quantum chemical modeling of enzymatic methyl transfer...
TRANSCRIPT
Quantum Chemical Modeling of
Enzymatic Methyl Transfer
Reactions
Polina Georgieva
Department of Theoretical Chemistry
Royal Institute of Technology
Stockholm, Sweden, 2008
Thesis for Philosophy Doctor degree
Department of Theoretical Chemistry
School of Biotechnology
Royal Institute of Technology
Stockholm
Sweden 2008
c© Polina Georgieva
ISBN 978-91-7415-171-8 pp i-ix, 1-51
ISSN 1654-2312
TRITA-BIO-Report 2008:26
Printed by Universitetsservice US-AB,
Stockholm
Sweden 2008
Abstract
In this thesis, quantum chemistry, in particular the B3LYP density functional method,
is used to investigate a number of methyl transfer enzymes. Quantum chemical
methodology is today a very important tool in the elucidation of properties and reac-
tion mechanisms of enzyme active sites. The enzymes considered in this thesis are the
S-adenosyl L-methionine-dependent enzymes - glycine N-methyltransferase, guanidi-
noacetate methyltransferase, phenylethanolamine N-methyltransferase, and histone
lysine methyltransferase. In addition, the reaction mechanism of the DNA repairing
enzyme O6-methylguanine methyltransferase is studied.
Active site models of varying sizes were designed and stationary points along the
reaction paths were optimized and characterized. Potential energy surfaces for the
reactions were calculated and the feasibility of the suggested reaction mechanisms
was able to be judged. By systematically increasing the size of the models, deeper
insight into the details of the reactions was obtained, the roles of the various active
site residues could be analyzed, and, very importantly, the adopted modeling strategy
was evaluated.
i
Acknowledgements
I have spent a few wonderful years in the Theoretical Chemistry group in Stockholm,
Sweden. During those lovely years I had the fortune that Docent Fahmi Himo was
my supervisor. Under his guidance, helpfulness, and encouragement I have learned a
lot about science and life. I cannot find words to express my gratitude and happiness
of working with him. Thank you Fahmi, thank you for everything.
I would like to express my gratitude to Prof. Hans Agren for accepting me in the
Theoretical Chemistry group and for providing a warm working environment. Grati-
tude goes also to all the people who took care of the computer technical problems,
administrative staff and the librarians without whom this thesis would not have been
possible.
I would like to acknowledge the people with whom I had the pleasure to work
- Kathrin, Jing Dong, Tommaso, Robin, Peter, Chen, and Liao. Thank you for
the discussions during our internal seminars and for the fun we had together. Over
the years the theoretical chemistry group guided by Prof. Hans Agren expanded a
lot. Right now there are people who I do not know personally, but I would like to
thank some present and former colleagues who made my time here more pleasant -
Yasen, Oscar, Luca, Viviane, Katja, Freddy, Elias, Emil. I had a special connection
and friendship to a former member of the group, Ivo. Thank you for the joys and
inspirations, thank you for changing my life, thank you..... Rest in peace.
I would like to thank the person who believed in me some years ago, opened my
eyes for science and gave me much support, Prof. Alia Tadjer from the University
of Sofia, Bulgaria. I will be forever grateful for everything I have learned and for the
positive and valuable life guidings.
Thanks to the friends that have been by my side during the years. Warm thanks
to my family - Valentin and Monica. Thank you Valentin for your love and support,
thank you for being by my side every day through the last years. Special thanks to
my daughter Monica for her love. I am infinitely grateful to you, with all my heart.
Thanks to my father, mother, and my brother who love me, and have been on my
side all the time. Thank you!
iii
List of Papers
• I. Methyl Transfer in Glycine N-Methyltransferase. A Theo-
retical Study.
Polina Velichkova, Fahmi Himo,
J. Phys. Chem. B 109 (16), 8216 –8219 (2005).
• II. Theoretical Study of The Methyl Transfer in Guanidinoac-
etate Methyltransferase.
Polina Velichkova, Fahmi Himo,
J. Phys. Chem. B 110 (1), 16 –19 (2006).
• III. The Reaction Mechanism of Phenylethanolamine N-Methyl-
transferase: A Density Functional Theory Study.
Polina Georgieva, Michael McLeish, Fahmi Himo,
Manuscript
• IV. Quantum Chemical Modeling of Enzymatic Reactions: The
Case of Histone Lysine Methyltransferase.
Polina Georgieva, Fahmi Himo,
Manuscript
• V. Density Functional Theory Study of The Reaction Mecha-
nism of The DNA Repairing Enzyme Alkylguanine Alkyltrans-
ferase.
Polina Georgieva, Fahmi Himo,
Chem. Phys. Lett. 463, 214 – 218 (2008).
v
Abbreviations and Acronyms
Abbreviation Description Definition
DFT Density Functional Theory page 3
SAM S-adenosyl L-methionine page 4
GNMT Glycine N-methyltransferase page 4
GAMT Guanidinoacetate methyltransferase page 4
HKMT Histone methyltransferase page 4
PNMT Phenylethanolamine N-methyltransferase page 4
MGMT O6-methylguanine methyltransferase page 4
TST Transition State Theory page 4
QC Quantum Chemistry page 5
QM Quantum Mechanics page 5
B3LYP Becke 3 parameter Lee-Yang-Parr Func-
tional
page 7
PDB Protein Data Bank page 16
ZPVE Zero-Point Vibrational Effect page 18
PCM Polarizable continuum model page 18
SAH S-adenosylhomocysteine page 20
ATP Adenosine triphosphate page 20
GAA Guanidinoacetate page 26
PEA Phenylethanolamine page 31
AGT Alkylguanine alkyltransferase page 39
vii
Amino Acids Abbreviations
1-Letter Symbol 3-Letter Symbol Full Name
A Ala Alanine
C Cys Cysteine
D Asp Aspartate
E Glu Glutamate
F Phe Phenylalanine
G Gly Glycine
H His Histidine
I Ile Isoleucine
K Lys Lysine
L Leu Leucine
M Met Methionine
N Asn Asparagine
P Pro Proline
Q Gln Glutamine
R Arg Arginine
S Ser Serine
T Thr Threonine
V Val Valine
W Trp Tryptophan
X Xaa Any Residue
Y Tyr Tyrosine
ix
Contents
Contents 1
1 Theoretical Background 5
1.1 Density Functional Theory, DFT . . . . . . . . . . . . . . . . . . . 51.2 B3LYP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 Performance of B3LYP . . . . . . . . . . . . . . . . . . . . 71.3 Reaction Rates and Transition State Theory . . . . . . . . . . . . . 91.4 Enzyme Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Modeling of Enzymatic Reactions 15
2.1 Modeling Methodology . . . . . . . . . . . . . . . . . . . . . . . . . 152.2 Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Modeling of Methyl Transfer Reactions 19
3.1 SAM-dependent Enzymes . . . . . . . . . . . . . . . . . . . . . . . 213.1.1 Glycine N-methyltransferase, GNMT (Paper I) . . . . . . . 213.1.2 Guanidinoacetate Methyltransferase, GAMT (Paper II) . . 263.1.3 Phenylethanolamine N-methyltransferase, PNMT (Paper III) 303.1.4 Histone Lysine Methyltransferase, HKMT (Paper IV) . . . 35
3.2 Other Methyl Transfer Enzymes . . . . . . . . . . . . . . . . . . . 393.2.1 O6-Methylguanine Methyltransferase, MGMT (Paper V) . 39
4 Conclusions 45
Bibliography 47
1
Introduction
Enzymes are the catalytic machines of life. Basically all chemical transformations
that take place in the organisms are catalyzed and controlled by specific enzymes.
Thus, failure of the function of certain key enzymes can lead to serious diseases. In
fact, many medicines work by blocking the function of some target enzyme. Due to
their specificity and selectivity properties, enzymes are today also used as synthetic
tools in large-scale industry processes to produce chemicals. It is thus of enormous
importance that one develops a detailed understanding of how these fascinating ma-
chines are able to catalyze their reactions. This understanding can, in the long run,
have far-reaching consequences. It can, for instance, allow for the rational design of
new drug compounds that will have improved selectivity and potency properties. It
can also make possible the design of bio-mimetic catalytic complexes for industrial
applications.
To this end, theoretical chemistry is today providing a number of computational
techniques capable of addressing and solving a wide range of problems in many di-
verse fields. It has proved its usefulness in many different research areas, ranging
from traditional chemistry, such as synthesis and analysis, to modern fields, such as
material science and drug development. Because the field heavily depends on compu-
tational power, it has, like other computational fields, benefited from the enormous
advancement in computer speed. Computers are becoming faster and cheaper every
day, and thus more and more complicated problems can be tackled.
Application of theoretical chemistry techniques to study enzyme catalysis is not
a recent developments. However, the use of accurate quantum chemical methods to
model enzyme active sites and reactivities is a quite recent field. The most important
reason for this development is the advances that density functional theory (DFT) has
made. DFT, and in particular hybrid functionals such as B3LYP, has made it possible
to treat, quite accurately, far larger systems than is possible with ab initio methods.
This has paved the way for wider applications, such as enzyme catalysis.
Theoretical studies of enzyme reactions provide excellent complement to the ex-
perimental studies. It also provides some advantages. For example, it is possible to
locate and characterize short-lived intermediates (and transition states), something
3
4 CONTENTS
that is very challenging to do with the current experimental techniques. Also, it
is possible to investigate hypothetical mechanistic scenarios and to provide detailed
information and break-down of various effects contributing to the reactivity of the
system.
The field of computational enzymology is a highly interdisciplinary one, on the in-
terface between several traditional research fields, such as biochemistry, organic chem-
istry, inorganic chemistry, physical chemistry, molecular biology, quantum physics, and
computational sciences. Knowledge in all these areas is a prerequisite for successful
research.
In this thesis, DFT is used to study a number of methyl transfer enzymes. This
class of enzymes catalyzes a number of very vital reactions, in many cases related to
brain and nerve activity. The enzymes considered in this thesis are the S-adenosyl L-
methionine-dependent (SAM) enzymes glycine N-methyltransferase (GNMT), guani-
dinoacetate methyltransferase (GAMT), phenylethanolamine N-methyl- transferase
(PNMT), and histone lysine methyltransferase (HKMT). Another enzyme, that is not
SAM-dependent, is studied also, namely the DNA repairing enzyme O6-methylguanine
methyltransferase (MGMT).
The thesis is organized in the following way: First, the theoretical methods, in
particular DFT, are briefly described in Chapter 1. Transition state theory (TST) and
enzyme kinetics are also briefly outlined in this chapter, due to their fundamental role
in making connections between theory and experimental measurements. Chapter 2
introduces the modeling approach used in this thesis to investigate enzyme reac-
tions. In Chapter 3, the main results of our studies on the methyltransferases are
summarized. Finally, in Chapter 4 some general conclusions are given.
Chapter 1
Theoretical Background
Quantum chemistry (QC) is a wide research field concerned with the general use
of the laws of quantum physics to study chemical problems. It encompasses many
diverse sub-fields, ranging from pure mathematical development och tools, to high-
end applications. This thesis deals with studying enzyme reaction mechanisms using
the tools of quantum chemistry.
One major reason for this becoming possible is the qualitative and quantitative
advances in density functional theory (DFT). With help of DFT methods, it is now
possible to treat quite large systems with reasonable accuracy and speed, which has
lead to more and more realistic models of enzyme active sites.
This chapter provides a very brief background to DFT. In particular, the hybrid
B3LYP functional and its performance are discussed. Also, brief accounts of transition
state theory and enzyme kinetics are given. These topics are very important in order
to relate the theoretical calculations to the experimental measurements.
