Chapter 29

QM/MM Methods for Studying Enzymatic Reactions

of Glycosyltransferases
Igor Tvaroka
Abstract
Hybrid quantum mechanics and molecular mechanics (QM/MM) methods have become a powerful tool
to provide an accurate and effective description of complex biological systems. The QM treatment of the
electronic structure of an active site region and the rest of the enzyme by molecular mechanics allows
enzymatic reaction to being modeled with including the impact of environment. Different reaction pathways of the enzymatic mechanism can be testedtransition states (TS) and intermediates characterized
using QM/MM methods, leading to significant advances in understanding enzymatic reactions. This
chapter discusses the ideas and the setting up of the structural and computational models for calculations
with QM/MM software. The use of QM/MM methodology is also illustrated using the case of the inverting glycosyltransferase GnT-I.
Key words QM/MM methods, GAMESS, MOZYME, Gaussian, ADF, Qsite, Enzymatic reactions,
Potential energy surface, Transition state, Glycosyltransferases

Introduction
Interest in carbohydrates has rapidly increased in recent years and
reflects the essential functions that oligosaccharides and glycoconjugates play in living cells and organisms. Glycosyltransferases that
are involved in selective biosynthesis of glycan structures compose
a group of enzymes that transfer glycosyl residues from a nucleoside
phosphate to other molecules [14]. The outcome of the reaction,
catalyzed by these enzymes, is the creation of a new glycosidic linkage. However, the catalytic mechanism of glycosyltransferases is still
not fully understood [59].
Understanding the mechanism of glycosyltransferases at the
atomic level is of vital importance in biomedical research.
Experimental studies provide essential and necessary information
concerning the reaction mechanism, thermodynamics, and kinetics; however, experimental data are often inadequate for determining the reaction mechanism at the atomic level. Most importantly

Thomas Ltteke and Martin Frank (eds.), Glycoinformatics, Methods in Molecular Biology, vol. 1273,
DOI 10.1007/978-1-4939-2343-4_29, Springer Science+Business Media New York 2015

489

490

Igor Tvaroka

experimental studies cannot directly determine the transition state

(TS) structure, which is essential for development of inhibitors or
drug design. Complementary to experimental study, the QM/
MM computational modeling offers extremely useful tool to a
deeper understanding of mechanistic strategy employed by
glycosyltransferases.

Theory
A description of the bond forming and bond breaking processes
during the catalytic reaction requires quantum mechanical methods. However, QM methods are restricted to systems of up to a
few hundred atoms and are unsuitable for large systems such as
enzymes. In fact, they were used only on small (cluster) models of
the enzymatic reaction. Cluster models provided valuable information of the mechanistic characteristics of a reaction catalyzed by
glycosyltransferases [1013]. However, a more realistic description
of reaction mechanism requires including the complete enzyme
environment in the computations. Thus, there is a need for the
method capable of treating up to several thousand atoms.
The field of mixed QM/MM methods has emerged during the
last few decades driven by needs of nanomaterials, condensed-phase
reactions, catalytic systems, and structural biology. A practical and
attractive solution to the energetic characterization of enzymatic
reaction is to combine both methods and utilize a QM treatment
for the active site (reaction) region and an MM method for the
remainder of the enzymatic environment. The QM/MM methods
combine the accuracy of a QM description with the low cost of
MM. In this chapter, we begin with a brief introduction to the basic
characteristics of the QM/MM approach. Then we introduce some
programs available for these calculations but do not review the
details. The following section describes a modeling process from
preparing structural and computational models to calculations of
reaction pathways. Finally, we illustrate this process showing the
key insights gained from applying the QM/MM method on the
catalytic mechanisms of the inverting glucosyltransferase.

