: 3.1.3.2; EMBL-Bank: AY496073.1[38]; PDB Id:
1DKP[39]), 3) N-Acetylneuramine acid aldolase from Escherichia coli (EC: 4.1.3.3; EMBL-
Bank: X03345.1[40] ; PDB Id: 1NAL[41]). The sequence composition of the enzymes: 1) D-
48

PART I: MAP 2.0 3D
amino acid oxidase (1107 bases: A 19.96 %; T 17.52 %; G 31.17 %; C 31.35 %; 369
residues), 2) Phytase (1299 bases: A 24.25 %; T 22.09 %; G 27.25 %; C 26.40 %; 433
residues) and 3) N-acetylneuraminic acid aldolase (894 bases: A 24.38 %; T 23.60 %; G
27.07 %; C 24.94 %; 298 residues). Secondary structure of the enzymes: 1) D-amino acid
oxidase (30 % helical, 28 % beta sheet), 2) Phytase (42 % helical, 15 % beta sheet), 3) N-
acetylneuraminic acid aldolase (50 % helical, 13 % beta sheet).
2.4. Results and discussions
The use of MAP 2.0 3D server is illustrated by performing the analysis of three
different enzymes evolved for different properties by using directed protein evolution. The
first example describes how to decrease effects of mutational bias and to generate a mutant
library with a higher fraction of active clones. The second and third examples show the
usability of the server to analyze the influence of mutational preferences on the evolution
of desirable property. Outputs of the complete MAP 2.0 3D analysis are provided as examples
in the instruction link of the server (http://map.jacobs-university.de/instruction.html).
2.4.1. D-amino acid oxidase
D-amino acid oxidase (DAAO) is a flavin adenine dinucleotide (FAD) dependent
flavoenzyme. DAAO catalyses the dehydrogenation of D-amino acid to the corresponding α-
keto acids, producing ammonia and hydrogen peroxide.[42,43] The high turnover rate, the
stable FAD-binding and the broad substrate specificity of DAAO from Rhodotorula gracilis
(RgDAAO) make it an attractive catalyst for biotechnological application as the biosensing
(i.e. the rapid and reliable detection of D-amino acid content in food specimens or of the
neurotransmitter D-serine in the brain).[43] We performed MAP 2.0 3D analysis on the
RgDAAO to evaluate the amino acid diversity generated by random mutagenesis methods.
49

PART I: MAP 2.0 3D
Table 2.1: Summary of the MAP 2.0 3D analysis for the oxidase, the phytase and the aldolase,
targeting different epPCR methods for random mutagenesis.
RgDAAO (1 st
RgDAAO (2 nd
Phytase
N-acetylneura-
round)
round)
minic acid
aldolase
epPCR method
Average amino
acid substitution a
Preserved amino
acid substitution b
Codon diversity
coefficient c
Stop codons
frequency d
Gly/Pro
frequency e
Charged amino
acid diversity f
Neutral amino
acid diversity g
Aromatic amino
acid diversity h
Aliphatic amino
acid diversity i
Taq
(+,G=A=C=T)
Taq
(+,G=A,C=T)
Taq
(+,G=A,C=T)
Taq
(+,G=A=C=T)
7.40 7.40 7.45 7.20
24.53 % 23.38 % 25.40 % 28.47 %
42.48 34.04 36.49 43.70
2.30 % 4.38 % 4.69 % 2.12 %
20.58 % 13.23 % 11.60 % 16.26 %
-0.34 %
(25.00 %)
-2.62 %
(25.00 %)
1.39 %
(19.21 %)
5.00 %
(22.22 %)
3.37 %
4.47 % -4.14 % 1.73 %
(27.99 %) (27.99 %) (35.65 %) (26.94 %)
-3.19 % -0.23 % 0.91 % -3.14 %
(7.34 %) (7.34 %) (5.79 %) (8.08 %)
-2.13 % -6.00 % -2.86 % -5.72 %
(39.67 %) (39.67 %) (39.35 %) (42.76 %)
aaverage number of amino acid substitutions per residue, b Iα→pr: fraction of variants with
preserved amino acid substitutions, c codon diversity coefficient, d Iα→st: fraction of variants with
stop codons, e Iα→gp: fraction of variants with Gly/Pro and chemical diversity generated by the
mutagenesis methods presented as f Iα→ch: charged, g Iα→ne: neutral, h Iα→ar: aromatic and I Iα→al:
50

PART I: MAP 2.0 3D
aliphatic amino acid diversity with the amino acid composition of the target protein sequence (in
parenthesis) and deviation from this composition after mutagenesis.
MAP 2.0 3D analysis
The sequence based MAP 2.0 3D analysis was performed using the following
descriptors: i) protein structure indicators, ii) amino acid diversity indicator with codon
diversity coefficient and iii) chemical diversity indicator.
Figure 2.2: Statistical analysis of stop codons frequencies (a) and Gly/Pro substitutions (b) for
RgDAAO. The random mutagenesis methods enclosed in the black rectangles (epPCR (Taq (MnCl 2,
G=A=C=T)) and epPCR (Taq (MnCl 2, G=A, C=T))) are used for the MAP 2.0 3D analysis.
In Figure 2.2, the values for the stop codon indicator (Iα→st)) and the Gly/Pro
indicator (Iα→gp)) for different random mutagenesis methods are reported. The two methods
show opposite trend in the generation of stop codons (sequence truncation) and Gly/Pro
51

PART I: MAP 2.0 3D
(α-helix destabilizers), i.e. higher the stop codons frequency lower the Gly/Pro
substitutions and vice versa.[17] The two epPCR methods (indicated in the Figure 2.2 with
the black rectangles) were found to be more appropriate for the RgDAAO with the balanced
frequencies of stop codons and Gly/Pro in comparison to other mutagenesis methods. In
Table 2.1, the sequence-based analysis of the server for selected random mutagenesis
methods is summarized. The first method, the balanced epPCR Taq-Pol (Mn 2+ , balanced
dNTP)[44] has strong preference for specific nucleotide exchange ~32 % AT → GC
(transition mutations). While second method, the unbalanced epPCR Taq-Pol (Mn 2+ ,
unbalanced dNTP)[45] is expected to produce more transversion (21.41 % AT → TA) than
transition (14.45 % AT→GC) mutations. Balanced epPCR was expected to generate lower
fraction of stop codons (Iα→st = 2.30 %) and higher Gly/Pro (Iα→gp = 20.58 %) content than
the unbalanced epPCR (Iα→st = 4.38 % and Iα→gp = 13.23 %) (see Table 2.1). For both
methods, an average 7.4 amino acid substitutions per residue was calculated.
In Figure 2.3, cartoon representations of the RgDAAO crystallographic structure
colored accordingly to Iα→gp using the Jmol[46] visualization feature of the new server, are
shown. Out of 30 % of the residues involved in helix formation, 51 % has a higher Iα→gp
value (if α is equal to S, L, E and D) with a prevalence of negative charged residues (E and D,
highlighted in stick format in Figure 2.3). In comparison to the unbalanced epPCR, the
balanced epPCR method was observed with a higher probability of the charged residues
substitution into Gly/Pro (represented by the color code used to define amino acid
substitution probability in Figure 2.3). The mapping of charged amino acid substitution
patterns on the structure of RgDAAO are reported in Figure 2.4 and found to be consistent
with the latter observation of Gly/Pro substitution patterns. The balanced epPCR (Figure
2.4a) shows lower probability for charged amino acid substitutions than unbalanced epPCR
(Figure 2.4b) that is found to be opposite to the Gly/Pro substitution patterns for both
methods (Figure 2.3). Hence, the amino acid substitutions of charged residues into residues
unfavorable for forming molecular interactions result in destabilization of protein. For
example, charged residues were found to be involved in molecular interactions like salt
bridges (15 out of 21 show more than 0.5 probability to be substituted in glycine) and side
chain H-bonds (5 out of 26 with more than 0.5 probability for glycine substitutions). In
52

PART I: MAP 2.0 3D
Figure 2.5, charged residues involved in salt bridge formation with the amino acid diversity
generated by the balanced (a1
and a2) and unbalanced epPCR (b1 and b2) methods are
reported. The balanced epPCR method shows lower probabilities for substitution into
charged residues when compared to the unbalanced epPCR method. The unbalanced
epPCR is less transition biased (AT → GC) that results in higher probability of substitutions
to charged residues (for E coded by GAG and D coded by GAC; a transition mutation leads
often to a substitution into glycine (GGC, GGG)). These effects of mutagenesis methods due
to mutational preferences might be minimized by codon optimization like for E using GAA
and for D using GAT codon.
Figure 2.3: Gly/Pro amino acid substitutions mapping on RgDAAO structure for (a) epPCR
(Taq (MnCl2, G=A=C=T)) and (b) epPCR (Taq (MnCl2, G=A, C=T)). For the balanced epPCR
method (a) the red colored regions of RgDAAO structure indicate an overall higher
probability of charged residues substitutions, mainly for negatively charged residues (in
stick representation), into Gly/Pro than the unbalanced epPCR (b).
53

PART I: MAP 2.0 3D
Figure 2.4: Amino acid substitutions mapping of charged residues (E, D, R, K, H) on RgDAAO with
for (a) epPCR (Taq (MnCl 2, , G=A=C=T)) and (b) epPCR (Taq (MnCl 2, G=A, C=T)).
54

PART I: MAP 2.0 3D
Figure 2.5: Chemical diversity and mutability of charged amino acid positions of D-amino acid
oxidase (E, D, R, K, H) that are involved in salt bridges formation (a1) and (a2) for epPCR (Taq
(MnCl 2, G=A=C=T)) and (b1) and (b2) for epPCR (Taq (MnCl 2, G=A, C=T)). Y-axis shows residue (i)
sequence id, (ii) PDB id, (iii) residue name, (iv) secondary structure elements (H: alpha helix; B:
beta bridge and extended strand; T: hydrogen bonded turn and bend; *: loop or irregular structure)
and (v) Amino acid category according to the chemical property of its side chain (P: charged, Y:
neutral, C: aromatic and B: aliphatic) with stop codon (R) and Gly/Pro (G) as separate classes.
Using the new focused analysis feature of the server, the amino acid substitutions
patterns for active site residues (Y223, Y238, and R285) were also evaluated. Y223 and
Y238 are involved in substrate binding and product release while R285 forms a pair with
carboxylate portion of the substrate (arginine) in RgDAAO.[37] R285 has a very low
residue mutability indicator (μ(285) < 0.3) (i.e. low probability of substitution leading to
amino acid change) for both methods. Y223 and Y238 have µ(223/238) = 0.9 and therefore
they have a higher probability to be substituted into another amino acid. For the balanced
epPCR, Y223 and Y238 are preferentially substituted into charged (δ(223/238)Y→ch = 0.37)
and neutral (δ(223/238)Y→ne = 0.46) amino acids. In the unbalanced epPCR, the chemical
diversity at Y223/238 is more preserved (δ(223/238)Y→ne = 0.44 and δ(223/238)Y→ar =
0.31). The tendency to the substitution of active site aromatic residues into chemically
different amino acid might result in the increased number of inactive clones in the mutant
library. In summary, MAP 2.0 3D provides qualitative indication that the balanced epPCR
method might be less beneficial (or of lower quality) than the unbalanced one in the
directed evolution of RgDAAO.
RgDAAO directed evolution
In one directed evolution study by Pollegioni et al.[47], the substrate specificity of
RgDAAO was altered to formulate it as biosensor for analytical determination of D-amino
acid in biological samples. Two rounds of directed evolution were performed employing
epPCR mutant libraries (balanced dNTP) followed by another round of directed evolution
55

PART I: MAP 2.0 3D
employing epPCR (unbalanced dNTP) for diversity generation. In the first round (1 st set of
epPCR: balanced), 91 % and in the second round (2 nd set of epPCR: unbalanced), 63 %
clones were reported to be inactive. The results of these experiments are in agreement
with the predictions of MAP 2.0 3D server. In fact, mutational preferences of the balanced
method induce more structural destabilizing substitutions and resulted in a higher number
of inactive clones than balanced epPCR. In addition, MAP 2.0 3D analysis suggests that most
of the inactive clones should be a result of substitutions into Gly/Pro (destabilizing amino
acids), which can destabilize the secondary structure of a helix or weaken intra-molecular
interactions.
The best variant obtained from the experiments was the triple mutant (T60A Q144R
K152E) with broader substrate specificity. Amino acid substitution patterns calculated by
the MAP 2.0 3D server at these positions were also found in agreement with experimental
results. All mutated positions were assigned by MAP 2.0 3D with high residue mutability
value (μ(60/144/152) > 0.8), i.e. within mutagenic hotspots generated by mutagenesis
methods. Q144R substitution was identified in the first round of the balanced epPCR. Q144
has a high probability to substitute into charged residue (δ(144)Q→ch = 0.67) and
experimentally the Q144R (φ(144)Q→R = 0.58) substitution was found. In second round of
random mutagenesis with unbalanced epPCR, T60A (φ(60)T→A = 0.58) and K152E
(φ(152)K→E = 0.36) were substituted. Both residues have a high preference to be
substituted into aliphatic (δ(60)T→al = = 0.6) and charged residues (δ(152)K→ch = = 0.7),
respectively.
In summary, the RgDAAO case illustrates how the MAP 2.0 3D server can be used in
developing efficient mutagenesis strategies before and during directed evolution
experiments by, for instance, the selection of the most efficient mutagenesis method for the
target gene with least unfavorable effects on its protein structure or function and codon
engineering. In this way, the gene can be synthesized prior to the directed evolution
experiment to reduce highly destabilizing substitutions at key amino acid positions.
56

PART I: MAP 2.0 3D
2.4.2. Phytase
Phytase is a class of phosphatase enzymes that catalyses the hydrolysis of phytic
acid (myoinositol hexakisphosphate) to release inorganic phosphorus in a usable form.
Phytases have been used as a feed supplement since decades.[48] Application of phytases
in industrial feed pelleting process requires high temperatures. For this reason, directed
evolution methods have been used to increase thermal resistance of phytases while
maintaining high activity at ambient temperature.[49]
MAP 2.0 3D analysis
MAP 2.0 3D analysis was performed on the phytase appA2 (full analysis is given in
MAP 2.0 3D server as an example). In comparison to other 18 random mutagenesis methods,
epPCR Taq (+, G=A, C=T) was found to be the preferred choice for directed appA2
evolution. In fact, as reported in Table 2.1, the sequence based MAP 2.0 3D analysis shows
frequency of stop codons Iα→st = 4.69 % and substitutions into Gly/Pro Iα→gp = 11.60 %. The
average 7.45 amino acid substitutions per residue were calculated. The value of codon
diversity coefficient was 36.49 % and resulted in preserved amino acid substitutions Iα→pr =
25.40 %. Charged (19.21 %) and aromatic (5.69 %) residues were overrepresented with
1.39 % and 0.91 % deviation from their chemical distribution, respectively. The aliphatic
(39.35 %) and neutral (35.65 %) residues were underrepresented with -2.86 % and -4.14
% deviation, respectively.
By using the structural data a different conclusion emerge in contrast to the
sequence analysis alone. One of the rule of thumb, used to enhance the thermostability of
an enzyme, is to increase the number of charged residues in the loop regions at the protein
surface. The reduction of mobility of these flexible regions by strengthening with
electrostatic and hydrogen bonding interactions usually has a stabilizing effect on the
thermal stability.[50] Hence, the amino acid substitution patterns of charged residues were
analyzed using the residue mutability indicator, the normalized B-factors (B´) as a residue
flexibility indicator and the relative solvent accessibility (RSA) to differentiate exposed and
57

PART I: MAP 2.0 3D
buried residues. In Figure 2.6, the mapping of amino acid substitution patterns, generated
by epPCR Taq (+, G=A, C=T), for different amino acid substitution classes (charged, neutral,
aromatic and aliphatic), stop codon and Gly/Pro on the phytase appA2 is reported with
charged residues represented in stick representation. The high probability of charged
residues substitutions into Gly/Pro, aliphatic and neutral residues were observed in
MAP 2.0 3D analysis. In Figure 2.7 the detailed information of amino acid substitution
patterns for charged residues is reported with three MAP 2.0 3D structural indicators for the
epPCR Taq (+, G=A, C=T) method. The experimentally determined mutations are
highlighted with black rectangles in Figure 2.7. Most of the charged residues were found
with mutability value µ > 0.6 i.e. high substitution probability to change into another amino
acid. In Figure 2.7, the high probabilities were evident to substitute from charged residues
into glycine or proline (alpha helix destabilizers), aliphatic and neutral residues (less
favorable to improve thermostability).
58

PART I: MAP 2.0 3D
Figure 2.6: MAP 2.0 3D analysis of amino acid substitutions probability of phytase appA2 after being
subjected to epPCR (Taq (MnCl 2, G=A, C=T) in cartoon representation; charged residues (D, E, H, K,
R) shown in stick representation. The probability values increase from blue (lowest probability) to
red (highest probability). Amino acids were grouped according to the chemical nature of their side
chain: charged (c), neutral (d), aromatic (e) or aliphatic (f) with sequence interrupting (stop codons
(a)) and structure destabilizing amino acids (glycine and proline (b)).
Phytase directed evolution
In one example, Kim et al. performed directed evolution on phytase appA2 from E.
coli to generate variants with increased thermostability by using epPCR with unbalanced
dNTPs.[51] Two variants (K46E and K65E K97M S209G) with 20 % improved
thermostabilty were found after screening 5000 clones. Out of four positions, three were
resulted from charged residue substitutions occurred at lysine residues.
MAP 2.0 3D analysis of amino acid substitution pattern for these positions was found
in agreement with experimental findings with, all four positions having a high mutability
indicator value (µ > 0.8) and relative solvent accessibility (RSA > 0.4). Furthermore, all
lysine residues in the mutated positions have a probability to a nucleotide exchange that
results in a stop codon. K46 and K97 have the same amino acid substitution patterns with
substitution preference for stop codon (δ(46/97)K→st = 0.24) but different for charged
(δ(46)K→ch = 0.40; φ(46)K→E = 0.16) and neutral residues (δ(97)K→ne = 0.35; φ(97)K→M =
0.24). K65 has different amino acid substitution values to change into residues with
aliphatic (δ(65)K→al = 0.18), charged (δ(65)K→ch = 0.36; φ(65)K→E = 0.12) and neutral
(δ(65)K→ne = 0.27) side chains and δ(65)K→st = 0.18 for stop codon. S209 has a high
probability to preserve the chemical property of its side chain and has high preference to
neutral substitution (δ(209)S→ne = 0.60). S209 substitution into glycine alone has
probability φ(209)S→G = 0.24. The mutations generated by using the epPCR Taq (+, G=A,
=T) mutagenesis method experimentally resulted in only 20 % active clones in the library
and only 80 were found improved in thermal stability. Phytase appA2 has high C helical
59

PART I: MAP 2.0 3D
content (42%) and substitutions into Gly/Pro residues might reduce thermal stability by
destabilizing the structure and increasing the number of inactive clones. In general, amino
acid substitutions of charged residues into aliphatic or neutral residues are less favorable
to improve thermal stability.
60

PART I: MAP 2.0 3D
Figure 2.7: Amino acid substitution patterns for charged residues in phytase with performance
the parameters residue mutability, residue flexibility and relative solvent accessibility of amino
acids. The experimentally determined mutations are highlighted in black boxes. Y-axis shows
sequence id, PDB id, amino acid name and in (a) secondary structure elements (H: alpha helix; B:
beta bridge and extended strand; T: hydrogen bonded turn and bend; *: loop or irregular structure),
(b) normalized Cα B-factor to differentiate between flexible: F and rigid: R residues and (c) relative
solvent associability to identify exposed: E or buried: B residues.
2.4.3. N-acetylneuraminic acid aldolase
N-acetylneuraminic acid aldolase (Neu5Ac aldolase) catalyses the aldol
condensation of N-acetyl-D-mannosamine and pyruvate to give N-acetyl-D-neuraminic acid
(D-sialic acid).[52] Neu5Ac aldolase is used in the synthesis of sialic acid, a complex sugar
with many pharmaceutical applications.
MAP 2.0 3D analysis
Based on the sequence based analysis of MAP 2.0 3D server, the balanced epPCR
method (Taq (MnCl2, G=A=C=T) was found suitable for directed evolution of Neu5Ac
aldolase (summarized in Table 2.1). For this method, the value of codon diversity
coefficient was 43.12, which is resulted in Iα→pr = 28.47 % preserved amino acid
substitutions with an average 7.20 amino acid substitutions per residue. The frequency for
stop codons was Iα→st = 2.12% and for Gly/Pro substitutions Iα→gp = 16.26% were reported.
The structure based analysis was focused on active site residues (A11, S47, T48, Y137,
I139, k165, T167, G189, Y190) using the new option of the MAP 2.0 3D server to restrict the
analysis to selected amino acids. Figure 2.8 shows the expected amino acid substitutions
for active site residues and, highlighted in boxes, experimentally determined mutation
positions[52] (G70, T84, Y98, F115, V251, E282). With the exception of residues A11 and
G189, the other active site residues have a residue mutability value (μ > 0.6). The values of
the RSA and B´ indicate that A11 and G189 are buried in the protein active site and highly
rigid. The residue I139, another aliphatic residue of active site, resulted in a moderately
61

PART I: MAP 2.0 3D
high preference of substitution into neutral amino acid (δ(139)I→ne = 0.36). The active site
residues, Y137 and Y190 have high residue mutability value (µ(137/190) = 0.94) and
substitute into charged (δ(137/190)K→ch = 0.37) or neutral (δ(137/190)K→ne = 0.46) amino
acids. S47 has mutability value µ = 0.88 with a substitution probability φ(47)S→G = 0.6 to
change into glycine. K165 has preference (µ(165) = 0.73) to substitute into charged
residues (φ(165)K→R/K/E = 0.26).
Figure 2.8: Amino acid substitution patterns for active site residues (A11, S47, T48, Y137, I139,
K165, T167, G189, Y190) of Neu5Ac aldolase and experimentally determined mutations (I st
generation: Y98H, P115L, II nd generation: V251I, III rd generation G70A, T84S, Q282L) in the boxes
for random mutagenesis method: epPCR (Taq (MnCl 2, G=A=C=T). Y-axis representations are same
as described in Figure 2.7.
Figure 2.9 and 2.10 show the analysis of hydrophobic contacts and hydrogen bonds
for active site residues and experimentally determined mutation positions, respectively.
The results of the analysis highly suggest an involvement of A11 in hydrophobic
interactions with Y43 or I206 (see Figure 2.9) and a side-chain hydrogen bond formation
with Y43 or G207 or N211 (see Figure 2.10). In short, the substitution spectra analysis of
62

PART I: MAP 2.0 3D
active site residues (A11, S47, T48, Y137, I139, K165, T167, G189, Y190) indicates that the
chemical environment of active site residues is not substantially modified by the epPCR
random mutagenesis method.
Figure 2.9: Amino acid substitution patterns for active site residues (A11, Y137, I139) of Neu5Ac
aldolase and mutations (1 st generation: Y98H and P115L; highlighted in the box frames) involved in
hydrophobic interactions. Figure (a) and (b) shows the interaction partners for hydrophobic
interaction. Y-axis representations are same as described in Figure 2.5.
Figure 2.10: Amino acid substitution patterns for active site residues (A11, S47, T48, Y137) of
Neu5Ac aldolase and mutation (I st generation: Y98H; highlighted in a black box) involved in side
chain hydrogen bond. Figure (a) and (b) shows the interaction partners for side chain hydrogen
bond. Y-axis representations are same as described in Figure 2.5.
63