1.1 Density Functional Theory, DFT
Explaining enzymatic reaction mechanisms that involve bond making and breaking
requires good description of the enzyme system at atomic level. The mathemati-
cal expressions that describe a molecular system are defined in quantum mechanics
(QM) [1, 2, 3]. The most important parameter is the wave function from the time-
independent Schrodinger equation, HΨ = EΨ, where Ψ is the wave function that
completely determine given physical system, H is the Hamiltonian operator that rep-
resents the energy of the system as a sum of kinetics and potential energy, and E is
the energy of the system obtained as an eigenvalue to the Hamiltonian. The equa-
tion cannot be solved exactly for systems larger than two particles. Since the real
chemical systems are quite complicated, solving the Schrodinger equation for them
5
6 CHAPTER 1. THEORETICAL BACKGROUND
requires approximations.
Depending on the desired accuracy, the approximations made in wave function-
based methods can be more or less sophisticated. In general, to obtain the high
accuracy needed to describe reaction energy profiles one has to use methods that are
quite expensive from computational point of view. Therefore, one is limited to treat
only relatively small systems.
An alternative way to express the energy of the system is by its electron den-
sity instead of the wave function [4, 5]. The method that uses electron density to
determine molecular properties is DFT.
The foundations of this theory lie on the two fundamental Hohenberg-Kohn the-
orems [6], which state that the ground state electron density of a system uniquely
determines the external potential, and that the density-dependent functional obeys
the variational principle. The first theorem implies that the ground state energy is
uniquely defined by the electron density, while the second means that the calculated
energy is higher than or equal to the true ground state energy. Mathematically, one
can thus express the total energy as a functional of the density:
Etot[ρ] = T [ρ] + Vee[ρ] + Vne[ρ], (1.1)
where ρ is electron density, T is the kinetic energy of the electrons, Vee is the
electron-electron repulsion, and Vne is the nuclear electron attraction. The first two
terms are independent of the nuclear position and represent the density functional,
F [ρ] = T [ρ] + Vee[ρ].
Etot[ρ] = F [ρ] + Vne[ρ], (1.2)
The problem with the Hohenberg-Kohn formalism is that the exact form of the
density functional F [ρ] is not known. It is known that it exists and connect energy
with a given electron density. The Kohn-Sham formalism gives a practical solution to
this problem by introducing an orbital-based scheme [7]. This approach divides the
kinetic energy of the system in two parts, one is the kinetic energy of non-interacting
system TS [ρ] and the rest is a residual kinetic part, TC [ρ], T [ρ] = TS [ρ] + TC [ρ].Introduction of the reference system gives the possibility larger part of the terms in
equation 1.2 to be calculated exactly. For example the kinetic energy functional of
non-interacting electrons is known exactly and this energy should be the same as in the
real interacting system. The second term in equation 1.2 contains classical Coulomb
interaction J [ρ] and residual non-classical part such as exchange, correlation, and
self-interaction Encl[ρ], Vee[ρ] = J [ρ] + Encl[ρ]. With this approaches equation 1.2
is reformulated as:
F [ρ] = TS [ρ] + J [ρ] + EXC [ρ] (1.3)
where EXC represents the difference in kinetic energy between the real system
and a system of non-interacting electrons plus the non-classical part of electron-
1.2. B3LYP 7
electron interaction. This term is called exchange-correlation functional and contains
all unknowns, and must be approximated. This is the biggest challenge in DFT
method, finding sufficiently accurate density functionals.
1.2 B3LYP
The most popular functionals in DFT are the hybrid functionals [5, 4] which introduce
parts of the Hartree-Fock exchange in the functional. Usually the parameters of the
functionals are fitted to reproduce some set of observables. The most widely used
functional in this category is Becke’s three parameter functional (B3LYP) [3], which
in general form can be written as:
FB3LY PXC = (1−A)FSlater
X + AFHFX + BFBecke
X + (1−C)F V WNC + CFLY P
C (1.4)
where FSlaterX is the Dirac-Slater exchange, FHF
X is the Hartree-Fock exchange
term, FBeckeX is the gradient part of the exchange functional of Becke [8, 9, 10,
11], F V WNC and FLY P
C are the correlation functionals of Vosko, Wilk, and Nusair
[12] and Lee, Yang, and Parr [13, 14] respectively. The parameters A, B and
C are related to the Hartree-Fock exchange and Coulomb correlation. They were
determined empirically by Becke [8, 9, 10, 11], and have the values of A = 0.20,
B = 0.72, and C = 0.81.
1.2.1 Performance of B3LYP
The use of approximations in the methods directly leads to errors in the obtained
results. In the present thesis we apply B3LYP functional to study enzyme reaction
mechanisms. The reason for this choice of B3LYP is the accuracy of the method
combined with its computational cost. Comparison between computational results
and experimental data assess the accuracy of a certain theoretical method. In case
of enzyme reaction mechanism, errors in the energies and geometrical parameters are
of interest.
Basis Set
The accuracy of a method depends, apart from the approximation of the methods
itself, also on the choice of a basis set used in the calculations [2, 3]. A basis set
is a set of functions centered on the different atoms in the molecule. In DFT the
basis set is used to describe the Kohn-Sham orbitals. Most of the commonly used
basis sets are composed of Gaussian functions. Common additions to the basis sets
are polarization and diffuse functions. Usually, a medium-sized basis set is used for
geometry optimization. In this thesis all geometry optimizations are performed by
using Pople double-ζ basis set, 6-31G(d,p). The core orbitals are presented by one
8 CHAPTER 1. THEORETICAL BACKGROUND
function composed of 6 Gaussians. The valence orbitals are treated with two func-
tions, one of which is composed of 3 and the other one of 1 Gaussians. Additionally p-
and d- polarization functions are added. These functions introduce some additional
flexibility important while considering accurate representation of bonding between
the atoms. For accurate evaluation of the energy a larger triple-ζ basis set is used,
6-311G+(2d,2p).
Accuracy on Geometries
The accuracy of B3LYP with respect to geometrical parameters was tested against the
standard G2 benchmark set of molecules [15, 16]. The G2 set includes 55 molecules
for which very accurate experimental data is available. As mentioned above, the
accuracy of the B3LYP calculations depends on the choice of basis set. The smallest
basis set which is recommended for calculations is 6-31G* or equivalent of double-ζ
quality. Comparison with experimental data reveals that this basis set works quite
well for geometry optimizations, see Table 1.1.
error 6-31G* 6-311+G(3df,2p)
average 0.013 0.008Bond Lengths,(A)maximum 0.055 0.039
average 0.62◦ 0.61◦Bond Angles, (◦)maximum 1.69◦ 1.85◦
average 0.35◦ 3.66◦Dihedral Angles, (◦)maximum 0.63◦ 6.61◦
average 5.18 2.20Atomization Energy, (kcal/mol)maximum 31.50 8.40
Table 1.1: Mean absolute errors of B3LYP on the G2 benchmark test set [15, 16, 17].
Using a very large basis set, as 6-311+G(3df,2p) [15, 16], contributes very little
to the geometry accuracy. The B3LYP functional performs thus very well in terms
of geometries.
1.3. REACTION RATES AND TRANSITION STATE THEORY 9
Accuracy on Energies
The average error in atomization energies of B3LYP on the G2 test set using the
6-31G* basis set is calculated to be 5.18 kcal/mol (Table 1.1) [17], while using
the much larger 6-311+G(3df,2p) basis set yields a much smaller error of only 2.2
kcal/mol. This is an outstanding performance, comparable to the most accurate ab
initio methods.
Another evaluation of B3LYP has been performed on the G3 test set [18, 19],
which includes many more molecules, see Table 1.2. The calculations are single-
points using B3LYP/6-311+G(3df,2p) on MP2/6-31G(d) geometries, and corrected
for zero-point vibrations by HF/6-31G(d) scaled by 0.89. The results show, again,
that B3LYP achieves quite high accuracy.
Energies Mean Absolute Deviation, (kcal/mol)
Enthalpies of Formation 4.63
Ionization Energies 3.83
Electron Affinities 2.99
Proton Affinities 1.39
All 4.11
Table 1.2: Mean absolute deviation of B3LYP on the G3 test set [19].
This, together with the favorable scaling of the method, make B3LYP an at-
tractive tool in the study of enzymes reaction mechanism, which is the topic of the
present thesis [20, 21].
1.3 Reaction Rates and Transition State Theory
The present thesis deals with studying the catalytic reaction mechanism in enzymes.
This process involves continuous chemical and potential energies changes and it is
characterized by reaction rates. A reaction rate defines how fast particular reaction
takes place, i.e. how quickly reactants change into products. Some of the important
quantities used in the characterization of reactions, in general and for the case of
enzymes, are defined in Figure 1.1.
10 CHAPTER 1. THEORETICAL BACKGROUND
E + S
ES
TS
EP
E + P
reactantsreactants
productsproducts
TS
S
P
uncatalyzed chemical reaction enzyme catalyzed chemical reaction
ΔG �
ΔG �
ΔG0
Relative
Energies
R e a c t i o n C o o r d i n a t e
Figure 1.1: Schematic free energy profile for uncatalyzed and enzyme catalyzed chemicalreactions.
For the uncatalyzed reaction, the barrier of the reaction is the relative free energy,
∆G6= of the reactant and the transition state. In the enzyme-catalyzed reaction, the
substrate, S, binds to the enzyme active site, E, and forms an enzyme-substrate
complex, ES. The ES goes through chemical steps to form an enzyme-product
complex, EP , which then releases the product, P , and free the enzyme. In this case
the reaction barrier is relative free energy between the enzyme-substrate complex,
ES and the transition state, TS.
Usually reaction rates are determined experimentally while potential energy sur-
faces and the relative energies between different states can be computed using the
quantum chemistry. A very powerful way to connect these two concepts is in terms
of classical transition state theory (TST)[22, 23, 24, 25]. TST assumes that once a
reaction passes through its reaction barrier it cannot go back again. Its postulates an
equilibrium (Boltzmann) energy distribution at all stable and unstable states along
the reaction coordinates, leading to the following expression for the rate constant k:
k =kBT
hexp
(−∆G6=
RT
), (1.5)
where k is the rate constant (s−1), kB is the Boltzmann’s constant (1.38 x
10−23J/K); h is the Planck’s constant (6.626 x 10−34Js); T is the absolute tempera-
1.4. ENZYME KINETICS 11
ture (298.15 K at room temperature); R is the universal gas constant (8.314JK−1mol−1);
and ∆G 6= corresponds to the Gibbs free energy of activation. The Gibbs free energy
of activation consists enthalpic, and entropic terms, ∆G 6= = ∆H 6= − T∆S 6=. The
rate constant becomes:
k =kBT
hexp
(∆S 6=
R
)exp
(−∆H 6=
RT
), (1.6)
Based on the Eyring equation 1.6, one can for example estimate that a reaction
with a rate constant of 1 s−1 at room temperature corresponds to a barrier of ca
18 kcal/mol. When the energy barrier decreases or increases with ca 1.4 kcal/mol,
the rate constant increases or decreases with one order of magnitude, respectively.
These are quite useful relations that one can use to quickly convert between rates
and energies. It is important to note that, due to the exponential nature of the
relationship, a calculated energy barrier with an error bar of ca 3 kcal/mol, which is
considered to be good, leads to an error in the rate of two orders of magnitude. It
is therefore not possible with the methods used in this thesis to determine accurate
reaction rates. Rather, the computed energies are used as a way to judge the energetic
feasibility of reaction mechanisms.
Considering the exponential dependence of activation energy and the accuracy of
B3LYP functional it is not possible to predict the rate constant accurately from the
computed barriers. However, comparison with experimentally defined rate constants
is suitable in order to evaluate the feasibility of a certain reaction mechanism. In
case of multistep mechanism, the reaction rate is defined by the transition state with
highest activation barrier. This step is a rate-limiting step.