2.1

QM/MM Method

In this section, we briefly describe only basic principles. The theoretical development and applications of QM/MM methods have
been recently discussed in a number of papers [1419] and readers
should refer to those reviews for more details. The general idea of
the QM/MM method is partitioning of the entire system into
localized regions, namely, QM and MM, as illustrated in Fig. 1.
For enzymatic reactions studies, the QM region is usually centered
on the active site where a chemical reaction proceeds. The QM
region includes structural moieties involved in the catalytic
reactions and amino acids involved in interactions with the reacting

QM/MM Methods for Studying Enzymatic Reactions of Glycosyltransferases

491

Fig. 1 Schematic representation of partitioning of the whole complex into QM and

MM regions

substrate(s). The MM region consists of the remaining part of

substrates, enzyme, and, if necessary, also solvent molecules.
The total internal energy of the QM/MM system can be
formally written as
EQM/MM = EQM+EMM+ EQM-MM

(1)

Where EQM is the QM energy of the QM region, EMM is the MM

energy of the MM region, and EQM-MM is the interaction energy
between these two regions. Several different schemes were designed
[18] to calculate the total energy (EQM/MM) of the system. The
QM/MM algorithm can utilize virtually any combination of QM
and MM methods, and the choice of the QM method depends on
the goal of the application. Recently, the QM region has been routinely treated using density functional theory (DFT) methods. For
the MM region, there are several available force fields parameterized for proteins, nucleic acids, and carbohydrates.
The QM/MM methods differ also in handling of interactions
(coupling) between the QM and MM region and different
approaches have been developed to explore these interactions. The
energy term for the coupling consists of both bonded (bond
stretching, bond bending, and torsion terms) and non-bonded
(van der Waals and electrostatic) interactions. The key component
of the coupling is the electrostatic interactions between the QM
charge density and the atomic charges in the MM region.
Depending on their treatment, QM/MM methods can be divided
into two basic models: mechanical embedding (ME) and electrostatic embedding (EE). The groups differ, essentially, by the degree
of mutual polarization of two regions. In the ME model, electrostatic interactions between the QM and MM regions are treated at
the MM level, with the same charge model applied to the QM
region. In the ME model, the QM charges are not polarized by the
electrostatic environment. In contrast, in the EM model, the QM

492

Igor Tvaroka

calculations are performed in the presence of MM charges by

incorporating the MM charges as one-electron terms in QM
Hamiltonian. Thus, the QM charges in the EE model are polarized
by charges in the MM region and the electrostatic interactions
between the QM and MM regions are treated at the QM level.
It is noteworthy that recently it has been shown that electrostatic
embedding approaches seem to overestimate electrostatic interactions [17, 20], which may affect results.
Another key issue is the treatment of the boundary region, particularly when the boundary goes through one or more covalent
bonds. Depending on the QM/MM scheme, the boundary region
may include localized orbitals (frozen orbitals) between the QM
and MM regions that are parameterized by calculations on small
molecules. Another approach uses additional atoms (link atoms;
usually hydrogen atoms) that cap the QM system and are not part
of the entire system, or it may consist of atoms with specific features
that are calculated by use of both the QM and MM methods.
As summarized in a recent review article [16], there are several
programs available for calculations using the QM/MM methods.
They were developed using three basic strategies. In the first, a
QM package was modified by incorporating MM functionalities
(examples of this approach are Gaussian [21], ADF [22],
MOZYME [23], GAMESS [24], and QSite [25]). In the second,
an MM code was extended by adding a QM code; examples include
AMBER [26] and CHARMM [27]. In the third, a core engine is
used to interface QM and MM modules. ChemShel [28] is an
example of this architecture. Each of these approaches has its merits and disadvantages. The third approach is the most flexible, but
its effectiveness may be decreased by a transfer of large amounts of
data between both modules. In contrast, the first two include functionalities from the original QM or MM programs, e.g. the capability of handling large biological systems by MM program and
transition state search module in QM methods.

Modeling Catalysis of Glycosyltransferases

An investigation of the catalytic mechanism of glycosyltransferases
using a QM/MM method is not a trivial task. The modeling
approach involves three steps: (1) preparing a structural model;
(2) building of a QM/MM model; and (3) modeling (mapping)
an enzymatic reaction. In the first step, the enzyme has to be prepared for calculations and the enzymesubstrates complex should
be assembled. The available structures of glycosyltransferase complexes usually have only one substrate (donor or acceptor) bound
in the active site. Therefore, the docking procedures are used to
accommodate a missing substrate into the active site. The choice of
the most probable binding position is guided by experimental data.

