Professional Documents
Culture Documents
COMPUTATIONAL INVESTIGATION OF
PEPTIDE BINDING AFFINITY AND COMPLEX
STABILITY OF MAJOR
HISTOCOMPATIBILITY COMPLEX (MHC)
ASUMAN BUNSUZ
M.Sc. THESIS
Department of Bioengineering
ADVISOR
Assoc. Prof. Dr. PEMRA ÖZBEK
ISTANBUL, 2018
MARMARA UNIVERSITY
INSTITUTE FOR GRADUATE STUDIES
IN PURE AND APPLIED SCIENCES
COMPUTATIONAL INVESTIGATION OF
PEPTIDE BINDING AFFINITY AND COMPLEX
STABILITY OF MAJOR
HISTOCOMPATIBILITY COMPLEX (MHC)
ASUMAN BUNSUZ
524214012
M.Sc. THESIS
Department of Bioengineering
ADVISOR
Assoc. Prof. Dr. PEMRA ÖZBEK
ISTANBUL, 2018
i
ACKNOWLEDGEMENTS
I would like to present my greatfullness to my thesis supervisor Assoc. Prof. Dr. Pemra
ÖZBEK who accepted me as a MSc student for her project. Her vision, encouragement,
patience and support are valuable sources in developing my professional skills. I would like
to acknowledge The Scientific and Technological Council of Turkey TUBITAK in the frame
work of the Project No. 113M293 for the financial support during my thesis.
I would also thank to my colleague, Research Assistant Onur SERÇİNOĞLU who taught me
computational methods patiently, respondend to my questions so promptly and that
throughout these years had always encouraging words.
I wish to thank to my team member Elif Naz BİNGÖL for all the meals and laughs we had.
Thank you for love, friendship and encouragement.
Special thanks are going to Hüsniye GENÇ who encouraged me to apply for a master
program at Marmara University and for her love and friendship, Nilay YÖNET for her
positive energy and emotional support, Ekin UZUNHASANOĞLU for offering me advice,
and supporting me through the thesis process.
I owe my greatest thanks to my parents Selma and İhsan BUNSUZ, my brother Abdulkerim
BUNSUZ and my sister Reyhan BUNSUZ for love, constant support, encouragement, never
ending patience and inspiring me to follow my childhood dream. Thank you.
i
TABLE OF CONTENT
PAGE
ACKNOWLEDGEMENTS ..................................................................... ii
TABLE OF CONTENT ........................................................................... ii
ABSTRACT ............................................................................................ iv
ÖZET .........................................................................................................v
LIST OF SYMBOLS .............................................................................. vi
LIST OF ABBREVATIONS ................................................................. vii
CHAPTER 1 AIM OF THE STUDY ..................................................13
CHAPTER 2 INTRODUCTION.........................................................15
2.1 Structure of Major Histocompatibility Complex (MHC) ............................... 15
2.2 HLA peptide binding affinity and stability ..................................................... 18
2.3 Molecular Dynamics (MD) Simulations......................................................... 21
2.3.1 Theoretical Background of Molecular Dynamics Simulations................... 22
2.3.2 Molecular Dynamics (MD) Simulations of peptide-HLA Class I
Complexes............................................................................................................... 24
2.4 Molecular Docking of HLA Class I Molecules .............................................. 26
2.4.1 DockTope Server ........................................................................................ 28
CHAPTER 3 MATERIALS AND METHODS ..................................30
3.1 HLA-A*02:01 Structures ................................................................................ 30
3.2 Molecular Docking of HLA-A*02:01 Structures ........................................... 33
3.3 Preparations of Structures for MD Simulations .............................................. 33
3.4 Analysis of the MD Simulation Trajectories .................................................. 34
3.4.1 Root Mean Square Deviation (RMSD) Calculations ................................. 34
3.4.2 Root Mean Square Fluctuation (RMSF) calculations ................................. 35
3.5 Calculation of pairwise amino acid non-bonded interaction energies via
gRINN tool.................................................................................................................. 35
CHAPTER 4 RESULTS AND DISCUSSIONS ...................................36
4.1 Analysis of RMSD profiles................................................................................... 36
4.2 Analysis of RMSF profiles ................................................................................... 39
4.3 Analysis of interaction energies using gRINN ..................................................... 42
ii
CHAPTER 5 CONCLUSION ............................................................67
CHAPTER 6 FUTURE WORK .........................................................69
REFERENCES ....................................................................................... 70
AUTOBIOGRAPHY .............................................................................. 83
iii
ABSTRACT
iv
ÖZET
v
LIST OF SYMBOLS
V Volume
N Number of Atoms
E Energy
μ Chemical Potential
T Temperature
P Pressure
H Enthalpy
F Force
m Mass
a Acceleration
r Atomic Position
v Velocity
p Potential Energy
kB Boltzmann Constant
vi
LIST OF ABBREVATIONS
TI Thermodynamic Integration
MD Molecular Dynamics
Cα Carbon Alpha
TOP Topology
vii
ALA Alanine
ARG Arginine
ASN Asparagine
CYS Cysteine
GLN Glutamine
GLY Glycine
HIS Histidin
ILE Isoleucine
LEU Leucine
LYS Lysine
MET Methionine
PHE Phenylalanine
PRO Proline
SER Serine
THR Threonine
TRP Tryptophan
TYR Tyrosine
VAL Valine
viii
LIST OF FIGURES
Figure 2.1 Schematic representation of MHC Class I (A) and MHC Class II (B)[40]. .............. 15
Figure 2.2 Structures of peptide binding groove for MHC Class I (A) and Class II (B)[40]. .... 16
Figure 3.1 Peptide binding groove pockets of HLA-A*02:01 (PDB ID : 3HLA [51]) are shown
in PyMOL representation. Colored dots represent the pocket residues interacting with the peptide
residues (A: red, B: green, C: teal, D: pink, E: yellow, F: turquouse) ......................................... 31
Figure 3.2 Peptide binding groove pockets of HLA-A*02:01 (PDB ID : 1HHH [119]) and the
interacting peptide residues are shown in PyMOL representation. Macromolecule is shown in
gray and peptide is shown in lightgreen sticks. Colored dots represent the pocket residues
interacting with the peptide residues shown in blue stick. (A) P1, (B) P2, (C) P6, (D) P3, (E) P7,
(F) P9 ............................................................................................................................................ 32
Figure 4.1 Root mean square deviation (RMSD) profile of the modeled complexes during the
100 ns simulations at 310 K. ......................................................................................................... 36
Figure 4.2 Root mean square deviation (RMSD) profile of the modeled complexes during the
100 ns simulations at 473 K. ......................................................................................................... 36
Figure 4.3 Distribution of alpha carbon root mean square deviation (RMSD) values of the
modeled complexes during the 100 ns simulations at 310 K........................................................ 37
Figure 4.4 Distribution of alpha carbon root mean square deviation (RMSD) values of the
modeled complexes during the 100 ns simulations at 473 K........................................................ 38
Figure 4.5 Root mean square fluctuations (RMSF) values of the modeled complexes during the
100 ns simulations at 310 K. ......................................................................................................... 39
Figure 4.6 Root mean square fluctuations (RMSF) values of the modeled complexes during the
100 ns simulations at 473 K. ......................................................................................................... 40
Figure 4.7 Peptide RMSF profiles of the modeled complexes during the 100 ns simulations at
310 K. ............................................................................................................................................ 41
Figure 4.8 Peptide RMSF profiles of the modeled complexes during the 100 ns simulations at
473 K. ............................................................................................................................................ 41
Figure 4.9 Correlation between the experimental stability measurements and our computational
ix
energy values obtained from gRINN. ........................................................................................... 43
Figure 4.10 Motion of peptides and peptide binding grooves throughput the all trajectory.