PART I: MAP 2.0 3D
Neu5Ac Aldolase directed evolution
Neu5Ac aldolase was engineered applying epPCR with balanced dNTP for a
complete reversal of enantioselectivity by Wada et al.[52] Three rounds of random
mutagenesis resulted in a variant more effective toward both D/L enantiomeric substrates
(3-deoxy–L/D-manno-2-octulosonic acid). The two mutation positions (Y98H and F115L,
see Figure 2.8) from the first round of random mutagenesis were found to be involved in
hydrophobic interaction (Figure 2.9, from Y98 to L63/A67/F100 and from F115 to
L147/L155) and side chain hydrogen bonds (Figure 2.10, from E64 to Y98) formation in
wild type. Y98H and F115L were present outside the active site, partially exposed to the
solvent with relative solvent accessibility value (RSA = 0.26), moderately flexible with
normalized B-factor (B´ = 0.91) and “variable” amino acid substitutions with residue
mutability indicator µ(98/115) = 0.75. The substitutions at these positions into
comparatively more hydrophobic residues resulted in increased activity of the wild type
aldolase. In MAP 2.0 3D, Y98 is preferably preserved or substitute into charged (δ(98)Y→ch =
0.27; φ(98)Y→H = 0.25) or neutral (δ(98)Y→ne = 0.33) residues while F115 shows a slightly
higher preference toward aliphatic substitution (δ(115)F→al = 0.42; φ(115)F→L = 0.33) and
cannot be substituted into charged residues (δ(115)F→ch = 0.00). The substitution at V251
residue was obtained in the second round of directed evolution experiment and resulted in
partially inverted enantiomeric preference of the enzyme. The position was found to be
more conserved in MAP 2.0 3D analysis with µ(251) = 0.54 or substituted into more
hydrophobic residues. The third generation mutations (G70A, T84S, Q282L) resulted into a
complete reversal of enzymatic enantioselectivity for use in the synthesis of both D- and L-
sugars. G70 has a high flexibility (B´ = 2.59) and low probability to be substituted into an
aliphatic residue (δ(70)G→al = 0.10; φ(70)G→A = 0.03). Thr84 is a part of a turn with a high
flexibility (B´ = 0.77) and exposure to the solvent (RSA = 0.25) with the residue mutability
µ(84) = 0.87. T84 has a high preference for being substituted by aliphatic residues
(δ(84)T→al = 0.58) and with less extend by a “neutral” amino acid (δ(84)T→ne = 0.34;
φ(84)T→S = 0.13). Q282 is a part of a helix, is rigid (B´ = -1.01) but partially exposed to the
solvent (RSA = 0.23). Q282 has high preference to be substituted into a charged residue
64

PART I: MAP 2.0 3D
(δ(282)Q→ch = 0.67) and very low for aliphatic (δ(282)Q→al = 0.13; φ(282)Q→L = 0.13) or
neutral (δ(282)Q→ne = 0.08) substitution.
In the case of aldolase, the MAP 2.0 3D analysis shows also a good agreement with
experimental results. The variability in amino acid substitution patterns for active site
residues resulted in exploring more sequence space for catalytic activity of the enzyme and
resulted in getting a high fraction of beneficial mutations in first generation.
2.5. Conclusions
In this manuscript, we introduced MAP 2.0 3D server and its use to assist the design of
directed evolution experiments. MAP 2.0 3D correlates the traditional sequence based MAP
indicators with the structural information of the target protein. The combined information
can help to improve the chances to find functional and stable enzyme variants. MAP 2.0 3D
helps to guide the directed evolution experiments by focusing the analysis on a set of
residues that are important for specific enhancement of enzymatic properties such as to
improve substrate specificity by targeting residues located in or near the active site, or to
enhance thermal stability or water solubility of proteins by increasing the number of
charged amino acid substitutions. The new structure oriented features of the MAP 2.0 3D
server have been applied to the analysis of three different proteins (phytase, oxidase and
aldolase) and the predicted results were compared with the experimental results. The
results of RgDAAO analysis indicate that the selection of the random mutagenesis method
by the pre-screening of the generated library can help to elucidate the effects of mutational
bias on the structural environment of the protein and how these effects can be optimized.
The analysis of phytase and Neu5Ac aldolase illustrate that how the structural analysis
features included in MAP 2.0 3D server can now assist to correlate the effect of mutational
biases with protein structural environment and to evolve desired property. In this way,
MAP 2.0 3D server facilitates the ‘in-silico’ pre-screening of the target gene and can also
promote an increase of the active population in random mutagenesis libraries, thereby
65

PART I: MAP 2.0 3D
decrease screening efforts and increase probability for obtaining desirable mutations even
in the small mutant library.
2.6. References
1. Bornscheuer UT, Pohl M (2001) Improved biocatalysts by directed evolution and
rational protein design. Curr Opin Chem Biol 5: 137-143.
2. Brakmann S (2001) Discovery of superior enzymes by directed molecular evolution.
Chembiochem 2: 865-871.
3. Wong TS, Arnold FH, Schwaneberg U (2004) Laboratory evolution of cytochrome
p450 BM-3 monooxygenase for organic cosolvents. Biotechnol Bioeng 85: 351-358.
4. Roccatano D, Wong TS, Schwaneberg U, Zacharias M (2006) Toward understanding
the inactivation mechanism of monooxygenase P450 BM-3 by organic cosolvents: a
molecular dynamics simulation study. Biopolymers 83: 467-476.
5. Wong TS, Zhurina D, Schwaneberg U (2006) The diversity challenge in directed
protein evolution. Comb Chem High T Scr 9: 271-288.
6. Wong TS, Roccatano D, Schwaneberg U (2007) Steering directed protein evolution:
strategies to manage combinatorial complexity of mutant libraries. Environ
Microbiol 9: 2645-2659.
7. Smith JM (1970) Natural Selection and Concept of a Protein Space. Nature 225: 563-
564.
8. Olsen M, Iverson B, Georgiou G (2000) High-throughput screening of enzyme
libraries. Curr Opin Biotech 11: 331-337.
9. Tawfik DS, Bershtein S (2008) Advances in laboratory evolution of enzymes. Curr
Opin Chem Biol 12: 151-158.
10. Shivange AV, Marienhagen J, Mundhada H, Schenk A, Schwaneberg U (2009)
Advances in generating functional diversity for directed protein evolution. Curr
Opin Chem Biol 13: 19-25.
66

PART I: MAP 2.0 3D
11. Turner NJ (2009) Directed evolution drives the next generation of biocatalysts. Nat
Chem Biol 5: 567-573.
12. Wong TS, Roccatano D, Loakes D, Tee KL, Schenk A, et al. (2008) Transversionenriched
sequence saturation mutagenesis (SeSaM-Tv+): a random mutagenesis
method with consecutive nucleotide exchanges that complements the bias of errorprone
PCR. Biotechnol J 3: 74-82.
13. Dennig A, Shivange AV, Marienhagen J, Schwaneberg U (2011) OmniChange: The
Sequence Independent Method for Simultaneous Site-Saturation of Five Codons.
PLoS ONE 6: e26222.
14. Chica RA, Doucet N, Pelletier JN (2005) Semi-rational approaches to engineering
enzyme activity: combining the benefits of directed evolution and rational design.
Curr Opin Biotech 16: 378-384.
15. Zumarraga M, Camarero S, Shleev S, Martinez-Arias A, Ballesteros A, et al. (2008)
Altering the laccase functionality by in vivo assembly of mutant libraries with
different mutational spectra. Proteins 71: 250-260.
16. Vanhercke T, Ampe C, Tirry L, Denolf P (2005) Reducing mutational bias in random
protein libraries. Anal Biochem 339: 9-14.
17. Wong TS, Roccatano D, Zacharias M, Schwaneberg U (2006) A statistical analysis of
random mutagenesis methods used for directed protein evolution. J Mol Biol 355:
858-871.
18. Wong TS, Roccatano D, Schwaneberg U (2007) Are transversion mutations better? A
Mutagenesis Assistant Program analysis on P450 BM-3 heme domain. Biotechnol J
2: 133-142.
19. Rasila TS, Pajunen MI, Savilahti H (2009) Critical evaluation of random mutagenesis
by error-prone polymerase chain reaction protocols, Escherichia coli mutator strain,
and hydroxylamine treatment. Anal Biochem 388: 71-80.
20. Ditursi MK, Kwon SJ, Reeder PJ, Dordick JS (2006) Bioinformatics-driven, rational
engineering of protein thermostability. Protein Eng Des Sel 19: 517-524.
21. Shoichet BK, Beadle BM (2002) Structural bases of stability-function tradeoffs in
enzymes. J Mol Bio 321: 285-296.
67

PART I: MAP 2.0 3D
22. Berman H, Henrick K, Nakamura H (2003) Announcing the worldwide Protein Data
Bank. Nat Struct Biol 10: 980-980.
23. Zhang H, Zhang T, Chen K, Shen S, Ruan J, et al. (2009) On the relation between
residue flexibility and local solvent accessibility in proteins. Proteins 76: 617-636.
24. Teilum K, Olsen JG, Kragelund BB (2009) Functional aspects of protein flexibility.
Cell Mol Life Sci 66: 2231-2247.
25. Gromiha MM, Oobatake M, Kono H, Uedaira H, Sarai A (1999) Role of structural and
sequence information in the prediction of protein stability changes: comparison
between buried and partially buried mutations. Protein Eng Des Sel 12: 549-555.
26. Gromiha MM, Selvaraj S (2004) Inter-residue interactions in protein folding and
stability. Prog Biophys Mol Bio 86: 235-277.
27. Kabsch W, Sander C (1983) Dictionary of protein secondary structure: pattern
recognition of hydrogen-bonded and geometrical features. Biopolymers 22: 2577-
2637.
28. Chothia C (1976) The nature of the accessible and buried surfaces in proteins. J Mol
Biol 105: 1-12.
29. Peisajovich SG, Tawfik DS (2007) Protein engineers turned evolutionists. Nat
Methods 4: 991-994.
30. Karplus PA, Schulz GE (1985) Prediction of Chain Flexibility in Proteins - a Tool for
the Selection of Peptide Antigens. Naturwissenschaften 72: 212-213.
31. Yuan Z, Zhao J, Wang ZX (2003) Flexibility analysis of enzyme active sites by
crystallographic temperature factors. Protein Eng Des Sel 16: 109-114.
32. Kumar S, Nussinov R (2002) Close-range electrostatic interactions in proteins.
Chembiochem 3: 604-617.
33. Kyte J, Doolittle RF (1982) A simple method for displaying the hydropathic
character of a protein. J Mol Biol 157: 105-132.
34. Burley SK, Petsko GA (1985) Aromatic-aromatic interaction: a mechanism of protein
structure stabilization. Science 229: 23-28.
35. Overington J, Johnson MS, Sali A, Blundell TL (1990) Tertiary structural constraints
on protein evolutionary diversity: templates, key residues and structure prediction.
P Roy Soc Lond B Bio 241: 132-145.
68

PART I: MAP 2.0 3D
36. Liao GJ, Lee YJ, Lee YH, Chen LL, Chu WS (1998) Structure and expression of the D-
amino-acid oxidase gene from the yeast Rhodosporidium toruloides. Biotechnol
Appl Bioc 27 ( Pt 1): 55-61.
37. Pollegioni L, Diederichs K, Molla G, Umhau S, Welte W, et al. (2002) Yeast D-amino
acid oxidase: structural basis of its catalytic properties. J Mol Biol 324: 535-546.
38. Rodriguez E, Han Y, Lei XG (1999) Cloning, sequencing, and expression of an
Escherichia coli acid phosphatase/phytase gene (appA2) isolated from pig colon.
Biochem Bioph Res Co 257: 117-123.
39. Lim D, Golovan S, Forsberg CW, Jia Z (2000) Crystal structures of Escherichia coli
phytase and its complex with phytate. Nat Struct Biol 7: 108-113.
40. Ohta Y, Watanabe K, Kimura A (1985) Complete nucleotide sequence of the E. coli N-
acetylneuraminate lyase. Nucleic Acids Res 13: 8843-8852.
41. Izard T, Lawrence MC, Malby RL, Lilley GG, Colman PM (1994) The threedimensional
structure of N-acetylneuraminate lyase from Escherichia coli. Structure
2: 361-369.
42. Pilone MS (2000) D-Amino acid oxidase: new findings. Cell Mol Life Sci 57: 1732-
1747.
43. Pollegioni L, Molla G (2011) New biotech applications from evolved D-amino acid
oxidases. Trends Biotechnol 29: 276-283.
44. Lin-Goerke JL, Robbins DJ, Burczak JD (1997) PCR-based random mutagenesis using
manganese and reduced dNTP concentration. Biotechniques 23: 409-412.
45. Vartanian JP, Henry M, Wain-Hobson S (1996) Hypermutagenic PCR involving all
four transitions and a sizeable proportion of transversions. Nucleic Acids Res 24:
2627-2631.
46. Jmol: an open-source Java viewer for chemical structures in 3D.
http://www.jmol.org/
47. Sacchi S, Rosini E, Molla G, Pilone MS, Pollegioni L (2004) Modulating D-amino acid
oxidase substrate specificity: production of an enzyme for analytical determination
of all D-amino acids by directed evolution. Protein Eng Des Sel 17: 517-525.
69

PART I: MAP 2.0 3D
48. Rao DE, Rao KV, Reddy TP, Reddy VD (2009) Molecular characterization,
physicochemical properties, known and potential applications of phytases: An
overview. Crit Rev Biotechnol 29: 182-198.
49. Garrett JB, Kretz KA, O'Donoghue E, Kerovuo J, Kim W, et al. (2004) Enhancing the
thermal tolerance and gastric performance of a microbial phytase for use as a
phosphate-mobilizing monogastric-feed supplement. Appl Environ Microb 70:
3041-3046.
50. Fields PA (2001) Review: Protein function at thermal extremes: balancing stability
and flexibility. Comp Biochem Phys A 129: 417-431.
51. Kim MS, Lei XG (2008) Enhancing thermostability of Escherichia coli phytase AppA2
by error-prone PCR. Appl Microbiol Biotechnol 79: 69-75.
52. Wada M, Hsu CC, Franke D, Mitchell M, Heine A, et al. (2003) Directed evolution of
N-acetylneuraminic acid aldolase to catalyze enantiomeric aldol reactions. Bioorgan
Med Chem 11: 2091-2098.
Part of this chapter is adapted with permission from ‘Verma R, Schwaneberg U,
Roccatano D. ACS Synthetic Biology 2012, 1 (4), 139-150.’
70

PART II: MD Simulation
Chapter 3
Introduction to Molecular Dynamics Simulation of
Biomolecules
This chapter provides the brief introduction of molecular dynamics (MD) simulation
followed by the description about the system preparation for MD simulation and the
analysis of generated trajectories. In the following chapters of this thesis, the same
procedure is used to perform MD simulation using P450BM-3 monooxygenase as model
system and the analysis of trajectories.
3.1. Background
Last decades witnessed the rapid development in the field of MD simulations for
biological molecules to study their dynamic processes at atomic level. Far from its infancy,
the computer simulation methods can nowadays provide an important insight into the
molecular basis of protein structure, function and dynamics relationships.[1-4]
MD is a computational chemistry method that describes the dynamics of a molecular
system by integrating Newton’s equations of motion for a system of N interacting atoms. In
MD simulation, the force acting on the i th particle (Fi) are calculated as negative derivatives
of a potential energy function (V) (equation 3.1), called force field that describes the atomic
interactions in an approximate way. In equation 3.1, ri represents the position of i th atom.
71

PART II: MD Simulation
F
i
∂V
( r
= −
1
,r3
…r
∂r
i
N
)
(3.1)
The dynamics of the system is calculated according to the Newton’s law by
numerically integrating the differential equations of motion (equation 3.2). In this way, a
new set of atomic positions and velocities (vi) can be generated at successive integration
time step dt.
∂v i =
∂t
Fi
m
i
(3.2)
The so-called Leap-frog algorithm[5] is commonly used in MD simulation to
integrate the equation of motion. It updates velocities (equation 3.3) and positions
(equation 3.4) of i th atom of mass mi using force F(t) at position ri(t).
1 1 dt
vi ( t + dt)
= vi
( t − dt)
+ F(
t)
2 2 m
(3.3)
1
ri ( t dt)
= ri
( t)
+ dtvi
( t + dt)
2
+
(3.4)
The simulation generates an ensemble of molecular configurations (trajectory) that
describes the evolution of the coordinates and velocities of the system as a function of time.
To generate equilibrium ensemble consistent with the experimental conditions, at which
the system was studied temperature and pressure of the simulated system are controlled
and keep constant during the simulation.[3,4]
72

PART II: MD Simulation
Force field equation
In MD simulation, the force field characterizes the different terms of atomic
interactions as bonded and non-bonded interaction. Bonded interactions include bonds,
angles, dihedrals and improper interaction terms and non-bonded interactions have van
der Waals (vdW) and electrostatic terms (equation 3.5).
V = V + V + V + V + V + V
bond
angle
dihedral
improper
vdW
es
(3.5)
Bonded interactions
Bond stretching between covalently bound atoms (i and j having bond length b) is
calculated by covalent bond potential (Vbond) in GROMOS-96 (equation 3.6) using
b
k
ij
force
constant and equilibrium bond length b0. Angle vibrations between triplets of atoms (i, j
and k) having bond angle θ are represented by cosine based angle potential (Vangle)
(equation 3.7) using
θ
k
ijk
force constant and equilibrium bond angle cosθ0.
V
bond
1
4
b 2 2 2
= kij
( b − b0
)
(3.6)
V
1
=
2
θ
angle
k ijk
(cosθ
− cosθ
)
o
2
(3.7)
Torsional interactions (equation 3.8) involve four atoms (i, j, k, and l) and define the
dihedral angle Φ as the angle present between two planes constituted by first three (i, j and
k) and last three (j, k and l) atoms. Torsional potential define the interactions arising by the
73

PART II: MD Simulation
rotation of two functional groups connected with a bond and defined in equation 3.8,
where
φ
k
ijkl
is the force constant.
V
dihedral
=
k
1+
cos( nφ
−φ
))
φ
ijkl( o
(3.8)
Improper dihedrals are used to define the planarity of the four atoms (i, j, k and l)
defined by harmonic interaction potential in equation 3.9 where
ξ
k
ijkl
is the force constant
and ξ is the dihedral angle on four atoms to keep them in special configuration. For
example, ξ0 will be equal to 0° to keep the four atoms in planar but also tetrahedral
configuration (i, j, k and l)).
V
1
2
ξ
ξ
ξ
2
improper
= k ijkl
( −
o)
(3.9)
Non-bonded interactions
vdW interactions are resulted from the induced atomic dipoles and excluded
volumes of atom pairs. They are attractive at long-range distance but become repulsive at
short-range distance between the atom pairs. In GROMACS, they are defined as Lennard-
Jones (LJ) potential terms (VLJ) (see equation 3.10). In equation 3.10, εij and σij are empirical
parameters where ε is the depth of potential well and σ is the distance at which the
potential is zero and rij is the distance between i th and j th atoms.
V
LJ
⎛
⎜⎛σ
ij
= 4ε
⎜
ij⎜
⎝⎝
rij
⎞
⎟
⎠
12
⎛σ
ij
− ⎜
⎝ rij
⎞
⎟
⎠
6
⎞
⎟
⎟
⎠
(3.10)
74

PART II: MD Simulation
Electrostatic potential (Ves) terms define the Coulombic interaction between two
charged atoms (i and j) calculated by equation 3.11, where ε0 and εr are the dielectric
constants, qi and qj are the atomic charges on i th and j th atoms and rij is the distance between
them.
V
es
=
1
4πε 0
q q
i
ε r
r
j
ij
(3.11)
The topology of a simulated biomolecule is the ordered list of all these interactions
and their predefined parameters for the selected force field. In this thesis all MD
simulations were performed using GROMOS96 43a1 force field.[4]
3.2. Setup of the simulated systems
The crystal structure of the protein was used as the starting coordinates of the MD
simulation (in this thesis PDB ID: 1BVY [6] is used for the simulation of P450BM-3
domains). The proteins were centered in a cubic periodic box and set to have at least a
minimal distance between the protein and any side of the box larger than 0.80 nm so that
the protein cannot see its periodic image across the boundary of the box. They were
solvated by stacking equilibrated boxes of solvent molecules to fill completely the
simulation box. All solvent molecules with any atom within 0.15 nm from the atoms of
protein were removed. SPC water model, a simple three atoms model was used for water
molecule. [7] Sodium counter ions were added by replacing solvent molecules at the most
negative electrostatic potential to provide a total charge of the box equal to zero. The
protonation state of residues in the protein was assumed to be the same as of the isolated
amino acids in solution at pH 7. Hence, the water molecules closest to the charges in the
protein structures are replaced by the counter ions to neutralize the system. The LINCS
(Linear Constraints Solver)[8] algorithm was used to constrain all bond lengths. LINCS [8]
75

PART II: MD Simulation
algorithm use the stable and fast way to reset the bond length after an unconstrained
update. SETTLE [9] algorithm was used for solvent molecules to constrain them as rigid
body.
Electrostatic interactions were calculated by using Particle Mesh Ewalds
method.[10] For the calculation of the long-range interactions, a grid spacing of 0.12 nm
combined with a fourth-order B-spline interpolation were used to compute the potential
and forces between grid points. A non-bonded pair-list cutoff of 1.3 nm was used and
updated at every 5 time-steps.
3.3. Equilibration procedure
Simulated systems were first energy minimized, using the steepest descent
algorithm [11], for at least 2000 steps in order to remove clashes between atoms that were
too close. After energy minimization, all atoms were given an initial velocity obtained from
a Maxwell-Boltzmann velocity distribution at 300 K to start MD simulations.
All systems were initially equilibrated by 100 ps of MD run with position restraints
on the heavy atoms of the solute to allow relaxation of the solvent molecules. In position
restraint, the protein is fixed in the reference position using force constants in each spatial
dimension and let the solvent relax around protein. Berendsen’s thermostat[12] was used
to keep the temperature at 300 K by weak coupling the systems to an external thermal bath
with a relaxation time constant τ = 0.1 ps. The pressure of the system was kept at 1 bar by
using the Berendsen’s barostat[12] with a time constant of 1 ps. After the equilibration
procedure, position restraints were removed and the system was gradually heated from 50
K to 300 K during 200 ps of simulation. Finally, a production run was performed at 300 K.
The analysis of trajectories were performed by using the GROMACS software package
(http://www.gromacs.org/).[13]
76

PART II: MD Simulation
3.4. Structural and dynamical analysis
The structural stability and convergence of protein were examined by analyzing
root mean square deviation (RMSD), radius of gyration (Rg) and secondary structure
elements with respect to its crystal structure as a function of time. Residual root mean
square fluctuation (RMSF) was used to access the dynamics of the target protein during the
simulation.
3.5. Cluster analysis
Cluster analysis was performed to characterize the conformational diversity of the
structures generated during MD simulations. Cluster analysis was performed using the
Gromos clustering algorithm[14] on the backbone atoms (Cα, C and N) of protein. The
analysis is performed on conformations extracted at regular time interval from the
generated trajectory. The resulting conformations were superimposed with respect to
backbone atoms of reference structure after removing overall translation and rotation of
protein in space. The similarity (RMSD-distance) matrix is prepared for all the pairs of
selected conformers.
In Gromos clustering algorithm, RMSD cutoff criteria is used to add similar atomic
coordinates in the cluster having RMSD value less than the defined cutoff. The atomic
coordinates having smallest RMSD from other members of the cluster is known as the
representative structure of that cluster. The same process is repeated until all the selected
structures for the analysis are assigned to the clusters. In this way, the large clusters
represent the ensemble of frequently populated configurations in conformational space
during MD simulation.
77