It is also important to point out that the barriers calculated in this work correspond
only to the enthalpy part of equation 1.6. The entropic part is not calculated due to
the coordinate locking scheme used to model enzyme active sites, see Chapter 2.
1.4 Enzyme Kinetics
Enzyme kinetics studies the speed or rate of an enzyme-catalyzed reaction, and
factors that affect reaction rates [26]. Kinetics experiments can be analyzed in terms
of Michaelis–Menten formalism. In this framework, the enzyme action starts with
binding of the substrate, S, to the free enzyme, E, by forming an enzyme-substrate
complex, ES. This complex undergoes chemical transformation (passing through
a transition state, TS) to form enzyme-product complex, EP . The EP releases
the products, P and frees the enzyme for another round of catalysis. Laboratory
procedure traces changes in the concentration of either substrates or products to
determined the reaction rate constant.
The Michaelis–Menten framework gives a mathematical description that links
the reaction rate, kcat or v, and the substrate concentration, [S] using the following
scheme:
12 CHAPTER 1. THEORETICAL BACKGROUND
E + S
k+1
Àk−1
ES
kcat
−→ E + P (1.7)
The terms k+1, k−1 and kcat are rate constants. k+1 describes the association
of substrate and enzyme. k−1 characterizes the dissociation of the enzyme-substrate
complex, ES. kcat characterizes the conversion of ES to product, P . This gives the
following expression for the velocity, v:
v =vmax[S]
KM + [S](1.8)
where [S] is the substrate concentration, vmax is the velocity at maximum con-
centration of substrate when [S] À [E] and all the enzyme molecules are in form
ES, KM is the Michaelis – Menten constant presented as equilibrium constant:
KM =k−1 + kcat
k+1(1.9)
The important parameters from equation 1.8 for the interpretation of the results
are: the catalytic constant kcat, the equilibrium constant, KM , and the specificity
constant, kcatKM
[26].
According to equation 1.8, kcat measures the number of substrate molecules
turned over per enzyme molecule per second. Thus, kcat is sometimes called the
turnover number. This is a first order rate constant that relates the reaction rate to
the concentration of enzyme-substrate complex, [ES].The significance of KM becomes obvious if one considers the case when the rate
of reaction, v, is exactly half of the maximal reaction rate, vmax. Applying this turns
the Michaelis – Menten equation 1.8 in:
vmax
2=
vmax[S]KM + [S]
, KM = [S] (1.10)
KM of an enzyme represents thus the substrate concentration at which the re-
action occurs at half of the maximum rate. KM can be used as an indicator of
the affinity of an enzyme to substrate. Enzymes with a high KM require a higher
substrate concentration to achieve a given reaction velocity.
The ratio kcatKM
is often referred to as ”specificity constant”. It is used for com-
paring the relative rates of enzyme acting on alternative, competing substrates. The
meaning of kcatKM
becomes clear when is assumed that [S] ¿ KM . Then equation 1.8
transforms to:
v ≈ kcat
KM[E][S] (1.11)
1.4. ENZYME KINETICS 13
This is a second-order rate constant that relates the reaction rate to the concen-
tration of free enzyme and free substrate molecules, and can function as an indicator
of catalytic efficiency.
Chapter 2
Modeling of Enzymatic Reactions
Quantum chemical studies of enzymes come with many obstacles. The main problem
is that the systems are too large for a quantum mechanical treatment. As mentioned
previously, DFT methods allow us to treat ca 100 atoms, at a reasonable level of
accuracy, quite routinely with the computer power of today. Although this is a very
big progress compared to a few years ago, it allows us to treat only a very small portion
of the enzyme. How can we then study the reaction mechanism of an enzyme that
consists of thousands of atoms? Obviously, a number of approximations have to be
made. In the last 5-10 years, a quite powerful methodology has been developed that
deals with these issues [27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38]. The basic
idea is to use the DFT methods to treat the critical parts of the active site of the
enzyme and make much cruder approximations about the rest of the enzyme. Here,
this methodology will be briefly outlined. A more detailed discussion and evaluation
of the various aspects of this approach is offered in Paper IV, for the case of the
histone lysine methyltransferase (HKMT).
2.1 Modeling Methodology
An essential part of the theoretical study of an enzyme is the choice of the active site
model. A good model should reflect the chemistry that takes place, i.e. the essential
groups of the active site that take part in the reaction have to be properly accounted
for. However, from a computational point of view it is important to keep the size
of the model relatively small to reduce the computational cost. For every enzyme
studied, these issue have to be taken into consideration.
Figure 2.1 represents the basic steps applied in this thesis for constructing active
site models. The starting point for modeling is usually the X-ray crystal structure of
the investigated enzyme. If such a structure is not available, a homologue structure
15
16 CHAPTER 2. MODELING OF ENZYMATIC REACTIONS
can many times be useful as a starting point. These structures are deposited at
different free databases, one of the most popular being the Protein Data Bank (PDB),
http://www.rcsb.org. Figure 2.1 shows an example of X-ray crystal structure of
GNMT enzyme with highlighted active site.
Since the chemical events that take place at the active site, such as bond making
or breaking, have to be treated with some accurate quantum chemical methods, in
our case DFT, the size of the model has to be limited. Consequently, a part of
the active site is usually cut out from the crystal structure. Based on the available
knowledge about the enzyme and the importance of various parts, a number of
groups are selected and the coordinates are extracted from the pdb-file. The groups
are truncated and hydrogen atoms are added manually. In some cases it is not
obvious which protonation state a certain group holds, and several scenarios have to
be tested.
What does one miss by this procedure? Two obvious effects are missing in this
kind of models of active sites. The enzyme surrounding that is not explicitly included
in the model could provide electrostatic polarization effects that could affect the
calculated energetics, and could also impose steric restraints on the active site groups,
limiting thus their movements in during the reactions. In the methodology adopted
in this thesis, these effects are takes care of in a quite simple way, which has turned
out to be quite powerful and robust.
As for the steric effects, they have to be treated in a different way. The enzyme
matrix around the active site can prevent groups from moving in a certain direction,
or from rotating in a certain way. This could be important for the reactivity and se-
lectivity of the enzyme. The quantum chemical model has to be able to reflect this.
One simple way to mimic these features is by using a coordinate locking scheme.
Some centers, typically where the truncation from the surrounding has been made,
are kept fixed during the geometry optimizations of the model. In metalloenzymes,
this is less of a problem since the metal ions help to keep things in place. However,
groups that are not bound to a metals, or not connected somehow by bonds or hy-
drogen bonds to other groups at the active site, they can move a lot in the geometry
optimization and form thus artificial structures that lead to the wrong description of
the active site. Therefore, the coordinate locking scheme is very helpful to ensure
that the model does not deviate a lot from the experimental structures, yet allowing
for some flexibility of various groups. If the model is too small, there is a risk that
the coordinate locking scheme yields a too rigid active site model, with severe conse-
quences on the calculated energies. However, the error made by this approximation
becomes smaller as the size of the quantum model increases, because the locked
points move further away from the active site.
The combination of continuum solvation and the coordinate locking scheme has
proved to be quite powerful. It has been applied to a wide spectrum of enzymes with
quite successful outcome in elucidating the reaction mechanisms. The calculated
2.1. MODELING METHODOLOGY 17
energies are often been enough to distinguish between different reaction pathways
and to support or dismiss proposed mechanisms.
GNMT enzyme
f
f
f
f
f
f
f
Rest of the enzyme
treated as a continuum solvation model
f
f
f
f
f
ff
GNMT active site
QM active site model
Figure 2.1: Construction of enzyme active site model using GNMT enzyme as an example.First, the coordinates of the X-ray crystal structure are retrieved from databases. The importantgroups of the active site are identified, and a model is constructed. Groups are truncated andhydrogen atoms are added manually. A number of centers are kept fixed during the geometryoptimizations (indicated here by arrows), and the missing enzyme surrounding is modeled by acontinuum solvation model.
18 CHAPTER 2. MODELING OF ENZYMATIC REACTIONS
2.2 Computational Details
All the calculations discussed in this thesis were performed using the hybrid DFT func-
tional B3LYP as implemented in Gaussian03 program package [39]. Unless stated oth-
erwise, all geometry optimizations were performed in gas phase with the 6-31G(d,p)
basis set. As discussed in Section 1.2.1, this double-ζ basis set is good enough to
yield reasonable geometric structures (see Table 1.1), but the errors in the energies
could be large (see Table 1.1). Therefore, to obtain more accurate energies (see
Table 1.2), single point calculations with a large basis set, 6-311+G(2d,2p), were
done.
Zero-point vibrational effect (ZPVE) were obtained by performing frequency cal-
culations on the optimized structures at the same level of theory as used in geometry
optimization. The frequency calculations were also used to confirm the nature of the
stationary points. Minimum have no imaginary frequencies, while transition states
have only one imaginary frequency. Locking the coordinates of some atoms during
the geometry optimizations (see discussion above) can lead to a few small imag-
inary frequencies, typically on the order of 10i - 20i cm−1. These frequencies do
not contribute significantly to the ZPVE and should thus not affect the accuracy of
the energies. However, they render the evaluation of the entropy effects unreliable.
Therefore, entropy was not considered in this thesis.
The solvation effects were added by performing single-point calculations using
the polarizable continuum model (PCM) or some variation of it [40, 41, 42, 43].
Usually, a dielectric constant equal to 4 is used to model the enzyme surrounding.
This value corresponds to an average of ε = 3 for the protein itself and ε = 80 for
the water medium surrounding the protein [44].
Chapter 3
Modeling of Methyl Transfer
Reactions
Methyl transfer, or methylation, represents a simple process of adding or removing a
CH3 group. In chemistry methylation refers to the alkylation process of transferring
one alkyl group from one molecule to another. The alkyl group (methyl group)
transfers as an alkyl carbocation (CH+3 ), a free radical (•CH3), a carbanion (CH−
3 )
or a carbene (••CH2).
Methylation is the most common type of alkylation undergoing in cells and it
has a huge number of effects in the body. The methylation process in cells requires
methyl donor agents. The most widely used methyl donor in the body is S-adenosyl
L-methionine (SAM). So far more than 120 different SAM-dependent methyltrans-
ferases are known to exist in the cell, and each of them catalyzes the synthesis of an
essential product. For example: creatine is important for muscle energy metabolism,
melatonine is the so called sleep hormone, acethylcholine is a neurotransmitter, car-
nitine is valuable for fat burning in mithochondria, choline is essential for fat mobi-
lization and cell membrane fluidity, etc. Hence, irregular functioning of the methyl
transferase enzymes can cause severe diseases, such as brain diseases, mental retar-
dation, epilepsy etc. Understanding the reaction mechanisms of these enzymes can
thus aid in the development of treatment for these diseases.
The present thesis is focused on studying the methyl transfer reaction mechanisms
mainly in the SAM-dependent enzymes glycine N-methyltransferase (GNMT), guani-
dinoacetate methyltransferase (GAMT), phenylethanolamine N-methyltransferase (PN-
MT), and histone lysine methyltransferase (HKMT). The HKMT enzyme has been
studied in order to test the advantages and limitations of the modeling aspects used
in this work to examine the enzyme reaction mechanism. We also did perform a
study on a DNA repair enzyme, O6-methylguanine methyltransferase (MGMT) to
shed more light on the methyl transfer reaction mechanism in this class of enzymes.