Macromolecule is shown in gray, peptide binding groove is shown in military green, peptides are
shown in colored tubes. (ILDDNLYKV: red, GLFDFVNFV: orange, SLSAYIIRV: yellow,
YLPEVISTI: purple, RLYDYFTRV: magenta, LMYDIINSV: brown, VLYDEFVTI: pink,
RVYEALYYV: green, YLYFCSSDV: cyan, YLYQPCDLL: cream, HVDGKILFV: blue,
VLPFDIKYI: black) ..................................................................................................................... 44
Figure 4.11 Interaction energies between residues of HLA and β2-microglobulin. .................... 46
Figure 4.12 Key residues for the interaction between HLA and β2-microglobulin (ILDDNLYKV
peptide) are shown in PyMOL representation. Colored dots represent the important residues
interacting with β2-microglobulin (ARG48: magenta GLY119: red, ARG202: green, peptide:
orange, macromolecule: gray ) ..................................................................................................... 47
Figure 4.13 Interaction energies of residue GLY119. .................................................................. 47
Figure 4.14 Interaction energies of residue ARG48..................................................................... 48
Figure 4.15 Interaction energies of residue ARG202................................................................... 49
Figure 4.16 Interaction energies between residues of HLA and peptide residues. ...................... 50
Figure 4.17 Interaction energies between region 1 (residues 60-80) and peptide residues.. ...... 51
Figure 4.18 Interaction energies between region 2 (residues 140-170) and peptide residues. .. 52
Figure 4.19 Important residues for interaction between HLA and peptide (ILDDNLYKV
peptide) are shown in PyMOL representation. Colored dots represent the important residues
interacting with β2-microglobulin (GLU163: red, LYS66: blue, TRP146: cyan, ASP77: yellow,
peptide: orange, macromolecule: gray )........................................................................................ 53
Figure 4.20 Interaction energies of residue GLU63..................................................................... 54
Figure 4.21 Interaction energies of residue GLU66..................................................................... 54
Figure 4.22 Interaction energies of residue TRP146. .................................................................. 55
Figure 4.23 Interaction energy profiles of the peptide residues. .................................................. 56
Figure 4.24 Interaction energies of residue P1............................................................................. 57
Figure 4.25 Interaction energies of residue P2............................................................................. 58
Figure 4.26 Interaction energies of residue P3............................................................................. 59
Figure 4.27 Interaction energies of residue P4............................................................................. 60
Figure 4.28 Interaction energies of residue P5............................................................................. 60
x
Figure 4.29 Interaction energies of residue P6............................................................................. 61
Figure 4.30 Interaction energies of residue P7............................................................................. 61
Figure 4.31 Interaction energies of residue P8............................................................................. 62
Figure 4.32 Interaction energies of residue P9............................................................................. 63
Figure 4.33 Interacting residues of ILDDNLYKV, GLFDFVNFV, SLSAYIIRV, YLPEVISTI,
RLYDYFTRV and LMYDIINSV peptides. ................................................................................. 65
Figure 4.34 Interacting residues of VLYDEFVTI, RVYEALYYV, YLYFCSSDV,
YLYQPCDLL, HVDGKILFV and VLPFDIKYI peptides. ......................................................... 66
xi
LIST OF TABLES
Table 3.1 Dataset of HLA-A*02:01 alleles selected for the study. ............................................. 30
Table 3.2 Peptide binding groove pockets and the interacting peptide residue positions of HLA-
A*02:01 (PDB ID : 3HLA [51] and PDB ID : 1HHH [119]) ....................................................... 31
Table 4.1 Interaction energies of 12 peptide HLA-A*02:01 complexes obtained from gRINN
....................................................................................................................................................... 42
xii
CHAPTER 1 AIM OF THE STUDY
Major histocompatibility complex (MHC) molecules are heterodimeric cell surface glycoproteins
playing a novel role in the regulation of the adaptive immune response by binding antigenic
peptides and presenting them at the cell surface for T cell recognition [1]–[3].
In recent years, studies have focused on designing peptide based vaccines, hence the factors
affecting the peptide immunogenicity has become a popular subject for detailed studies [4]–[6].
Therefore, many experimental approaches have been developed to measure the peptide binding
Early experimental studies reported that peptide immunogenicity is highly related to peptide
binding affinity [17], [18]. However, several studies proposed that stability is a more accurate
demonstrated that although having similar binding affinities, immunogenic peptides form more
experimental studies have revealed that highly stable peptide-MHC complexes have longer half-
lives on the cell surface to activate CTLs, while complexes having low stabilities are released
from the cell surface in a shorter time duration [12], [23]. Therefore, stability of peptide-MHC
Although there are several experimental methods in literature, molecular dynamics (MD)
simulation methods have been widely used to understand structural, kinetic and thermodynamics
properties of peptide-MHC complexes at the atomic level. MD simulations provide the most
detailed information related to intermolecular and intramolecular forces that are important for
13
protein stability [24], [25].
Furthermore, high temperature MD simulations have been extensively used in order to estimate
folding processes and the protein stability comparisons can be conducted within shorter
simulation times. Therefore, in order to explore the stability of peptide-MHC complexes, high
In humans, MHC molecules are called as Human Leukocyte Antigens (HLA). HLA-A2 is the
most common HLA-A allele in different ethnic populations. Among the HLA-A2 allelic variants,
HLA-A*02:01 is the most frequent one and it is commonly used to study HLA-A2-restricted
In this study, we have investigated peptide binding affinity and complex stability on 12 peptide-
HLA-A*02:01 modeled complexes. Firstly, molecular docking studies were performed for these
peptides using Docktope web server [33]. Then, MD simulations were carried out on the docked
structures using GROMACS software [34] with the OPLS-AA/L all-atom force field [35] for 100
ns at 310 K and 473 K. RMSD, RMSF and PCA analysis were performed on the equilibrated
MD simulation trajectories. In the final step, pairwise amino acid non-bonded interaction
We compared our computational results with previous experimental measurements [10], [13]
and investigated the correlations among them. We have also determined the critical residues
which play a role in the observed differences in the stability of these 12 peptide- HLA-A*02:01
complexes.
14
CHAPTER 2 INTRODUCTION
Major histocompatibility complex (MHC) molecules are heterodimeric cell surface glycoproteins
playing a novel role in regulation of the adaptive immune response by binding antigenic
peptides and presenting them at the cell surface for T cell recognition [2], [3], [37].
MHC molecules are classified in to two categories according to their structure and function;
MHC Class I and MHC Class II [38]. MHC Class I molecules are composed of a heavy chain (α
chain) consisting of three domains (α1, α2 and α3) and a light chain (β2-microglobulin), while
MHC Class II molecules are made up of an α (α1 and α2) and a β (β1 and β2) chain (Figure 1.1)
[39].
A B
Figure 2.1 Schematic representation of MHC Class I (A) and MHC Class II (B)[40].
In MHC Class I, α1 and α2 helices constitute the sides of the peptide binding groove while the
base of the binding groove is formed by the eight stranded β pleated sheet (Figure 1.2) [41] .
Binding groove is closed at terminal regions and an endogenous peptide between 8 to 11 amino
15
acids in length is accommodated. After peptide binding, peptide-MHC Class I complex is
A B
Figure 2.2 Structures of peptide binding groove for MHC Class I (A) and Class II (B)[40].
In MHC Class II, the sides of the peptide binding groove are constituted by α1 and β1 helices
and the base of the binding groove is constructed by a β sheet (Figure 1.2) [41]. Peptide binding
groove of MHC Class II is open at both ends, allowing peptides to extend out of the groove and
generally, an exogenous peptide of 13-25 amino acids in length is held in the binding groove via
In humans, MHC molecules are called as Human Leukocyte Antigens (HLA). Genes encoding
HLAs located on the short arm of chromosome 6 (6p21.3) are the most polymorphic genes in the
human genome. HLA Class I molecules are encoded by HLA-A, HLA-B and HLA-C genes,
while HLA Class II molecules are encoded by HLA-DP, HLA-DQ and HLA-DR genes [42].
16
Earlier experimental studies focused on identifying sequence of peptides presented by HLA
molecules. These studies revealed allele specific peptide binding motifs for various alleles of
HLA Class I consisting of HLA-A1 [43], A11 [44], A2.1 [45], A3 [46], A68 [47], B27 [48] and
B8 [49]. In 1987, three dimensional structure of HLA-A2 allele belonging to HLA Class I was
detailed by Bjorkman using X-ray crystallography [50]. Since the determination of the first
crystal structure of HLA Class I molecule, a variety of three dimensional structures of HLA
molecules have been elucidated. Experimental and structural analyses have revealed that
polymorphic residues in the peptide binding groove play essential roles in determining allele
To date, several computational algorithms depending on peptide sequence and structure of the
MHC molecules have been proposed for the prediction of possible T cell epitopes for specific
MHC alleles [53]. Sequence based prediction algorithms such as; sequence motif, motif
matrices, machine learning motifs (Artificial Neural Network and Support Vector Machine),
Hidden Markov models and Quantitative matrices make use of the knowledge related to peptides
known to bind HLA molecules [54]–[63]. On the other hand, structure based algorithms such as
molecular docking, threading and molecular dynamics simulations require the three dimensional
Moreover, sequence based databases and web servers have been developed to facilitate
However, sequence based predictions are dependent on the knowledge related to peptides known
to bind HLA molecules [53]. On the other hand, structure based predictions can provide more
detailed information at the molecular level. Therefore, structure based predictions have become
17
2.2 HLA peptide binding affinity and stability
In protein-ligand interactions, the term ‘binding affinity’ is used to describe the force of
attraction between the ligand and the receptor protein [77]. On the other hand, definition of
stability can be classified into two classes; thermodynamic stability and kinetic stability. The
thermodynamic protein stability is defined as the difference in free energies between the folded
and the unfolded states, whereas the kinetic stability of a protein is based on the energy barrier
between the folded and the unfolded states of the protein [78].