PART II: MD Simulation
3.6. Principal component analysis
Principal component analysis (PCA, also called essential dynamics analysis) was
performed to access the conformational space by identifying collective motions in the
biomolecules during MD simulation. PCA correlates the atomic positional fluctuations in
proteins and can enhance the molecular level understanding of protein function.[15] The
covariance matrix (C) of atomic coordinates (3N x 3N) is used to construct to identify
collective motions in the biomolecules[16,17] The backbone atoms (Cα, C and N) of the
target proteins were used to explore the conformational subspace in solution. For PCA
analysis first the translational and rotational motions are eliminated from the trajectory (to
consider internal motions only) by the least square fitting of atomic coordinates (using
backbone atoms) to crystal structure. The resulted set of atomic coordinates is used to
construct C of positional deviations using equation 3.12.[17]
C
=
( x − x )( x − x ) T
(3.12)
where x is the subset of atoms in the trajectory x(t) and
represents the ensemble
average over time. The eigenvectors or essential modes can be identified by the
diagonalization of the symmetric matrix (C) using orthogonal transformation matrix T. The
displacements along different eigenvectors were calculated by projecting the atomic
coordinates on eigenvectors. The comparison of eigenvectors obtained from different
simulations was performed using the root-mean-square inner product (RMSIP)[18], which
is defined in equation 3.13.
RMSIP
=
1
N
m
m
∑ ∑
i= 1 j=
1
2
( v
i
⋅ u
j
)
(3.13)
78

PART II: MD Simulation
where vi and uj are i th and j th eigenvectors of the two different m dimensions essential
subspaces of the two systems. RMSIP gives a simple measure to assess the dynamical
similarity of eigenvectors.[18]
3.7. References
1. Dror RO, Dirks RM, Grossman JP, Xu H, Shaw DE (2012) Biomolecular simulation: a
computational microscope for molecular biology. Annu Rev Biophys 41: 429-452.
2. Mccammon JA, Gelin BR, Karplus M (1977) Dynamics of Folded Proteins. Nature
267: 585-590.
3. Karplus M, McCammon JA (2002) Molecular dynamics simulations of biomolecules.
Nat Struct Biol 9: 646-652.
4. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH, Kruger P, et al. (1996)
Biomolecular Simulation: The GROMOS96 Manual and User Guide. VdF:
Hochschulverlag AG an der ETH Zurich and BIOMOS bv, Zurich, Groningen.
5. Hockney RW, Goel SP, Eastwood JW (1974) Quiet high-resolution computer models
of a plasma. J Comp Phys 14: 148-158.
6. Sevrioukova IF, Li HY, Zhang H, Peterson JA, Poulos TL (1999) Structure of a
cytochrome P450-redox partner electron-transfer complex. P Natl Acad Sci USA 96:
1863-1868.
7. Berendsen HJC, Postma JPM, van Gunsteren WF, Hermans J (1981) Interaction
models for water in relation to protein hydration. Intermolecular Forces: 331-342.
8. Hess B, Bekker H, Berendsen HJC, Fraaije JGEM (1997) LINCS: A linear constraint
solver for molecular simulations. J Comput Chem 18: 1463-1472.
9. Miyamoto S, Kollman PA (1992) Settle - an Analytical Version of the Shake and
Rattle Algorithm for Rigid Water Models. J Comput Chem 13: 952-962.
10. Darden T, York D, Pedersen L (1993) Particle Mesh Ewald - an N.Log(N) Method for
Ewald Sums in Large Systems. J Chem Phys 98: 10089-10092.
79

PART II: MD Simulation
11. Cauchy A (1847) Méthode générale pour la résolution des systèmes d'équations
simultanées. C R Acad Sci Paris 25: 536-538.
12. Berendsen HJC, Postma JPM, Vangunsteren WF, Dinola A, Haak JR (1984) Molecular-
Dynamics with Coupling to an External Bath. J Chem Phys 81: 3684-3690.
13. Hess B, Kutzner C, van der Spoel D, Lindahl E (2008) GROMACS 4: Algorithms for
highly efficient, load-balanced, and scalable molecular simulation. J Chem Theory
Comput 4: 435-447.
14. Daura X, Gademann K, Jaun B, Seebach D, van Gunsteren WF, et al. (1999) Peptide
folding: When simulation meets experiment. Angew Chem Int Edit 38: 236-240.
15. Berendsen HJ, Hayward S (2000) Collective protein dynamics in relation to function.
Curr Opin Struct Biol 10: 165-169.
16. Ichiye T, Karplus M (1991) Collective motions in proteins: a covariance analysis of
atomic fluctuations in molecular dynamics and normal mode simulations. Proteins
11: 205-217.
17. Amadei A, Linssen AB, Berendsen HJ (1993) Essential dynamics of proteins. Proteins
17: 412-425.
18. Amadei A, Ceruso MA, Di Nola A (1999) On the convergence of the conformational
coordinates basis set obtained by the essential dynamics analysis of proteins'
molecular dynamics simulations. Proteins 36: 419-424.
80

PART II: P450BM-3 Reductase Domain
Chapter 4
Conformational Dynamics of the FMN-binding Reductase
Domain of Monooxygenase P450BM-3
4.1. Abstract
In the cytochrome P450BM-3, flavin mononucleotide (FMN) binding domain is an
intermediate electron donor between the flavin adenine dinucleotide (FAD) binding
domain and the HEME domain. Experimental evidence has shown that different redox
states of FMN cofactor were found to induce conformational changes in the FMN domain.
Herein, molecular dynamics (MD) simulation is used to gain insight into the latter
phenomenon at atomistic level. We have studied the effect of FMN cofactor and its redox
states (oxidized and reduced) on the structure and dynamics of FMN domain. The results of
our study show significant differences in the atomic fluctuation amplitude of FMN domain
in both holo- and apo-protein. The change in the protonation state of FMN cofactor mostly
affects its binding in holo-protein. In particular, the loops involved in the binding of
isoalloxazine ring (Lβ4) and ribityl side chain (Lβ1) adopt different conformations in both
reduced and oxidized states. In addition, the reduced FMN cofactor mainly induces a
conformational change in Trp574 residue (Lβ4) that is essential to control electron
transfer (ET) within P450BM-3 domains. The structure of the apo-protein in solution
remains mostly unchanged with respect to the crystal structure of the holo-protein.
However, FMN binding loops were more flexible in apo-protein that might favor the
81

PART II: P450BM-3 Reductase Domain
rebinding of FMN cofactor. In the holo-protein simulation, the largest conformational
changes in FMN cofactor are caused by ribityl side chain. The isoalloxazine ring of FMN
cofactor remains almost planar (~177°) in oxidized state and bends along N5 — N10 axis
at the angle of ~160° in reduced state. The collective modes of isoalloxazine ring were
identical in both the protonation states of FMN cofactor except the first eigenvector. In
reduced state, the isoalloxazine ring attains the butterfly motion as a dominant collective
motion in first eigenvector due to the bending along N5 — N10 axis.
4.1. Introduction
Cytochrome P450 monooxygenases, the largest superfamily of heme-containing
soluble proteins, spread widely in almost all domains of life e.g. bacteria, yeast, insects,
mammalian tissues, and plants.[1-3] They catalyze the oxidation of wide variety of
substrates involved in biosynthesis and biodegradation pathways, or in xenobiotics
metabolism.[4] Cytochrome P450BM-3, isolated from Bacillus megaterium, is a
multidomain self-sufficient NADPH dependent flavoenzyme (class III bacterial P450).[5] As
a pivotal member of its super family it has been deeply studied as an important model
system for the comprehension of structure/function relationships and many structural and
kinetic data are available in literature.[6,7] The peculiar catalytic properties of this enzyme
towards industrial applications has also been successfully enhanced by protein
engineering.[8,9]
This enzyme is composed by two reductase domains (with FAD and FMN cofactors)
and a P450 HEME domain (with a HEME cofactor) are arranged on a single polypeptide
chain as HEME-FMN-FAD from N- to C-terminus. The transfer of two successive electrons
from NADPH to HEME cofactor is essential for oxygenation reaction.[10-12] During the
oxygenation reaction, the enzyme is reduced by NADPH, with electrons first transferred to
FAD cofactor of FAD-binding domain then to the FMN cofactor of FMN-binding domain and
finally to the HEME iron in the substrate bound HEME domain. In this ET process, FMN
82

PART II: P450BM-3 Reductase Domain
domain serves as one or two electrons mediator from the FAD cofactor to the heme
iron.[13] FMN cofactor switches between fully oxidized and semiquinone state during
catalytic turnover. The thermodynamically unstable anionic semiquinone state can reduce
HEME iron. However, other P450s utilize FMN hydroquinone as the reduction species.[11-
13] In the substrate free P450BM-3, FMN cofactor stays in a thermodynamically stable
hydroquinone state which is not able to reduce HEME iron.[11]. The use of the anionic
semiquinone state of FMN cofactor as reduction species makes P450BM-3 more efficient
with a high turnover rate in comparison to other members of this family.[14] Therefore, the
thermodynamics properties of the FMN moiety are mainly responsible for the unusual
redox properties of P450BM-3. The protein environment has a strong influence on the
latter mechanism by changing the redox potential of FMN and HEME cofactors. The
mutagenesis studies have shown that the conformation of FMN binding loops plays a
critical role in stabilizing the different redox states of FMN cofactor in the protein
environment.[14,15] The insertion of a glycine residue in the re-face (inner-FMN binding)
loop able to stabilize neutral semiquinone state in P450BM-3 as observed in other diflavin
reductases.[16]
Although, experimental data of the isolated FMN domain in solution and the
crystallographic structure[17,18] are available, molecular dynamics (MD) study of this
protein has not been reported. In this paper, the structural and dynamics properties of the
FMN domain as holo-protein, with FMN cofactor in oxidized and reduced states and as apoprotein
are investigated using classical MD simulations. The aim of this study is to
understand the structural and dynamics properties of the protein in solution and the effect
of the protonation state of FMN cofactor on the conformational dynamics of FMN cofactor
and the whole protein. The paper is organized as follows. In the Methods Section, the
details of the simulations, in particular the refinement of the original GROMOS96 FMN
parameters for the oxidized and reduced states of the FMN cofactors are reported. In the
results part, the analysis of the simulation trajectories for the apo- and the holo-protein in
the oxidized and reduced states are reported. The analysis will be focused on the structural
and dynamics properties of the overall protein structure and of the FMN binding side as
well as the FMN cofactor. Finally, in the discussion and conclusions, the results of the
83

PART II: P450BM-3 Reductase Domain
simulations will be discussed in the context of the experimental knowledge of the FMN
domain and a summary of the paper will be provided.
4.2. Methods
4.2.1. Starting coordinates
The starting coordinates of the FMN domain (residue 479 - 630) were taken from a
non-stoichiometric complex of P450BM-3 (PDB ID: 1BVY, 2.03 nm resolution) which has
one FMN domain with two HEME domains.[18] The crystallographic water (within a
distance of 0.6 nm from the FMN domain) was also kept during system preparation for MD
simulations.
4.2.2. Molecular dynamics simulation
Table 4.1 summarizes the systems used to perform MD simulations with GROMOS96
43a1 force field.[19] The crystal structure[18] has FMN cofactor in oxidized state (FOX)
(see Figure 4.1a). The protonation of FOX isoalloxazine ring at N1 and N5 position
represents the reduced state of FMN cofactor (FHQ) as indicated in Figure 4.1b. For the
preparation of apo-protein simulation, FMN cofactor was removed from the crystal
structure of FMN domain. GROMOS96 force field[20] was used for FMN cofactor. Additional
improper dihedrals were introduced to adopt the conformation of isoalloxazine ring as
observed in crystallographic structure and molecular geometry optimization of flavin in
both redox states, reported in Table S4.1 (for FOX) and Table S4.2 (for FHQ) of supporting
information (SI).[21,22]
In the isoalloxazine ring of FMN cofactor, bending angle (δ) and puckering angle (ρ)
were calculated along N5—N10 axis using v1, v2, a1, a2 and c vectors as shown in Figure 4.2.
84

PART II: P450BM-3 Reductase Domain
In FOX, isoalloxazine ring was kept close to planar while the bending angle of ~160º was
used for FHQ.
Table 4.1: Summarizing MD Simulation of P450BM-3 FMN domain in water.
FMN domain
No. of atoms
No. of solvent No.
of Simulation
molecules counter ions length (ns)
Oxidized (FOX) 33483
10650 14
50
Reduced (FHQ) 33491
10652 14
50
Apo protein
(APO)
33482
10662 13
50
*The abbreviations FOX, FHQ and APO are used in the rest of the paper for FMN domain in oxidized
and reduced states, and as apo-protein, protein, respectively.
Figure 4.1: The schematic representation of FMN cofactor in oxidized (a) and reduced (b) states
with atomic numbering of isoalloxazine ring.[23] N1 and N5 atoms, in blue ovals highlight the
protonation positions. ChemSketch[24] was used to draw the figures.
85

PART II: P450BM-3 Reductase Domain
Figure 4.2: The schematic structure of the isoalloxazine ring to define the bending angle (δ) and
puckering angle (ρ) using the vectors v 1, v 2, a 1, a 2 and c.
4.2.3. FMN binding site analysis
The FMN binding site of holo-protein and apo-protein protein were tracked throughout the
MD simulation using the MDpocket method.[25] The analysis was performed on total 5000
snapshots after taking every 50 th frame from the trajectories. The pocket volume analysis
was performed on aligned MD snapshots with the minimum alpha sphere size of 3 Å, the
minimum number of alpha sphere close to each other for clustering of alpha sphere equal
to The number of iteration to perform pocket volume calculation using Monte Carlo
algorithm was set to 5000. The grid file generated by the first MDpocket run (iso-value =
0.7) was used to extract the grid points for FMN binding pocket. The grid file of FMN
binding pocket was edited manually by deleting some grid points using PyMOL[26] for
better representation of FMN binding site in FMN domain. The FMN binding grid file was
used with the aligned snapshots to track the changes in FMN binding site during the
simulation.
4.2.4. Multiple structural alignment of FMN domain
The homologous structures of FMN domain were obtained by performing BlastP[27]
against PDB database.[28] The protein sequences with identity greater than 20% were
86

PART II: P450BM-3 Reductase Domain
taking in account for further analysis. Six structures were selected as homologous to FMN
domain of P450BM-3 after manually removing the redundant entries from BlastP results
(summarized in Table 2). Multiple structural alignment (MSA) was performed on selected
structures taking the FMN domain as reference structure with maximum RMSD cutoff of
0.5 nm using UCSB chimera.[29]
4.3. Results
4.3.1. FMN domain: structural and dynamical properties
Figure 4.3 shows the backbone root mean square deviation (RMSD) of the FMN
domain as apo- and holo- protein with FMN cofactor in oxidized and reduced states. In FOX,
RMSD curve stabilizes to a plateau with an average value of 0.24 ± 0.04 nm after a rapid
increase in the first 15 ns simulation. In the first 10 ns simulation, FHQ follows the same
trend as observed in FOX. However, after 15 ns, RMSD of FHQ stabilizes to an average value
of 0.16 ± 0.01 nm lower than the one in FOX. In APO, FMN domain remains stable
throughout the simulation with the average RMSD value of 0.19 ± 0.01 nm after the short
equilibration of ~5 ns. For all the simulations, the average radius of gyration (Rg) did not
show appreciable variations from the crystal structure value (1.45 nm).
87

PART II: P450BM-3 Reductase Domain
Figure 4.3: Backbone RMSD with respect to crystal structure as a function of time for APO (in
green) and holo-protein in FOX (in black) and FHQ (in red).
FMN domain has a highly classical flavodoxin fold with five parallel β-sheets (β1 –
β5) that are surrounded by four α-helices (α1 – α4). The loop regions together with
irregular structures (coils and turns) are named according to the secondary structure
element, α(–helix) or β(–sheet), preceding them. FMN cofactor is surrounded by three
loops that succeed β sheets, hence named as Lβ1, Lβ3 and Lβ4 for ribityl binding loop
(residues 488 – 491), inner (re face residues 534 – 544) and outer (si face residuse 571 –
579) FMN binding loop, respectively. The crystallographic protein secondary structure,
calculated using the DSSP method[30], was preserved during the simulation of FOX, FHQ
and even in the APO simulations (see Figure S4.1 of SI).
Figure 4.4: Backbone RMSD (a) and RMSF (b) per residue with respect to crystal structure for
FOX (black), FHQ (red) and APO (green). Vertical bars in grey color show the loop regions. Loops
surrounding the FMN cofactor are shown in black horizontal bars. (c) FMN domain in pink and FMN
cofactor in cyan color with labeled helices and loop regions. FMN binding loops are labeled in
orange color. N and C represent the amino and carboxy terminus of FMN domain.
88

PART II: P450BM-3 Reductase Domain
In Figure 4.4a and 4.4b, backbone RMSD and RMSF per residue with respect to
crystal structure are reported, respectively. FMN domain with labeled helices and loops is
reported in Figure 4.4c. Residues with large RMSD and RMSF values corresponds to loop
regions (represented by grey colored bars in the Figure 4.4a and 4.4b) and to N- and C-
terminus. In all simulations, the largest deviations and fluctuations were observed in Lβ3
and Lβ4 FMN binding loops (black horizontal bars in Figure 4.4a and 4.4b) and Lβ2 and
Lα2 loops which are present opposite to FMN binding site. Lα2 loop shows the highest
deviation in FOX. In APO, the deviations and fluctuations are mainly observed in the FMN
binding loops (especially in the Lβ1 loop) that also occupy the FMN binding cavity during
simulation.
4.3.2. Cluster analysis of FMN domain
Total number of clusters observed in FOX, FHQ and APO are 11, 12 and 7,
respectively. The first two clusters accounts for 80.26 % and 79.25 % of the population for
FOX and FHQ, respectively. The first cluster contributes to 47.10 % in FOX and 60.07 % in
FHQ. While the second cluster represent the 33.16 % and 19.18 % of the total population in
FOX and FHQ, respectively. In APO, even the first cluster covers the 85.17 % of the
population. The apo-protein shows the least conformational diversity during the
simulation. However the difference in the population of clusters in holo-protein indicates
that the FMN protonation state notably influences the conformation of FMN domain.
In Figure 4.5a, 4.5b and 4.5c, the representative conformations of the first two
clusters of FOX and FHQ, and the first cluster of APO are superimposed with the crystal
structure (in sky blue) and shown in cartoon representation. Major differences occur in
loop regions as well as in N- and C- terminus. In particular, slightly larger deviations are
present in FMN binding loops in FOX and FHQ. On the contrary, FMN binding region in APO
is the most deviating part of FMN domain. The loop regions Lβ2 and Lα2 are also affected
by the presence and the protonation state of FMN cofactor. In FOX, Lβ2 and Lα2 show
larger deviations than in FHQ and APO as evidenced by the conformation of the first two
89

PART II: P450BM-3 Reductase Domain
clusters (see Figure 4.5). Lα2 loop flips inwards with higher deviation from crystal
structure in the first two clusters of FOX.
Figure 4.5: The conformation of first two clusters of (a) oxidized (black and gray), (b) reduced
(red and coral) and (c) apo-protein protein (green) superimposed with crystal structure (in sky blue).
Loops and helices (α1, α2, α3 and α4) are labeled. FMN binding loops are labeled in red color. The
labeling of loops belongs to the secondary structure element succeed them.
4.3.3. FMN binding site
Figure 4.6 represents FMN binding site in detail using the representative structure
of the first cluster of FOX (4.6a) and FHQ (4.6b), respectively. The hydrogen bonds between
FMN domain and cofactor were calculated for distance between acceptor and hydrogen
donor ≤ 0.35 nm and an angle among acceptor, donor and acceptor ≤ 30°.
The occurrence of hydrogen bonds that are observed in crystal structure between
FMN domain and the isoalloxazine ring of FMN cofactor are reported in Figure 4.7 for the
simulation of FOX (in blue) and FHQ (in red). NMR spectroscopy studies of the protein in
solution have also evidenced the hydrogen-bonding network involving N1, C2O, N3, C4O
and N5 atoms of the isoalloxazine ring.[14,31] In the crystal structure, the hydrogen atom
90

PART II: P450BM-3 Reductase Domain
from the backbone amino group (NH) of Asn537 was involved in a hydrogen bond
formation with N5 atom of isoalloxazine ring (NH – N5) with the distance of 0.175 nm. The
same hydrogen bond was observed in the first 15 ns of FHQ simulation (in Figure 4.7a).
However, being N5 atom in FHQ a hydrogen donor due to the protonation, a hydrogen
bond with the oxygen from the side chain carboxamide group (-CONH2) of Asn537 (-
(NH2)CO – HN5) was observed in last 25 ns simulation.
Figure 4.6: (a) and (b) shows the FMN binding site from the representative structures of the first
cluster for FOX and FHQ, respectively. The represented residues are within 0.4 nm from the FMN
cofactor (in black). FMN binding loops Lβ1, Lβ3 and Lβ4 are shown as the ribbon of yellow, pink
and cyan, respectively. The residues are labeled in red, blue and green as the part of Lβ1, Lβ3 and
Lβ4, respectively. Dashed lines show hydrogen bonds between isoalloxazine and surrounding
residues. The underlined labels indicate the residues that have major change in conformation after
the change in the redox state of FMN cofactor.
In the crystal structure and simulations, the stable hydrogen bonds were observed
at oxygen (O2) and nitrogen (N3H) of isoalloxazine ring with the hydrogen of backbone
amino group of Gln579 (NH – O2) and the oxygen from backbone carbonyl group (CO) of
91

PART II: P450BM-3 Reductase Domain
Thr577 (N3H – OC), respectively. O4 position was involved in hydrogen bond formation
with hydrogen of hydroxyl group of Thr577 (OH – O4) in the first 15 ns simulation of FOX
but it occurred throughout the whole FHQ simulation. Atom N1 was observed to form
hydrogen bond with backbone amino group of Asp571 (NH – N1) in the crystal structure
and FOX. However, in the FHQ simulation, the occurrence of this bond (NH – N1H) was very
low.
Figure 4.7: Hydrogen bond existence between FMN binding residues and a) isoalloxazine ring (a)
and ribityl side chain (b) of FMN cofactor throughout the MD simulations calculated using every
50 th ps frame. Blue and red color lines show hydrogen bond occurrences in FOX and FHQ,
respectively as a function of time. On Y-axis labeled the partners of hydrogen bond.
92

PART II: P450BM-3 Reductase Domain
The change in protonation state of FMN cofactor affects its binding in FMN domain.
FHQ strengthen the hydrogen-bonding network between ribityl side chain and phosphate
moiety of FMN cofactor and domain. In FHQ, the phosphate group of FMN cofactor forms
strong hydrogen bonds with the residues of Lβ1 loop than in FOX (shown in Figure 4.7b).
The first hydroxyl group of ribityl side chain was involved in strong hydrogen bonding with
the carbonyl oxygen of Ser537 (OH – OC) in FOX and FHQ. The third hydroxyl group of
ribityl side chain shows stronger hydrogen bond with Thr492 hydroxyl in FOX than FHQ.
The latter was also observed to form hydrogen bond with carbonyl oxygen of Cys569 in
FHQ.
In FHQ, the tighter binding of FMN cofactor with stronger hydrogen bonding
network between flavin and protein than FOX, induces the conformational change in FMN
binding loops Lβ1, Lβ3 and Lβ4. The major conformational change was observed in the
orientation of Trp574 residue due to the protonated N5 position in the isoalloxazine ring in
FHQ (Figure 4.6b). The indole ring of Trp574 was nearly coplanar to the isoalloxazine ring
of FMN cofactor in the crystal structure.[17] Experimentally, Trp574 conformation was
observed to be critical to the FMN binding and found to be involved in ET tunneling from
FMN to HEME. In FOX, the indole ring of Trp574 remains in the same conformation as in
the crystal. In FHQ, it rotates to another configuration not aligned to the isoalloxazine ring.
This rotation is a consequence of the steric hindrance induced by the conformation change
in the reduced isoalloxazine ring.
The change in the volume, hydrophobicity, solvent accessibility and polarity of FMN
binding site are reported in Figure 4.8a, 4.8b , 4.8c and 4.8d, respectively during the
simulation of FOX (in black), FHQ (in red) and APO (in green). In APO, the absence of FMN
cofactor promotes a rearrangement of FMN binding site. After rearrangement, the side
chains of amino acids of FMN binding loops replaced the initial water molecules and
occupied the cavity after ~5 ns of simulation. FOX shows larger variation in the geometric
properties of FMN binding pocket than FHQ. Major changes were observed in the volume of
FMN binding pocket after the change in protonation state of FMN cofactor with averages
429 ± 86 and 357 ± 45 for FOX and FHQ, respectively. In FHQ, the pocket volume showed
93