19
20 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
GNMT is responsible for the formation of sarcosine by methylation of glycine. So
far the product sarcosine has no known physiological role. Then why does this reaction
occur in cells? A lot of biochemical and structural studies show that the conversion
of glycine to sarcosine, and from sarcosine to glycine, play an important role in
the regulation of methyl group metabolism in the liver and pancreas, by regulating
the ratio between SAM and S-Adenosylhomocysteine (SAH). Furthermore, GNMT is
important for the folate metabolism [45].
In recent years there has been a huge interest in the product of the reaction
catalyzed by the enzyme GAMT – creatine. GAMT catalyzes the final step of the
creatine biosynthesis. Creatine is a natural energy compound used to supply en-
ergy to body muscles [46, 13]. It is produced in the liver, pancreas, and kidneys,
and then transported to the muscles through the bloodstream. Once it reaches the
muscles, it is converted into phosphocreatine (creatine phosphate), which is then
used to regenerate the muscles energy source adenosine triphosphate (ATP) [46].
GAMT deficiency may lead to creatine deficiency, which can cause different mental
diseases. Nowadays, ingesting creatine supplements has become a fashion because
it helps increasing fat-free mass and improves the anaerobic, and possibly the aer-
obic, performance. Creatine is one of the most popular and commonly used sports
supplements available today.
PNMT catalyzes the formation of adrenaline (also referred to as epinephrine)
by N-methylation of noradrenaline (also referred to as norepinephrine) [47, 48]. It
is produced in the cytosol of adrenergic neurons and cells of the adrenal medulla
(chromaffin cells). Adrenaline is a hormone and neurotransmitter that prepares the
body for action in emergency situations. Once its get into the bloodstream it increases
heart rate, blood pressure, boots the supply of oxygen and glucose to the brain and
muscles [49].
The HKMT enzymes are responsible for the regulation of chromatin structure
and control the access of a genomic DNA by transferring a single methyl group from
SAM to amino group in histone peptide [50, 51]. Breakdown this methylation process
can lead to abnormal gene regulation that usually causes cancers [52].
The MGMT enzyme repairs the alkylated DNA by directly removing the alkyl
group from the O6 position of the DNA guanine base [53, 54, 55]. These alkyl
adducts are highly mutagenic and carcinogenic because they results in transition
DNA mutations. Human MGMT is a target in cancer therapy because it repairs
damage induced by anticancer chemotherapies [56, 57].
In this chapter, the most important results of our investigations on these enzymes
will be summarized. For detailed discussions, see Papers I-V in the appendices.
3.1. SAM-DEPENDENT ENZYMES 21
3.1 SAM-dependent Enzymes
A natural methyl donor substance present in the cells and used by the methyl transfer
enzymes is SAM. The role of this compound is to donate a methyl group in a variety
of reactions catalyzed by the enzymes, Figure 3.1 [58].
S
O
CH3
SAM
HO
HO
adenine
COO
NH3+
-
S
O
SAH
HO
HO
adenine
COO
NH3+
-
R
X HB
substrate
enzy
me
R
X CH3
HB
product
enzy
me
Figure 3.1: Generic mechanism for SAM-dependent enzymes.
The substrate molecule enters the active site of the enzyme and binds there by
a number of hydrogen bonds to different protein residues. The nucleophilic entity,
having a lone pair of electrons or possessing partial negative charge, pulls the methyl
group, while the positively charged sulfur atom of SAM attracts electron density from
the methyl group. In this way the methyl transfer reaction can occur. Figure 3.1
shows a general reaction mechanism, in which a base is needed to abstract a proton
from the substrate.
3.1.1 Glycine N-methyltransferase, GNMT (Paper I)
As a SAM-dependent enzyme, GNMT follows the mechanism described in Section
3.1 to transfer a methyl group, see Figure 3.2, [59].
The reaction starts with the binding of SAM and the glycine substrates, in strict
order. In this particular case the glycine has a lone-pair of electrons and binds in such
a way that this lone-pair of the amino nitrogen is directed toward to the CE methyl
carbon of SAM. A single SN2 methyl transfer step occurs from SAM to glycine,
resulting in products SAH and sarcosine. In the case of GNMT, there is no base at
the active site that can abstract a proton, so one has to assume that the glycine
substrate is bound to the enzyme in a deprotonated amine form.
To study the reaction mechanism of GNMT described above we used the recent
X-ray crystal structure, solved in complex with SAM and an acetate molecule at a 2.0
22 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
S
O
CH3
SAM
HO
HO
adenine
COO
NH3+
-
S
O
SAH
HO
HO
adenine
COO
NH3+
-
N
H
HO
O
N
CH3
HO
O
H
glycine sarcosine
Figure 3.2: Reaction catalyzed by GNMT.
A resolution by Takata et al, Figure 3.3 [59]. As described in Chapter 2, one of the
most important parts of enzyme modeling is creating a good model for investigation.
Looking inside the GNMT enzyme it can be seen that a number of hydrogen bonds
are formed between the active site residues and the substrate molecule.
Figure 3.3: X-ray crystal structure of GNMT active site [59].
3.1. SAM-DEPENDENT ENZYMES 23
These bonds are non-covalent, elec-
trostatic interactions and they set up the
substrate molecule for the reaction that
the enzyme catalyzes. The positively char-
ged guanidino group of Arg175 forms a
pair of hydrogen bonds with the carboxy-
late group of the acetate. Other groups
that form hydrogen bonds to the sub-
strate are Tyr33, Asn138 and Gly137.
The largest model used for studying
the SN2 mechanism catalyzed by GNMT
consists of 98 atoms, Figure 3.4. This
model includes the SAM molecule – trun-
cated two carbons away in each direc-
tion from the sulfur center. The glycine
substrate, which was modeled based on
the structure of the acetate, to which an
amino group was added. The side chain
of Arg175 forms strong hydrogen bonds
to the carboxylate of glycine and is hence
essential to bind the substrate and sta-
bilize its charge. The phenol group of
Tyr21 was included to test the proposal
that this group polarizes the S −C bond
of SAM. Parts of Gly137 and Asn138, are
present as these groups are found to form
hydrogen bonds to both the amino and
the carboxylate groups of the glycine sub-
strate. The phenol group of Tyr194, was
included since this group forms hydrogen
bonds to both the glycine substrate and
to Gly137. Hydrogen atoms were added
manually.
The methyl group transfers to the gly-
cine molecule in a single SN2 reaction
that involves the displacement of the leav-
ing group (SAH), by the nucleophile (gly-
cine). The structure of the optimized
transition state is displayed in Figure 3.4B.
At the transition state the bond to the
glycine is partially formed while the bond
f
f
f
f
f
f
f
A
f
f
f
f
f
ff
B
f
f
f
f
f
f
f
C
Figure 3.4: Optimized reactant (A), transi-tion state (B), and product (C) structures ofthe largest model of GNMT active site. Ar-rows indicate atoms that are fixed during thegeometry optimizations.
24 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
to the SAM is partially broken. The critical SD−CE and CE−N bond distances
are 2.28 and 2.18 A, respectively. The barrier for the methyl transfer in GNMT is
calculated to be 15.0 kcal/mol. No experimental data for a rate constant is available,
but an energy barrier of 15.0 kcal/mol for enzyme reaction is considered energetically
feasible, see Section 1.3.
The reaction was found to be exergonic by 14.1 kcal/mol. The optimized structure
of the product is shown in Figure 3.4C.
f
f
f
f
f
f
f
f
f
f
f
f
f
f
f
f
f
A B C D
f
f
f
f
f
f
f
f
f
f
f
f
f
f
f
E F G
Figure 3.5: Optimized transition-state structures for models A-G.
We have investigated the role of the various active site residues on the energetics
by creating different active site models. The smallest model is composed of glycine,
a truncated model of SAM, and truncated residue Arg175, Figure 3.5A. The roles
of the other amino acid residues were tested by adding the residues to the small
3.1. SAM-DEPENDENT ENZYMES 25
Model A one at the time, in order to isolate the contribution of each group. The
optimized transition state structures are shown in Figure 3.5, and the energetic results
are summarized in Table 3.1.
model parts included barrier reaction energy
A SAM + Glycine + Arg175 11.2 -20.1
B A + Tyr21 13.5 -16.8
C A + Tyr21 + His 142 11.4 -21.2
D A + Gly137 9.9 -24.8
E A + Asn 138 15.1 -12.6
F A + Tyr194 17.5 -14.5
G A + Tyr194 + Gly137 10.5 -16.6
largest A + Tyr21 + Gly137 + Asn 138 + Tyr194 15.0 -14.1
Table 3.1: Calculated Barriers and Reaction Energies (kcal/mol) for the different models.
Model A has a barrier of 11.2 kcal/mol and is exothermic by 20.1 kcal/mol.
Adding the phenol group of Tyr21 to this model (Figure 3.5B) results in a slight
increase in the barrier, to 13.5 kcal/mol, and a decrease in the exothermicity, to
16.8 kcal/mol. When both Tyr21 and the imidazole ring of His142 are added, Figure
3.5C, the barrier is found to be almost identical to Model A, 11.4 kcal/mol. These
results speak against the suggestion that Tyr21 polarizes the SD−CE bond to cause
a decrease of the barrier [59]. As also seen for the largest model discussed above,
Figure 3.4, the phenolic proton of Tyr21 was found to point away from SAM, despite
attempts to make it to point toward the sulfur center.
Adding the peptide bond of Gly137, which forms a hydrogen bond to the amino
group of the substrate (Figure 3.5D), results in a decrease of the barrier by 1.2
kcal/mol to 9.9 kcal/mol. The hydrogen bond to the carbonyl moiety of Gly137
makes the nitrogen center of the substrate slightly more negative, which would make
the transfer of the positively charged methyl group slightly easier.
On the other hand, adding the side chain of Asn138 (Figure 3.5E), which forms
26 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
a hydrogen bond to the carboxylate moiety of the substrate, and hence makes the
substrate slightly less negative, leads to a higher barrier, calculated to 15.1 kcal/mol.
Also, adding the Tyr194 residue to Model A leads to a dramatic increase in the
barrier, from 11.2 to 17.5 kcal/mol (Figure 3.5F). This is easily explained if we
note that the phenolic proton forms a hydrogen bond to the nitrogen atom of the
substrate in the reactant species. This hydrogen bond will be lost when the methyl is
transferred to the nitrogen, resulting in the barrier raise. On the other hand, if both
Tyr194 and the peptide bond of Gly137 are added at the same time (Figure 3.5G),
the tyrosine will form a hydrogen bond to the carbonyl of the glycine instead and the
barrier is lowered to 10.5 kcal/mol.
Hence, the calculations on the GNMT enzyme have confirmed that the reaction
takes place in an SN2 fashion. Furthermore, by adding or eliminating various groups
at the active site, we showed that hydrogen bonds to the amino group of the sub-
strate lower the reaction barrier, whereas hydrogen bonds to carboxylate group of the
substrate raise the barrier.
3.1.2 Guanidinoacetate Methyltransferase, GAMT (Paper II)
The GAMT enzyme catalyzes the transfer of a methyl group from SAM to guani-
dinoacetate (GAA), resulting in the formation of creatine and SAH, see Figure 3.6
[60, 61, 62, 63, 64, 65]. The way how this reaction occurs is similar to the generic
mechanism described in Section 3.1.
S
O
CH3
SAM
HO
HO
adenine
COO
NH3+
-
S
O
SAH
HO
HO
adenine
COO
NH3+
-
NE
H
O
O
N
H
HH2N
OD1
O
NE
O
O
N
H
HH2N
OD1
O
CEH3
H
Asp134Asp134
guanidinoacetate creatine
Figure 3.6: Reaction mechanism of the GAMT enzyme.
3.1. SAM-DEPENDENT ENZYMES 27
Figure 3.7: X-ray crystal structure of the GAMT active site [66].
With information about the structure of GAMT, crystallized with SAH and GAA,
Figure 3.7 [66], the reaction mechanism was theoretically investigated.