Recently, studies have focused on designing peptide based vaccines, hence the factors affecting
the peptide immunogenicity, defined as the ability of a peptide to stimulate CTL mediated
immune response, has become a popular subject for detailed studies [4]–[6]. Therefore, many
experimental approaches have been developed to measure the peptide binding affinity and
Labeling peptides by radioactive, biotinyl or fluorescent markers is one of the most common
However, peptide binding features may change in response to alterations in the labeled peptides.
On the other hand, indirect competitive binding assay approaches have been performed with the
use of synthetic peptides and monoclonal antibodies [80]. Moreover, Khilko et al. [7] used
surface plasmon resonance (SPR) technology for the investigation of peptide-MHC binding. In
1992, Parker et al. [11] have introduced a new indirect in vitro assay approach based on the
dissociation of radiolabelled β2m from MHC-peptide complexes having unlabeled heavy chain
and peptide. With this approach, interactions between heavy chain and peptide have been
18
measured without labeling. The first assay approach based on dissociation of labeled β2m led to
Harndahl et al. [9] have developed high throughput homogenous binding assay to determine the
peptide binding affinity for MHC molecules via application of Alpha Screen technology.
Furthermore, the same group have introduced high throughput homogenous β2m dissociation
assay for calculating the stability of peptide-MHC complexes [79]. These two binding assays are
Besides these experimental approaches, computational methods for the calculation of binding
free energy are also extremely popular. These computational approaches can be divided into
two; endpoint and pathway approaches [81]. Endpoint approaches including linear interaction
energy (LIE) [82] and molecular mechanics Poisson-Boltzmann surface area (MM/PBSA)
methods [83] calculate the difference in binding free energy between the two extremes; unbound
perturbation (FEP) and thermodynamic integration (TI) methods [84] calculate binding free
energy along the transition pathway between the end point states [81], [85]. Although, pathway
approaches provide highly accurate results, the end-point approaches are more efficient and cost
effective [85].
Early experimental studies reported that peptide immunogenicity is highly related to peptide
binding affinity [17], [18]. However, several studies proposed that the stability is a more accurate
parameter for peptide immunogenicity [19]–[22]. Moreover, in 2012, Harndahl et al. [10]
demonstrated that immunogenic peptides form more stable peptide-MHC complexes than non-
immunogenic peptides having similar binding affinities. In addition, several experimental studies
have revealed that highly stable peptide-MHC complexes have longer half-lives on the cell
19
surface to activate CTLs, while complexes having low stabilities are released from the cell
surface in a shorter time duration [12], [23] . Moreover, high and intermediate affinity peptides
with high stability were classified as immunogenic peptides, while high affinity peptides with
low stability were classified as weak immunogenic or non-immunogenic peptides based on their
interaction with CTLs [23]. Therefore, stability of peptide-MHC complexes seems to be the
major determinant factor for immunogenicity when compared with the peptide binding affinity
Although there are several experimental and computational methods in literature, molecular
dynamics (MD) simulation methods have been widely used to understand structural, kinetic and
provide the most detailed information related to intermolecular and intramolecular forces that are
important for protein stability [24], [25] and can guide the future experiments in this field.
have been extensively used [26]–[31]. Through these simulations, differences in folding
processes and the protein stability comparisons can be conducted within shorter simulation
times. Therefore, in order to explore the stability of peptide-MHC complexes, high temperature
In 2004, Zacharias et al. [86] studied the changes in conformational dynamics of α1 and α2
simulations at two different temperatures 310 K and 355 K for 26 ns. Nevertheless, the short
simulations at 355 K did not achieve complete unfolding process and they observed partial
unfolding of α1 and α2 domains at high temperature. MD simulations of the peptide free HLA-A
at 355 K revealed that the binding of F-pocket region to C-terminus of the peptide is more
20
flexible than the binding to N terminus of the peptide. Since the peptide C and N terminus
provides the stabilization of the α1-α2 peptide binding domain, this study motivated us to carry
out high temperature MD simulations for the investigation of peptide-MHC complex stability.
In 1957, the first MD simulation was accomplished by Alder and Wainwright [87] in order to
investigate the equilibrium features of hard spheres and later in 1964, the first realistic MD
simulation was performed using Lennard-Jones potential for liquid argon by Rahman [88]. The
first application of MD simulations to proteins was carried out by McCommon [89] in 1977, in
order to study the folding dynamics of Bovine Pancreatic Trypsin Inhibitor (BPTI).
clarified using statistical mechanics. Thus, microscopic properties obtained from MD simulations
states are called ensembles that are a collection of variety of microscopic states [90]. The five
most common ensembles used in MD simulations are canonical (NVT), micro canonical (NVE),
In canonical ensemble (NVT), number of atoms, volume and temperature are held fixed. The
microcanonical ensemble (NVE) describes a system having constant number of atoms (N),
volume (V) and energy (E). Grand canonical ensemble (μVT) corresponds to fixed chemical
potential (μ), volume (V) and temperature (T). For isobaric-isothermal ensemble (NPT), number
of atoms (N), pressure (P) and temperature (T) are held constant. Isobaric-isoenthalpic ensemble
(NPH) has fixed number of atoms (N), pressure (P) and enthalpy (H) [85].
21
2.3.1 Theoretical Background of Molecular Dynamics Simulations
In MD simulations, Newton’s equations of motion [Eq. 2.1] is solved numerically for each atom
of the molecular system and trajectories. Each atoms position, velocity and energy is produced
asa function of time to give insight into time dependent development of the system.
2.1
Positions and velocities of all atoms are calculated via the [Eq. 2.2]
2.2
where F is force on an atom, m is the mass of the atom, r is the atoms position and V is the
potential energy. The force F on each atom is also computed using negative gradient of the
The starting positions of the molecules are taken from structures determined by X-ray or NMR
techniques. Starting velocities are assigned by Gaussian- Boltzman distribution [Eq. 2.3] where
pi denotes the momentum of the ith atom, vx represents the velocity of the atom, m is the mass, kB
mi 1 mi vi2, x
=p ( vi , x ) exp −
2π k B
T 2 k B T
2.3
The temperature is calculated based on the mean kinetic energy equation via the [Eq. 2.4]
mi ( vi2 )
N
2 1
T=
3 Nk B
∑
i =1 2 2.4
22
where N is overall number of unconstrained degrees of freedom, v represents velocity of an atom
using finite difference methods. The Verlet algorithm depending on Taylor series expansion is the
most widely preferred algorithm. In Verlet algorithm [Eq. 2.5, Eq. 2.6, Eq. 2.7], r represents
1
r ( t + ∆=
t ) r ( t ) + v ( t ) ∆t + a ( t ) ∆t 2 … 2.5
2
1
v ( t + ∆=
t ) v ( t ) + a ( t ) ∆t + b ( t ) ∆t 2 … 2.6
2
a ( t + ∆=
t ) a ( t ) + b ( t ) ∆t … 2.7
Velocities are computed non-explicitly via [Eq. 2.8, Eq. 2.9, Eq. 2.10]
1
r ( t + ∆=
t ) r ( t ) + v ( t ) ∆t + a ( t ) ∆t 2 2.8
2
1
r ( t − ∆=
t ) r ( t ) − v ( t ) ∆t + a ( t ) ∆t 2 2.9
2
r ( t +=
∆t ) 2r ( t ) − r ( t − ∆t ) + a ( t ) ∆t 2 2.10
Force field includes the parameters and the functions to calculate the potential energy of the
system. In this study, OPLS-AA/L force field is used. The OPLS-AA/L force field was first
introduced by Jorgensan [91] to study dynamics of organic liquids. Later, this force field has
been used for proteins, carbohydrates, nucleic acids and drug molecules. Moreover, OPLS-AA/L
force field has been developed for proteins by refitting the Fourier torsional coefficients [8]. The
total potential energy of the system is calculated by summing the non-bonded, bond strenging,
23
2.3.2 Molecular Dynamics (MD) Simulations of peptide-HLA Class I Complexes
There have been various MD studies on HLA molecules over the years. In 1992, Rognan et
al.[92] generated a peptide-HLA-A2 complex model for influenza virus matrix protein IMP 58-
66 peptide (GILGFVFTL) by manually docking the peptide into peptide binding cleft of HLA-
A2 crystal structure and carried out 100 ps long MD simulations of this modeled complex using
AMBER software [93]. This study revealed important residue positions for peptide binding,
including Tyr7, Tyr59, Glu63, Tyr84, Thr143, Tyr159 and Tyr171. Their results appeared to be
inline with the findings obtained from the crystal structure of the HLA-B*27 allele.
residues and the three residues 95, 97 and 99 within the binding groove of HLA-B*02:01-
ILKEPVHGV complex. Then, they performed 1000 ps long MD simulations using Discover 95.0
program. Although these three residues differ between HLA-B*02:17 and HLA-B*02:01, these
two alleles had distinct peptide motifs. For HLA-B*02:17 allele, prolin at P3 position seemed to
In 2007, Joseph et al. [95] generated a variety of HLA*02:01-peptide complex models using
0.2 ns long MD simulations using Molecular Operating Environment (MOE) 2001.01 program
to study the effect of amino acid replacements on GP2 peptide. This study revealed that
GLU7PHE mutant affects the peptide-HLA complex stability the most and peptide residues P5-
Pohlmann et al. [96] conducted 20 ns long MD simulations on the HLA-B*27:05 and HLA-
B*27:09 alleles bound to the m9 peptide using GROMACS [34] software in 2004. They
24
peptide-HLA-B*27:05 complex and revealed that the micropolymophism at residue 116 causes
In 2008, Fabian et al. [97] performed 40 ns long MD simulations of the HLA-B*27:05 and HLA-
B*27:09 alleles bound to pVIPR peptide using GROMACS software. For pVIPR-HLA-B*27:05
complex, the canonical and the non- canonical peptide conformations were simulated separately.