PART II: P450BM-3 Reductase Domain
less variation than in FOX (see Figure 4.8a). The hydrophobicity (Figure 4.8b) and polarity
(Figure 4.8d) of FMN pocket are slightly perturbed in the first 15 ns simulation for then
converge to the same values in both FQH and FOX simulations.
Figure 4.8: FMN binding pocket (a) volume, (b) hydrophobicity, y, (c) solvent accessibility
and (d) polarity for FOX (in black), FHQ (in red) and APO (in green).
4.3.4. Conservation profile of FMN binding site
In Table 4.2, the summary of MSA for FMN domain of P450BM-3 and its homologous
structures (in SI see Figure S4. 4.2 for MSA and Figure S4.3 for conservation patterns mapped
on FMN domain of P450BM-3) is reported. The ribityl side-chain binding region is the most
94

PART II: P450BM-3 Reductase Domain
Table 4.2: Summarizing MSA of FMN domain of P450BM-3 and its homologous structures
and the characterization of their FMN-binding pocket.
PDB
Id
1BVY
1B1C
1JA1
2BF4
1YKG
3HR4
1F4P
Max.
FMN binding pocket
RMSD sequence
Hydro Solvent
Protein
Pola
(nm) identity Volume phobicitbility
accessi-
-rity
(%)
CPR a 0.000 100.00 601.82 358.17 16.69 11
CPR a 0.134 30.26 508.28 299.54 25.26 11
CPR a 0.135 30.26 594.28 313.50 24.65 12
CPR a 0.135 25.00 554.08 354.37 11.29 12
SiR-FP b 0.172 19.86 639.03 354.39 20.88 15
NOS c 0.144 23.68 466.78 258.09 14.44 12
Fld d 0.214 23.13 558.02 340.19 10.82 15
Organism
Bacillus
megaterium
Homo
sapiens
Rattus
norvegicus
Saccharomyces
cerevisiae
Escherichia
coli
Homo
sapiens
Desulfovibrio
vulgaris
conserved region in FMN domain since it is responsible for the tight binding of FMN
cofactor. Among the cytochrome P450 reductases (CPR), the P450BM-3 one has the higher
volume of FMN binding site with higher hydrophobicity and lower polarity. The solvent
accessibility of the FMN binding pocket was found to be in the middle of other homologous
protein. Together all these differences in the properties of FMN-binding pocket of P450BM-
3 results into the better catalytic turnover than other P450 monooxygenases.[14,17]
*CPR a : cytochrome P450 reductase, SiR-FP b : sulfite reductase, NOS c : nitric oxide synthase, Fld d : flavodoxin
95

PART II: P450BM-3 Reductase Domain
4.3.5. Principal component analysis of FMN domain
The cumulative relative positional fluctuation of the first 20 eigenvectors accounts
for 82 %, 79 % and 75 % of the total RPF in the simulation of FOX, FHQ and APO,
respectively. The convergence of the trajectory has been analyzed by comparing the RMSIP
value for the first 20 eigenvectors obtained from the PCA of MD trajectories (50 ns). The
RMSIP values calculated by the two halves of the trajectories resulted in 0.563, 0.624 and
0.594 for the simulation of FOX, FHQ and APO, respectively. The relatively high values of
the RMSIP for the trajectories indicate the good convergence of the essential eigenvectors.
The first two eigenvectors cover the 50%, 41%, and 32% (with 32% and 29% and
20% contribution just from the first eigenvector) of the total fluctuations in FOX, FHQ, and
APO, respectively. The inner product (IP) values for the first two eigenvectors obtained
from the inner product matrix of two trajectories are reported in Table 4.3. The inner
product of the first eigenvector in all simulations was found to be less than 0.350, which
shows that the most important essential mode is different for the three systems.
Table 4.3: RMSIP values of the first 2 eigenvectors obtained from the last 50 ns trajectories
of two different simulations.
Inner product
FOX/FHQ
FOX/APO
FHQ/APO
1 st eigenvector 2 nd eigenvector
1 st eigenvector 0.143 0.400
2 st eigenvector 0.485 0.414
1 st eigenvector 0.268 0.156
2 st eigenvector 0.033 0.156
1 st eigenvector 0.351 0.281
2 st eigenvector 0.183 0.169
96

PART II: P450BM-3 Reductase Domain
Figure 4.9a represents RMSF of the backbone atoms in the first and second
eigenvector of FOX (black), FHQ (red) and APO (green). The corresponding tridimensional
representations obtained after the projection of first and second eigenvectors on MD
trajectories are reported in Figure 4.9b, 4.9c and 4.9d for FOX, FHQ and APO, respectively.
In all simulations, the residues of C terminal and long loop regions, Lβ2 and Lα2 that are
present opposite to FMN binding site show higher fluctuations and together constitute the
collective motions in the first eigenvector. In the first eigenvector of FOX the collective
motion is restricted to Lβ2, Lβ3 and Lα2 loops and C terminal region and have higher
fluctuations. In the second eigenvector of FOX, the higher fluctuation was shown by C
terminal residues and Lα1, Lβ2, Lα2 loops. FHQ shows higher fluctuations in Lα1, Lα2, Lβ4
loops, and C and N terminus in the first eigenvector. The second eigenvector of FHQ shows
higher fluctuations in Lα2, Lβ2 and Lβ4 loops. The first eigenvector of APO shows higher
fluctuations in Lβ1, Lβ2 and Lα2 loops and C terminus, while the second eigenvector shows
in Lα2 loop and in all the FMN binding loops. The higher fluctuations constitute the
collective motion in the first eigenvector were observed in inner FMN binding loop Lβ3 for
FOX, outer FMN binding loop Lβ4 for FHQ and Lβ1 for APO.
Figure 4.10 shows the crystallographic structure complex of HEME with FMN
domain with labeled helices, cofactors and FMN binding loops. In the crystal structure, the
α1 helix is involved in direct or water mediated contacts with HEME domain and the outer
FMN binding loop (Lβ4) interact with the peptide precede the HEME binding loop (K/L
loop). 18 These interaction sites are crucial for the ET from FMN to HEME. Higher
fluctuations observed in Lβ4 and α1 helix regions in the first eigenvector of FHQ might be
related to the inhibition of electron transfer from FMN to HEME by reduced state of FMN
cofactor as observed experimentally. 11 In the first eigenvector of FOX, the higher
fluctuation was restricted to the residues of inner FMN binding loop Lβ3 and loops,
opposite to FMN binding site Lα2 and Lβ2. So the latter defined region is found opposite to
the region of probable HEME binding surface in the crystal structure so the collective
motion constitute by this region in FOX might be related to the electron transfer from FAD
to FMN and it could be the probable binding site for FAD domain. In APO the local structure
97

PART II: P450BM-3 Reductase Domain
remain conserved with highly flexible FMN binding loops that helps to rebind FMN cofactor
in apo-protein protein and working again as holo-protein as found experimentally. 41
Figure 4.9: (a) RPF associated with eigenvectors. Vertical bars in grey color show the loop
regions. FMN binding loop are shown in black horizontal bars. Representation of the RMSF of the
98

PART II: P450BM-3 Reductase Domain
protein backbone atoms along first and second eigenvectors after projection of the trajectory of
FOX (black), FHQ (red) and APO (green) on the corresponding eigenvectors. The 10 sequential
frames representing the extension of the fluctuations in FOX (b), FHQ (c) and APO (d) trajectories
along the first and second eigenvector are reported. . The first extreme is shown in blue color and
last extreme in cyan. Loops and helices es are labeled. Labels in red show the FMN binding loops. N
and C indicate the N- and C-terminus of the protein.
Figure 4.10: FMN domain (in pink) complex with P450 HEME domain (in blue) in crystal
structure (1BVY[18]). HEME cofactor represented in orange and FMN cofactor in green. Helices and
N- and C-terminus are labeled in both domains. Labels Lβ1, Lβ3 and Lβ4 show the FMN cofactor
binding loops in the FMN domain.
4.3.6. FMN cofactor: structural and dynamical properties
The conformational changes of the FMN cofactor induced by the surrounding
protein environment were studied for both redox states. The RMSD and RMSF of phosphate
group atoms for both states show higher fluctuations and deviations than other FMN
cofactor heavy atoms (see Figure 4.11a and 4.11b). Furthermore, the phosphate group of
99

PART II: P450BM-3 Reductase Domain
Figure 4.11: (a) RMSF and RMSD of heavy atoms calculated with respect to crystal
structure. Vertical line shows the beginning of ribityl side chain. (b) Schematic diagram of
FMN cofactor with the numbering used in the plots (4.4a) for the atomic positions of heavy
atoms.
FMN cofactor in oxidized state deviates more from the crystal structure and with higher
fluctuations than it does in the reduced state. This is consistent with the observed
variations of the hydrogen bonding network between the FMN cofactor and the protein.
Figure 4.12a shows the distribution of the value of δ angles of the isoalloxazine ring
(see Figure 4.2) in FOX and FHQ. For the oxidized state of FMN cofactor the vales are
normal distributed in the range from 170º to 180º with the peak centered at 177º. In the
reduced state, the distribution of δ has a larger width. The reduced state of FMN shows a
distribution ranging from 154º to 171º with the peak at 162º. For the latter case, the
average value is consistent with quantum mechanical calculations in vacuum of the
isoalloxazine ring in the reduced state[22,32] that give a value for δ = ~160º.
100

PART II: P450BM-3 Reductase Domain
Figure 4.12: Distribution of (a)
angles calculated at N5 — N10 axis of isoalloxazine ring along the
50 ns simulation of FOX (in black color) and FHQ (in red color) with 0.3 bin width and (b) the
beginning to end distance of the ribityl side chain along 50 ns simulation for FOX and FHQ (0.007
bin width) and in X-ray (in green color) and NMR (in blue color) homologous structures of FMN
domain.
Figure 4.12b shows the distribution of the beginning to end distance for the ribityl side
chain of FMN cofactor in FOX, FHQ and the homologous structures of FMN domain with
FMN cofactor in oxidized state. For the crystallographic and NMR homologous structures
the distance range from 0.73 to 0.8 nm and 0.70 to 0.81 nm, respectively. The distance
observed ed in the crystal structure of FMN domain was 0.77 nm. The distances obtained from
the FOX and FHQ simulations are distributed in the range of 0.61 to 0.90 nm with the main
peaks at 0.78 nm and 0.80 nm, respectively. The beginning to end distance for ribityl side
chain in the simulations and in NMR and crystallographic studies was consistent and
distributed in the same range.
101

PART II: P450BM-3 Reductase Domain
4.3.7. Cluster analysis of FMN cofactor
The cluster analysis was performed on the heavy atoms of FMN cofactor using the
crystal structure as reference and a cutoff of 0.04 nm in the protein environment. The first
cluster comprises 87 % and 99 % of the total 13 and 8 clusters in FOX and FHQ,
respectively. The cumulative sum of the number of clusters obtained from different
simulations as a function of time is reported in Figure S4.4 of the SI. Both the simulations
reached a plateau, which indicates a sufficient sampling of conformational space along the
trajectories. The representative structure of first cluster of FOX (in black) and FHQ (in red)
are reported in Figure S4.4 in SI. The conformational flexibility of ribityl side chain was
observed to be mainly responsible for the conformational diversity of clusters in both the
simulations. The reduced number of clusters in the FHQ simulation is consistent with the
small RMSF fluctuations of the ribityl side chain. The preferred conformation for the FMN
cofactor in both redox states is characterized by a partially elongated ribityl side chain (see
Figure S4.4 in SI).
4.3.8. Principal component analysis of FMN cofactor
The first eight eigenvectors cover ~79 % of the total RPF in both the simulations. In
the protein environment, RPF covers ~25 % and ~27 % by the first eigenvector in FOX and
FHQ, respectively. The superimposition of the first and last extreme structures of
isoalloxazine ring generated by the projection of the first eight eigenvectors on the
trajectory of FOX is shown in Figure 4.13. In the first eigenvector, a symmetric bending
mode of the ring is present in both the simulations. However, the reduced state (Figure
4.13 (1b)) manifest a “butterfly wing” bending mode around the N5 — N10 axis. This type
of vibrational mode has been also reported in previous experimental and quantum
mechanics study.[21,22,32] The other seven modes show similar eigenvectors in both
states. The second most dominant motion is the twisting of the isoalloxazine ring along the
main isoalloxazine axis. The observed collective modes are in the qualitative agreement
with the vibrational normal modes obtained from resonance Raman spectroscopy
measurements and QM calculations for the Lumiflavin.[23,32] In addition, surface
102

PART II: P450BM-3 Reductase Domain
enhanced resonance Raman scattering studies of the free FMN cofactor and in FMN domain
indicate evidences the presence of vibrational modes resulted by atomic displacement of
atoms like C4, O, C4a, C10a , C5a and C9a.[33] These experimental observations are
consistent with the higher fluctuations in correspondence of these atoms observed in PC
modes from the first 8 eigenvectors of our simulations.
Figure 4.13: The superimposition of two extreme structures generated after projecting FOX
trajectory on the first eight eigenvectors (structures from 1 - 8). 1a and 1b show the different
collective motion of 1 st eigenvector in oxidized and reduced state, respectively.
4.4. Discussion and conclusions
MD simulations have been performed on FMN binding reductase domain of
monooxygenase P450BM-3 using FMN cofactor in oxidized and reduced state to
understand the effect of the change in protonation state of isoalloxazine ring on the
103

PART II: P450BM-3 Reductase Domain
conformation and dynamics of FMN domain and cofactor. The results of the simulations
showed that the change of the protonation state in the reduced FMN affect the overall
structure and dynamics of FMN domain in solution. In particular, the structural and
dynamic properties of the si-face FMN binding loop (Lβ3) are strongly influenced by the
change in protonation of FMN cofactor (in FOX). In the apo-protein, the overall local
structure of the protein remains reserved but higher fluctuations were observed in FMN
binding loops. The latter effect can explain the experimental finding of reversible rebinding
of FMN cofactor in apo-protein.[31,34,35] The loop Lβ2 were observed to contribute
mainly on the collective modes of the FMN domain as holo-protein or apo-protein that is
also in agreement with the solution structure of flavodoxin-like domain of E.coli
determined by NMR.[35] The inner FMN binding loop (Lβ3) contributed to the prominent
collective mode of FMN domain in oxidized state. While the outer FMN binding loop (Lβ4)
contribute to the prominent collective mode of FMN domain in reduced state and in apoprotein.
In FHQ, the major conformational change in the FMN binding site residue Trp574
was observed. Trp574 is critical to FMN cofactor binding and electron transfer in P450BM-
3. In FHQ, the latter do not remain coplanar to isoalloxazine ring to avoid the steric
hindrance induced by the conformation change in the FMN cofactor upon protonation as
also suggested by 15 N–NMR[31] and surface enhanced resonance Raman scattering[33]
experiments. Hence, the change in the conformation of Trp574 might be the major factor
that makes the reduced state kinetically unfavorable in P450BM-3 for transferring the
electron from FMN to HEME as observed in previous studies.[36]
The FMN cofactor during simulations acquires different conformations that are
mainly influenced by the movement of ribityl side chain. The binding region of the ribityl
side chain was evolutionary more conserved. In general the oxidized state was observed
more flexible to obtain different conformation in protein environment. The latter might be
the result of change in the FMN binding site properties and hydrogen bond environment in
the reduced state. The isoalloxazine ring of FMN cofactor exhibits mainly 8 collective
motions. Except first vibrational modes all modes were identical in both redox states. FMN
cofactor in reduced state constitutes to the so-called “butterfly motion” as the first
collective motion due to bending of isoalloxazine along the N5 — N10 axis.
104

PART II: P450BM-3 Reductase Domain
In summary, we have analyzed for the first time the dynamics of the FMN binding
domain of P450BM-3 in water. In particular, we have studied the effect of FMN binding on
the fluctuation modes of FMN domain. FMN cofactor is involved in the electron transfer in
P450BM-3 and its dynamics can play an important role in electron transfer. The results of
our study indicate a difference in the fluctuation amplitude of the FMN cofactor in the
different redox states. The latter effect was resulted by the change in the conformation of
FMN binding site due to the protonation state of isoalloxazine ring.
4.5. References
1. Chefson A, Auclair K (2006) Progress towards the easier use of P450 enzymes. Mol
Biosyst 2: 462-469.
2. Wong LL (1998) Cytochrome P450 monooxygenases. Curr Opin Chem Biol 2: 263-
268.
3. Guengerich FP (2001) Common and uncommon cytochrome P450 reactions related
to metabolism and chemical toxicity. Chem Res Toxicol 14: 611-650.
4. Kumar S (2010) Engineering cytochrome P450 biocatalysts for biotechnology,
medicine and bioremediation. Expert Opin Drug Metab Toxicol 6: 115-131.
5. Warman AJ, Roitel O, Neeli R, Girvan HM, Seward HE, et al. (2005) Flavocytochrome
P450 BM3: an update on structure and mechanism of a biotechnologically important
enzyme. Biochem Soc Trans 33: 747-753.
6. Munro AW, Leys DG, McLean KJ, Marshall KR, Ost TW, et al. (2002) P450 BM3: the
very model of a modern flavocytochrome. Trends Biochem Sci 27: 250-257.
7. Girvan HM, Waltham TN, Neeli R, Collins HF, McLean KJ, et al. (2006)
Flavocytochrome P450 BM3 and the origin of CYP102 fusion species. Biochem Soc
Trans 34: 1173-1177.
8. Jung ST, Lauchli R, Arnold FH (2011) Cytochrome P450: taming a wild type enzyme.
Curr Opin Biotechnol 22: 809–817.
105

PART II: P450BM-3 Reductase Domain
9. Whitehouse CJC, Bell SG, Wong L-L (2012) P450BM3 (CYP102A1): connecting the
dots. Chem Soc Rev 41: 1218-1260.
10. Sevrioukova I, Peterson JA (1996) Domain-domain interaction in cytochrome
P450BM-3. Biochimie 78: 744-751.
11. Sevrioukova I, Shaffer C, Ballou DP, Peterson JA (1996) Equilibrium and transient
state spectrophotometric studies of the mechanism of reduction of the flavoprotein
domain of P450BM-3. Biochemistry 35: 7058-7068.
12. Sevrioukova I, Truan G, Peterson JA (1996) The flavoprotein domain of P450BM-3:
Expression, purification, and properties of the flavin adenine dinucleotide- and
flavin mononucleotide-binding subdomains. Biochemistry 35: 7528-7535.
13. Hazzard JT, Govindaraj S, Poulos TL, Tollin G (1997) Electron transfer between the
FMN and heme domains of cytochrome P450BM-3. Effects of substrate and CO. J Biol
Chem 272: 7922-7926.
14. Narhi LO, Fulco AJ (1987) Identification and characterization of two functional
domains in cytochrome P-450BM-3, a catalytically self-sufficient monooxygenase
induced by barbiturates in Bacillus megaterium. J Biol Chem 262: 6683-6690.
15. Pylypenko O, Schlichting I (2004) Structural aspects of ligand binding to and
electron transfer in bacterial and fungal P450s. Annu Rev Biochem 73: 991-1018.
16. Chen HC, Swenson RP (2008) Effect of the Insertion of a Glycine Residue into the
Loop Spanning Residues 536-541 on the Semiquinone State and Redox Properties of
the Flavin Mononucleotide-Binding Domain of Flavocytochrome P450BM-3 from
Bacillus megaterium. Biochemistry 47: 13788-13799.
17. Sevrioukova IF, Hazzard JT, Tollin G, Poulos TL (1999) The FMN to heme electron
transfer in cytochrome P450BM-3 - Effect of chemical modification of cysteines
engineered at the FMN-heme domain interaction site. J Biol Chem 274: 36097-
36106.
18. Sevrioukova IF, Li HY, Zhang H, Peterson JA, Poulos TL (1999) Structure of a
cytochrome P450-redox partner electron-transfer complex. P Natl Acad Sci USA 96:
1863-1868.
106

PART II: P450BM-3 Reductase Domain
19. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH, Kruger P, et al. (1996)
Biomolecular Simulation: The GROMOS96 Manual and User Guide. VdF:
Hochschulverlag AG an der ETH Zurich and BIOMOS bv, Zurich, Groningen: 1.
20. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH, Kruger P, et al. (1996)
Biomolecular Simulation: The GROMOS96 Manual and User Guide. VdF:
Hochschulverlag AG an der ETH Zurich and BIOMOS bv, Zurich, Groningen.
21. Zheng Y-J, Ornstein RL (1996) A Theoretical Study of the Structures of Flavin in
Different Oxidation and Protonation States. J Am Chem Soc 118: 9402-9408.
22. Walsh JD, Miller AF (2003) Flavin reduction potential tuning by substitution and
bending. J Mol Struc-Theochem 623: 185-195.
23. Abe M, Kyogoku Y (1987) Vibrational Analysis of Flavin Derivatives - Normal
Coordinate Treatments of Lumiflavin. Spectrochim Acta A 43: 1027-1037.
24. (2010) ACD/ChemSketch Freeware, version 1201, Advanced Chemistry
Development, Inc, Toronto, ON, Canada, wwwacdlabscom.
25. Schmidtke P, Bidon-Chanal A, Luque FJ, Barril X (2011) MDpocket : Open Source
Cavity Detection and Characterization on Molecular Dynamics Trajectories.
Bioinformatics 27: 3276-3285.
26. (2012) The PyMOL Molecular Graphics System, Version 1504 Schrödinger, LLC,
http://wwwpymolorg/.
27. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ (1990) Basic local alignment
search tool. J Mol Biol 215: 403-410.
28. Bernstein FC, Koetzle TF, Williams GJ, Meyer EF, Jr., Brice MD, et al. (1977) The
Protein Data Bank: a computer-based archival file for macromolecular structures. J
Mol Biol 112: 535-542.
29. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, et al. (2004) UCSF
Chimera--a visualization system for exploratory research and analysis. J Comput
Chem 25: 1605-1612.
30. Kabsch W, Sander C (1983) Dictionary of protein secondary structure: pattern
recognition of hydrogen-bonded and geometrical features. Biopolymers 22: 2577-
2637.
107

PART II: P450BM-3 Reductase Domain
31. Kasim M, Chen HC, Swenson RP (2009) Functional characterization of the re-face
loop spanning residues 536-541 and its interactions with the cofactor in the flavin
mononucleotide-binding domain of flavocytochrome P450 from Bacillus
megaterium. Biochemistry 48: 5131-5141.
32. Nakai S, Yoneda F, Yamabe T (1999) Theoretical study on the lowest-frequency
mode of the flavin ring. Theor Chem Acc 103: 109-116.
33. Macdonald IDG, Smith WE, Munro AW (1999) Analysis of the structure of the flavin
binding sites of flavocytochrome P450BM3 using surface enhanced resonance
Raman scattering. Eur Biophys J 28: 437-445.
34. Wittung-Stafshede P (2002) Role of cofactors in protein folding. Acc Chem Res 35:
201-208.
35. Sibille N, Blackledge M, Brutscher B, Coves J, Bersch B (2005) Solution structure of
the sulfite reductase flavodoxin-like domain from Escherichia coli. Biochemistry 44:
9086-9095.
36. Klein ML, Fulco AJ (1993) Critical residues involved in FMN binding and catalytic
activity in cytochrome P450BM-3. J Biol Chem 268: 7553-7561.
Part of this chapter is adapted with permission from ‘Verma R, Schwaneberg U,
Roccatano D. Journal of Chemical Theory and Computation DOI: 10.1021/ct300723x.’
108