The substrate and the cofactor are attached to the active site by a number of
hydrogen bonds. Glu45 and Asp134 form hydrogen bonds with the guanidino group
of GAA, while the amide groups of Leu170 and Thr171 form hydrogen bonds with the
carboxylate group of GAA. These hydrogen bonds facilitate the orientation of GAA in
GAMT. In the GAMT:(SAH+GAA) structure, the distance between the sulfur center
of SAH (SD) and the NE of GAA is found to be 3.9 A.
The model system used to reproduce the GAMT active site and to elucidate the
reaction mechanism consisted of 92 atoms. This model includes part of the cofactor
SAM built on the basis of the SAH structure by adding a methyl group to the sulfur
atom. Furthermore, SAM was truncated in both directions relative to the sulfur
center – at the adenine group on one side and three carbons away on the other
side. This is sufficient to model the properties of the SD − CE bond and to grant
flexibility to the SAM-model. The GAA substrate molecule was included in the model
without any changes. Five amino acids were furthermore included in the model –
Glu45, Asp134, Thr135, Leu170, and Thr171 (Figure 3.7). The Glu45 group forms
strong hydrogen bonds to the GAA guanidino group, and is hence essential to bind
the substrate and stabilize the positive charge of the guanidino group. Asp134 and
Thr135 are found to form hydrogen bonds to the guanidino group of GAA. Leu170
and Thr171 form hydrogen bonds with the carboxylate group of GAA.
28 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
The optimized structure of the reac-
tant (Figure 3.8A) shows high resemblance
to the crystal structure. The hydrogen
bonding network around the substrate mo-
lecule orients it in such a way that there
is nearly a straight line between the sul-
fur center of SAM (SD) and the nitrogen
atom of GAA (NE). The distance be-
tween NE(GAA) −CE(SAM) is calculated
to be 2.98 A, and the angle SD −CE −NE is found to be 167.7◦. The distance
OD1(Asp134) − HNE(GAA) is also impor-
tant, and it is calculated to be 1.86 A in
the optimized reactant structure. The
guanidino group of GAA is planar and the
positive charge is delocalized over three
nitrogen atoms.
It was previously proposed that the
reaction mechanism starts with a proton
transfer from NE of GAA to the Asp134
base upon which the methyl group is trans-
ferred from SAM to the deprotonated GAA
[66].
To check this hypothesis, a linear tran-
sit scan of moving the proton from NE(GAA)
of the substrate to Asp134 was performed.
The HGAA − OD1(Asp134) distance was
kept fixed in steps between 1.86 A and
1.00 A, while all other degrees of free-
dom were optimized. As seen from Figure
3.9A the energy increases monotonously,
and no energy minimum could be found
corresponding to an intermediate where
the proton is transferred to the Asp134.
Transferring a proton from GAA costs in
gas phase 15.9 kcal/mol, while applying
protein environment in the form of a ho-
mogeneous medium yields a slightly lower
value of 12.6 kcal/mol. This results speaks
against the stepwise mechanism, where
the proton is transferred first.
f
f
f
f
f
f
A
f
f
f
f
f
f
B
f
f
f
f
f
f
C
Figure 3.8: Optimized reactant (A), transi-tion state (B), and product (C) structures ofthe GAMT active site model.
3.1. SAM-DEPENDENT ENZYMES 29
Instead, a linear transit scan to move the methyl group from SAM to GAA was
used to find the transition state structure. The CE −NE distance was kept fixed in
steps starting from 2.98 A, which is the distance in the reactant, to 1.46 A, which is
the distance in the product, see Figure 3.9B. As the methyl approaches the nitrogen
center, the energy increases up to a distance of 2.2 - 2.0 A, after which it starts to
drop. The NE proton moves toward Asp134 and the nitrogen center becomes more
pyramidal. At a CE −NE distance of 1.8 A the proton has transferred completely.
2.0 1.8 1.6 1.4 1.2 1.0-2
0
2
4
6
8
10
12
14
16
18
Rel
ativ
e E
nerg
y [k
cal/m
ol]
H - O distance [Å]
gas phase incl. solvation
3.2 3.0 2.8 2.6 2.4 2.2 2.0 1.8 1.6 1.4-40-36-32-28-24-20-16-12-8-4048
12162024
[1.02]
[1.01]
[1.02]
[1.47][1.79]
[1.84]
[1.86][1.86]
Rel
ativ
e en
ergy
[kca
l/mol
]
C - N distance [Å]
gas phase incl. solvation
A B
Figure 3.9: Linear transit scans for: A) moving the proton from NE of the substrate to OD1
of Asp134, and B) moving the methyl group from SAM to NE of GAA substrate, the H −OD1
distance is given in brackets. Solvation is added using ε = 4.
Based on this, the fully optimized transition state for this reaction was located,
see Figure 3.8B. At the transition state the critical SD−CE and CE −NE distances
are calculated to be 2.29 A and 2.16 A, respectively. The other two important
distances NE − H and H − OD1 in the transition state structure are found to be
1.05 A and 1.73 A, respectively. The nature of the transition state was confirmed
to have one imaginary frequency of -446i cm−1. It shows that the methyl and the
proton transfer take place in one concerted asynchronous step. The activation barrier
for this process was calculated to be 14.9 kcal/mol in gas phase, and 19.7 kcal/mol
in protein environment with ε = 4. This energy barrier is in good agreement with the
measured rate constant of 3.8 ± 0.2 min−1, which corresponds to ca 19 kcal/mol.
The reaction was calculated to be exergonic by as much as 36.2 kcal/mol in gas
phase and 24.0 kcal/mol using ε = 4.
The optimized structure of the product is displayed in Figure 3.8C. The protonated
Asp134 rotates in such a way that the hydrogen bond to the guanidino group of GAA
is broken, while a hydrogen bond to acetate of GAA is formed.
We tried to optimize the intermediate structure in which the substrate is methy-
lated, but the proton is not transferred to Asp134, i.e. corresponding to a stepwise
30 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
mechanism, but this was not possible. We tried also to optimize the intermediate in
which the proton is transferred to Asp134 before the methyl transfer, but it was not
possible either.
The calculations on the GAMT enzyme show thus that the methyl transfer from
SAM to the substrate takes place in a concerted asynchronous step with the proton
transfer from the substrate to the Asp134 base.
3.1.3 Phenylethanolamine N-methyltransferase, PNMT (Paper III)
The PNMT enzyme catalyzes the transfer of a methyl group from SAM to nora-
drenaline, resulting in the formation of adrenaline and SAH (Figure 3.10) following
the generic mechanism described in Section 3.1:
S
O
CH3
SAM
HO
HO
adenine
COO
NH3+
-
S
O
SAH
HO
HO
adenine
COO
NH3+
-
N
H
HHO
OH
OH
N
CH3
HHO
OH
OH
noradrenaline adrenaline
Figure 3.10: Reaction catalyzed by PNMT.
Several X-ray crystal structures of PNMT with various substrates and inhibitors
have been solved [67, 68, 69, 70]. To investigate the reaction mechanism, several
quantum chemical models of the PNMT active site were constructed on the basis
of the X-ray crystal structure (PDB code: 2AN4, Figure 3.11), which is in complex
with SAH and a methyl acceptor substrate, p-octopamine [69].
Model A is composed only of parts of cofactor and substrate molecule, as shown
in Figure 3.12. It is assumed that the proton of the amino group of the substrate is
lost, so the substrate was modeled in its neutral form. In the optimized structure of
the reactant, Figure 3.12A, it is found that the important SD−C and C−N distances
are 1.83 A and 3.40 A, respectively. Those distances are very similar to the distances
3.1. SAM-DEPENDENT ENZYMES 31
Figure 3.11: X-ray crystal structure of PNMT active site [69].
in GNMT enzyme where the methyl group is transferred from the same cofactor to
the amino group of the glycine residue. The transition state for the methyl transfer
was located (Figure 3.12B) and the distances SD − C and C − N are 2.36 A and
2.19 A, which are also quite similar to the distances found for GNMT (2.38 A and
2.20 A). The barrier for this transfer was calculated to be 16.7 kcal/mol which is
about 2 - 4 kcal/mol too low compared to the experimental rate constant of kPEAcat =
2.6 ± 0.1 min−1, corresponding to ca 19 -20 kcal/mol (PEA - phenyletanolamine ).
The reaction is calculated to be exothermic by 5.7 kcal/mol (13.2 and 15.8 kcal/mol
including solvation with ε = 4 and ε = 80, respectively).
A somewhat larger model consisting of 48 atoms was also used, Figure 3.13.
This model includes the important Glu185 and Glu219 (represented by acetates)
and a water molecule (W), which initially was bridging the substrate and the Glu185
residue. Both glutamate residues were initially modeled in the deprotonated form, and
the substrate was modeled as a cation, i.e. in its protonated state. The total charge
of the model is thus 0. In the optimization, a proton moved spontaneously from
the amino group of the substrate to Glu185, through the bridging water molecule,
32 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
O
N
C
C
C
C
C
C
S
C
3.40
1.83
SAM
substrate
molecule
N
O
C
CC
C
C
C
S
C
2.19
2.36
SAM/SAH
substrate
molecule
C
CN
C
O
S
CC
C
C
1.50
3.82
SAH
product
molecule
A B C
Figure 3.12: Optimized reactant (A),transition state (B), and product (C) structures of ModelA of PNMT.
which subsequently moved out to bridge the two glutamate groups instead, see Figure
3.13A. Effectively, in this model the substrate is in the neutral form and one of the
glutamates protonated.
Using this reactant structure, a transition state for the methyl transfer was lo-
cated, Figure 3.13B. The transition state is characterized by an imaginary frequency
of -359i cm−1, and the SD−C and C−N distances are calculated to be 2.22 A and
2.40 A, respectively. This model has a calculated barrier of 6.0 kcal/mol without
solvation effects, which increases to 11.8 and 13.7 kcal/mol, using ε = 4 and ε = 80,
respectively. These values are significantly lower than the ones found for Model A.
It is easy to rationalize this result if one recognizes that in the reaction a methyl
cation is transferred to the substrate. In the case of Model B a charged glutamate
is present that can stabilize the product better than the substrate, lowering thus the
barrier for the reaction.
This fact is more evident from the calculated exothermicity of the reaction.
Model B yields a reaction energy of -48.3 kcal/mol, which however is considerably
decreased to -28.5 and -21.8 kcal/mol when ε = 4 and ε = 80 are used, respectively.
For this model, we also found another product structure in which a proton from
the substrate nitrogen has transferred to the Glu219 residue (Figure 3.13D). This
structure has a very similar energy compared to the other product structure (-47.9,
-27.7 and -20.9 kcal/mol, without solvation, and with ε = 4 and ε = 80 respectively).
The biggest PNMT active site model used in the investigations consists of 93
atoms. In addition to Model B, this model contains parts of the amino acids Asp267,
Arg44, and Asn39, see Figure 3.14. A second water molecule (W2) observed in the
crystal structure is also included. Also, the full phenylethanolamine substrate was
used and one more carbon was kept in the SAM to grant even more flexibility to
the model. The total charge of the model is thus 0. As in the case of Model B,
3.1. SAM-DEPENDENT ENZYMES 33
O
C
C
C
C
C
C
O
N
C
O
O
S
C O
C
C
C
2.84
1.57
2.91
1.45
1.86
1.78
SAM
Glu185
Glu219
substrate
molecule
OW
O
C
C
C
C
C
C
O
N
O
C
O
S
CO
C
C
C
2.85
1.57
2.40
1.46
2.22
1.75
SAM/SAH
Glu185
Glu219
substrate
molecule
WO
A B
C
O
C
C
C
C
CO
O
O
S
O
O
C
N
C
C
C
C
1.71
1.99 Å1.741.14
1.40
1.48
SAH
Glu185
Glu219
W
productmolecule
O
C
C
C
C
C
C
O
O
O
S
OC
C
C
C
1.65
1.821.62
1.53
1.49
SAH
Glu185
Glu219
methylated
substrate
molecule
W
N
C
O
C D
Figure 3.13: Optimized reactant (A),transition state (B), and products (C, and D) structuresof Model B of PNMT.
the substrate was initially modeled in the protonated form and the two glutamates
(Glu185 and Glu219) were anionic. Also here, during the geometry optimization of
the reactant a proton was transferred spontaneously from the amino group of the
substrate to the Glu185 residue through a water bridge (Figure 3.14A).