The results demonstrated that the heavy chain of the canonical form of HLA-B*27:05 was more
flexible compared to the heavy chain of HLA-B*27:09. However, the canonical form of HLA-
B*27:05 was more thermostable than HLA-B*27:09. But, similar flexibilities were observed
Narzi et al. [98] investigated the impact of micropolymorphism and the effect of peptide binding
on the dynamics of the binding groove of HLA-B*27:05 and HLA-B*27:09 alleles bound to a
viral (pLMP2) and three self-peptides (pVIPR, pGR, and TIS) by performing 400 ns long MD
simulations of these peptide-HLA complexes using GROMACS software. All peptides exhibited
higher conformational flexibilities at alpha 1 and peptide binding groove regions when bound to
HLA-B*27:05. This study demonstrated that the dynamics of the peptide binding groove is
In 2015, Abualrous et al. [99] performed 50 ns long MD simulations of HLA-B*27:05 and HLA-
B*27:09 alleles bound to IRAAPPPLF peptide. As a result of their study, they found that the
polymorphic residue HIS116 neutralizes the negative charge by binding surrounding residues to
provide stability, while ASP116 causes charge repulsion and higher flexibility in the F pocket.
Thus, it is concluded that HLA-B*27:05 is dependent upon tapasin for peptide loading.
25
Ozbek [100] studied the dynamics of HLA-B*44:02, HLA-B*44:03 and HLA-B*44:05 alleles
that peptide binding stability is more dependent on the peptide sequence rather than the allele.
HLA-B*27:09 alleles bound to four different peptides (m9, pLMP2, pCatA, and TIS) in order to
study the effect of residue interactions and the micropolymorphism at residue 116 on the global
dynamics of peptide-HLA complexes. As a result, they observed the strong interaction between
the different domains of the peptide-HLA complexes and revealed that in the absence of peptide,
dynamic interaction within the residues of HLA-B*27:09 allele is stronger than the HLA-B*27:05
allele.
calculating peptide motion in order to predict peptide immunogenicity. They revealed that
rigidity and hydrophobicity are the two most significant factors which have an effect on the
immune cellular response. They have also concluded that the stability of the peptide can be
interaction mechanisms provide valuable information on the development of new drugs. Due to
the expensive and time consuming experimental methods, many protein structures have not yet
been defined experimentally, hence three dimensional structures of all these proteins are not
complexes with unknown three dimensional structures. One of the most important applications
Molecular docking methods generate a series of probable complex structures from its unbound
constituents by predicting the favored conformation of the ligand in the binding site of a protein.
Each probable complex structure is ranked depending on the global minimum free energy of
protein-ligand complex by scoring functions. Thus, the structure with the lowest score is ranked
During the last two decades, various docking programs have been developed for docking of
ligand molecules into proteins such as Autodock [108] , Glide [109], GOLD [110] and
HADDOCK [111]. Due to the peptides having large number of rotatable bonds that are more
flexible than the small ligand molecules, most of the popular docking programs are not
improvement have been achieved in the field of peptide-protein docking programs such as DINC
Liu et al. [113] carried out docking studies using FlexPepDock for HLA Class I alleles including
HLA-A2, HLA-B27, HLA-B35, HLA-B44, and HLA-E to generate a special protocol. In their
study, RMSD (root mean square deviation) values less than 2 Å and 1 Å were determined for the
near native and subangstrom models, respectively. As a result, 66% of docking studies were
scored as subangstrom models. Rigo et al. [33] developed DockTope and a generated refinement
protocol for HLA-A*02:01 and HLA-B*27:05 alleles by using structurally determined peptide-
HLA complexes. Their calculated RMSD values are less than 1 Å and 2 Å for Cα and all atoms,
27
respectively. In 2018, Antunes et al. [114] achieved redocking studies using DINC for HLA Class
I alleles and found RMSD values less than 2 Å for all atoms.
In this study, Docktope server [33] is used for the modeling of the HLA-peptid complexes.
DockTope is a web server [33], which aims to model the bound conformations of peptide-MHC
structural modeling of MHC Class I peptides, energy minimization of modeled peptides, first
molecular docking of peptides into MHC molecules, selection of the best pose, energy
minimization of the best pose generated in the first docking process, the final molecular docking
process, selection of the best pose created in final docking process and production of three
In the first step, target peptide sequence is provided in the form of a single aminoacid code by
the user and the target peptide is modeled using PyMol [115]. Then, energy minimization of the
modeled peptide is performed to eliminate unnecessary interactions. During the first molecular
docking process, Autodock tools and Autodock Vina [116] are employed. Previous to docking,
hydrogen and Gasteiger charges are added to both the peptide and MHC Class I molecule. For
the peptide, torsions are additionally specified. In a following step, search space is determined in
order to obtain the best pose in the peptide binding groove. Each run is compiled for 20 times.
Then, the best poses are selected from each run. After the first docking process, undesirable
interactions are removed by performing energy minimization using GROMACS [117]. After this
step, molecular docking is performed for a second time using the same procedure as above and
the best pose is selected based on the algorithm generated by Rigo et al [33]. In the final step,
28
three dimensional structure of peptide-MHC Class I complex is created and the structure file is
29
CHAPTER 3 MATERIALS AND METHODS
Among the HLA-A2 allelic variants, HLA-A*02:01 is the most frequent one and it is commonly
used to study HLA-A2-restricted CTL responses [32]. In this study, we have investigated the
complexes. Experimental studies were conducted on these peptide-HLA complexes [10], [118]
previously. Table 3.1 demonstrates the details of these structures along with the peptide
generated by Assarsson [118] and Harndahl [10] and a predicted affinity measurement by
Harndahl [10].
Immunogenic GLFDFVNFV 1 2 2 23
Immunogenic SLSAYIIRV 2 2 5 23
Immunogenic YLPEVISTI 2 1 5 19
11 4 3
Immunogenic LMYDIINSV 14
Immunogenic RLYDYFTRV 2 1 3 13
Immunogenic VLYDEFVTI 5 1 10 12
Immunogenic RVYEALYYV 2 1 2 11
Immunogenic HVDGKILFV 39 41 160 3
Nonimmunogenic YLYFCSSDV 2 1 7 10
Nonimmunogenic YLYQPCDLL 2 1 14 7
15 20 149
Nonimmunogenic VLPFDIKYI 1
30
For all HLA-A*02:01 alleles, peptide binding groove is divided into six binding pockets from A
to F [51]. Peptide binding motif is determined by the two primary anchors, P2 and P9. While P2
binds to B-pocket, P9 binds to F-pocket of the binding groove [119]. The pockets and the
interacting peptide residue positions are shown in Table 3.2, Figure 3.1 and Figure 3.2.
Table 3.2 Peptide binding groove pockets and the interacting peptide residue positions of HLA-
Interacting Positions of
Pockets Residues
the Peptide
5, 7, 59, 63, 66, 159, 163, 167,
A 171 [51] P1 [119]
31
Figure 3.2 Peptide binding groove pockets of HLA-A*02:01 (PDB ID : 1HHH [119]) and the
interacting peptide residues are shown in PyMOL representation. Macromolecule is shown in
gray and peptide is shown in lightgreen sticks. Colored dots represent the pocket residues
interacting with the peptide residues shown in blue stick. (A) P1, (B) P2, (C) P6, (D) P3, (E) P7,
(F) P9
32
3.2 Molecular Docking of HLA-A*02:01 Structures
Molecular docking studies for the peptides given in Table 3.1 were performed by DockTope web
1. An account is created to use the web server for modeling peptide-MHC Class I
complexes.
2. For each requested model, a special job name is entered to the text box provided.
5. Job is submitted.
6. When the job is completed, the resulting peptide-MHC Class I structure pdb file is sent to
MD simulations were performed for 100 ns at 310 K and 473 K by an eight step procedure
given below.