PART II: P450BM-3 Reductase Domain SI
Supporting Information
Conformational Dynamics of the FMN-binding Reductase
Domain of Monooxygenase P450BM-3
Table S4.1: Force field parameters for FMN cofactor in oxidized state for GROMOS96 43a1
force field.[1]
Atom number Atom type Atom name Charge group Partial charge
1 C FC9A 1 0.200
2 NR FN10 1 -0.200
3 C FC10A 2 0.360
4 NR FN1 2 -0.360
5 C FC2 3 0.380
6 O FO2 3 -0.380
7 NR FN3 4 -0.280
8 H FH3 4 0.280
9 C FC4 5 0.380
10 O FO4 5 -0.380
11 C FC4A 6 0.180
12 NR FN5 6 -0.280
13 C FC5A 6 0.100
14 CR1 FC6 7 0.000
15 C FC7 8 0.000
109

PART II: P450BM-3 Reductase Domain SI
16 CH3 FCM7 8 0.000
17 C FC8 9 0.000
18 CH3 FCM8 9 0.000
19 CR1 FC9 10 0.000
20 CH2 FCA 11 0.000
21 CH1 FCB 12 0.150
22 OA FOB 12 -0.548
23 H FHB 12 0.398
24 CH1 FCG 13 0.150
25 OA FOG 13 -0.548
26 H FHG 13 0.398
27 CH1 FCD 14 0.150
28 OA FOD 14 -0.548
29 H FHD 14 0.398
30 CH2 FCE 15 0.150
31 OA FOZ 15 -0.36
32 P FPH 15 0.630
33 OA FOH 15 -0.548
34 H FHH 15 0.398
35 OM FOT1 15 -0.635
36 OM FOT2 15 -0.635
Dihedral parameters
ai aj ak al function c0 c1 c2
13 2 1 3 1 180.000 33.5 2
1 2 3 11 1 180.000 33.5 2
3 11 12 13 1 180.000 33.5 2
11 13 12 1 1 180.000 33.5 2
2 20 21 24 1 0.000 5.86 3
20 22 21 23 1 0.000 1.26 3
20 21 24 27 1 0.000 5.86 3
110

PART II: P450BM-3 Reductase Domain SI
21 24 25 26 1 0.000 1.26 3
21 24 27 30 1 0.000 5.86 3
24 27 28 29 1 0.000 1.26 3
24 30 27 31 1 0.000 5.86 3
27 30 31 32 1 0.000 3.77 3
30 31 32 33 1 0.000 1.05 3
31 33 32 34 1 0.000 1.05 3
Improper dihedral parameters
ai aj ak al function c0 c1
1 2 19 13 2 0.0 167.42
1 12 14 13 2 0.0 167.42
1 13 14 15 2 0.0 167.42
1 19 17 15 2 0.0 167.42
1 2 3 4 2 180.0 167.42
1 12 13 11 2 0.0 167.42
1 14 13 12 2 180.0 167.42
1 2 3 11 2 0.0 167.42
1 17 19 18 2 180.0 167.42
2 1 13 12 2 0.0 167.42
2 11 3 12 2 0.0 167.42
2 1 13 14 2 180.0 167.42
2 3 11 9 2 180.0 167.42
2 19 1 17 2 180.0 167.42
2 4 3 5 2 180.0 167.42
2 1 3 20 2 0.0 167.42
3 2 1 19 2 180.0 167.42
3 5 4 7 2 0.0 167.42
3 4 5 6 2 180.0 167.42
3 11 9 7 2 0.0 167.42
3 2 4 11 2 0.0 167.42
111

PART II: P450BM-3 Reductase Domain SI
3 12 9 11 2 0.0 167.42
3 11 12 13 2 0.0 167.42
3 1 2 13 2 0.0 167.42
3 9 11 10 2 180.0 167.42
4 11 3 9 2 0.0 167.42
4 5 7 9 2 0.0 167.42
4 5 7 8 2 180.0 167.42
4 3 2 11 2 180.0 167.42
4 11 3 12 2 180.0 167.42
5 4 7 6 2 0.0 167.42
5 4 3 11 2 0.0 167.42
5 9 7 11 2 0.0 167.42
5 9 7 10 2 180.0 167.42
6 7 5 8 2 0.0 167.42
6 5 7 9 2 180.0 167.42
7 9 5 8 2 0.0 167.42
7 9 10 11 2 180.0 167.42
8 7 9 20 2 0.0 167.42
8 7 9 11 2 180.0 167.42
8 10 11 12 2 180.0 167.42
9 7 11 10 2 0.0 167.42
9 12 11 13 2 180.0 167.42
12 1 13 19 2 180.0 167.42
12 11 9 7 2 180.0 167.42
12 13 14 15 2 180.0 167.42
13 19 1 17 2 0.0 167.42
13 15 14 17 2 0.0 167.42
13 15 14 16 2 180.0 167.42
13 12 1 14 2 0.0 167.42
14 12 13 11 2 180.0 167.42
112

PART II: P450BM-3 Reductase Domain SI
14 1 13 19 2 0.0 167.42
14 17 15 19 2 0.0 167.42
14 15 17 18 2 180.0 167.42
15 17 14 16 2 0.0 167.42
16 15 17 18 2 0.0 167.42
17 15 19 18 2 0.0 167.42
19 17 15 16 2 180.0 167.42
20 24 22 21 2 35.0 167.42
21 27 25 24 2 35.0 167.42
24 30 28 27 2 35.0 167.42
Table S4.2: Force field parameters for FMN cofactor in reduced state for GROMOS96 43a1
force field.[1]
Atom number Atom type Atom name Charge group Partial charge
1 C FC9A 1 0.1
2 NR FN10 1 -0.2
3 C FC10A 1 0.1
4 NR FN1 2 -0.28
5 H FH1 2 0.28
6 C FC2 3 0.38
7 O FO2 3 -0.38
8 NR FN3 4 -0.28
9 H FH3 4 0.28
10 C FC4 5 0.38
11 O FO4 5 -0.38
12 C FC4A 6 0.00
13 NR FN5 7 -0.28
14 H FH5 7 0.28
113

PART II: P450BM-3 Reductase Domain SI
15 C FC5A 8 0.00
16 CR1 FC6 9 0.00
17 C FC7 10 0.00
18 CH3 FCM7 10 0.00
19 C FC8 11 0.00
20 CH3 FCM8 11 0.00
21 CR1 FC9 12 0.00
22 CH2 FCA 13 0.00
23 CH1 FCB 14 0.15
24 OA FOB 14 -0.548
25 H FHB 14 0.398
26 CH1 FCG 14 0.15
27 OA FOG 15 -0.548
28 H FHG 15 0.398
29 CH1 FCD 15 0.15
30 OA FOD 16 -0.548
31 H FHD 16 0.398
32 CH2 FCE 16 0.15
33 OA FOZ 17 -0.36
34 P FPH 17 0.63
35 OA FOH 17 -0.548
36 H FHH 17 0.398
37 OM FOT1 17 -0.635
38 OM FOT2 17 -0.635
Dihedral parameters
ai aj ak al function c0 c1 c2
15 1 3 2 1 180.00 33.5 2
1 12 2 3 1 180.00 33.5 2
3 15 12 13 1 180.00 33.5 2
12 13 1 15 1 180.00 33.5 2
114

PART II: P450BM-3 Reductase Domain SI
2 22 23 26 1 0.00 5.86 2
22 24 23 25 1 0.00 1.26 2
22 23 26 29 1 0.00 5.86 2
23 26 27 28 1 0.00 1.26 2
23 26 30 29 1 0.00 5.86 2
26 29 30 31 1 0.00 1.26 2
26 32 29 33 1 0.00 5.86 2
29 32 33 34 1 0.00 3.77 2
32 33 34 35 1 0.00 1.05 2
33 35 34 36 1 0.00 1.05 2
Improper dihedral parameters
ai aj ak al function c0 c1
1 15 2 21 2 5 167.42
1 13 16 15 2 0 167.42
1 15 16 17 2 0 167.42
1 21 19 17 2 0 167.42
1 2 3 4 2 160 167.42
1 13 15 12 2 50 167.42
1 16 15 13 2 180 167.42
1 3 2 12 2 50 167.42
1 19 21 20 2 180 167.42
2 1 15 13 2 60 167.42
2 12 3 13 2 60 167.42
2 15 1 16 2 180 167.42
2 3 12 10 2 180 167.42
2 21 1 19 2 180 167.42
2 4 3 6 2 180 167.42
2 3 1 22 2 5 167.42
3 1 2 21 2 160 167.42
3 6 4 8 2 0 167.42
115

PART II: P450BM-3 Reductase Domain SI
3 4 6 7 2 180 167.42
3 12 10 8 2 0 167.42
3 2 4 12 2 5 167.42
3 13 10 12 2 0 167.42
3 12 13 15 2 50 167.42
3 1 2 15 2 50 167.42
4 12 3 10 2 0 167.42
4 6 8 10 2 0 167.42
4 6 8 9 2 180 167.42
4 3 2 12 2 180 167.42
4 12 3 13 2 180 167.42
5 6 4 3 2 180 167.42
6 4 8 7 2 0 167.42
6 4 3 12 2 0 167.42
6 10 8 12 2 0 167.42
6 10 8 11 2 180 167.42
7 8 6 9 2 0 167.42
7 6 8 10 2 180 167.42
8 10 6 9 2 0 167.42
8 10 11 12 2 180 167.42
9 8 10 11 2 0 167.42
9 8 10 12 2 180 167.42
10 8 12 11 2 0 167.42
10 13 12 15 2 160 167.42
11 12 10 3 2 180 167.42
13 1 15 21 2 180 167.42
13 10 12 8 2 180 167.42
13 16 15 17 2 180 167.42
14 15 13 12 2 135 167.42
15 21 1 19 2 0 167.42
116

PART II: P450BM-3 Reductase Domain SI
15 17 16 19 2 0 167.42
15 16 17 18 2 180 167.42
15 13 1 16 2 0 167.42
16 15 13 12 2 160 167.42
16 1 15 21 2 0 167.42
16 19 17 21 2 0 167.42
16 17 19 20 2 180 167.42
17 19 16 18 2 0 167.42
18 17 19 20 2 0 167.42
19 17 21 20 2 0 167.42
21 19 17 18 2 180 167.42
22 26 24 23 2 35 334.84
23 29 27 26 2 35 334.84
26 32 30 29 2 35 334.84
117

PART II: P450BM-3 Reductase Domain SI
Figure S4.1: Secondary structure per residue calculated by DSSP[2] along the trajectory as a
function of time for (a) FOX, (b) FHQ and (c) APO. Color code represents different secondary
structure elements.
118

PART II: P450BM-3 Reductase Domain SI
Figure S4.2: (a) Multiple structure alignment of FMN domain (1BVY) and its homologous
structures (summarized in Table 2) with the conservation profile, root mean square deviation
(RMSD) and charge variation per residue created by using Chimera[3]. . Green and yellow color
boxes show the helixes and beta strands, respectively. However the purple color boxes represent
the residues involved in FMN binding. (b) The phylogenetic tree of FMN domain and its homologous
structures generated by ClustalW2[4].
119

PART II: P450BM-3 Reductase Domain SI
Figure S4.3: Evolutionary conservation profile on the FMN domain of P450BM-3 using
RWB color scheme. Red region shows the highly conserved residues in the domain. FMN
domain is shown in cartoon representation with FMN cofactor in green color and labeled
helixes and loop op regions. FMN binding loops are labeled in blue color. N and C represent
the amino and carboxy terminus of FMN domain.
Figure S4.4: Cumulative sum of the number of clusters obtained from the simulations. The
sampling of clusters was performed over 50 ns of FOX (black) and FHQ (red) using RMSD cutoff of
0.04 nm. . The representative conformations of FMN cofactor in the first cluster of FOX (black) and
FHQ (red) are shown.
120

PART II: P450BM-3 Reductase Domain SI
References
1. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH, Kruger P, et al. (1996)
Biomolecular Simulation: The GROMOS96 Manual and User Guide. VdF:
Hochschulverlag AG an der ETH Zurich and BIOMOS bv, Zurich, Groningen.
2. Kabsch W, Sander C (1983) Dictionary of protein secondary structure: pattern
recognition of hydrogen-bonded and geometrical features. Biopolymers 22: 2577-
2637.
3. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, et al. (2004) UCSF
Chimera--a visualization system for exploratory research and analysis. J Comput
Chem 25: 1605-1612.
4. Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA, et al. (2007) Clustal
W and Clustal X version 2.0. Bioinformatics 23: 2947-2948.
Part of this chapter is adapted with permission from ‘Verma R, Schwaneberg U,
Roccatano D. Journal of Chemical Theory and Computation DOI: 10.1021/ct300723x.’
121

PART II: P450BM-3 HEME/FMN Complex
Chapter 5
Insight into the redox partner interaction mechanism in
cytochrome P450BM-3 using molecular dynamics
simulation
5.1. Abstract
Flavocytochrome P450BM-3 is a soluble bacterial reductase composed by two flavin
(FAD/FMN) and one HEME domains. The understanding of atomic details of the inter
domain electron transfer (ET) mechanism is requisite for better exploitation of the enzyme
in biotechnological applications and to extend the knowledge of P450 proteins family in
general. In this paper, we have performed molecular dynamics (MD) simulations on both
FMN and HEME domains, isolated and their crystallographic complex to study their binding
modes and to garner insight on structural determinant for inter-domain ET. In the
simulation of the complex, we observed conformational rearrangements in both the
domains that reduce the separation between FMN and HEME cofactor. In particular,
FMN/HEME closest distance was decreased from 1.84 nm (in crystal structure) to an
average of 1.41 ± 0.09 nm during the simulation with a minimum distance of 1.02 nm.
These distance values are within the range of distance for ET tunneling between the two
redox centers. The analysis of the possible ET pathways in the crystal complex indicates
Met490 of ribityl tail binding loop of FMN domain, and Ala399 and Cys400 of HEME
domain as possible mediators for ET. However, during simulation, at ~1.41 nm FMN/HEME
122

PART II: P450BM-3 HEME/FMN Complex
distance, in spite of Ala399, Cys400, Phe393 along with Met490 take part in ET while, with
the minimum FMN/HEME distance of ~1.02 nm, only Met490 mediates the ET tunneling.
The results of the simulations are in agreement with previously proposed hypotheses that
the crystal complex of FMN/HEME domains is not in the optimal arrangement for favorable
needed electron transfer rate under physiological conditions.
5.2. Introduction
Cytochrome P450s are the regio- and stereo-selective monooxygenase of the family
oxidoreductase with a wide variety of substrates.[1-3] They have been studied as the
potential catalyst for the production of high value oxygenated organic molecules to
promote enzyme-mediated product formation.[4-6] In particular, cytochrome P450BM-3 is
a NADPH dependent fatty acid hydroxylase system, isolated from soil bacterium Bacillus
megaterium.[7,8] This enzyme is an attractive target and model system for biochemical and
biomedical applications for different reasons. First, it is a stable, catalytically self-sufficient
protein with a convenient multidomain structure that allows easier production and
handling than other monooxygenases of the same family. Second, it is a water soluble
enzyme with a high catalytic efficiency and oxygenase rate and readily expressed
recombinantly.[9,10] Third, it resembles to eukaryotic diflavin reductase such as human
microsomal P450s. As a pivotal member of its super family it has been widely studied as an
important model system for the comprehension of structure-function-dynamics
relationships with the wealth of structural and kinetic data.[11,12]
P450BM-3, being a multidomain protein, has two reductase flavin adenine
dinucleotide (FAD)- and flavin mononucleotide (FMN)- binding domains and a HEME
domain arranged as HEME-FMN-FAD on a single polypeptide chain.[13,14] The main
catalytic function of P450s is to transfer oxygen atom from molecular oxygen to their
substrates. During the reaction, the enzyme is reduced by NADPH, with electrons first
transferred to FAD cofactor of FAD-binding domain and then to HEME iron in the substrate
123

PART II: P450BM-3 HEME/FMN Complex
bound HEME domain mediated by FMN cofactor of FMN-binding domain. The
crystallization of the whole P450BM-3 protein has been proven difficult due to the
presence of flexible linker regions between domains. However, the crystallographic
structures of the isolated HEME domain[15], FAD domain[16] and a non-stoichiometric
complex with one FMN and two HEME domains[15] are available in the PDB database. In
the FMN/HEME complex (PDB ID: 1BVY) the smallest edge to edge distance between redox
centers is 1.81 nm.[15] However, it has been shown from the survey of electron transfer
(ET) in oxidoreductase protein structures that the latter should be less than 1.40 nm for an
efficient ET tunneling between redox centers in the protein environment.[17] Munro et al.
used modeling approach to rationalize the electron transfer between FMN to HEME and
postulated the movement of FMN domain is essential to decrease the distance between
FMN and HEME cofactors within the physiological range (less than 1.40 nm) for ET.[11]
In this study, we aim to extend our knowledge regarding structure-functiondynamics
relationships in P450BM-3 at atomistic level using Molecular dynamics (MD)
simulations of the isolated HEME and FMN domains and of their complex in water. It has
been proved experimentally that specific arrangement of HEME and FMN domain is
responsible for the catalytic efficiency and high oxygenase rate of P450BM-3.[18] In this
paper, for the first time, the dynamics in solution of the complex and the isolated HEME and
FMN domains will be comparatively investigated. In particular, the relative rearrangement
of FMN/HEME domains and how the latter affects the ET pathways from isoalloxazine ring
of FMN cofactor to HEME iron will be analyzed.
The chapter is organized as follows. The details of the MD simulations and the
analysis of the trajectories are reported in the Method section. The Results and Discussions
section is organized as follows. In the first part, the general structural and properties of the
simulated systems to assess the quality of the simulation are reported. Cluster analysis is
used to identify representative structures to evidence the difference of the domain in
solution and in the complex. The following paragraphs will focus on the ET pathways
between the FMN and HEME calculated on selected conformations from the cluster analysis
of the trajectory. The structural behavior of the substrate access channel will be also
124

PART II: P450BM-3 HEME/FMN Complex
reported. Hence, the collective dynamics of the system will be analyzed using the principal
component analysis of the trajectories. Finally, in the conclusion section a summary of the
outcome of the study is provided.
5.3. Methods
5.3.1. Starting coordinates
The non-stoichiometric FMN/(HEME)2 complex of one FMN domain without
substrate (PDB ID: 1BVY with resolution 0.203 nm)[15] were used as to obtain the starting
coordinate for MD simulation. Out of two HEME domains (chain A: 20 - 450) was in close
proximity of FMN domain (chain F: 479 - 630) in the crystal structure. Hence, These A and
F chains were extracted from crystal structure (including crystallographic water within
0.60 nm from the proteins) and used as starting coordinates for MD simulation. 1,2-
ethanediol molecules were removed from the crystallographic structure and replaced by
water molecules.
5.3.2. Molecular dynamic simulations
The GROMOS96 43a1 force field[19] was used for all simulations. The MD
simulations performed in this study are summarized in Table 5.1. Figure 5.1 shows the
FMN and HEME cofactors in stick representation. The parameters for the ferric iron of
HEME cofactor were adopted from Helms et al.[20] and has been employed already for the
MD simulation of P450BM-3 HEME domain by Roccatano et al..[21,22] The partial charges
were redistributed on porphyrin ring of HEME cofactor to adopt the parameters for
GROMOS96 43a1 force field[19] with hydrogen atoms bound to bridging carbon in
porphyrin ring (see Table S5.1 in Supplementary Information (SI)). FMN cofactor was in
oxidized state in the FMN domain. Additional improper dihedrals were introduced to adopt
125

PART II: P450BM-3 HEME/FMN Complex
the conformation of isoalloxazine ring as observed in crystallographic structure and
molecular geometry optimization of flavin in both redox states.[23-25]
Table 5.1: Summary of the MD simulations of P450BM-3 in water
Starting coordinates
No. of atoms No. of solvent No.
of
molecules counter ions
(Na + )
HEME Domain (A
65650 20365 16
chain)
FMN Domain (F chain) 33483 10650 14
Complex (AF chain) 86101 26671 30
Simulation
length (ns)
100
100
100
*The abbreviation A, F and AF chain will be used in rest of the paper for HEME domain, FMN domain and
HEME/FMN complex, respectively.
126

PART II: P450BM-3 HEME/FMN Complex
Figure 5.1: (a) HEME cofactor and (b) FMN cofactor are in stick representation, colored by
elements such as, oxygen in red, nitrogen in blue, hydrogen in green, iron or phosphorus in orange
and carbon in gray, with atomic labeling according to GROMOS96[19] topology.
5.3.2. Electron transfer tunneling
Electron tunneling (ET) from FMN to HEME cofactor was calculated by the program
Pathways.[26,27] The method calculates donor to acceptor partial electronic coupling
influenced by protein structure using graph theory to identify the electron transfer
pathways in biological electron transfer reactions.[26] FMN to HEME cofactor ET pathway
was identified in the crystal structure and P450BM-3 conformation after rearrangement by
taking C8 atom of isoalloxazine ring of FMN cofactor as donor and HEME iron as acceptor
with the default parameters of Pathways. The ET pathway was visualized by VMD.[28]
5.4. Results and discussion
5.4.1. Structural properties
The structural stability and convergence of P450BM-3 domains were examined by
analyzing root mean square deviation (RMSD), radius of gyration (Rg) and secondary
structure elements with respect to crystal structure during the MD simulation. Figure 5.2a
shows the backbone RMSD of the proteins as a function of time. The total RMSD curves for
both the AF chains and the single A chain reach to a plateau with an average RMSD of 0.41 ±
0.03 nm and 0.36 ± 0.03 nm, respectively. The RMSD of isolated A chain shows an average
plateau to a slightly lower value of 0.33 ± 0.02 nm. The F chain in the complex increases its
RMSD value to an average on the last 10 ns of simulation of 0.25 ± 0.02 nm. The RMSD of
isolated F chain increases rapidly to stabilize after 10 ns of simulation to an average value
of 0.26 ± 0.02 nm. In Figure 5.2b, the radius of gyrations is also reported. In the first 10 ns
127

PART II: P450BM-3 HEME/FMN Complex
of the simulation, the Rg of the complex decreases of ~3.7% from the crystallographic
value (2.42 nm) to the average value of 2.33 ± 0.01 nm. The variation of the single A
domain in the complex and in solution with respect the initial structure (2.16 nm) is less
than 1.8% (2.12 ± 0.01 nm and 2.14 ± 0.01 nm, respectively). F chain does not show
variation from the crystal structure (1.45 nm) with an average of 1.45 ± 0.01 nm and 1.46 ±
0.01 nm for isolated F chain and in complex simulation, respectively.
Figure 5.2: (a) Backbone RMSD and (b) Rg with respect to crystal structure as a function of time
for AF chain (black), A of AF chain (red), F of AF chain (green), A chain (blue) and F chain (orange).
In P450BM-3, A and F chain have structurally conserved P450 and flavodoxin like
protein fold, respectively. Figure 5.3c shows the structure with labeled helices of A (A to L)
and F (α1 to α4) chain and FMN binding loops (Lβ1, Lβ3 and Lβ4). The loop regions
together with irregular structures (coils and turns) are named according to the secondary
structure element (α helix or β sheet) preceding them. DSSP criteria[29]
were used to
follow the secondary structure of the P450BM-3 domains in isolated and complex MD
simulations (Figure S5.1 in SI). The secondary structure remains fairly conserved during
the simulations.
128