The transition state for the methyl transfer in Model C was located and is also
shown in Figure 3.14B. The SD − C and C − N distances are calculated to be
2.11 Aand 2.35 A, respectively. We find that the transition state now is even earlier
than in Model B, and the calculated reaction barrier is 4.8 kcal/mol, which increases
to 8.8 and 9.9 kcal/mol when ε = 4 and ε = 80 are used, respectively.
34 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
CC
C
O
C
C
CC
O
C
O
C
C
CC
N
O
C
C
N
N
O
O
O
O
O
C
N
C
C
C
C
C
O
C
N
C
S
C
C
1.47
1.61
2.02
1.53
1.80
1.70
2.93
SAM
Glu185
Glu219
Asp267
Arg44
W1
W2
phenylethanolamine
Asn39
A
CC
O
CC
C
CC
O
C
O
C
C
C
C
C
O
NC
O
N
O
O
N
C
N
C
O
O
CC
C
C
C
O
N
C
SC
C
1.63
1.46
2.01
1.82
1.50
2.35
1.69
2.11
SAM/SAH
phenylethanolamineGlu185
Glu219
Asp267
Arg44
W1W2
Asn39
B
CC
C
O
C
C
C
CC
O
C
O
O
C
C
C
O
O
C
N
N
N
C
O
N
O
O
O
C
C
C
C
C
N
C
C
C
S
C
C
1.92
1.84
1.75
1.00
1.66
1.56
1.74
SAH
methylated
phenylethanolamine
Glu185
Glu219
Asp267
Arg44
W1
W2
Asn39
C
Figure 3.14: Optimized reactant(A),transition state (B), and product(C) structures of Model C of PNMT.
C
C
C
C
O
C
CC
C
C
N
C
O
N
N
O
C
C
C C
O
O
C
C
C
O
O
O
O
C
C
N
C
O
N
C
C
S
C
C
2.75
1.71
1.70
1.841.70
1.56
1.59
1.83
3.18
2.19
SAM
phenylethanolamine
Glu185
Glu219
Asp267
Arg44
Asn39
W1 W2
A
C
C
C
O
C
C
C
C
C
O
C
C
C
NO
C C
C
N
O
N
O
O
OO
O
O
C
N
C
C
N
C
C
C
CC
SC
C
2.51
1.73
1.71
1.65 1.86
1.60
1.54
2.24
2.20
2.22
SAM/SAH
phenylethanolamine
Glu185
Glu219
Asp267
Arg44
Asn39
W1
W2
B
C
C
C
C
O
C
CC
C
C
C
O
N
C
C
O
C
N
N
C
O
O
O
O
O
O
C
O
N
C
C
N
C
C
C
CC
SC
C
2.39
1.64
1.93
1.65 1.621.92
1.55
1.82
1.49
SAH
methylated
phenylethanolamine
Glu185
Glu219
Asp267
Arg44
Asn39
W1
W2
C
Figure 3.15: Optimized reactant(A),transition state (B), and product(C) structures of Model C(H+) of PNMT.
3.1. SAM-DEPENDENT ENZYMES 35
The reaction is exothermic by 47.9 kcal/mol, which, as in the case of Model B,
is considerably decreased to -29.5 and -28.3 kcal/mol, when ε = 4 and ε = 80 are
used, respectively. We find that simultaneously with the methyl transfer, a proton is
transferred from the substrate to Glu219, Figure 3.14C.
Since inclusion of the negatively-charged Glu185 and Glu219 lowered the barrier
in both Model B (Figure 3.13) and Model C (Figure 3.14), we tried the possibility
of protonating one with them and added thus one proton to Glu219. The model,
called Model C(H+), now has a total charge of +1. The optimized geometries of the
reactant, transition state, and product are shown in Figure 3.15. The substrate was
also here modeled on its protonated state but a proton was spontaneously moved to
the Glu185 during the geometry optimization. The reaction barrier was calculated to
be 13.6 kcal/mol (15.8 and 16.4 kcal/mol including ε = 4 and ε = 80 respectively).
These values are in much better agreement with the experimental rate compared to
the Model B and Model C, in which only one of the glutamates is protonated.
These results suggest thus that either the two active site glutamates are in the
protonated form prior to the binding of a neutral substrate, or that one of them is
protonated and the other receives a proton from a charged substrate.
3.1.4 Histone Lysine Methyltransferase, HKMT (Paper IV)
HKMT catalyzes the transfer of a methyl group from SAM to the amino group
of the lysine of the histone peptide (Figure 3.16) [50, 51, 52]. By doing this it
is responsible for the regulation of chromatin structure and control the access of a
genomic DNA. The reaction mechanism of this enzyme is well-understood and follows
the same pattern as for the precious cases discussed in this thesis (GNMT, GAMT,
and PNMT). In particular, there has been a QM/MM study by Hu and Zhang that has
addressed the details of the mechanism HKMT [71, 72]. Instead, we chose to study
it as a test case to evaluate the quantum chemical methodology used throughout the
thesis.
Several models of HKMT were constructed based on the X-ray crystal structure
(PDB code: 1O9S), which is in complex with SAH and a methylated histone peptide
[73]. First we designed a very small model of the enzyme active site, consisting of
the truncated SAM and substrate molecules as shown in Figure 3.17A. This model,
called Model A, consists of 29 atoms and has a total charge of +1. The next
model, called Model B consists of 46 atoms and includes larger parts of SAM and
substrate molecules as shown in Figure 3.17B. Model C consists of 72 atoms and
includes two tyrosine molecules, Tyr245 and Tyr305, forming hydrogen bonds with
the amino group of the substrate, Figure 3.17C, while the largest model, Model D,
consists of 132 atoms and includes additional parts of the active site, such as Tyr335,
Asn265, Thr266, Lys294, Ala295, and an active site water molecule, Figure 3.17D.
For everyone of these models, several dielectric constants were used to investigate
the influence of the choice of this parameter on the calculated energies. The results
36 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
S
O
CH3
SAM
HO
HO
adenine
COO
NH3+
-
S
O
SAH
HO
HO
adenine
COO
NH3+
-
N
H
H
CH3
NH
H
Lys4(Histone Peptide)
Methylated Lys4(Histone Peptide)
Figure 3.16: Reaction catalyzed by HKMT.
are listed in Table 3.2.
Let us first look at some geometrical parameters. For Model A (Figure 3.17A), at
the transition state the critical distances Sδ−CE , CE−Nζ , and the angle Sδ−CE−Nζ
of this SN2 reaction are 2.44 A, 2.18 A, and 175.3◦ respectively. In the slightly
larger Model B those parameters are 2.41, 2.16, and 177.1◦ respectively, Figure
3.17B. The transition state for Model C is characterized by Sδ −CE , CE −Nζ , and
Sδ−CE−Nζ parameters of 2.31 A, 2.26 A, and 173.8◦, respectively (Figure 3.17C).
Very similar distances and angle are obtained for the largest model, Model D (Figure
3.17D), 2.37 A, 2.23 A, and 175.6◦, respectively. The transition state geometries are
thus internally consistent among the various models. Also, quite importantly, those
parameters agree quite well with the available QM/MM geometrical data for HKMT
enzymes [71, 72], which are 2.32 ± 0.02 A, 2.30 ± 0.02 A, and 173.3◦± 1.4◦
respectively for the Sδ − CE , CE −Nζ , and Sδ − CE −Nζ .
The solvation corrections (i.e. the effects of the applied dielectric medium on the
barrier and reaction energy) are quite different for the different models, see Table 3.2.
For the calculated barriers, the solvation effects of all models are quite small, less
than 3 kcal/mol in going from gas phase to the largest dielectric constant (ε = 80).
However, for the reaction energies, the solvation effects are quite large for Models A
and B, while for Models C and D they are much smaller. When the model size
increases, and more and more groups are added, it is thus seen that the effects of
the solvation model are diminished. In this case, more groups are added around
the substrate, which in the reactant is neutral, but in the product is methylated
3.1. SAM-DEPENDENT ENZYMES 37
C
C
C
CS
C
C
N
C
2.44
2.18
ii
i
C
C
C
OC
N
C
C
C
C
C
C
S
C
C
C
2.16
2.41
i
i
i
A B
C
C
C
C
O
C
C
C
C
C
CO
C
N
C
C
C
C
S
C
C
C
C
C
C
C
C
C
O
C
2.23
2.26
2.31
2.06
i
i
i
i
i
CC
C
C
C
C
CC
C
C
O
C
C
O
C
C
C
CC
O
C
C
C
C
C
O
N
C
C
O
NC
O
C
O
O
C
S
C
C
C
C
C
C
O
C
O
C C
N
N
C
CN
C
C
C
C
O
O
C
C
2.44 2.02
2.23
2.37
2.29
2.12
2.29
1.75
Tyr335
Asn265
Thr266
Ty305
H2O
Lys294
Ala295
Tyr245
i
i
i
i
i
i
i
i
i
i
C D
Figure 3.17: Optimized transition states for methyl transfer reaction in the different HKMTactive site models. A) ModelA, B) ModelB, C) ModelC, and D) ModelD.
(cation). Proper description of the active site around this area is quite important to
obtain stable results. As seen for Model D the barrier is essentiallt unaffected by the
solvation, while the reaction energy changes by up to 3 kcal/mol. As a consequence
of this, one can conclude that the particular choice of the dielectric constant becomes
less critical as the model size increases. This is a very important result to realize for
future quantum chemical studies of enzyme active sites.
In this study (Paper IV), we also investigated the validity of some other technical
38 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
Model A Model B Model C Model D29 atoms 46 atoms 72 atoms 132 atoms
barrier reaction barrier reaction barrier reaction barrier reaction
energy energy energy energy
no solvation 18.8 -2.9 21.7 +0.5 15.4 -16.7 18.9 -9.2
ε = 2 18.2 -9.3 20.6 -7.2 16.7 -17.5 19.0 -10.6
ε = 4 17.8 -12.7 19.9 -11.4 17.3 -18.0 19.1 -11.5
ε = 8 17.6 -14.5 19.5 -13.6 17.6 -18.3 19.1 -12.0
ε = 16 17.4 -15.4 19.3 -14.8 17.8 -18.5 19.1 -12.2
ε = 80 17.3 -16.1 19.1 -15.7 17.9 -18.6 19.1 -12.4
Table 3.2: Calculated energetics for the various active site models of HKMT enzyme.
assumptions. For example, it is a very common procedure to first optimize the
geometries in the gas phase and then perform a single-point solvation correction.
To check the validity of this strategy, we performed the geometry optimization for
Model A in PCM with ε = 4 and ε = 80 with two different basis sets (6-31G(d,p),
and 6-311+G(2d,2p)). As seen from Table 3.3, optimization in PCM gives barriers
and reaction energies that are almost the same as the ones calculated with the above-
mentioned methodology. The differences are smaller than 1 kilocalory/mol. It can
thus be concluded that it is a safe procedure, at least for this kind of reactions, to
first optimize the geometry in the gas phase and then apply solvation.