1. All hydrogen and water molecules were removed from the peptide-MHC Class I
2. Structure pdb file was converted to GROMACS [34] topology file using the OPLS-AA/L
3. Protein was solvated in a cubic box placed at a distance of 2 nm using the Simple Point
4. To neutralize the net charge of the system, an appropriate number of sodium and chloride
33
5. The potential energy of the solvated system was minimized using the steepest descent
6. Minimization step was followed by an NVT equilibration, where the number of atoms,
volume and temperature were all fixed. This process was performed in 100 picoseconds.
7. The system was then equilibrated using a NPT ensemble, in which the pressure is
8. In final step, MD simulation was performed for 50.000.000 steps with an integration time
MD simulations were carried out through TUBITAK TR GRID computing infrastucture and
University.
Root mean square deviation (RMSD) is a quantative measurement between the two atomic
particles. Atom positions are identified with 3-dimensional vectors. Each vector is considered in
the RMSD calculation.
∑(r − r0,i )
1 2
=
RMSDt t ,i
N i =1 3.1
where N is the number of the Cα atoms in the system, rt,i indicates ith atom atomic position at any
time t and r0,i indicates the Cα position in the minimized structure.
34
In this study, RMSD values were calculated only for the Cα atoms of residues after
Root mean square fluctuation (RMSF) is a measurement of the deviation between the
position of particle i and the reference position. RMSF is calculated using the [Eq. 3.2].
3.2
where T is the number of frames obtained by the MD simulation trajectory, is the position of
the atom i at time t and represents the average atomic position throughout the MD simulation
trajectory.
In this study, RMSF values were calculated from the average position of each Cα atom in the
equilibrated MD trajectory.
3.5 Calculation of pairwise amino acid non-bonded interaction energies via gRINN tool
Pairwise residue interaction energy was calculated using of the gRINN tool [36].
1. gRINN tool was downloaded from web page generated by Serçinoğlu et al.
2. For each calculation, special folder including topology (TOP), run input (TPR) and MD
3. For each folder, gRINN interface was started and new calculation mode was selected for
6. Pairwise residue interaction energy was displayed using view results interface.
35
CHAPTER 4 RESULTS AND DISCUSSIONS
All MD simulations were carried out for 100 ns and the structures are observed to reach
equilibrium after 40 ns. RMSD and RMSF analysis were performed on the last 60 ns of the MD
simulation trajectories. The RMSD figures are given in Figure 4.1-4.2 for 310 K and 473 K,
respectively.
Figure 4.1 Root mean square deviation (RMSD) profile of the modeled complexes during the
100 ns simulations at 310 K.
Figure 4.2 Root mean square deviation (RMSD) profile of the modeled complexes during the
100 ns simulations at 473 K.
36
When the simulations are conducted at a higher temperature, the RMSD values are observed to
be at a higher scale as well. Figure 4.3 demonstrates the distributions of the RMSD values at 310
K. In the presence of ILDDNLYKV peptide, HLA-A*02:01 allele displays the narrowest RMSD
distribution, while the widest RMSD distribution is observed in the presence of YLYYQPCDLL
and HVDGKILFV peptides. In addition, for ILDDNLYKV peptide, RMSD distribution profile is
observed at the leftmost compared to all other peptides. However, in presence of VLPFDIKYI
peptide, RMSD distribution profile is seen at the rightmost, meaning fluctuations at higher
RMSD values.
Figure 4.3 Distribution of alpha carbon root mean square deviation (RMSD) values of the
modeled complexes during the 100 ns simulations at 310 K.
37
Figure 4.4 Distribution of alpha carbon root mean square deviation (RMSD) values of the
modeled complexes during the 100 ns simulations at 473 K.
Figure 4.4 shows the distributions of the RMSD values at 473 K. We observe that the stability of
peptide-HLA*02:01 complex, RMSD distribution profile is seen at the leftmost with respect to
all other peptides, while for complexes with VLPFDIKYI and HVDGKILFV peptides, RMSD
According to the results of our RMSD analysis at 310 K and 473 K, ILDDNLYKV peptide forms
the most stable complex with the HLA-A*02:01 allele, while VLPFDIKYI and HVDGKILFV
38
4.2 Analysis of RMSF profiles
The RMSF figures are given in Figure 4.5 and Figure 4.6 for 310 K and 473 K, respectively.
Figure 4.5 Root mean square fluctuations (RMSF) values of the modeled complexes during the
100 ns simulations at 310 K.
For 310 K, the RMSF values are in the range of 0 to 4 Å as can be observed in Figure 4.5.
However at a higher temperature value (473 K), all the structures display higher flexibility
increased, the flexibility of the residues is increased as expected. The maximum flexibilities are
observed in the loop regions (residues 18, 30, 41, 56, 58, 83, 89, 106, 138, 177, 181, 196, 221,
226, 252, 256, 268, 275, 294, 318, 334, 351, 364, 374), while the minima is observed at 9, 26,
37, 48, 97, 99, 114, 158, 164, 187, 206, 242, 264, 286, 303, 315, 330, 342, 356, 371, 378.
39
Figure 4.6 Root mean square fluctuations (RMSF) values of the modeled complexes during the
100 ns simulations at 473 K.
In addition to the RMSF of the overall structure, we have also analysed the fluctuation profile of
the peptide residues as well, given in Figure 4.7. At 310 K, the residues of the peptides
demonstrate a very stable behavior as their alpha carbon atoms have rekatively small RMSF
values in the range of 1Å to 2Å. However, they display higher RMSF values (in the range of 2Å
to10Å) at 473 K as given in Figure 4.8. The residues of ILDDNLYKV, SLSAYIIRV and
GLFDFVNFV peptides have relatively smaller RMSF values with respect to other peptides at
both temperature values. These are also the top three most stable peptides that form complexes
with HLA-A*02:01 allele according to the experimental studies. The stability of these peptides is
40
reflected on their fluctuation profiles investigated in terms of RMSF profiles. The least stable
ones on the other hand, display a highly unstable profile at both temperatures.
Figure 4.7 Peptide RMSF profiles of the modeled complexes during the 100 ns simulations at
310 K.
Figure 4.8 Peptide RMSF profiles of the modeled complexes during the 100 ns simulations at
473 K.
41
4.3 Analysis of interaction energies using gRINN
In order to reveal the important interactions between the residues, pairwise residue interaction
energies at 310 K are computed via gRINN tool [36]. Total energy, posititive energy and negative
energy values obtained from gRINN are given in Table 4.1 and the correlation between
experimental stability measurements and our computational energy values are demonstrated in
Figure 4.9.
Table 4.2 Interaction energies of 12 peptide HLA-A*02:01 complexes obtained from gRINN.
42
Figure 4.9 Correlation between the experimental stability measurements and our computational
energy values obtained from gRINN.
Although an exact correlation can not be obtained among the experimental stability
measurements and the interaction energy results computed via gRINN, a general tendency can
still be captured. Especially the last three peptides (the least stable) and the first peptide (the
most stable) can be distinguished from the rest of the peptides in terms of having the highest and
The most stable complex (ILDDNLYKV peptide-HLA-A*02:01 complex) has the highest total
energy (-36400 kcal mol), while the least stable complex (VLPFDIKYI peptide-HLA-A*02:01
complex) has the lowest total energy (-35950 kcal mol) (Figure 4.9).
43
Figure 4.10 Motion of peptides and peptide binding grooves throughput the all trajectory.
Macromolecule is shown in gray, peptide binding groove is shown in military green, peptides are
shown in colored tubes. (ILDDNLYKV: red, GLFDFVNFV: orange, SLSAYIIRV: yellow,
YLPEVISTI: purple, RLYDYFTRV: magenta, LMYDIINSV: brown, VLYDEFVTI: pink,
RVYEALYYV: green, YLYFCSSDV: cyan, YLYQPCDLL: cream, HVDGKILFV: blue,
VLPFDIKYI: black)
44
Moreover, motion of peptides and peptide binding grooves throughout the trajectories are
analysed. According to our analysis, differences in stability are also consistent with the motion
of the peptides and the peptide binding grooves (Figure 4.10). The last for peptides show a much
more flexible profile when compared with the rest. In all the others, peptides display a much
Further residue based analyses are also conducted in order to gain insight about the nature of the
observed behavior. Through these extensive analyses, residues that play a key role in the stability
of the complex could be extracted. At first, the interaction energies between the residues of HLA
and β2-microglobulin are analysed and given in Figure 4.11. The important residues are also
In previous studies, the residues of HLA-A2 that interact with β2-microglobulin were stated as
PHE8, ARG48, GLN96, ALA117, ASP122, ARG202, GLU232, P235, ALA236 and GLY237 by
Hee [120].
As a general observation, according to our results, GLY119 has the highest energy value in all
residues ALA117 and ASP122 that are mentioned in the previous study [120]. This residue can
be considered as the key residue of interaction between the HLA and the β2-microglobulin
sections in the complex structure. In all the peptides, without any exception, this residue displays
the highest interaction energy values as shown in Figure 4.11 and Figure 4.13.