PART II: P450BM-3 HEME/FMN Complex
Figure 5.3a and 5.3b show residual RMSD and RMSF with respect to crystal
structure, respectively. The regions involved in cofactor binding show smaller deviation
and fluctuation from the crystal structure in isolated (in red color) and complex (black
color) simulations. For both domains, the loop regions and N- and C- terminus show higher
deviation. The isolated domains deviate more than the one in complex except the region
between helices, A - B, B’ - C, H - I, and K - L and in G helix in A chain and Lβ3 in F chain.
Isolated F chain shows largest deviation in Lβ2 and Lβ4 regions. In both systems, F chain
shows higher fluctuation in Lβ2 and Lα2 loops. In complex simulation, the loop regions
A/B, and F/G fluctuate more. While in isolated F chain simulation, inner FMN cofactor
binding loop Lβ3 fluctuate slightly more.
Figure 5.3: Backbone RMSD (a) and RMSF (b) per residue with respect to crystal structure for
isolated domains (in red) and in complex (in black) MD simulations. The green vertical line
separates HEME and FMN domains. Horizontal bars, in blue and orange color represent helices
(labeled) and beta sheets, respectively. The regions involved in cofactor binding are represented by
horizontal bars in purple color. (c) HEME and FMN domain are in cartoon representation in sky
blue and tan color, respectively. HEME and FMN cofactor are in red and green color, respectively.
Helices, cofactors, FMN binding regions and, N- and C- terminus are labeled.
129

PART II: P450BM-3 HEME/FMN Complex
5.4.2. Cluster analysis
The first two clusters account for 46.16 % and 27.12 % for AF chain, respectively.
For isolated domain simulation, A chain and F chain have 6 clusters and in complex
simulation 7 and 8 clusters, respectively. The first two clusters of A chain in complex covers
76.87 % and 10.99 % and as isolated domains they account for 46.53 % and 30.12 %,
respectively. The first cluster of F chain covers ~64 % and second cluster covers 21.05%
and 23.05 % in complex and isolated simulation, respectively. A chain is more liable for
conformational change in isolated simulation than in complex, while F chain shows the
negligible difference in conformation space in both the simulations.
Figure 5.4: The representative conformation of first cluster of (a) AF, (b) A and (c) F chain in
cartoon representation superimposed with crystal structure. In the crystal structure, A and F chain
are in sky blue and tan color, respectively. For the complex simulation, A and F chain are in dark
blue and brown color, respectively. For the isolated domain simulation, A and F chain are in orange
130

PART II: P450BM-3 HEME/FMN Complex
and purple color, respectively. HEME and FMN cofactors are in green, red and blue color in crystal
structure, isolated domain and in complex structure, respectively. The helices and FMN cofactor
binding loops are labeled. Amino and carboxy terminal of the domain are labeled in red color.
Figure 5.4a, 5.4b and 5.4c show the crystal structure superimposed with the
representative conformation of the first cluster of AF chain and, A and F chain in isolated
and complex simulation, respectively. Major differences were observed in the loop regions
of the domains in both simulations. N terminus region (residue 20 – 82, including A, B and
B’ helices) of A chain deviates more from crystal structure in both the simulations. In
complex simulation, larger deviation in G helix and, H/I and K/L (residue 380 - 390) loop
region of A chain. Residues 380 – 390 in K/L loop region precedes HEME cofactor binding
region. H/I and K/L loops are involved in the binding of FMN domain. α2 helix of F chain
shows larger deviation in isolated F chain simulation and resulted in a compact
conformation of FMN domain in solution than in complex. In complex simulation, the
representative structure of first cluster of AF chain represents the conformational
rearrangement in both domains to increase compactness of AF chains complex. The
deviations in both the domains from crystal structure mainly involve G helix and H/I and
K/L loops of A chain and displacement of F chain towards HEME domain that resulted into
the decrease in the minimum distance between both the cofactors.
5.4.3. Substrate access channel
Pro45 and Ala191 were found to be at the mouth of substrate access channel. In the
crystal structure of P450BM-3 complex, P45Cα - A191Cα is 1.61 nm apart (0.87 nm in A
chain of 1BU7). Chang et. al. observed that the substrate binding was not dramatically
affected by the closeness of substrate access channel in P450BM-3 using MD simulation
and docking approach.[30] The behavior of substrate access channel has been accessed by
monitoring the distance between these two residues by Roccatano et al..[22] P45Cα -
A191Cα minimum distance was calculated and reported in Figure 5.5 during the isolated
domain and complex simulations. Both simulations show higher variations in P45Cα -
131

PART II: P450BM-3 HEME/FMN Complex
A191Cα distance in the first 20 ns simulation. After that in isolated A chain, an average
distance of 1.11 ± 0.10 nm was observed with slight variations. In A chain of AF chain, the
P45Cα - A191Cα distance continues decreasing till 32 ns simulation and reaches to an
average distance of 0.59 ± 0.10 nm. In comparison to isolated A chain substrate access
channel was partially closed in A chain in complex that might be the result of more
deviation of F/G loop in complex than in isolated domain.
Figure 5.5: Minimum distance between P45Cα and A191Cα as a function of time for A chain in
isolated (in red color) and complex (in black color) simulations.
In the crystal structure, the crystallographic water molecule was not ligated to heme
iron (Fe) (distance in 1BVY > 6 nm and 1BU7 0.24 nm). When A chain was solvated in
water (crystal structure), the water molecule was present at the distance of 0.47 nm and
0.34 nm from heme iron in isolated and complex simulation. Figure S5.2 in SI shows the
minimum distance between Fe and water molecules (every 100 ps). An average distance of
0.28 ± 0.13 nm and of 0.34 ± 0.14 nm was observed for A chain in complex and isolated
domain simulation, respectively.
132

PART II: P450BM-3 HEME/FMN Complex
5.4.4. ET tunneling pathways
The minimum distance between heavy atoms of isoalloxazine ring of FMN and
HEME cofactors is represented in Figure 5.6 (the AF chain simulation was extended to 150
ns to check the distance convergence). During the simulation, FMN/HEME distance is
decreased from 1.81 nm (in crystal structure) to an average distance of 1.41 ± 0.09 nm with
the minimum distance of 1.02 nm that is within the range for expected ET between redox
centers[17] (1.40 - 1.50 nm) and proposed by Munro et al. 11 The decreased distance might
result into the ET rate of 10 8 to 10 11 s -1 , that is consistent with experimental and theoretical
observations.[11,17]
Figure 5.6: Minimum distance between heavy atoms of isoalloxazine ring of FMN and HEME
cofactor as a function of time. Red color horizontal line shows the distance observed in crystal
structure.[15]
133

PART II: P450BM-3 HEME/FMN Complex
Figure 5.7a, 5.7b and 5.7c show the ET pathway identified by Pathways VMD
plugin[27] in the crystal structure (min. dist 1.80 nm), representative of first cluster of AF
chain (minimum distance 1.41 nm) and AF chain with minimum distance (minimum
distance 1.10 nm) between FMN to HEME cofactor, respectively. In Table 5.2, the results of
the analysis are summarized. In the crystal structure, FMN to HEME ET tunneling is
mediated by solvent molecules as well but after rearrangement in AF conformation, FMN
cofactor come close to HEME cofactor and eliminate the involvement of water molecules in
ET tunneling. In Figure 5.7c, when the FMN to HEME distance is ~ 1 nm, ET tunneling is
mediated by the Met490 residue only and the ET pathway length decrease from 2.7 nm (in
crystal structure) to 1.8 nm and electronic coupling from 4.00 x 10 -10 to 2.68 x 10 -8 ,
respectively.
Table 5.2: Electron transfer tunneling pathway in AF chain complex calculated by
Pathways[27] VMD plugin.
Coordinates
FMN/HEME
Max. Distance
minimum
coupling along ET
distance
(a.u.) pathway (nm)
(nm)
Crystal
structure
1.8 4.00 x10 -10 2.70
First cluster 1.4 9.07 x10 -9 1.96
Minimum
FMN/HEME 1.1 2.68 x10 -8 1.79
distance
Amino acids involved in
the ET pathway
FMN(C8) → M490 →
Sol → Sol → A399 →
C400 → HEME(FE)
FMN(C8) → M490 →
→ F393 → HEME(FE)
FMN(C8) → M490 →
→ HEME(FE)
134

PART II: P450BM-3 HEME/FMN Complex
Figure 5.7: ET tunneling from the isoalloxazine ring (C8 atom) of FMN cofactor (in gray color) to
iron center of HEME cofactor (in black color) represented by red color tubes in a) crystal structure,
b) conformation of first cluster and c) conformation with minimum distance between HEME to FMN
cofactor. The amino acids with in the distance of 0.50 nm from both the cofactors are labeled and
shown in licorice representation colored by element type (oxygen in red, carbon in cyan and
nitrogen in blue color) and their associated secondary structure in cartoon representation in sky
blue for HEME domain and in orange color for FMN domain. The residues involved in electron
tunneling are represented and labeled in green color.
5.4.5. Essential dynamics
The cumulative sum of relative positional fluctuation (RPF) of first 50 eigenvectors
of A and F chain in isolated and complex simulation is greater than 69% and reported in
135

PART II: P450BM-3 HEME/FMN Complex
Figure S5.3 of SI. RMSIP for first twenty eigenvectors of A chain and F chain in both
simulations was less than 0.53. The inner product value of the first three eigenvectors for A
and F chain were less than 0.25 and 0.43, respectively. The overlap and inner product
analysis indicate the existence of different set of collective motions in the eigenvectors of
same time windows of both the trajectories.
Figure 5.8a, 5.8b and 5.8c represent RPF associated with first three eigenvectors of
A and F chain in isolated (in red color) and complex (in black color) simulation. Figure 5.9
show the RMSF associated with first three eigenvector (a, b and c) of A (in sky blue) and F
(in tan color) chain in isolated (a1, b1 and c1) and complex (a2, b2 and c2) simulation,
respectively.
Figure 5.8: RPF for (a) first, (b) second and (c) third eigenvector of A and F chain in isolated (red
color) and complex (black color) simulation. The green vertical line separates HEME and FMN
domain. Horizontal bars, in blue and orange color represent helixes (labeled) and beta sheets,
136

PART II: P450BM-3 HEME/FMN Complex
respectively. The regions involved in cofactor binding are represented by horizontal bars in purple
color.
In complex simulation, the first collective motion (Figure 5.9a1) of A chain involves
the turn succeeds beta sheet 1 (residues 44 – 48, highest RPF for Arg47 that is involved in
substrate binding), D/E loop (residues 130 – 138), F/G loop (residues 190 – 196), K/L loop
(residue 385 – 390) and C- terminus loop (residues 425 – 432 and 452 – 458). The
cooperative motion in the turn succeeds beta sheet 1 and F/G loop related to the
movement of substrate access channel closing and opening. Residue F393 of the latter
region of K/L loop was involves FMN domain binding and found to be involved in ET
tunneling in the average structure of first cluster of AF chain. The first collective mode of F
chain in complex involves the major contribution of Lα2 loop with slightly higher RPF of
Lβ2 and Lβ3 (inner FMN binding loop). The cooperative motion of Lα2 and Lβ3 might
facilitate ET tunneling from FMN to HEME cofactor. In complex the collective motions of
both the domains were synchronized to relate ET tunneling and change in substrate
binding. The effect was clearly seen when the first eigenvectors of AF chain was compared
with that of A and F chain (reported in Figure S5.4a and S5.5a in SI). In both AF and, A and F
chain, the first eigenvector show fluctuations in the same regions with higher fluctuations
in the collective mode of AF chain and cooperative effect due to their binding. The second
collective mode in A chain involve mainly the motion in D/E and G/H loops, beta sheets in
K/L regions and A/B region and the third collective motion was restricted to D/E and G/H
loops and C-terminus loop (residues 425 – 432). F chain shows involvement of Lα2 and Lβ2
loops and C- terminus region in the second collective mode and Lα2, Lβ3 and Lβ5 in the
third eigenvector. In AF chain, the collective motion associated with the first two
eigenvectors belongs to the movement of F chain towards A chain to decrease the distance
between FMN and HEME cofactor and show slightly higher fluctuation than in the
individual chains. In the third eigenvector the major difference was observed mainly in Lβ3
and Lα2 loop of F chain with higher fluctuations. The collective motion associated with the
first three eigenvectors of AF chain is reported in Figure S5.5a, S5.5b and S5.5c,
respectively in SI.
137

PART II: P450BM-3 HEME/FMN Complex
Figure 5.9: RMSF of protein backbone atoms along first (a), second (b) and third (c)
eigenvector after projection of the trajectory on the corresponding eigenvector of A and F
138

PART II: P450BM-3 HEME/FMN Complex
chain in complex simulation (a1, b1 and c1) and in isolated simulation (a2, b2 and c2). The
10 sequential frames represent the extension of the fluctuations in trajectories along the
eigenvectors. The first extreme conformation is shown in green color and last extreme in
violet color. Other conformations of A and F chain are in sky blue and tan color,
respectively. Helices and loops in FMN domain are labeled. N and C indicate the N- and C-
terminus of the protein (labeled in red color).
The isolated A chain have the first collective mode (Figure 5.9a2) have higher RPF at
the end of C helix (residue 103 – 107) and C- terminus (residues 452 – 458). Other region
involves in first collective mode were D/E, E/F, F/G and K/L loop (residue 385 – 390).
Together the motion in related to the change in substrate binding region and FMN domain
binding region. The first collective mode of the isolated F chain shows higher RPF in Lα2
and Lβ2, and slightly high RPF in Lβ3 and Lβ4. In the isolated domains, the collective
motions were more independent i.e. in F chain related to binding of FMN cofactor and in A
chain restricted to substrate binding region. The second collective mode in A chain involve
mainly the motion in D/E, E/F and F/G loops and only in F/G region in the third collective
motion. F chain shows involvement of Lα2 and Lβ2 in the second collective mode and
Lα2 and Lβ3 in the third eigenvector.
5.5. Conclusions
We performed MD simulation on HEME and FMN domains as isolated domain or in
complex. Structure remains conserved in both the systems throughout the simulation.
During simulation, HEME/FMN complex undergoes into the conformational rearrangement
in the first 10 ns simulation (with decrease in Rg from 2.42 nm to 2.33 nm) and resulted
into the compactness of the complex with decrease in FMN/HEME distance from 1.81 nm
to an average 1.41 nm. FMN domain in solution show major conformational change in Lα2
loop in the absence of HEME domain. In isolated HEME domain major conformational
139

PART II: P450BM-3 HEME/FMN Complex
change were observed in FMN binding region especially in C helix and H/I and K/L (residue
385 – 395) loops. G helix and inner FMN cofactor loop (Lβ3) fluctuate more in both the
simulations. Both domains differ in the atomic fluctuation amplitude in isolated and
complex simulation. In complex the collective motion was dominated by the interaction
mechanism between HEME and FMN domain and associated change in substrate access
channel. The movement of FMN domain over HEME domain might be related to ET
mechanism in P450BM-3 as proposed earlier and responsible to the ET rate between both
the domains in the range from 10 8 to 10 11 s -1 under physiological condition as observed
experimentally and proposed theoretically earlier.[11]
5.6. References
1. Chefson A, Auclair K (2006) Progress towards the easier use of P450 enzymes. Mol
Biosyst 2: 462-469.
2. Wong LL (1998) Cytochrome P450 monooxygenases. Curr Opin Chem Biol 2: 263-
268.
3. Guengerich FP (2001) Common and uncommon cytochrome P450 reactions related
to metabolism and chemical toxicity. Chem Res Toxicol 14: 611-650.
4. Urlacher VB, Eiben S (2006) Cytochrome P450 monooxygenases: perspectives for
synthetic application. Trends biotechnol 24: 324-330.
5. Bernhardt R (2006) Cytochromes P450 as versatile biocatalysts. J Biotechnol 124:
128-145.
6. Coon MJ (2005) Cytochrome P450: nature's most versatile biological catalyst. Annu
Rev Pharmacol Toxicol 45: 1-25.
7. Narhi LO, Fulco AJ (1986) Characterization of a catalytically self-sufficient 119,000-
dalton cytochrome P-450 monooxygenase induced by barbiturates in Bacillus
megaterium. J Biol Chem 261: 7160-7169.
140

PART II: P450BM-3 HEME/FMN Complex
8. Narhi LO, Fulco AJ (1987) Identification and Characterization of 2 Functional
Domains in Cytochrome-P-450bm-3, a Catalytically Self-Sufficient Monooxygenase
Induced by Barbiturates in Bacillus-Megaterium. J Biol Chem 262: 6683-6690.
9. Munro AW, Lindsay JG, Coggins JR, Kelly SM, Price NC (1994) Structural and
Enzymological Analysis of the Interaction of Isolated Domains of Cytochrome-P-450
Bm3. Febs Letters 343: 70-74.
10. Warman AJ, Roitel O, Neeli R, Girvan HM, Seward HE, et al. (2005) Flavocytochrome
P450 BM3: an update on structure and mechanism of a biotechnologically important
enzyme. Biochem Soc Trans 33: 747-753.
11. Munro AW, Leys DG, McLean KJ, Marshall KR, Ost TW, et al. (2002) P450 BM3: the
very model of a modern flavocytochrome. Trends Biochem Sci 27: 250-257.
12. Girvan HM, Waltham TN, Neeli R, Collins HF, McLean KJ, et al. (2006)
Flavocytochrome P450 BM3 and the origin of CYP102 fusion species. Biochem Soc
Trans 34: 1173-1177.
13. Peterson JA, Sevrioukova I, Truan G, GrahamLorence SE (1997) P450BM-3: A tale of
two domains - Or is it three? Steroids 62: 117-123.
14. Munro AW, Daff S, Coggins JR, Lindsay JG, Chapman SK (1996) Probing electron
transfer in flavocytochrome P-450 BM3 and its component domains. Eur J Biochem
239: 403-409.
15. Sevrioukova IF, Li HY, Zhang H, Peterson JA, Poulos TL (1999) Structure of a
cytochrome P450-redox partner electron-transfer complex. P Natl Acad Sci USA 96:
1863-1868.
16. Joyce MG, Ekanem IS, Roitel O, Dunford AJ, Neeli R, et al. (2012) The crystal
structure of the FAD/NADPH-binding domain of flavocytochrome P450 BM3. FEBS
Journal 279: 1694-1706.
17. Page CC, Moser CC, Chen X, Dutton PL (1999) Natural engineering principles of
electron tunnelling in biological oxidation-reduction. Nature 402: 47-52.
18. Hazzard JT, Govindaraj S, Poulos TL, Tollin G (1997) Electron transfer between the
FMN and heme domains of cytochrome P450BM-3. Effects of substrate and CO. J Biol
Chem 272: 7922-7926.
141

PART II: P450BM-3 HEME/FMN Complex
19. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH, Kruger P, et al. (1996)
Biomolecular Simulation: The GROMOS96 Manual and User Guide. VdF:
Hochschulverlag AG an der ETH Zurich and BIOMOS bv, Zurich, Groningen.
20. Helms V, Deprez E, Gill E, Barret C, Hui Bon Hoa G, et al. (1996) Improved binding of
cytochrome P450cam substrate analogues designed to fill extra space in the
substrate binding pocket. Biochemistry 35: 1485-1499.
21. Roccatano D, Wong TS, Schwaneberg U, Zacharias M (2005) Structural and dynamic
properties of cytochrome P450 BM-3 in pure water and in a
dimethylsulfoxide/water mixture. Biopolymers 78: 259-267.
22. Roccatano D, Wong TS, Schwaneberg U, Zacharias M (2006) Toward understanding
the inactivation mechanism of monooxygenase P450 BM-3 by organic cosolvents: a
molecular dynamics simulation study. Biopolymers 83: 467-476.
23. Verma R, Schwaneberg U, Roccatano D Conformational Dynamics of the FMNbinding
Reductase Domain of Monooxygenase P450BM-3. Unpublished.
24. Walsh JD, Miller AF (2003) Flavin reduction potential tuning by substitution and
bending. J Mol Struc-Theochem 623: 185-195.
25. Zheng Y-J, Ornstein RL (1996) A Theoretical Study of the Structures of Flavin in
Different Oxidation and Protonation States. J Am Chem Soc 118: 9402-9408.
26. Beratan DN, Betts JN, Onuchic JN (1991) Protein electron transfer rates set by the
bridging secondary and tertiary structure. Science 252: 1285-1288.
27. Balabin IA, Hu X, Beratan DN (2012) Exploring biological electron transfer pathway
dynamics with the Pathways Plugin for VMD. J Comput Chem 33: 906-910.
28. Humphrey W, Dalke A, Schulten K (1996) VMD: Visual molecular dynamics. J Mol
Graphics 14: 33-&.
29. Kabsch W, Sander C (1983) Dictionary of protein secondary structure: pattern
recognition of hydrogen-bonded and geometrical features. Biopolymers 22: 2577-
2637.
30. Chang YT, Loew GH (1999) Molecular dynamics simulations of P450 BM3--
examination of substrate-induced conformational change. J Biomol Struct Dyn 16:
1189-1203.
142

PART II: P450BM-3 HEME/FMN Complex SI
Supporting Information
Insight into the redox partner interaction mechanism in
cytochrome P450BM-3 using molecular dynamics
simulation
Table S5.1: Partial charge on HEME cofactor with ferric iron.[1-3]
Atom number Atom type Atom name Charge group Partial charge
1 FE FE 1 1.0
2 NR NA 1 -0.4
3 NR NB 1 -0.4
4 NR NC 1 -0.4
5 NR ND 1 -0.4
6 C CHA 2 -0.2
7 HC HHA 2 0.2
8 C C1A 3 0.2
9 C C2A 3 -0.1
10 C C3A 3 0.0
11 C C4A 3 0.1
12 CH3 CMA 4 0.0
13 CH2 CAA 5 0.0
14 CH2 CBA 5 0.0
143

PART II: P450BM-3 HEME/FMN Complex SI
15 C CGA 6 0.27
16 OM O1A 6 -0.635
17 OM O2A 6 -0.635
18 C CHB 7 -0.2
19 HC HHB 7 0.2
20 C C1B 8 0.05
21 C C2B 8 0.05
22 C C3B 8 -0.1
23 C C4B 8 0.2
24 CH3 CMB 9 0.0
25 CR1 CAB 10 0.0
26 CH2 CBB 10 0.0
27 C CHC 11 -0.2
28 HC HHC 11 0.2
29 C C1C 12 0.2
30 C C2C 12 0.0
31 C C3C 12 -0.1
32 C C4C 12 0.2
33 CH3 CMC 13 0.0
34 CR1 CAC 14 0.0
35 CH2 CBC 14 0.0
36 C CHD 15 -0.2
37 HC HHD 15 0.2
38 C C1D 16 0.2
39 C C2D 16 0.1
40 C C3D 16 -0.2
41 C C4D 16 0.2
42 CH3 CMD 17 0.0
43 CH2 CAD 18 0.0
44 CH2 CBD 18 0.0
144

PART II: P450BM-3 HEME/FMN Complex SI
45
46
47
C CGD 19
OM O1D 19
OM O2D 19
0.27
-0.635
-0.635
Figure S5.1: Secondary structure per residue calculated by DSSP[4] along the trajectory as a
function of time for HEME domain and FMN domain (a) in complex simulation and (b) isolated.
Color code represents different secondary structures.
145

PART II: P450BM-3 HEME/FMN Complex SI
Figure S5.2: Minimum distance between water molecules and HEME iron as a function of
time (every 100 ps) in isolated (in red color) and complex (in black color) simulation.
146

PART II: P450BM-3 HEME/FMN Complex SI
Figure S5.3: Relative positional fluctuation of first 50 eigenvectors of the A and F chains in
isolation and complex simulation. In AF chain, the first 50 eigenvectors account for 80.45 % of total
RPF with 25.96 % contribution by the first eigenvector. A chain has 79.28 % and 86.77 %
cumulative RPF with 27.19 % and 48.54 % contribution by first eigenvector in complex and
isolated domain simulation, respectively. For F chain, cumulative RPF of first 50 eigenvectors was
90.96 % and 89.19 % with 33.98 % and 35.02 % RPF of first eigenvector in complex and isolated
domain simulation.
147

PART II: P450BM-3 HEME/FMN Complex SI
Figure S5.4: RPF for (a) first, (b) second and (c) third eigenvector of AF chain (cyan color), and A
and F chain in complex (black color) simulation. The green vertical line separates Heme and FMN
domain. Horizontal bars, in blue and orange color represent helixes (labeled) and beta sheets,
respectively. The regions involved in cofactor binding are represented ed by horizontal bars in purple
color.
148