QM/MM Methods for Studying Enzymatic Reactions of Glycosyltransferases

493

The next step is to address the issue on the choice of a computational

model that includes dividing the total system into QM and MM
regions and the choice of a proper QM and MM method, respectively. It is recommended that the QM region could be as large as
possible. The third step is to map the catalytic reaction. Here, we
discuss only so-called stationary search; molecular dynamics calculations are beyond the scope of this chapter. In the stationary
approach, a scan (one- or two-dimensional) is performed along
selected reaction coordinates. This requires performing of hundreds of energy minimizations that hopefully lead to transition
states and others stationary points on PES.
3.1 Preparing
Structure

Preferably, the coordinates of a GT-substrate complex are obtained

from experimentally solved structures (X-ray or NMR). The crystal
structure coordinates can be found in the PDB database [29].
If these are not available, also homology models can be used,
although the quality of the homology model biases the reliability
of results. Nevertheless, all x-ray crystal structures have to be prepared for calculations using available software (e.g., Modeller [30],
Chimera [31], Maestro [32], and Protein Preparation Wizard
[33]). Preparation of the complex includes several steps. One can
start with the determination of amino acid residues with two possible conformations localized in the X-ray and a selection of only
one conformation. We found appropriate at this step to remove all
water oxygen atoms from the structure, and the necessary water
molecules may be included later. The next step is to identify missing amino-acid residues. These can be added by an appropriate
program (e.g., Modeller) and their orientation in a loop refined by
constraining coordinates of all atoms from the original PDB file.
Then, in the complex structure, it is necessary to add hydrogen
atoms and assign protonation states, based on a residue pKa value.
Orientation of added hydrogen atoms and protonation of histidine
residues can be determined using the method implemented in
available programs (e.g., Chimera). Of course, the above-described
steps and their order are based on our experience and the user may
choose a different approach. The obtained structure can be used
for the docking of the missing active substrate into the active site
of the enzyme. An application of docking methods has been
recently described [34] and is not discussed here. However, it is
noteworthy that an employment of restraints that drive the substrate into the active site is useful since it decreases a number of
solutions. From the obtained poses, the position that has a right
position with respect to other substrate in the active site is chosen.
Then relevant water molecules can be added.

3.2

Regardless of which QM/MM software is selected for calculations,

the user has to decide how to divide the total enzyme with all substrates into the QM and MM region, respectively. It was shown

QM/MM Partition

494

Igor Tvaroka

that the accuracy of QM/MM calculations depends on the size of

the QM region [20, 35, 36]. The selection of atoms included in
the QM region is, therefore, particularly important. Obviously, if
more atoms (residues) are included in the QM region the description is better. Usually, the reactive part of the active site is treated
by QM methods, whereas the remaining part of the system is modeled by MM methods. In this respect, a definition of the boundary
between QM and MM region requires a particular attention. It is
also recommended moving the boundary between the QM and
MM region further from the active residues [35]. A choice of the
QM and MM method is the next step. In both cases, the choice
follows the same criteria as in QM or MM studies, and, therefore,
is not further elaborated here. We just point out that a choice of
DFT functional is essential and may affect the results due to their
known deficiencies [37, 38]. When a model is determined it is
necessary to perform a geometry optimization of the whole QM/
MM system due to inaccuracies of the initial coordinates obtained
from X-ray, NMR, or homology modeling or docking. In glycosyltransferases, the active site is big and usually consists of the donor
(nucleoside diphosphate, e.g., UDP-GlcNAc), acceptor (monosaccharide, oligosaccharide, polysaccharide, peptides, lipids, etc.),
metal ion, some water molecules, and surrounding amino acids.
Thus, the QM region for the modeling of the catalytic reaction of
GTs must be carefully selected and is because of a large number of
atoms also computationally quite demanding.
3.3 Potential Energy
Surface

The first requirement for modeling a catalytic reaction of glycosyltransferases is the calculation of a potential energy surface (PES)
that accurately describes the chemical changes with the redistribution of electrons. In principle, the PES of the entire system (enzyme
and substrates) describes many TS and other stationary states
because there are many degrees of freedom that are not directly
related to the chemical changes but determine the shape of the
PES. Therefore, the progress of a catalytic reaction is often
described by the structural changes of a reduced number of reaction coordinates. The choice of reaction coordinate(s) is extremely
important. An improper choice of the reaction coordinate can bias
the calculation and yield slower convergence.
The catalytic reaction of glycosyltransferases can be characterized as the creation of one new glycosidic linkage between the
acceptor hydroxyl group oxygen and the anomeric carbon C1 of
the donor, cleavage of the donor glycosidic linkage between a
sugar and nucleotide diphosphate, and removal of a proton H
from the acceptor hydroxyl OH by a catalytic base. Thus, the reaction mechanism can be monitored by three reaction coordinates.
The first reaction coordinate r1 represents the nucleophilic attack
of the hydroxyl oxygen from the acceptor on the anomeric carbon
of the donor leading to the formation of a new -glycosidic linkage