45
Figure 4.11 Interaction energies between residues of HLA and β2-microglobulin.
46
Figure 4.12 Key residues for the interaction between
HLA and β2-microglobulin (ILDDNLYKV peptide) are
shown in PyMOL representation. Colored dots represent
the important residues interacting with β2-microglobulin
(ARG48: magenta GLY119: red, ARG202: green,
peptide: orange, macromolecule: gray )
We observe that residue ARG48 have lower interaction energy values in the least stable
VLPFDIKYI peptide-HLA-A*02:01) when compared with the rest. This is further shown in
Figure 4.14, where the interaction energy of this residue is analysed alone for all the peptides.
47
Additionally, residue ARG202, which is also a residue mentioned in the previous study [120] has
the lowest energy value in the least stable complex (VLPFDIKYI peptide-HLA-A*02:01
For all the residues that interact with β2-microglobulin, there does not exist a specific difference
that can lead to the observed stability differences among the complexes of this study. It is already
recognized that the interaction between HLA and β2-microglobulin is crucial for the stability of
the peptide-HLA complexes [121][122]. Therefore, following our results, residue GLY119,
which displays an equally high interaction energy value in all the complexes, is an important
residue for HLA-A2 molecules in terms of HLA and β2-microglobulin connection. On the other
hand, residues ARG48 and ARG202 can be considered to have a role in the stability of these
complexes since they display lower interaction energy profiles for the least stable complexes.
48
Figure 4.15 Interaction energies of residue ARG202.
Secondly, interaction energies between residues of HLA and peptide residues are calculated and
given in Figures 4.16 - 4.18. HLA residues that interact with peptide residues are widely known
from the previous studies [119], [123]. Accordingly, GLU63 (interacts with P1 and P2), LYS66
(interacts with P1, P2, P3 and P4), ASP77 (interacts with (P8 and P9), ARG97 (interacts with P3
and P6) and TRP146 (interacts with P9) are important sites for peptide binding of HLA-A*02:01
alleles.
Using our computational methods, we have also verified these residues and observed that
especially GLU63, LYS66, ASP77 and TRP146 are the four most important residues for stable
binding of HLA-A*02:01 alleles to their peptides (Figure 4.19). The peptide stabilities can be
dependant on the interaction energy differences observed at these sites. Figures 4.16 display the
interaction energy values of the binding groove residues with the peptide residues. Further
detailed representation of the important regions are given in Figures 4.17 and 4.18 to zoom into
the differences observed. The region between residues 60-80 and 140-170 seem to be the key
49
Figure 4.16 Interaction energies between residues of HLA and peptide residues.
50
Figure 4.17 Interaction energies between region 1 (residues 60-80) and peptide residues.
51
Figure 4.18 Interaction energies between region 2 (residues 140-170) and peptide residues.
52
Figure 4.19 Important residues for
interaction between HLA and peptide
(ILDDNLYKV peptide) are shown in
PyMOL representation. Colored dots
represent the important residues interacting
with β2-microglobulin (GLU163: red,
LYS66: blue, TRP146: cyan, ASP77: yellow,
peptide: orange, macromolecule: gray )
Especially residues GLU63, LYS66 in the B pocket appear to have weak interactions for
peptides. This is further detailed in the Figure 4.20 and Figure 4.21. The interactions of
HVDGKILFV and VLPFDIKYI peptides with residue TRP146 is relatively weaker than the
interactions of the other peptides (Figure 4.22). That might be the underlying reason of the low
stability value for VLPFDIKYI peptide-HLA-A*02:01 complex since residue TRP146 binds the
anchor residue P9, and the loss of interaction energy at this site might be the indicator of this
phenomena.
53
Figure 4.20 Interaction energies of residue GLU63.
54
Figure 4.22 Interaction energies of residue TRP146.
In order to investigate the peptide residues’ point of view, the interaction energy profiles of the
peptide residues are plotted (Fig 4.23). Individually, each residue’s role on the varying peptides
55
Figure 4.23 Interaction energy profiles of the peptide residues.
56
P1 and P9 display very high interaction energies for all the structures (Figure 4.24 and Figure
4.32). In most of the highly stable structures (top 7 out of 8), P4 has an emerging effect as well
(Figure 4.27). P9, which is an anchor residue, seem to be more effective than the other anchor
residue, P2, in terms of interaction energies due to having higher values (Figure 4.125 and figure
4.32).
Although in all the peptides, there is a high interaction energy profile observed for P1; the highly
stable peptides display a slightly higher interaction energy profile for this residue (Figure 4.24).
On the contrary, in the last four complexes in terms of stability (YLYFCSSDV peptide-HLA-
57
P2 has lower energies for the last five peptide complexes (RVYEALYYV peptide-HLA*02:01,
complexes (Figure 4.25). It is highly possible that these last five peptides form less stable
complexes with the HLA-A*02:01 allele due to the poor binding of anchor P2.
HVDGKILFV peptides are the only ones having ASP in position P3. Hence, the observed
58
Figure 4.26 Interaction energies of residue P3.
For P4, there is a higher interaction energy in almost all the highly stable peptides with an
exception of SLSAYIIRV peptide-HLA-A*02:01 complex (Figure 4.27). For the ones having
high interaction energies, there is either ASP (D) or GLU (E) at P4 location. Both of these are
negatively charged side groups. Whereas, for SLSAYIIRV, there is ALA at this position, which is
a nonpolar aminoacid. This might be the reason of exception for this peptide. But generally, it
can be concluded that having a negatively charged group (ASP or GLU) at this location has a
definite stabilizing factor, since all the peptides having these residues at this location have high
stability values.
59
For P5 to P8, there are no significant energy value differences observed (Figure 4.28 - 4.31).
60
Figure 4.29 Interaction energies of residue P6.
61
Figure 4.31 Interaction energies of residue P8.
P9 has relatively lower energy values the least stable two complexes, VLPFDIKYI peptide-
RVYEALYYV peptide-HLA-A*02:01 complex also has a low energy. This peptide is binding to
the HLA structure through the first peptide residue, P1, much more tightly when compared with
the rest (Figure 4.24). Hence, it the energy seems to switch to that location rather than this
section.
62
Figure 4.32 Interaction energies of residue P9.
As a final step, further detailed analysis is conducted and for each peptide, the residues involved
in interaction with each peptide residue throughout the MD simulation is calculated and shown in
We have revealed all the interactions between residues through the MD simulations (Figures 4.33
and 4.34). According to our results, residue P1 can interact with 5, 7, 9, 26, 27,33, 45, 51, 52, 54,
55, 58-67, 70, 97, 99, 101, 114, 159, 160, 162-168, 170, 171, residue P2 can interact with 7, 9,
24, 25, 45, 58- 60, 62- 67, 69, 70, 97- 99, 114, 159, 163, 164, 167, 171, residue P3 can interact
with 6, 7, 9, 63, 64, 66, 68- 71, 97, 99, 113-115, 126, 133, 152, 153, 155- 160, 162- 164, residue
P4 can interact with 9, 22, 58, 61- 74, 77, 97, 99, 113-115, 126, 133, 150- 152, 154- 156, 158-
160, 165, 167, residue P5 interact with 62, 63, 65- 70, 72, 73, 77, 96, 97, 99, 112, 114, 116, 132,
133, 146, 147, 150- 156, 159, 160, residue P6 can interact with 9, 62, 63, 65-67, 69- 77, 97, 99,
114-117, 133, 147, 149-153, 155, 156, 159, residue P7 can interact with 9, 11, 22, 66- 70, 73,
63
74, 77-81, 84, 97, 99, 112, 114-116, 118, 123-125, 133, 139, 140, 142-157, residue P8 can
interact with 65, 66, 69, 70, 72- 81, 84, 97, 114- 116, 122-124, 133, 140, 142- 152, residue P9
can interact with 70, 72- 85, 96, 97, 115- 118, 122- 124, 133, 138-140- 152 .
From these heat maps, we can one more time acknowledge that in the most stable complexes,
residues P1-P4 have strong interactions with residues GLU63 and LYS66 compared to the least
64
Figure 4.33 Interacting residues of ILDDNLYKV, GLFDFVNFV, SLSAYIIRV, YLPEVISTI,
RLYDYFTRV and LMYDIINSV peptides.
65
Figure 4.34 Interacting residues of VLYDEFVTI, RVYEALYYV, YLYFCSSDV,
YLYQPCDLL, HVDGKILFV and VLPFDIKYI peptides.