PART II: P450BM-3 HEME/FMN Complex SI
Figure S5.5: RMSF of protein backbone atoms along first, second and third eigenvector after
projection of the trajectory on the corresponding eigenvector of AF chain in complex simulation in
(a), (b) and (c), respectively. The 10 sequential frames represent the extension of the fluctuations in
trajectories along the eigenvectors. The first extreme conformation is shown in green color and last
extreme in violet color. Other conformations of Heme and FMN domain are in sky blue and tan
color, respectively. Helixes and loops are labeled. N and C indicate the N- and C-terminus of the
protein (labeled in red color).
149

PART II: P450BM-3 HEME/FMN Complex SI
References:
1. Helms V, Deprez E, Gill E, Barret C, Hui Bon Hoa G, et al. (1996) Improved binding of
cytochrome P450cam substrate analogues designed to fill extra space in the
substrate binding pocket. Biochemistry 35: 1485-1499.
2. Roccatano D, Wong TS, Schwaneberg U, Zacharias M (2005) Structural and dynamic
properties of cytochrome P450 BM-3 in pure water and in a
dimethylsulfoxide/water mixture. Biopolymers 78: 259-267.
3. Roccatano D, Wong TS, Schwaneberg U, Zacharias M (2006) Toward understanding
the inactivation mechanism of monooxygenase P450 BM-3 by organic cosolvents: a
molecular dynamics simulation study. Biopolymers 83: 467-476.
4. Kabsch W, Sander C (1983) Dictionary of protein secondary structure: pattern
recognition of hydrogen-bonded and geometrical features. Biopolymers 22: 2577-
2637.
150

PART II: P450BM-3 HEME/FMN & CoSep
Chapter 6
A molecular dynamics study of the effect of
cobalt(II)sepulchrate as an electron transfer mediator
on the conformational and dynamics of P450BM-3
6.1. Abstract
The major limitation of the exploitation of P450BM-3 in the industrial processes is
the consumption of expensive NADPH as a reduction equivalent in the catalytic cycle.
Experimentally NADPH has also been found to inactivate the enzyme in the absence of
substrate. The use of alternative cost effective cofactor like cobalt(III)sepulchrate (CoSep)
with zinc dust as the source of electron has been proposed as a possible alternative
solution to overcome the latter limitation. The mechanism of interaction of cobalt(III)
sepulchrate with the protein has not yet elucidated at molecular level. In this paper, we
propose a novel model of CoSep and use to study using molecular dynamic simulations its
interaction with isolated HEME domain and the HEME/FMN complex of the P450BM-3. The
aim of the study is to identify the putative binding modes of the CoSep on P450BM-3
domains and their effect on their conformation, dynamics and electron transfer (ET)
tunneling. The results of this study indicates that CoSep preferentially bind to negative
charged residue on the surface exposed regions of P450BM-3 domains. Two ET tunneling
pathways were observed for HEME/FMN complex in the presence of CoSep. First one is
from CoSep to FMN isoalloxazine ring involving Trp574 then to HEME iron mediated by
151

PART II: P450BM-3 HEME/FMN & CoSep
water molecule on interface and Met490 (ribityl tail of FMN cofactor binding loop) and
Phe394, the same residues were involved in ET tunneling in water simulation of P450BM-3
domains. The second ET pathways was from CoSep to HEME iron via Ile102 and Leu103 (C
helix residues) and Ile401 and Cys400. In isolated HEME domain ET tunneling involved the
residues of B’/C loop, Asp84, Gly85, Leu86 and Phe87. The collective motions of different
amplitude were observed in both the systems and were found to facilitate the ET tunneling
from CoSep to HEME iron.
6.2. Introduction
Cytochrome P450 monooxygenases, the largest superfamily of heme-containing
soluble proteins, spread widely in almost all domains of life e.g. bacteria, yeast, insects,
mammalian tissues, and plants.[1-3] They catalyze the oxidation using oxygen molecules of
wide variety of substrates involved in biosynthesis and biodegradation pathways, or in
xenobiotics metabolism[4] in the presence of reduction equivalents. The high
stereoselectivity and large variety of possible substrates make these enzymes particular
interesting for industrial applications. However, their complexity, low solubility, low
catalytic turnover and in particular the utilization of expensive source of electron have so
far limited their use.[4] Cytochrome P450BM-3, from the soil bacterium Bacillus
megaterium, is one of the most widely studied members of this family.[5,6] Being soluble
and self sufficient (P450 and reductase domains linked together on a single polypeptide
chain), P450BM-3 has higher catalytic turnover with easy expression and purification in
cell free medium.[7] Protein engineering approaches have been used successfully to
increase technologically viability of P450BM-3 by fine-tuning its catalytic parameters and
substrate recognition.[8,9] In past years, fast advancements are also made towards the cost
effectiveness of the P450BM-3 catalytic reaction by the regeneration or substitution of
expensive cofactor (NADPH or NADH) as a source of electrons.[7] The electrochemistry of
P450BM-3 received considerable attention and various methods have allowed direct
electron transfer system (from electrode to protein via conducting polymer films like
152

PART II: P450BM-3 HEME/FMN & CoSep
BaytronP)[10] or mediated electron transfer system (to shuttle electrons from electrode to
protein via small electro active compounds like Zn dust (as a source of electron) with
Co(III)sepulchrate (as electron mediator) for driving the catalytic cycle.[11,12] Protein
engineering via directed evolution and rational design offers an attractive solution to
improve the enzymatic properties and to enhance the electrochemical performance of the
enzyme.[11-13] In this paper, we performed molecular dynamic simulation to gain insight
into the interaction mechanism of P450BM-3 domains with cobalt(II)sepulchrate (CoSep)
as an electron transfer mediator. The results will help to investigate the effect of CoSep
binding on conformation, dynamics and ET tunneling in P450BM-3 domains.
The chapter is organized as follows. The details of MD simulations and force field
modeling of the CoSep are reported in Method section. The Results and Discussion section
is organized as follows. The preferential binding sites of CoSep on P450BM-3 domains are
reported. The following paragraph provides information about the ET tunneling from
CoSep to P450BM-3 domains. Hence, the collective dynamics of the system will be analyzed
using the principal component analysis of the trajectories. Finally, in the conclusion section
provides a summary of the outcome of the study.
6.3. Methods
6.3.1. Starting coordinates
The non- stoichiometric complex of one FMN domain to two HEME domains without
substrate were used as a starting coordinate (PDB ID: 1BVY with resolution 0.203 nm).[14]
For MD simulation, HEME domain (chain A: 20 - 450) associated with FMN domain (chain
F: 479 - 630) was extracted from the starting coordinates including crystallographic water
(within 0.60 nm from the protein was extracted using VMD software[15]). 1,2-ethanediol
molecules were removed and replaced by water molecules from the crystallographic
153

PART II: P450BM-3 HEME/FMN & CoSep
structure. The MD simulation was set up for isolated HEME-binding domain and
HEME/FMN complex in water- CoSep mixture.
6.3.2. Molecular dynamics simulation and modeling
The GROMOS96 43a1 force field[16] was used for all simulations. The MD
simulations performed in this study are summarized in Table 6.1. The HEME cofactor
parameters for ferric iron was adopted from Helms et al.[17], that was already employed
for the MD simulation of P450BM-3 HEME domain by Roccatano et al..[18,19] The partial
charges were redistributed on porphyrin ring of HEME cofactor to adopt the parameters
for GROMOS96 43a1 force field[16] with hydrogen atoms bound to bridging carbon in
prophyrin ring.[20] FMN cofactor was in oxidized state in the FMN domain. Additional
improper dihedrals were introduced to adopt the conformation of isoalloxazine ring as
observed in crystallographic structure and molecular geometry optimization of flavin in
both redox states. [21,22] Detail of the modified force field for FMN are reported in a
previous paper.[23]
For CoSep, (schematically represented in Figure 6.1) the force field parameters for
bond and bond angles are adapted from Dehayes et al.[24] (the values are reported in
Table S6.1 in Supporting Information (SI)). The non-bonded parameters are adopted from
GROMOS96 43a1 force field.[16]
Density functional theory calculation using Becke3LYP method[25] with LanL2DZ
basic set[26] was used for the geometry optimization. Atomic partial charges were derived
using CHelpG scheme[27] after constraints them to reproduce dipole moment (partial
charges are reported in Table S6.2 in SI). A ionic radius for Co +2 of 0.075 nm was used to fit
electrostatic potentials. All the calculations were performed using Gaussian09 package.[28]
Fourty molecules of CoSep were randomly placed in the simulation box and solvated
by stacking equilibrated boxes of solvent molecules to fill the simulation box. The CoSep
154

PART II: P450BM-3 HEME/FMN & CoSep
concentration was equal to ~0.5 mM and it corresponds to the one used experimentally for
the fastest biotransformation in P450BM-3 using Zn dust and cobalt(III)sepulchrate as
alternative electron transfer system.[12]
Figure 6.1: CoSep is in ball and stick representation, colored by elements such as, nitrogen
in blue, hydrogen in green, and carbon in gray with labeled atom name and number (except
hydrogen).
Table 6.1: Summarizing the MD simulations of P450BM-3 domains in water and CoSep
solution.
No. of
Starting
No. of counter Simulation
No. of atoms No. of CoSep solvent
coordinates
ions
length (ns)
molecules
Heme
domain (A 65650 - 20365 16 Na + 100
chain)
FMN domain 33483 - 10650 14 Na + 100
155

PART II: P450BM-3 HEME/FMN & CoSep
(F chain)
Complex (AF
chain)
A chain &
CoSep
AF chain &
CoSep
86101 - 26671 30 Na + 100
64597 40 19638 64 Cl - 100
85275 40 26029 50 Cl - 100
*The abbreviation A, F and AF are used in the rest of the paper for HEME domain, FMN domain and
HEME/FMN complex, respectively.
6.4. Results and discussion
The difference in conformation and dynamics of isolated FMN and HEME domain
and HEME/FMN complex has been discussed in our previous paper.[20,23] Herein, we will
focus on the effect of CoSep binding on the conformation, dynamics and ET tunneling in
P450BM-3 domains in isolated HEME domain and in HEME/FMN complex. The presence of
CoSep does not affect the structure of P450BM-3 domains significantly. The structural
stability and convergence of P450BM-3 domains in CoSep solution were compared with the
one in water and reported in SI through backbone root mean square deviation (RMSD)
(Figure S6.1), radius of gyration (Rg) (Figure S6.2) and backbone RMSD and RMSF per
residue (Figure S6.3a and S6.3a, respectively) using crystal structure as reference. In CoSep
solution, both HEME domain and the complex show the same behavior as in pure water.
The backbone RMSD of the HEME domain in the CoSep solution reaches a plateau with an
average value of 0.25 ± 0.01 nm after 10 ns of simulation (Figure S6.1 in SI) and it shows
the lowest RMS deviations and fluctuations (Figure S6.3a and S6.3b in SI). On the contrary,
the AF complex shows the largest deviation in the residues of H helix (Figure S6.3a in SI).
156

PART II: P450BM-3 HEME/FMN & CoSep
6.4.1. CoSep binding on P450BM-3 domains
At the end of the simulations of both the isolated HEME domain and the complex,
CoSep molecules were found bounded mainly at the surface exposed loop regions of the
protein (see Figure S6.4 of SI). The average minimum distances between the CoSep
molecules and HEME iron and the isoalloxazine ring along the simulations are reported in
Figure S6.5 of SI. After 20 ns of simulation, CoSep molecules approach the HEME domain
within an average distance of 1.72 ± 0.44 nm and 1.94 ± 0.19 nm in the isolated protein and
in the complex, respectively. Figure 6.2a, 6.2b and 6.2c shows the minimum distance
between CoSep and residues of isolated A chain and, of A and F chain in complex,
respectively.
The cluster analysis was used to select representative structure for the isolated
HEME domain and for the complex. The first cluster of A chain and complex accounts for
more ~83 % and 99 %, respectively in CoSep solution. The binding of CoSep in isolated A
chain and AF chain complex is shown in Figure 6.2c and 6.2d, respectively. In isolated A
chain, CoSep molecules bind mainly at HEME/FMN interface in contact with C (94 – 107)
and H (233 – 238) helix and, B’/C (82 – 94), H/I (239 – 251), K/L (359 – 367) C-terminus
turn (441 – 445) regions. Other regions of CoSep binding on isolated chain was A/B (32 –
38 and 51 – 55), B helix (55 – 61) and B/B’ loop (61 – 68) and E helix (139 – 143) and F
helix (181, 182) and F/G (192 – 198) loop. In AF chain, the binding of F chain slightly
influence the distribution of CoSep on A chain. CoSep were more abundant at F chain, two
of them were present near (≤ 0.50 nm) to FMN cofactor binding loop Lβ3 and Lβ4 at
FMN/HEME interface. In P450BM-3 domains, the regions of CoSep binding were found to
be rich in charged residues especially negative charged polar residues (aspartic acid and
glutamic acid) i.e. obtained from the analysis of number of contacts between CoSep and
P450BM-3 residues within the distance of 0.50 nm (reported in Figure S6.6 and also shown
in Figure 6.2d and 6.2e by the abundance of oxygen (red color surface) in CoSep binding
region).
157

PART II: P450BM-3 HEME/FMN & CoSep
Figure 6.2: Minimum distance (≤ 1.0 nm) between CoSep and residues of (a) isolated A chain, (b)
A chain and (c) F chain in complex. Horizontal bars, in blue and orange color represent helices
(labeled) and beta sheets, respectively. The regions involved in cofactor binding are represented by
horizontal bars in purple color. (d) and (e) show binding site for CoSep in the structure of first
158

PART II: P450BM-3 HEME/FMN & CoSep
cluster of the isolates A chain and AF chain, respectively. CoSep molecules are in ball and stick
representation and colored by element type (nitrogen in blue color, hydrogen in white color, and
carbon in black color). HEME and FMN domain are in cartoon representation in sky blue and tan
color, respectively with surface colored according to element type. FMN and HEME cofactors are in
green and red, respectively. Helices, cofactors, loops and, N- and C- terminus (in red color) are
labeled.
6.4.2. Effect of CoSep binding on substrate access channel
The accessibility of active site has been monitored by the dynamics behavior of
residues Pro45 and A191 that line the substrate access channel by Roccatano et al..[19]
P45Cα - A191Cα minimum distance (1.61 nm in crystal structure) calculated and reported
in Figure S6.7 of SI. After 30 ns of simulation, least variation in the distances was observed
in all the simulations. CoSep binding in A chain of AF complex induce larger deviation in G
helix and F/G loop region and resulted in wider substrate access channel with an average
distance of 1.87 ± 0.15 nm than in A chain of AF complex in water (0.59 ± 0.10 nm). Isolated
A chain was less affected by CoSep binding and show slightly higher P45Cα - A191Cα
distance (1.50 ± 0.14 nm) in it CoSep solution than in water with an average distance of
1.11 ± 0.10 nm. In isolated A chain, reverse effect of CoSep binding was observed with 0.22
± 0.03 nm average distance between water and HEME iron that was observed to be 0.34 ±
0.14 nm in water simulation. Hence, CoSep binding in isolated HEME domain make its
structure slightly compact and decreased the size of substrate access channel.
6.4.3. Effect of CoSep binding on ET tunneling
In CoSep solution, the distance between FMN and HEME cofactor was as average
1.35 ± 0.01 nm (with the minimum distance of 0.95 nm), lower than the one in water (1.41
± 0.09 nm) (reported in Figure S6.8 in SI). The ET tunneling in AF chain in crystal structure
and in the simulation has been discussed in detail in our previous paper.[20] In CoSep
159

PART II: P450BM-3 HEME/FMN & CoSep
solution, the ET tunneling was identified form CoSep to HEME iron in representative
structures of isolated and complex simulation obtained via cluster analysis and reported in
Table 6.2.
Table 6.2: Electron transfer tunneling in AF chain and isolated A chain in CoSep solution
calculated by Pathways[29] VMD plugin.
Coordinates Redox Max. Distance Amino acids involved in
partners coupling along ET the ET pathway
(a.u.) pathway (nm)
A chain CoSep/HEME 6.39 x10 -9 3.08 CoSep → D84 →
G85 → L86 → F87 →
HEME (FE)
AF chain CoSep/HEME 2.25 x10 -9 2.93 CoSep → I102 →
L103 → I401 → C400 →
HEME (FE)
AF chain CoSep/FMN 4.00 x10 -6 1.72 CoSep → W574 →
→ FMN (C7)
AF chain FMN/HEME 1.38 x10 -9 2.08 FMN (C7) → SOL →
→ M490 → F393 →
HEME (FE)
Figure 6.3a and 6.3b shows the possible ET tunneling from CoSep to HEME iron of A
chain in isolated and complex simulation. In isolated A chain, ET was mediated by the
residues of B’/C loop, Asp84, Gly85, Leu86 and Phe87. In A chain of AF complex, ET
tunneling can be mediated by two pathways. The first ET pathway is from CoSep to HEME
iron mediated by Iso102, Leu103, Ile401 and Cys400. Figure 6.3c and 6.3d shows the
160

PART II: P450BM-3 HEME/FMN & CoSep
second possible pathway, first from CoSep to isoalloxazine ring of FMN (C7 atom) and then
from C7 atom to HEME iron mediated by water molecule involving Met490 of Lβ1 FMN
binding loop and Phe393 of K/L loop. The same FMN/HEME ET pathway was observed in
water simulation of AF complex without the involvement of water molecule.[20]
161

PART II: P450BM-3 HEME/FMN & CoSep
Figure 6.3: ET tunneling from CoSep (in purple color) to HEME iron in AF complex (a) and
isolated A chain (b) in CoSep solution. (c) ET from CoSep to the isoalloxazine ring (C7 atom) of FMN
cofactor and d) from C7 atom of FMN cofactor to HEME iron in AF complex. ET is represented by
red color tubes. HEME and FMN cofactors are in black and pink color, respectively. The
conformation of first cluster with minimum distance between HEME to FMN cofactor is used. The
amino acids with in the distance of 0.50 nm from both the cofactors are labeled and shown in
licorice representation colored by element type (oxygen in red, carbon in cyan and nitrogen in blue
color) and their associated secondary structure in cartoon representation in sky blue for HEME
domain and in orange color for FMN domain. The residues involved in electron tunneling are
represented and labeled in green color.
6.4.4. Effect of CoSep binding on P450BM-3 dynamics
The subspace overlap and inner product of first ten eigenvectors A chain (together
account for ~60% of total residue position fluctuation) in isolated and complex simulation
was less than 0.20 and 0.34, respectively. The latter indicate the existence of different set of
collective motions in the eigenvectors of same time windows of both the trajectories. The
first three eigenvectors of A chain together represent ~41 % of the total relative positional
fluctuation (RPF). Figure 6.4a, 6.4b and 6.4c represents RPF associated with first three
eigenvectors of A chain in isolated (in green color) and complex (in orange color)
simulation (comparison to P450BM-3 domain in water is reported in Figure S6.9). Figure
6.5 shows RMSF associated with first three eigenvector of A chain in isolated (a1, a2 and
a3) and complex (b1, b2 and b3) simulation, respectively in CoSep solution.
In isolated A chain, the first collective motion (Figure 6.4a and 6.5a1) involves
mainly N-terminus region (residue 20 – 26), turn (residue 35 – 38) between A helix and
beta sheet 1, C-terminus (residue 450 – 458) and in K/L loop region (residue 325 – 380,
involved in HEME cofactor binding) and slight motion in B helix and D/E, F/G and G/H
loops. In the second eigenvector involve the collective motion (Figure 6.4b and 6.5a2) along
the turn (residue 35 – 38) between A helix and beta sheet, F/G loop, K/L loop (residue 366
162

PART II: P450BM-3 HEME/FMN & CoSep
– 385) and C-terminus (residue 425 – 430 and 450 – 458). In the third eigenvector (Figure
6.4c and 6.5a3), mainly at C-terminus (residue 425 – 458) and slightly D/E, F/G and K/L
loop (residue 390 – 402). The first three eigenvectors show that the substrate channel
remains open (also found in P45Cα - A191Cα
Figure 6.4: RPF for (a) first, (b) second and (c) third eigenvector of A chain in isolated (green
color) and complex (orange color) simulation. Horizontal bars, in blue and orange color represent
helixes (labeled) and beta sheets, respectively. The regions involved in cofactor binding are
represented by horizontal bars in purple color.
distance) and the collective motion is related to the interaction of residues to HEME
cofactor and to facilitate ET tunneling from CoSep to HEME iron through B’/C loop in
isolated A chain.
163

PART II: P450BM-3 HEME/FMN & CoSep
In complex simulation, the first eigenvector of A chain (Figure 6.4a and 6.5b1)
involve RPF in B’/C loop (residue 83 – 94), C and D helix and C/D loop (residue 100 – 130),
E/F loop (residue 159 – 171), G helix and G/H loop (residue 198 – 230), I helix (residue
250 – 268) and K/L loop (residue 335 – 340 and 392 – 400). The first collective motion
involved residues in contact with HEME cofactor and is related to interaction of A chain
with F chain with the largest RPF the regions on interface of A chain mainly constituted by
C – D helix (found to be involved in ET tunneling from CoSep to HEME iron) and K/L loop
(F393 is involved in water mediated ET tunneling from FMN to HEME iron). In the second
eigenvector (Figure 6.4b and 6.5b2), the collective motion was involve mainly G helix and
slightly in B’/C loop, G/H loop and K/L loop (residue 495 - 400). Larger RPF in G helix is
resulted by CoSep binding the slight kink formation in G helix. The collective motion in
third eigenvector (Figure 6.4c and 6.5b3) involve mainly G/H loop only and slightly in G
helix and K/L loop regions.
AF chain in CoSep solution shows the collective motion of different amplitude than
the one in water. RPF of first three eigenvectors of AF chain in both water and CoSep
solution is reported in Figure S6.10 of SI. The collective motion associated with the first
two eigenvectors of AF chain in the presence of CoSep does not belongs solely to the
movement of F chain towards A chain as observed in water simulation. The collective
motion associated with the first three eigenvectors of AF chain in the presence of CoSep is
reported in Figure S6.11 in SI. In the first eigenvector in CoSep, A chain of AF complex show
higher fluctuation in C helix, D helix, beginning of F helix (residue 170 – 175) and G/H loop
and lower fluctuation for F chain than the one in water. In the second eigenvector, the
collective motion in A chain in the presence of CoSep mainly involve G helix and slightly
higher RMSF for B’/C loop and K/L loop (residue 390 – 400), both loops are involved in
HEME cofactor binding. Third collective motion in A chain of AF complex involves N-
terminus residues (A – C helix), G helix and K/L loop. Third eigenvector of AF chain show
slight collective motion of F chain towards A chain.
164

PART II: P450BM-3 HEME/FMN & CoSep
Figure 6.5: RMSF of protein backbone atoms along first (a), second (b) and third (c) eigenvector
after projection of the trajectory on the corresponding eigenvector of isolated A chain in water (a1,
a2 and a3) and in CoSep solution (b1, b2 and b3). The 10 sequential frames represent the extension
of the fluctuations in trajectories along the eigenvectors. The first extreme conformation is shown
in green color and last extreme in violet color. Other conformations of A chain are in sky blue.
Helices and loops are labeled. N-terminus of the protein is labeled in red color.
165