6-311+G(2d,2p) 6-31G(d,p)gas phase PCM gas phase PCM
barrier reaction barrier reaction barrier reaction barrier reaction
energy energy energy energy
no solvation 18.8 -2.9 16.3 -6.0
ε = 4 17.8 -12.7 18.4 -12.5 15.4 -15.8 15.9 -15.8
ε = 80 17.3 -16.1 18.2 -16.0 14.8 -19.2 15.6 -19.6
Table 3.3: Calculated energetics for ModelA using different geometry optimization schemes.
It is commonly argued that the B3LYP functional underestimates reaction barriers
[71, 72]. Therefore, we decided to study the reaction of Models A using different
methods, such as BLYP, MPW1K [74, 75], and MP2, following the same strategy
described above. The results are presented in Table 3.4. For BLYP, which has
no Hartree-Fock exchange, the barrier lowered by around 5 kcal/mol compared to
3.2. OTHER METHYL TRANSFER ENZYMES 39
the B3LYP, while using MPW1K, which includes more Hartree-Fock exchange than
B3LYP, the barrier is raised by 7-8 kcal/mol. MP2 gives barriers that are around 10
kcal/mol higher compared to B3LYP. Clearly, the BLYP barrier is too low compared
to experiments (ca 13 vs. 20.9 kcal/mol), and the MPW1K and MP2 are too high
(ca 26 and 27 vs. 20.9 kcal/mol). B3LYP seems to yield the good agreement with
the experimental data, which is also consistent with the results of our previous studies
on the other SAM-dependent methyl transfer enzymes.
B3LYP BLYP MPW1K MP2barrier reaction barrier reaction barrier reaction barrier reaction
energy energy energy energy
no solvation 18.8 -2.9 13.5 -1.7 26.4 -3.3 28.3 -6.4
ε = 4 17.8 -12.7 12.8 -11.5 25.7 -13.2 27.0 -16.3
ε = 80 17.3 -16.1 12.4 -14.9 25.3 -16.6 26.4 -19.7
Table 3.4: Calculated energetics for ModelA using different methods. Experimental barrier is20.9kcal/mol. Barriers calculated by Zhang et al with QM/MM are: 14.6±1.7 (B3LYP/6-31G*/ MM), 21.9± 1.9 (MP2/6-31G* / MM), and 21.5± 1.9 (MP2/6-31+G* / MM) [71, 72].
In conclusion, the results of Paper IV, provide confidence in the methodology
used in this thesis for studying enzyme reactions. Both the size of the model, and
the adopted technical details seem reliable enough to provide useful information about
the reaction mechanism of the considered enzymes.
3.2 Other Methyl Transfer Enzymes
3.2.1 O6-Methylguanine Methyltransferase, MGMT (Paper V)
MGMT, also called alkylguanine alkyltransferase (AGT) repairs the alkylated DNA
by directly removing the alkyl group from the O6 position of the guanine. It works
by transferring the alkyl lesion to an active site cysteine residue (Cys145 in human
MGMT) in an irreversible stoichiometric suicide reaction, Figure 3.18.
To study the repair reaction mechanism of MGMT we have used a recent X-ray
crystal structure (PDB code: 1T38) of the protein in complex with DNA [76], Figure
3.19, in which the critical Cys145 is mutated into a serine residue. The hydroxyl group
of the Cys145Ser moiety is hydrogen bonded to a Water-His146-Glu172 triad that
is believed to be a charge-relay system to shuttle the proton of the thiol. Several
interactions help binding the flipped-out O6-methylguanine base. Tyr114 forms a
hydrogen bond to the N3 atom and the peptide bond of Ser159 forms a hydrogen
bond to the O6 atom.
Based on the information from X-ray crystal structure we have created an active
40 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
CO
O
Cys145
S
His146
N
N
N
N N
N
O
H2N
DNA
H3C
H
O
H
H
H
O6-methylguanine
Cys145
S
His146
N
N
N
N N
N
O
H2N
DNA
H3C
H
O
H
H
H
O6-methylguanine
Glu172
Cys145
S
His146
N
N
N
NN
N
O
H2N
DNA
CH3
H
O
H
H
H
guanine
CO
O
Glu172
CO
O
Glu172
Figure 3.18: Reaction mechanism of the MGMT enzyme.
Figure 3.19: X-ray crystal structure of the MGMT active site.
3.2. OTHER METHYL TRANSFER ENZYMES 41
site model consisting of 74 atoms, see Figure 3.20. In the model, the Ser145 residue
was modified to cysteine by changing the oxygen atom from hydroxyl group of serine
to a sulfur atom. The methylated guanine base, which in the crystal structure is
already flipped out from the DNA strand and positioned in the enzyme active site, is
cut on the place where the ribose start. The glycosidic bond is kept in the chemical
model as an N − CH3 bond. Glu172 from the H2O −His146−Glu172 hydrogen
bonding network is modeled as an acetate molecule. Tyr114 (modeled as a phenol
group) was also included in the model. This residue makes a hydrogen bond with
N3 of the guanine base and it is proposed to assist the reaction by stabilizing the
negative charge developing at the guanine.
The optimized structure of the reactant species shows high resemblance to the
crystal structure, Figure 3.20. The hydrogen bonding network connecting the thiol
group of Cys145, the ordered water molecule, His146, and Glu172 is well reproduced.
Because Glu172 is in the deprotonated form, the hydrogen bond to the histidine is
rather short, with O −H and H −N distances of 1.41 A and 1.14 A, respectively.
The substrate NH2 is forming a hydrogen bond to the carbonyl oxygen of the peptide
bond connecting Cys145 and His146. The methyl lesion is pointing toward the sulfur
atom of the Cys145 residue, with a C − S distance of 4.07 A.
C
C
C
C
C
S
C
C
O
C
O
C
N
C
C
O
C
C
N
N
N
C
O
C
C
C
N
O
C
NC
C
O
C
C
N
C
N
C
4.072.37
1.411.85
1.81
1.77
Glu172
His146
Cys145
Tyr114
Methylguanine
f
f
f
f
Figure 3.20: Optimized reactant structure of the MGMT active site model.
Starting from this structure a transition state for proton transfer from Cys145
through the Glu172-His146-water network was located, Figure 3.21A. At the transi-
tion state, the critical SCys145−H and H−Owater distances are 1.57 A and 1.32 A,
respectively, and the Owater−H and H−NHis146 distances are 1.22 A and 1.28 A, re-
spectively. We noticed also that a proton has transferred from His146 to the Glu172,
in a charge relay fashion. This could be an artefact of the model used in the current
study, since in our calculations, we have completely neglected the surrounding of
the Glu172 residue, which makes the ion pair (protonated His146 and deprotonated
42 CHAPTER 3. MODELING OF METHYL TRANSFER REACTIONS
Glu172) relatively unfavored, leading therefore to the proton transfer. Whether the
proton transfers to Glu172 or not, it does not affect the energetics of the reaction
significantly.
The activation barrier for the first step is calculated to be 9.3 kcal/mol, and the
product of this step, i.e. the thiolate intermediate, is found to have an energy of
+5.9 kcal/mol above the reactant structure. The optimized structure is shown in
Figure 3.21B.
The barrier is not affected when the surrounding is included as a homogenous
dielectric medium with ε = 4. The energy of the intermediate, however, is lowered
somewhat, from +5.9 to +2.0 kcal/mol, relative to the reactant. This is a result
of the fact that the Cys145 its charge state from neutral to anionic it is located at
the edge of the quantum chemical model. Solvation will thus stabilize the charged
species more than the neutral one.
At the first step a thiolate ion is generated and Cys145 and can act as a nuclephile
in the dealkylation reaction. In the second step the methyl group is transferred from
O6-methylguanine to Cys145 residue of the enzyme active site.
The unconstrained transition state of methyl group transfer was located and the
optimized structure is shown on the Figure 3.21C. At the TS, the critical bond dis-
tance for methyl group transfer Omethylguanine−Cmethylguanine and Cmethylguanine−SCys145 are 1.97 A and 2.44 A, respectively. We note that the hydrogen bond to the
Tyr114 is tightened somewhat in this step, from 1.73 A to 1.62 A. This is a result
of the negative charge created at the guanine ring and delocalized to the N3 center.
The accumulated activation barrier for the methyl transfer (i.e. the barrier for this
step added to the endothermicity of the previous step) was calculated to be 24.1
kcal/mol in the cluster model and 23.2 kcal/mol using ε = 4.
Once the methyl is transferred a guanine base is repaired and a negative charge
is delocalized between the N1 and N3 atoms of a substrate molecule, Figure 3.21D.
The overall reaction is found to be exothermic by 0.3 kcal/mol in the cluster model
and 6.4 kcal/mol using ε = 4.
In conclusion, the calculations on the MGMT enzyme give support to the sug-
gested reaction mechanism and provide detailed characterization of the structures
and energies of the various stationary points along the reaction path.
3.2. OTHER METHYL TRANSFER ENZYMES 43
A
Glu172
His146
Cys145
Tyr114
Methylguanine
f
f
f
f
C
C
S
O
C
C
C
CNN
C
N
C
O
O
C
O
CC
C
C
N
C
C
N
C
C
O
C
C
C
N
O
C
C
N
N
C
C
1.57
3.79
1.05 1.55
1.32
1.28
1.22
1.85
1.75
B
Glu172
His146
Cys145
Tyr114
Methylguanine
f
f
f
f
C
C
S
O
C
C
C
C
N
C
N
C
O
N
O
C
O
CC
C
C
N
C
N
C
C
C
O
C
C
C
N
C
O
C
N
N
C
C
2.053.61
1.03
1.621.86
1.05
1.64
1.45
1.36
1.33
1.36
1.33
1.73
C
Glu172
His146
Cys145
Tyr114
Guanine
f
f
f
f
C
C
S
C
O
C
C
NO
C
C
N
C
N
O
C
O
C
CC N
C
C
N
C
C
C
O
C
C
N
C
C
C
O
N
N
C
C
2.16 2.44
1.02
1.662.01
1.04
1.71
1.97
1.34
1.36
1.36
1.28
1.67
D
Glu172
His146
Cys145
Tyr114
Guanine
f
f
f
f
C
C
S
C
O
C
N
O
C
C
C
C
N
C
O
N
CNC
C
C
C
O
N
C
C
O
C
C
N
C
C
C
C
O
N
N
C
C
1.83
2.25
1.71
2.00
1.83
1.33
1.39
1.36
1.25
1.62
Figure 3.21: Optimized stationary points along the reaction path of MGMT.
Chapter 4
Conclusions
This thesis has dealt with the quantum chemical modeling of enzyme reactions.
More specifically, five methyl transfer enzymes were considered, namely glycine N-
methyltransferase (GNMT), guanidinoacetate methyltransferase (GAMT), phenyl-
ethanolamine N-methyltransferase (PNMT), histone lysine methyltransferase (HKMT),
and O6-methylguanine methyltransferase (MGMT). The four first enzymes use the
S-adenosyl L-methionine (SAM) cofactor to a transfer methyl group to their specific
substrates, while the last one is a suicide enzyme that repairs methylated DNA by
abstracting a methyl group from a guanine base.
Active site model of these enzymes were developed systematically in order to
develop detailed understanding of the reactions under consideration and also to eval-
uate the methodology used in the investigations. Points along the reaction paths
were optimized and characterized, and potential energy surfaces for the reactions
were calculated. By comparing the results of the calculations with available exper-
imental data, the plausibility of the suggested reaction mechanisms was judged. In
most cases, the presented calculations give support to the reaction mechanisms that
are proposed based on experiments. The calculations help working out the details
of the reactions, and similarities and differences among the various enzymes are dis-
cussed.