66
CHAPTER 5 CONCLUSION
In the present study, molecular docking studies were performed for 12 peptide-HLA-A*02:01
complexes by DockTope web server [33] and then 100 ns parallel MD simulations were carried
out on the docked structures using GROMACS software [34] with the OPLS-AA/L [91] all-atom
The RMSD and RMSF results of MD simulations demonstrate differences between the 12
with the experimental measurements [10]. All modeled peptide-HLA-A*02:01 complexes show
higher fluctuations at 473 K compared to 310 K. But, three relatively most stable complexes still
remained stable throughout 100 ns at 473 K. However, two relatively less stable complexes lost
interactions with the HLA-A*02:01 allele. Therefore, high temperature MD verified the previous
complexes [10].
Additionally, pairwise amino acid non-bonded interaction energies at 310 K were calculated via
gRINN tool[36] to determine the critical residues which play a role in the observed difference in
the stability of these 12 peptide-HLA-A*02:01 complexes. We observed that our total energy
values correlate with the experimental stability measurements [10]. Therefore, we proved that the
stability is the most accurate parameter for peptide immunogenicity. Also, our results verified the
equally high interaction energy value in all the complexes, is an important residue for HLA-A2
molecules. On the other hand, residue ARG48 and ARG202 can be considered to have a role in
the stability of these complexes since they display lower interaction energy profiles for the least
stable complexes.
67
Moreover, we analyzed the interaction energies between residues of HLA and peptide residues
and observe dramatic energy differences at residue positions GLU63 and LYS66 in the B pocket.
Our results prove that GLU63 and LYS66 interact with residues P2-P4 and are important for the
stable binding of HLA-A*02:01 alleles [123]. Moreover, these results are also inline with
Furthermore, we observed that P1 is the most critical residue for binding since it displays the
highest interaction energy values for all the peptides. Additionally, P4 is found to be important
for stable binding. In general the interaction between P4 and LYS66 seem to be crucial for the
stability. When P4 is ASP or GLU, there is a high stability observed for all the peptides because
As a result, it can be concluded that the computational approaches used in this study provides
detailed information for the prediction of peptide-HLA-A*02:01 complex stabilities, for which
the three dimensional structures are unknown. Our methodology can be used to guide future
68
CHAPTER 6 FUTURE WORK
analyzed using our computational approaches. For peptide binding affinity prediction, new
methods can be developed. In addition, the effect of mutations on the stability of peptide HLA-
69
REFERENCES
[1] M. Wieczorek et al., “Major Histocompatibility Complex (MHC) Class I and MHC Class
II Proteins: Conformational Plasticity in Antigen Presentation,” Front. Immunol., vol. 8, p.
292, Mar. 2017.
[2] J. B. Rothbard and M. L. Gefter, “Interactions between Immunogenic Peptides and MHC
Proteins,” Annu. Rev. Immunol., vol. 9, no. 1, pp. 527–565, Apr. 1991.
[4] M. Schumacher et al., “Restricted T Cell Epitopes − HLA-A2 Design through Chemically
Modified Altered Peptide Ligands Revisited: Vaccine Altered Peptide Ligands Revisited:
Vaccine Design through Chemically Modified HLA-A2–Restricted T Cell Epitopes,”
DCSupplemental.html J. Immunol. Med. Libr. Vrije Univ. Novemb. J. Immunol., vol. 193,
no. 9, pp. 4803–4813, 2014.
70
Binding to HLA Class I Molecules: Homogenous, High-Throughput Screening, and
Affinity Assays,” J. Biomol. Screen., vol. 14, no. 2, pp. 173–180, Feb. 2009.
[10] M. Harndahl et al., “Peptide-MHC class I stability is a better predictor than peptide
affinity of CTL immunogenicity,” Eur. J. Immunol., vol. 42, no. 6, pp. 1405–1416, Jun.
2012.
[12] D. H. Busch and E. G. Pamer, “MHC class I/peptide stability: implications for
immunodominance, in vitro proliferation, and diversity of responding CTL.,” J. Immunol.,
vol. 160, no. 9, pp. 4441–8, May 1998.
[13] E. Assarsson et al., “A quantitative analysis of the variables affecting the repertoire of T
cell specificities recognized after vaccinia virus infection.,” J. Immunol., vol. 178, no. 12,
pp. 7890–901, Jun. 2007.
[17] a Sette et al., “The relationship between class I binding affinity and immunogenicity of
potential cytotoxic T cell epitopes.,” J. Immunol., vol. 153, no. 12, pp. 5586–92, 1994.
[18] M. E. Ressing et al., “Human CTL epitopes encoded by human papillomavirus type 16 E6
71
and E7 identified through in vivo and in vitro immunogenicity studies of HLA-A*0201-
binding peptides.,” J. Immunol., vol. 154, no. 11, pp. 5934–5943, 1995.
[26] A. Krammer et al., “Forced unfolding of the fibronectin type III module reveals a tensile
molecular recognition switch.,” Proc. Natl. Acad. Sci. U. S. A., vol. 96, no. 4, pp. 1351–6,
Feb. 1999.
72
[27] W. F. GUNSTEREN and A. E. MARK, “On the interpretation of biochemical data by
molecular dynamics computer simulation,” Eur. J. Biochem., vol. 204, no. 3, pp. 947–961,
Mar. 1992.
[28] V. Daggett and M. Levitt, “A model of the molten globule state from molecular dynamics
simulations.,” Proc. Natl. Acad. Sci. U. S. A., vol. 89, no. 11, pp. 5142–6, Jun. 1992.
[29] V. Daggett and M. Levitt, “Protein Unfolding Pathways Explored Through Molecular
Dynamics Simulations,” J. Mol. Biol., vol. 232, no. 2, pp. 600–619, Jul. 1993.
[30] M. Karplus and D. L. Weaver, “Protein folding dynamics: the diffusion-collision model
and experimental data.,” Protein Sci., vol. 3, no. 4, pp. 650–68, Apr. 1994.
[32] K. Y. Chen, J. Liu, and E. C. Ren, “Structural and functional distinctiveness of HLA-A2
allelic variants,” Immunol. Res., vol. 53, no. 1–3, pp. 182–190, Sep. 2012.
[33] M. Menegatti Rigo et al., “DockTope: a Web-based tool for automated pMHC-I
modelling,” Sci. Rep., vol. 5, no. 1, p. 18413, Nov. 2016.
[34] M. J. Abraham et al., “Gromacs: High performance molecular simulations through multi-
level parallelism from laptops to supercomputers,” SoftwareX, vol. 1–2, pp. 19–25, 2015.
[36] O. Serçinoğlu and P. Ozbek, “gRINN: a tool for calculation of residue interaction energies
and protein energy network analysis of molecular dynamics simulations,” Nucleic Acids
Res., May 2018.
[37] J. M. Werner et al., “Major Histocompatibility Complex (MHC) Class i and MHC Class ii
Proteins: Conformational Plasticity in Antigen Presentation,” Front. Immunol, vol. 8, no.
8, pp. 2923389–292, 2017.
[38] K. Maenaka and E. Y. Jones, “MHC superfamily structure and the immune system.,”
Curr. Opin. Struct. Biol., vol. 9, no. 6, pp. 745–53, Dec. 1999.
73
[39] M. Wieczorek et al., “Major Histocompatibility Complex (MHC) Class I and MHC Class
II Proteins: Conformational Plasticity in Antigen Presentation,” Front. Immunol., vol. 8, p.
292, Mar. 2017.
[43] R. T. Kubo et al., “Definition of specific peptide motifs for four major HLA-A alleles.,” J.
Immunol., vol. 152, no. 8, pp. 3913–24, Apr. 1994.
[49] J. Sutton et al., “A sequence pattern for peptides presented to cytotoxic T lymphocytes by
74
HLA B8 revealed by analysis of epitopes and eluted peptides,” Eur. J. Immunol., vol. 23,
no. 2, pp. 447–453, Feb. 1993.
[57] K. Gulukota, J. Sidney, A. Sette, and C. DeLisi, “Two complementary methods for
predicting peptides binding major histocompatibility complex molecules,” J. Mol. Biol.,
vol. 267, no. 5, pp. 1258–1267, Apr. 1997.
[58] P. Dönnes and A. Elofsson, “Prediction of MHC class I binding peptides, using
SVMHC.,” BMC Bioinformatics, vol. 3, p. 25, Sep. 2002.
[59] C.-W. Tung and S.-Y. Ho, “POPI: predicting immunogenicity of MHC class I binding
peptides by mining informative physicochemical properties,” Bioinformatics, vol. 23, no.
75
8, pp. 942–949, Apr. 2007.
[60] H. Mamitsuka, “Predicting peptides that bind to MHC molecules using supervised
learning of hidden Markov models.,” Proteins, vol. 33, no. 4, pp. 460–74, Dec. 1998.
[62] K. C. Parker, M. A. Bednarek, and J. E. Coligan, “Scheme for ranking potential HLA-A2
binding peptides based on independent binding of individual peptide side-chains.,” J.
Immunol., vol. 152, no. 1, pp. 163–75, Jan. 1994.