PART II: P450BM-3 HEME/FMN & CoSep
6.5. Conclusions
We performed the simulation of isolated HEME domain and HEME/FMN complex in
CoSep solution. Structure remains conserved in both the systems throughout the
simulation. CoSep was found to bound mainly on surface exposed loop regions (richer in
charged amino acid mainly negative charged, E and D) in both the systems. CoSep binding
affects the substrate access channel, found to be relatively more open in compassion to the
one in water simulation.[20] Isolated HEME domain adopts ET tunneling from CoSep to
HEME iron mediated by residues of B’/C loop. HEME/FMN complex has two possible ET
tunneling pathways. First one was from CoSep to FMN and then from FMN to HEME by the
involvement of same residues (as observed in water simulation, Met490 and Phe393) but
mediated by water molecule. However, the average distance between FMN/HEME was
lesser (1.35 ± 0.11 nm) than the one observed in water simulation (1.41 ± 0.09 nm and 1.81
nm in crystal structure). Second ET tunneling was directly from CoSep to HEME iron. Both
the system shows atomic fluctuations of different amplitude in CoSep solution and in water
simulation. Hence, the presence of CoSep does not affect dramatically the conformation of
P450BM-3 domains but mainly it results into the stabilization on the loops on surface.
Except in HEME/FMN complex CoSep binding induced the conformational change in G helix
and resulted in higher fluctuation in F/G and G/H loop regions during the simulation. In
HEME/FMN complex, the preferable ET pathway is from CoSep to FMN and then to HEME
iron and in this process surface water molecule plays an important role. However, in
isolated HEME domain direct ET from CoSep to HEME iron might fasten the ET tunneling
and hence the performance of enzyme as observed in protein engineering experiment of
P450BM-3 that isolated HEME domain perform better in the presence of Zn/Co(III)sep. The
results of this study provide indication of the mechanism of ET by the CoSep. These results
are in agreement with the findings of directed evolution and side directed mutagenesis
experiments on the whole P450BM-3 and the HEME domain.
166

PART II: P450BM-3 HEME/FMN & CoSep
6.6. References
1. Chefson A, Auclair K (2006) Progress towards the easier use of P450 enzymes. Mol
Biosyst 2: 462-469.
2. Wong LL (1998) Cytochrome P450 monooxygenases. Curr Opin Chem Biol 2: 263-
268.
3. Guengerich FP (2001) Common and uncommon cytochrome P450 reactions related
to metabolism and chemical toxicity. Chem Res Toxicol 14: 611-650.
4. Kumar S Engineering cytochrome P450 biocatalysts for biotechnology, medicine and
bioremediation. Expert Opin Drug Metab Toxicol 6: 115-131.
5. Narhi LO, Fulco AJ (1986) Characterization of a catalytically self-sufficient 119,000-
dalton cytochrome P-450 monooxygenase induced by barbiturates in Bacillus
megaterium. J Biol Chem 261: 7160-7169.
6. Narhi LO, Fulco AJ (1987) Identification and Characterization of 2 Functional
Domains in Cytochrome-P-450bm-3, a Catalytically Self-Sufficient Monooxygenase
Induced by Barbiturates in Bacillus-Megaterium. J Biol Chem 262: 6683-6690.
7. Whitehouse CJC, Bell SG, Wong L-L (2012) P450BM3 (CYP102A1): connecting the
dots. Chem Soc Rev.
8. Warman AJ, Roitel O, Neeli R, Girvan HM, Seward HE, et al. (2005) Flavocytochrome
P450 BM3: an update on structure and mechanism of a biotechnologically important
enzyme. Biochem Soc Trans 33: 747-753.
9. Girvan HM, Waltham TN, Neeli R, Collins HF, McLean KJ, et al. (2006)
Flavocytochrome P450 BM3 and the origin of CYP102 fusion species. Biochem Soc
Trans 34: 1173-1177.
10. Schuhmann W (2002) Amperometric enzyme biosensors based on optimised
electron-transfer pathways and non-manual immobilisation procedures. J
Biotechnol 82: 425-441.
11. Nazor J, Dannenmann S, Adjei RO, Fordjour YB, Ghampson IT, et al. (2008)
Laboratory evolution of P450 BM3 for mediated electron transfer yielding an
167

PART II: P450BM-3 HEME/FMN & CoSep
activity-improved and reductase-independent variant. Protein Eng Des Sel 21: 29-
35.
12. Schwaneberg U, Appel D, Schmitt J, Schmid RD (2000) P450 in biotechnology: zinc
driven omega-hydroxylation of p-nitrophenoxydodecanoic acid using P450 BM-3
F87A as a catalyst. J Biotechnol 84: 249-257.
13. Wong TS, Schwaneberg U (2003) Protein engineering in bioelectrocatalysis. Curr
Opin Biotechnol 14: 590-596.
14. Sevrioukova IF, Li HY, Zhang H, Peterson JA, Poulos TL (1999) Structure of a
cytochrome P450-redox partner electron-transfer complex. P Natl Acad Sci USA 96:
1863-1868.
15. Humphrey W, Dalke A, Schulten K (1996) VMD: Visual molecular dynamics. J Mol
Graphics 14: 33-&.
16. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH, Kruger P, et al. (1996)
Biomolecular Simulation: The GROMOS96 Manual and User Guide. VdF:
Hochschulverlag AG an der ETH Zurich and BIOMOS bv, Zurich, Groningen.
17. Helms V, Deprez E, Gill E, Barret C, Hui Bon Hoa G, et al. (1996) Improved binding of
cytochrome P450cam substrate analogues designed to fill extra space in the
substrate binding pocket. Biochemistry 35: 1485-1499.
18. Roccatano D, Wong TS, Schwaneberg U, Zacharias M (2005) Structural and dynamic
properties of cytochrome P450 BM-3 in pure water and in a
dimethylsulfoxide/water mixture. Biopolymers 78: 259-267.
19. Roccatano D, Wong TS, Schwaneberg U, Zacharias M (2006) Toward understanding
the inactivation mechanism of monooxygenase P450 BM-3 by organic cosolvents: a
molecular dynamics simulation study. Biopolymers 83: 467-476.
20. Verma R, Schwaneberg U, Roccatano D Insight into the redox partner interaction
mechanism in cytochrome P450BM-3 using molecular dynamics simulation.
Unpublished.
21. Walsh JD, Miller AF (2003) Flavin reduction potential tuning by substitution and
bending. J Mol Struc-Theochem 623: 185-195.
22. Zheng Y-J, Ornstein RL (1996) A Theoretical Study of the Structures of Flavin in
Different Oxidation and Protonation States. J Am Chem Soc 118: 9402-9408.
168

PART II: P450BM-3 HEME/FMN & CoSep
23. Verma R, Schwaneberg U, Roccatano D Conformational Dynamics of the FMNbinding
Reductase Domain of Monooxygenase P450BM-3. Unpublished.
24. Dehayes LJ, Busch DH (1973) Conformational Studies of Metal-Chelates .1. Intra-
Ring Strain in 5-Membered and 6-Membered Chelate Rings. Inorganic Chemistry 12:
1505-1513.
25. Becke AD (1993) Density-functional thermochemistry. III. The role of exact
exchange. The Journal of Chemical Physics 98: 5648-5652.
26. Hay PJ, Willard RW (1985) Ab initio effective core potentials for molecular
calculations. Potentials for the transition metal atoms Sc to Hg. The Journal of
Chemical Physics 82: 270-283.
27. Breneman CM, Wiberg KB (1990) Determining Atom-Centered Monopoles from
Molecular Electrostatic Potentials - the Need for High Sampling Density in
Formamide Conformational-Analysis. Journal of Computational Chemistry 11: 361-
373.
28. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, et al. (2009) Gaussian 09,
Revision B.01. Gaussian 09, Revision B01, Gaussian, Inc, Wallingford CT. Wallingford
CT.
29. Balabin IA, Hu X, Beratan DN (2012) Exploring biological electron transfer pathway
dynamics with the Pathways Plugin for VMD. J Comput Chem 33: 906-910.
169

PART II: P450BM-3 HEME/FMN & CoSep SI
Supporting Information
Insight into the redox partner interaction mechanism in
cytochrome P450BM-3 using molecular dynamics
simulation
Table S6.1: Force field parameters for cobalt(II)sepulchrate adopted from the force
constants calculated on the energy minimized geometries by Dehayes et al.[1]
Bond stretching parameters
Bond type Force constant (kJ mol -1 nm -1 )
Co – N 885.234
N – C 2342.558
N – H 2962.824
C – C 2709.900
C – H 2740.010
Angle bending parameters
Angle type Force constant (kJ mol -1 rad -2 )
N – Co – N 240.88
Co – N – C 167.41
Co – N – H 167.41
C – C – C 417.92
C – C – H 292.06
170

PART II: P450BM-3 HEME/FMN & CoSep SI
C – N – H 167.41
C – N – C 417.92
N – C – C 417.92
N – C – H 292.06
N – C – N 334.22
H – N – H 251.11
Dihedral parameters
Dihedral type Force constant (kJ mol -1 )
H – C – C – H 2.729
H – C – N – Co 2.729
H – C – N – C 2.729
H – N – C – C 2.729
H – C – N – H 1.807
H – C – C – C 4.103
H – C – C – N 4.103
Co – N – C – C 1.373
N – Co – N – C 0.000
N – C – C – N 2.060
N – C – C – C 2.060
C – C – C – C 2.060
171

PART II: P450BM-3 HEME/FMN & CoSep SI
Table S6.2: Partial charges on cobalt(II)sepulchrate calculated by DFT calculations and
adopted for GROMOS96 43a1 force field.[2]
Atom number Atom type Atom name Charge group Partial charge
1 CH2 C 1 0.514
2 N N 2 -0.544
3 CH2 C 3 0.149
4 CH2 C 4 0.149
5 N N 5 -0.542
6 CH2 C 6 0.511
7 N N 7 -0.945
8 CH2 C 8 0.639
9 N N 9 -0.785
10 CH2 C 10 0.143
11 CH2 C 11 0.253
12 N N 12 -0.944
13 CH2 C 13 0.697
14 N N 14 -0.947
15 CH2 C 15 0.639
16 N N 16 -0.785
17 CH2 C 17 0.143
18 CH2 C 18 0.253
19 N N 19 -0.944
20 CH2 C 20 0.697
21 CO CO 21 1.682
22 H H 22 0.252
23 H H 23 0.381
24 H H 24 0.351
25 H H 25 0.381
26 H H 26 0.351
172

PART II: P450BM-3 HEME/FMN & CoSep SI
27 H
H 27
0.251
Figure S6.1: Backbone root means square deviation (RMSD) with respect to reference structure
as a function of time for AF chain (black), A of AF chain (red), F of AF chain (green), A (blue) and F
(orange) chain in water (as dotted line), and CoSep solution (straight line). P450BM-3 domains
deviated less in CoSep solution than the one in water only. Major difference was observed for
isolated A chain (green color solid line) in the presence of CoSep with lower deviation than the one
in water and it reached to a plateau after 25 ns with less variation.
173

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.2: Radius of gyration with respect to reference structure as a function of time for AF
chain (black), A of AF chain (red), F of AF chain (green), A chain (blue) and F chain (orange) in
water (as dotted line), and CoSep solution (straight line).
174

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.3: Backbone RMSD (a) and RMSF (b) per residue with respect to crystal structure for
isolated A and F chain (in red color), AF complex (in black color) in water, and A chain (in green
color) and AF complex (in orange color) in CoSep solution. The maroon vertical line separates
HEME and FMN domains. Horizontal bars, in blue and orange color represent helices (labeled) and
beta sheets, respectively. The regions involved in cofactor binding are represented by horizontal
bars in purple color.
175

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.4: The binding of CoSep (in ball and stick representation) on a) A chain and b) AF chain
of P450BM-3 in cartoon representation with surface colored by element type (carbon in gray,
oxygen in red, nitrogen in blue and hydrogen in white). FMN and HEME cofactors are in licorice
representation in red and green color, respectively. Helices, loops and N- and C- terminus (in red)
are labeled.
176

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.5: Average over the minimum distance (less than 2.3 nm) between CoSep and
isoalloxazine ring of FMN cofactor (in green color), and HEME iron of A chain in isolated domain (in
black color) and complex simulation (in red color) as a function of time.
Figure S6.6: Number of contacts between CoSep and amino acids of P450BM-3 domains within
the distance of 0.50 nm.
177

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.7: Minimum distance between P45C α and A191C α (1.61 nm in crystal structure) as a
function of time for isolated A and F chain (in red color), AF complex (in black color), and A chain
(in green color) and AF complex (in orange color) with CoSep.
178

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.8: Minimum distance between heavy atoms of isoalloxazine ring of FMN and HEME
cofactor as a function of time in AF complex in water (in black color) and AF complex in CoSep
solution (in red color). Green color horizontal line shows the distance observed in crystal
structure.[3]
179

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.9: RPF for first, second and third eigenvector of isolated A and F chain (in red color), AF
complex (in black color) in water, and A chain (in green color) and AF complex (in orange color) in
CoSep solution. . The maroon vertical line separates HEME and FMN domain. Horizontal bars, in blue
and orange color represent helices es (labeled) and beta sheets, respectively. The regions involved in
cofactor binding are represented by horizontal bars in purple color.
180

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.10: RPF for first, second and third eigenvector of AF complex in water (black color) and
in CoSep solution (orange color). The maroon vertical line separates HEME and FMN domain.
Horizontal bars, in blue and orange color represent helices (labeled) and beta sheets, respectively.
The regions involved in cofactor binding are represented by horizontal bars in purple color.
181

PART II: P450BM-3 HEME/FMN & CoSep SI
Figure S6.11: RMSF of protein backbone atoms along first (a), second (b) and third (c)
eigenvector after projection of the trajectory on the corresponding eigenvector of AF complex in
CoSep solution. The 10 sequential frames represent the extension of the fluctuations in trajectories
along the eigenvectors. The first extreme conformation is shown in green color and last extreme in
violet color. Other conformations of A and F chain are in sky blue and tan color, respectively. Helices
and loops in FMN domain are labeled. N and C indicate the N- and C-terminus of the protein
(labeled in red color).
182

PART II: P450BM-3 HEME/FMN & CoSep SI
References
1. Dehayes LJ, Busch DH (1973) Conformational Studies of Metal-Chelates .1. Intra-
Ring Strain in 5-Membered and 6-Membered Chelate Rings. Inorganic Chemistry 12:
1505-1513.
2. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH, Kruger P, et al. (1996)
Biomolecular Simulation: The GROMOS96 Manual and User Guide. VdF:
Hochschulverlag AG an der ETH Zurich and BIOMOS bv, Zurich, Groningen.
3. Sevrioukova IF, Li HY, Zhang H, Peterson JA, Poulos TL (1999) Structure of a
cytochrome P450-redox partner electron-transfer complex. P Natl Acad Sci USA 96:
1863-1868.
183

Summary and outlook
In the first part of the thesis, the importance of the combined computational and
directed evolution methods have been reviewed as a winning strategy for protein
engineering. The computational approaches can assist the design of protein engineering
experiments and holds particular promise to tailor proteins for specific functions.
MAP 2.0 3D server has been introduced to assist the development of directed evolution
experiments for generating sequence libraries with the highest chance to have variants
with desired enzymatic properties. This task is accomplished by correlating the generated
amino acid substitution patterns for a specific random mutagenesis method to the
structural information of the target protein. The combined information can help to select
an experimental strategy that improves the chances to obtain functional efficient and/or
stable enzyme variants. Hence, MAP 2.0 3D server facilitates the ‘in-silico’ pre-screening of
the target gene by predicting the amino acid diversity population in random mutagenesis
libraries. Currently, MAP 2.0 3D server provides sequence/structure based analysis using the
protein sequence/structure (crystallographic structure or homology model) provided by
the user. In future, the capability of the server can further be extended by (1) dynamically
identifying the functionally important regions e.g. active site residues and trans-membrane
regions in the target protein and focusing the analysis only on those regions, (2) by
providing MAP 2.0 3D results of structural analysis in the absence of crystallographic or
model structure using the predicted secondary structure elements from protein sequence,
and (3) predicting the flexible regions in protein structure using e.g. Gaussian network
model (GMN) and correlate them with MAP 2.0 3D analysis.
184

In the second part of the thesis, molecular dynamics simulations were used to
understand the interaction mechanism in the HEME and FMN domains of P450BM-3 in
solution and in the presence of electron mediator cobalt(II)sepulchrate (CoSep).
Cytochrome P450BM-3 is the pivot member of cytochrome P450 monooxygenase
superfamily particularly for being bacterial P450, fused with its eukaryotic like P450s
redox partners (FMN and FAD binding domains). This structural feature makes the enzyme
catalytically self-sufficient. In addition, being soluble in water, it has high catalytic
efficiency and monooxygenase rate. These characteristics make the enzyme particularly
interesting for possible biotechnological application. For this reason, the comprehension of
structure-function-dynamics relationships in P450BM-3 is relevant. In this thesis we have
analyzed different dynamic and structural properties of the HEME domain, FMN domain
and their complex in solution.
In the first study, the effect of protonation states (oxidized and reduced) of FMN
cofactor on conformation and dynamics of FMN-binding domain of P450BM-3 was
analyzed by performing MD simulations of holo- and apo- protein in solution. In holoprotein,
the protonation state of isoalloxazine ring influences the conformation and
dynamics of FMN cofactor and resulted in change in FMN binding site. In particular, the
dynamics of FMN domain showed significant differences in the atomic fluctuation
amplitude in oxidized and reduced states. In apo-protein, the overall structure remained
conserved but high fluctuations were observed in FMN binding region that can promote the
feasible rebinding of FMN cofactor as observed experimentally.
The MD simulation of HEME and FMN domains were performed to gain insight into
the interaction mechanism and inter domain electron transfer in HEME/FMN complex. The
simulations of isolated HEME and FMN domains were also performed to compare their
behavior in solution and in HEME/FMN complex. The HEME/FMN complex undergoes
conformational rearrangement during the simulation and decrease the distance between
FMN and HEME cofactor within the range for expected ET between both the redox centers.
185

In complex the main collective motion was dominated by the interaction mechanism
between HEME and FMN domain.
The MD simulations of HEME/FMN complex and isolated HEME domain were
performed to investigate the binding modes between CoSep and P450BM-3 domains and
their effect on ET pathway. CoSep prefers to bind on surface exposed loop regions mainly
having negative charged residues. CoSep binding on HEME domain was observed to affect
the substrate access channel and keep it more open in comparison to the one observed in
solution. Putative ET pathways were proposed between CoSep and HEME iron in
HEME/FMN complex and isolated HEME domain.
The results of P450BM-3 simulations can enhance our basic understanding with the
possible applications in enzyme catalysis toward (1) the effect of protonation state on
dynamics of P450BM-3 reductase domain, (2) the interaction mechanism of redox partners
and its effect on ET tunneling between redox centers, and (3) the effect of the presence of
ET mediator on redox partner interaction and ET tunneling. The study can be further
extended by performing the MD simulation of HEME/FMN complex with FMN cofactor in
reduced state. The modeling of linker regions connecting HEME and FMN domains followed
by its simulation will help to further enhance our understanding toward HEME/FMN
binding interaction mechanism and ET tunneling. Recently the release of FAD domain of
P450BM-3 with NADPH also offers a chance to perform the simulation of the whole
complex.
186

Curriculum vitae
Personal Details
Name:
Rajni Verma
Address:
School of Engineering and Science,
Jacobs University Bremen,
28759 Bremen, Germany
Tel.: +49 421 200 3208
Email:
ra.verma@jacobs-university.de
Date of Birth: 15 th April 1984
Nationality
Indian
Linguistic skills:
Hindi, English
__________________________________
Employment & Education
_______________________________
Since 06/09
PhD Fellow in Computational Chemistry and Bioinformatics,
Jacobs University Bremen, Bremen, Germany
04/08 – 03/09 Project Assistant, Bioinformatics,
Institute of Genomics and Integrative Biology, New Delhi, India
07/05 – 03/08 Master of Science in Bioinformatics,
CCS CS University, Meerut, India
06/04 – 06/05 Advanced Diploma in Computer Application,
CCSCS University, Meerut, India
07/01 – 06/05 Bachelor of Science in Life Sciences,
CCSCS University, Meerut, India
Publications
1. Verma R, , Schwaneberg U, Roccatano D. Conformational dynamics of the FMN-binding reductase
domain of monooxygenase P450BM-3. J Chem Theory and Comput 2012, DOI: 10.1021/ct300723x.
2. Verma R, , Schwaneberg U, Roccatano D. Computer-aided protein directed evolution: a review of
web servers, databases and other computational tools for protein engineering. Computational and
Structural Biotechnology Journal 2012, 2 (3), e201209008.
3. Ruff AJ, Marienhagen J, Verma R, , Roccatano D, Genieser HG, Niemann P, Shivange AV,
Schwaneberg U. dRTP and dPTP a complementary nucleotide couple for the Sequence Saturation
Mutagenesis (SeSaM) method. J Mol Catal B-Enzym 2012, 84, 40-47.
4. Verma R, Schwaneberg U, Roccatano D. MAP 2.0 3D: a sequence/structure based
server for protein
engineering. ACS Synth Bio. 2012, 1 (4), 139-150.

Curriculum vitae
5. Ramachandran S, Chaudhuri R, Verma R, Shah AR, Sen R, Paul C. Systems Immunology: Data
modeling and scripting in R Book Chapter: Encyclopedia of Systems Biology, Springer Science &
Business Media, LLC 2011. Edited by W. Dubitsky, O. Wolkenhauer, K. Cho, & H. Yokota.
6. Zhu L, Verma R, Roccatano D, Ni Y, Sun Z, Schwaneberg U. A potential antitumor drug (arginine
deiminase) reengineered for efficient operation under physiological conditions. ChemBioChem 2010,
11, 2294-2301. [inside cover page]
Conferences/Abstracts
1. Verma R, Schwaneberg U, Roccatano D. Molecular dynamics simulations of P450BM-3 reductase
domain. Computer simulation and theory of macromolecules 2012, Hunfeld, Germany.
2. Verma R, Schwaneberg U, Roccatano D. Protein and cofactor conformational dynamics of FMNbinding
reductase domain of monooxygenase P450BM-3. 5th Meeting of the North German
Biophysicist 2012, Borstel, Germany.
3. Verma R, Schwaneberg U, Roccatano D. MAP 2.0 3D: a structure based substitution spectra analyses
of mutagenesis methods. 10 th International Symposium on Biocatalysis- Biotrans 2011, Sicily, Italy.
4. Ruff AJ, Marienhagen J, Verma R, Roccatano D, Mundhada H, Shivange AV, Schwaneberg U.
Ribavarin: A complementary universal base to P for Sequence Saturation Mutagensis Method
(SeSaM). 10 th International Symposium on Biocatalysis- Biotrans 2011, Sicily, Italy.
5. Verma R, Schwaneberg U, Roccatano D. Conformational dynamics of oxidized and reduced FMN in
water and methanol. MoLife Center Jacobs University Bremen 2011, Seefeld, Germany.
6. Verma R, Schwaneberg U, Roccatano D. MAP2.0: Evolution of Mutagenesis Assistant Program. 5 th
International Congress on Biocatalysis- Biocat 2010, Hamburg, Germany.
7. WE-Heraeus Summer School June 2009, ‘Quantum and classical simulation of biological systems and
their interaction with technical materials’, Bremen, Germany. (Participation)
References
Prof. Dr. Danilo Roccatano
Assistant Professor
School of Engineering and Science,
Jacobs University Bremen,
Campus Ring 1, Research II,
28759 Bremen, Germany
Tel: +49-421 200-3144
Fax: +49-421-200-3249
Email: d.roccatano@jacobs-university.de
Web: http://ses.jacobs-university.de/ses/droccatano
Prof. Dr. Ulrich Schwaneberg
Head of the Institute
Department of Biotechnology,
RWTH Aachen University,
Worringer Weg 1,
52056 Aachen, Germany
Tel.: +49-241-80-24176
Fax: +49-241-80-22387
E-Mail: u.schwaneberg@biotec.rwth-aachen.de
Web: www.biotec.rwth-aachen.de

Statutory Declaration
I, RAJNI VERMA, hereby declare that I have written this PhD thesis independently,
unless where clearly stated otherwise. I have used only the sources, the data and the
support that I have clearly mentioned. This PhD thesis has not been submitted for
conferral of degree elsewhere.
Bremen, December 19, 2012
Signature ____________________________________________________________