Apart from the conclusions concerning the specific enzymes, the overall results
of this thesis prove that the DFT methods, in particular the B3LYP functional, are
a very useful tool in the study of enzyme reactions. The effectiveness of the use of
relatively small active site models in the elucidation of reaction mechanisms is clearly
demonstrated.
45
Bibliography
[1] Szabo A.; Ostlund N.;. Modern Quantum Chemistry. McGraw - Hill, 1982.
[cited at p. 5]
[2] Cramer C.;. Essentials of Computational Chemistry. Theories and Models. John
Wiley & Sons Ltd, England, 2002. [cited at p. 5, 7]
[3] Jensen F.;. Introduction to Computational Chemistry. John Wiley & Sons Ltd,
England, 1999. [cited at p. 5, 7]
[4] Parr R.; Yang W.;. Density Functiona Theory of Atoms and Molecules. Oxford
University Press, 1989. [cited at p. 6, 7]
[5] Koch W.; Holthausen M.;. A Chemist’s Guide to Density Functiona Theory.
Wiley - VCH Verlag GmbH, 2001. [cited at p. 6, 7]
[6] Hohenberg P.; Kohn W.;. Phys. Rev., 136(3B):B864, 1964. [cited at p. 6]
[7] Kohn W.; Sham L.;. Phys. Rev., 140(4A):A1133, 1965. [cited at p. 6]
[8] Becke A.;. Phys. Rev. A, 38:3098, 1988. [cited at p. 7]
[9] Becke A.;. J. Chem. Phys., 96:2155, 1992. [cited at p. 7]
[10] Becke A.;. J. Chem. Phys., 97:9173, 1992. [cited at p. 7]
[11] Becke A.;. J. Chem. Phys., 98:5648–5652, 1993. [cited at p. 7]
[12] Vosko S.; Wilk L.; Nusair M.;. Can. J. Phys., 58:1200, 1980. [cited at p. 7]
[13] Lee C.; Yang W.; Parr R.;. Phys. Rev. B, 37:785, 1988. [cited at p. 7, 20]
[14] Miehlich B.; Savin A.; Stoll H.; Preuss H.;. Chem. Phys. Lett., 157:200, 1989.
[cited at p. 7]
[15] Curtiss L.; Raghavachari K.; Trucks G.; Pople J.;. J. Chem. Phys., 94:7221,
1991. [cited at p. 8]
47
48 BIBLIOGRAPHY
[16] Bauschlicher C.;. Chem. Phys. Lett., 246:40, 1995. [cited at p. 8]
[17] Bauschlicher C.; Partridge H.;. Chem. Phys. Lett., 240:533, 1995. [cited at p. 8, 9]
[18] Curtiss L.; Raghavachari K.; Redfern P.; Pople J.;. J. Chem. Phys., 109:7764,
1998. [cited at p. 9]
[19] Curtiss L.; Redfern P.; Raghavachari K.;. J. Chem. Phys., 123:124107, 2005.
[cited at p. 9]
[20] Siegbahn P.;. Inorg. Chem., 39(13):2923, 2000. [cited at p. 9]
[21] Siegbahn P.;. Quarterly Review of Biophysics, 36:91, 2003. [cited at p. 9]
[22] Anderson J.;. Adv. Chem. Phys., 91, 1995. [cited at p. 10]
[23] Eyring H.; Stern A.;. Chem. Rev., 24(2):253, 1939. [cited at p. 10]
[24] Truhlar D.; Hase W.; Hynes J.;. J. Phys. Chem., 87(26):5523, 1983. [cited at p. 10]
[25] Truhlar D.; Garrett B.; Klippenstein S.;. J. Phys. Chem., 100(31):12771, 1996.
[cited at p. 10]
[26] Fersht A.;. Structure and Mechanism in Protein Science : A Guide to Enzyme
Catalysis and Protein Folding. W. H. Freeman and Company, New York, 1998.
[cited at p. 11, 12]
[27] Blomberg M.; Siegbahn P.;. J. Phys. Chem. B, 105:9375, 2001. [cited at p. 15]
[28] Himo F.; Siegbahn P.;. Chem. Rev., 103:2421, 2003. [cited at p. 15]
[29] Noodleman L.; Lovell T.; Han W.-G.; Li J.; Himo F.;. Chem. Rev., 104:459,
2004. [cited at p. 15]
[30] Himo F.;. Biophys. Biochim. Acta, 1707:24, 2005. [cited at p. 15]
[31] Siegbahn P.; Borowski T.;. Acc. Chem. Res., 39:729, 2006. [cited at p. 15]
[32] Ramos M.; Fernandes P.;. Acc. Chem. Res., 41:689, 2008. [cited at p. 15]
[33] Himo F.; Guo J.; Rinaldo-Matthis A.; Nordlund P.;. J. Phys. Chem. B,
109:20004, 2005. [cited at p. 15]
[34] Hopmann K.; Himo F.;. Chem. Eur. J., 12:6898, 2006. [cited at p. 15]
[35] Hopmann K.; Guo J.; Himo F.;. Inorg. Chem., 46:4850, 2007. [cited at p. 15]
[36] Chen S.; Marino Fang W.; Russo N.; Himo F.;. J. Phys. Chem. B, 112:2494,
2008. [cited at p. 15]
BIBLIOGRAPHY 49
[37] Liao R.; Yu J.; Raushel F.; Himo F.;. Chem. Eur. J., 14:4287, 2008. [cited at p. 15]
[38] Hopmann K.; Himo F.;. J. Chem. Theor. Comp., 4:1129, 2008. [cited at p. 15]
[39] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R.
Cheeseman, J. A. Montgomery, Jr., T. Vreven, K. N. Kudin, J. C. Burant, J. M.
Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scal-
mani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota,
R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai,
M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo,
J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi,
C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Sal-
vador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C.
Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Fores-
man, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov,
G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith,
M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill,
B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, and J. A. Pople. Gaussian
03, Revision C.02. Gaussian, Inc., Wallingford, CT, 2004. [cited at p. 18]
[40] Cammi R.; Mennucci B.; Tomasi J.;. J. Phys. Chem. A, 103:9100, 2000.
[cited at p. 18]
[41] Cammi R.; Mennucci B.; Tomasi J.;. J. Phys. Chem. A, 104:5631, 2000.
[cited at p. 18]
[42] Cossi M.; Rega N.; Scalmani G.; Barone V.;. J. Chem. Phys., 114:5691, 2001.
[cited at p. 18]
[43] Cossi M.; Scalmani G.; Rega N.; Barone V.;. J. Chem. Phys., 117:43, 2002.
[cited at p. 18]
[44] Blomberg M.; Siegbahn P.; Babcock G.;. J. Am. Chem. Soc., 120(34):8812,
1998. [cited at p. 18]
[45] Rowling M.; Schalinske K.;. J. Nutr., 133:3392, 2003. [cited at p. 20]
[46] Wyss M.; Kaddurah-Daouk R.;. Physiological Reviews., 80:1107, 2000.
[cited at p. 20]
[47] Gee G.; Tyndall J.; Grunewald G.; Wu Q.; McLeish M.; Martin J.;. Biochemistry,
44:16875, 2005. [cited at p. 20]
[48] Martin J.; Begun J.; McLeish M.; Caine J.; Grunewald G.;. Structure, 9:977,
2001. [cited at p. 20]
[49] Pohorecky L.; Wurtman R. Pharmacological Reviews, 21(1):1, 1971. [cited at p. 20]
50 BIBLIOGRAPHY
[50] Strahl B.; Allis C.;. Nature, 403:41, 2000. [cited at p. 20, 35]
[51] Sims R.; Nishioka K.; Reinberg D.;. Trends. Genet., 19:629, 2003. [cited at p. 20,
35]
[52] Schneider R.; Bannister A.; Kouzarides T.;. Trends. Biochem. Sci., 27:396,
2002. [cited at p. 20, 35]
[53] Mitra S. DNA Repair, 6:1064, 2007. [cited at p. 20]
[54] Pegg A.; Fang Q.; Loktionova N.;. DNA Repair, 6:1071, 2007. [cited at p. 20]
[55] Tubbs J.;Pegg A.; Tainer J.;. DNA Repair, 6:1100, 2007. [cited at p. 20]
[56] Gerson S.;. J. Clin. Oncol., 20:2388, 2002. [cited at p. 20]
[57] Fang Q.; Gerson S.;. Clin. Cancer Res., 12:328, 2006. [cited at p. 20]
[58] Silverman R.;. The organic chemistry of enzyme-Catalyzed reactions. Academic
press, An Elsevier Science Imprint, 2002. [cited at p. 21]
[59] Takata Y.; Huang Y.; Komoto J.; Yamada T.; Konishi K.; Ogawa H.; Gomi T.;
Fujioka M.; Takusagawa F.;. Biochemistry, 42:8394, 2003. [cited at p. 21, 22, 25]
[60] Fujioka M.; Konishi K.; Takata Y.;. Biochemistry, 27:7658, 1988. [cited at p. 26]
[61] Fujioka M.; Konishi K.; Gomi T.;. Biochem. Biophys., 285:181, 1991.
[cited at p. 26]
[62] Takata Y.; Fujioka M.;. Int. J. Biochem., 22:1333, 1990. [cited at p. 26]
[63] Takata Y.; Date T.; Fujioka M.;. Biochem. J., 277:399, 1991. [cited at p. 26]
[64] Takata Y.; Fujioka M.;. Biochemistry, 31:4369, 1992. [cited at p. 26]
[65] Takata Y.; Konishi K.; Gomi T.; Fujioka M.;. J. Biol. Chem., 269:5537, 1994.
[cited at p. 26]
[66] Komoto J.; Yamada T.; Takata Y.; Konishi K.; Ogawa H.; Gomi T.; Fujioka
M.; Takusagawa F.;. Biochemistry, 43:14385–14394, 2004. [cited at p. 27, 28]
[67] Martin J.; Begun J.; McLeish M.; Caine J.; Grunewald G.;. Structure, 9:977,
2001. [cited at p. 30]
[68] McMillan F.; Archbold J.; McLeish M.; Caine J.; Criscione K.; Grunewald
G.; Martin J.;. J. Med. Chem., 47:37, 2004. [cited at p. 30]
[69] Gee C.; Tyndall J.; Grunewald G.; Wu Q.; McLeish M.; Martin J. Biochemistry,
44:16875, 2005. [cited at p. 30, 31]
BIBLIOGRAPHY 51
[70] Wu Q.; Gee C.; Lin F.; Tyndall J.; Martin J.; Grunewald G.; McLeish M.;. J.
Med. Chem., 48:7243, 2005. [cited at p. 30]
[71] Hu P.; Zhang Y.;. J. Am. Chem. Soc., 128:12728, 2006. [cited at p. 35, 36, 38, 39]
[72] Whang S.; Hu P.; Zhang Y.;. J. Phys. Chem. B, 111:3758, 2007. [cited at p. 35,
36, 38, 39]
[73] S.; Xiao B.; Jing C.; Wilson J.; Walker P.; Vasisht N.; Kelly G.; Howell S.;
Taylor I.; Blackburn G.; Gamblin. Nature, 421:625, 2003. [cited at p. 35]
[74] Lynch B.; Fast P.; Harris M.; Truhlar D.;. J. Phys. Chem. A, 104:4811, 2000.
[cited at p. 38]
[75] Lynch B.; Zhao Y.; Truhlar D.;. J. Phys. Chem. A, 107:1384, 2003. [cited at p. 38]
[76] Daniels D.; Woo T.; Luu K.; Noll D.; Clarke N.; Pegg A.; Tainer J.;. Nat.
Struct. & Mol. Biol., 11:714, 2004. [cited at p. 39]