[63] B. Peters, W. Tong, J. Sidney, A. Sette, and Z. Weng, “Examining the independent
binding assumption for binding of peptide epitopes to MHC-I molecules,” Bioinformatics,
vol. 19, no. 14, pp. 1765–1772, Sep. 2003.
[64] R. Rosenfeld, Q. Zheng, S. Vajda, and C. DeLisi, “Flexible docking of peptides to class I
major-histocompatibility-complex receptors.,” Genet. Anal., vol. 12, no. 1, pp. 1–21, Mar.
1995.
[67] D. Rognan, “Molecular dynamics simulations: a tool for drug design,” Perspect. Drug
Discov. Des., vol. 9/11, no. 0, pp. 181–209, 1998.
[68] M. Nielsen et al., “NetMHCpan, a method for quantitative predictions of peptide binding
to any HLA-A and -B locus protein of known sequence.,” PLoS One, vol. 2, no. 8, p.
e796, Aug. 2007.
[69] I. Hoof et al., “NetMHCpan, a method for MHC class I binding prediction beyond
humans.,” Immunogenetics, vol. 61, no. 1, pp. 1–13, Jan. 2009.
76
[70] E. Karosiene, C. Lundegaard, O. Lund, and M. Nielsen, “NetMHCcons: a consensus
method for the major histocompatibility complex class I predictions,” Immunogenetics,
vol. 64, no. 3, pp. 177–186, Mar. 2012.
[76] B. Knapp and B. Knapp, “Structural Bioinformatics and Immuoinformatics in the T Cell
Receptor / Peptide / Major Histocompatibility Complex Interface,” pp. 1–7, 2011.
[78] J. M. Sanchez-Ruiz, “Protein kinetic stability,” Biophys. Chem., vol. 148, no. 1–3, pp. 1–
15, May 2010.
77
Jul. 1979.
[81] M. K. Gilson and H.-X. Zhou, “Calculation of Protein-Ligand Binding Affinities *,”
Annu. Rev. Biophys. Biomol. Struct, vol. 36, pp. 21–42, 2007.
[82] J. Aqvist, C. Medina, and J. E. Samuelsson, “A new method for predicting binding affinity
in computer-aided drug design.,” Protein Eng., vol. 7, no. 3, pp. 385–91, Mar. 1994.
[83] P. A. Kollman et al., “Calculating structures and free energies of complex molecules:
combining molecular mechanics and continuum models.,” Acc. Chem. Res., vol. 33, no.
12, pp. 889–97, Dec. 2000.
[86] M. Zacharias and S. Springer, “Conformational flexibility of the MHC class I alpha1-
alpha2 domain in peptide bound and free states: a molecular dynamics simulation study.,”
Biophys. J., vol. 87, no. 4, pp. 2203–14, Oct. 2004.
[87] B. J. Alder and T. E. Wainwright, “Phase Transition for a Hard Sphere System,” J. Chem.
Phys., vol. 27, no. 5, pp. 1208–1209, Nov. 1957.
[88] A. Rahman, “Correlations in the Motion of Atoms in Liquid Argon,” Phys. Rev., vol. 136,
no. 2A, pp. A405–A411, Oct. 1964.
[91] W. L. Jorgensen and J. Tirado-Rives, “The OPLS [optimized potentials for liquid
simulations] potential functions for proteins, energy minimizations for crystals of cyclic
peptides and crambin,” J. Am. Chem. Soc., vol. 110, no. 6, pp. 1657–1666, Mar. 1988.
78
complex between the human histocompatibility antigen HLA-A2 and the IMP58-66
nonapeptide from influenza virus matrix protein.,” Eur. J. Biochem., vol. 208, no. 1, pp.
101–13, Aug. 1992.
[93] D. A. Pearlman et al., “AMBER, a package of computer programs for applying molecular
mechanics, normal mode analysis, molecular dynamics and free energy calculations to
simulate the structural and energetic properties of molecules,” Comput. Phys. Commun.,
vol. 91, no. 1–3, pp. 1–41, Sep. 1995.
[97] H. Fabian et al., “HLA-B27 Subtypes Differentially Associated with Disease Exhibit
Conformational Differences in Solution,” J. Mol. Biol., vol. 376, no. 3, pp. 798–810, Feb.
2008.
[99] E. T. Abualrous et al., “F pocket flexibility influences the tapasin dependence of two
differentially disease-associated MHC Class I proteins,” Eur. J. Immunol., vol. 45, no. 4,
pp. 1248–1257, Apr. 2015.
[103] H. M. Berman et al., “The Protein Data Bank.,” Nucleic Acids Res., vol. 28, no. 1, pp.
235–42, Jan. 2000.
[104] S.-Y. Huang, “Search strategies and evaluation in protein–protein docking: principles,
advances and challenges,” Drug Discov. Today, vol. 19, no. 8, pp. 1081–1096, Aug. 2014.
[107] D. L. Mobley and K. A. Dill, “Binding of small-molecule ligands to proteins: ‘what you
see’ is not always ‘what you get’.,” Structure, vol. 17, no. 4, pp. 489–98, Apr. 2009.
[109] M. P. Repasky, M. Shelley, and R. A. Friesner, “Flexible Ligand Docking with Glide,” in
Current Protocols in Bioinformatics, vol. Chapter 8, Hoboken, NJ, USA: John Wiley &
Sons, Inc., 2007, p. Unit 8.12.
[110] G. Jones, P. Willett, R. C. Glen, A. R. Leach, and R. Taylor, “Development and validation
of a genetic algorithm for flexible docking 1 1Edited by F. E. Cohen,” J. Mol. Biol., vol.
267, no. 3, pp. 727–748, Apr. 1997.
80
docking approach based on biochemical or biophysical information.,” J. Am. Chem. Soc.,
vol. 125, no. 7, pp. 1731–7, Feb. 2003.
[116] O. Trott and A. J. Olson, “AutoDock Vina: improving the speed and accuracy of docking
with a new scoring function, efficient optimization, and multithreading.,” J. Comput.
Chem., vol. 31, no. 2, pp. 455–61, Jan. 2010.
[118] E. Assarsson et al., “A quantitative analysis of the variables affecting the repertoire of T
cell specificities recognized after vaccinia virus infection.,” J. Immunol., vol. 178, no. 12,
pp. 7890–901, Jun. 2007.
[120] C.-S. Hee et al., “Dynamics of free versus complexed β2-microglobulin and the evolution
of interfaces in MHC class I molecules,” Immunogenetics, vol. 65, no. 3, pp. 157–172,
Mar. 2013.
81
1994.
[122] M. Shiroishi et al., “Structural basis for recognition of the nonclassical MHC molecule
HLA-G by the leukocyte Ig-like receptor B2 (LILRB2/LIR2/ILT4/CD85d).,” Proc. Natl.
Acad. Sci. U. S. A., vol. 103, no. 44, pp. 16412–7, Oct. 2006.
82
AUTOBIOGRAPHY
E-Mail : asumanbunsuz@gmail.com
Education Statue
University /
Degree Section / Program Graduation Year
Highschool
Scientific Works
Conference Publications:
1. Bunsuz, A., Sercinoglu, O., Ozbek, P, Effect of temperature on the molecular dynamics
simulations of HLA-A*02 alleles, International Symposium on Chemistry via
Computation Applications on Molecular Nanoscience, Istanbul,Turkey, 30 October 2017.
(Poster Presentation)
2. Bunsuz, A., Sercinoglu, O., Ozbek, P, Immunogenic Peptide - HLA-A*02:01 Complexes
Studied by Comparative Molecular Dynamics Simulation, 5th International BAU-Drug
Design Congress, İstanbul, Turkey, 19-21 October 2017. (Poster Presentation)
3. Bunsuz, A., Sercinoglu, O., Ozbek, P, Investigation of peptide binding affinity and
thermal stability of Human Leukocyte Antigens (HLAs), 19th IUPAB and 11th EBSA
83
Congress, Edinburgh, UK, 16-20 July 2017. (Poster Presentation)
4. Bunsuz, A., Sercinoglu, O., Ozbek, P, Investigation of peptide binding affinity and
complex stability of Human Leukocyte Antigens (HLAs), 4th International BAU-Drug
Design Congress, İstanbul, Turkey, 12-15 October 2016. (Poster Presentation)
5. Bunsuz, A., Sercinoglu, O., Ozbek, P, HLA-B*44 Alellerinin Peptid Bağlanma Davranış
Mekanizmalarının Hesaplamalı Olarak araştırılması, Biruni Üniversitesi Bilgisayar
Destekli İlaç Tasarımı, İstanbul, Türkiye, 16-17 Mayıs 2016 (Poster Award)
6. Bunsuz A., Sercinoglu, O., Ozbek, P., Computational Study on the Binding Behaviour of
HLA-B44 Alleles, 3rd International BAU-Drug Design Congress, Istanbul, Turkey,
October 1-3, 2015. (Poster Presentation)
Project Assignment:
84