QM/MM Methods for Studying Enzymatic Reactions of Glycosyltransferases

495

and is defined as the distance between the anomeric carbon C1 and

the nucleophile oxygen O of the acceptor hydroxyl group. The
second reaction coordinate r2 characterizes the dissociation of
the C1O1 glycosidic bond and is defined as the distance between
the anomeric carbon C1 and glycosidic oxygen O1 of the donor. The
third reaction coordinate r3 represents the transfer of proton from
nucleophile hydroxyl to the catalytic base and is defined as the distance between the proton of the acceptor hydroxyl group, and the
accepting atom of the catalytic base. The reaction potential energy
surface (PES) is usually determined by a so-called adiabatic mapping. The reaction coordinates may be varied by for example 0.2
increments, between 3.2 and 1.4 for r1 and r2, and between
2.0 and 1.0 for r3, respectively. Even in this arrangement, the
calculation of PES requires a large amount of geometry optimizations (10 10 6 = 600). Therefore, PES is usually scanned by calculating individual two-dimensional sections of three-dimensional
reaction energy surface. Usually, three two-dimensional PESs,
namely, (r1, r2), (r1, r3), and (r2, r3) are sufficient for exploring all
possible reaction pathways. In these calculations, all the geometrical variables are optimized except two reaction coordinates.
We note that the third reaction coordinate is also optimized during
these calculations, e.g., for each point on the (r1, r3) map, the r2
reaction coordinate (the C1O1 distance) is included in geometry
optimization. The QM/MM calculated sections of PES show
possible reaction pathways and energetic information about the
Michaelis complex (ES), transition state(s) (TS), and products.
The lowest energy pathway found on PES characterizes the
reaction pathway that connects ES with products via TS.

The Illustrative Example: N-acetylglucosaminyltransferase-I

In this section, we illustrate the application of the QM/MM on
the case of the N-acetylglucosaminyltransferase I (UDP-N-acetyld-glucosamine:-3-D-mannoside -1,2-N-acetylglucosaminyltransferase I, GnT-I), the first glycosyltransferase, which catalytic
mechanism was investigated using QM/MM method [38]. The
inverting glycosyltransferase GnT-I catalyzes the transfer of a
2-acetamido-2-deoxy--D-glucopyranose residue (GlcNAc) from
uridine 5-(2-acetamido-2-deoxy--D-glucopyranosyl pyrophosphate) (UDP-GlcNAc) to the C2-hydroxyl group of a mannose
residue in the Man5GlcNAc2-Asn-X oligosaccharide [39] (Fig. 2).
The structural model was build using a similar procedure as
described in Subheading 3.1 using the crystal structure of the
catalytic domain of rabbit GnT-I in complex with UDP-GlcNAc/
Mn2+ [40]. The trisaccharide Man1-3(3,6-OMe-Man1-6)Man
was used as the acceptor since this oligosaccharide represents the
minimal acceptor substrate [41]. The trisaccharide was docked

496

Igor Tvaroka

Fig. 2 Schematic diagram of the enzymatic reaction catalyzed by GnT-I

into the X-ray crystallographic structure of GnT-I, in complex with

UDP-GlcNAc, using the Glide program from Schrdinger Inc.
[42]. To build the computational model, the whole complex system was divided into two regions (Fig. 3). The QM region includes
the UDP-GlcNAc portion of the sugar-donor molecule; the
Man1-3 mannose residue of the trisaccharide-acceptor, aspartate
D291, the divalent metal cofactor Mn2+ fully coordinated by three
water molecules, and aspartate D213. The MM region consists of
the remaining part of the substrates and the enzyme, altogether
5,633 atoms. The QM region made up of 88 atoms was treated
with DFT method using BP functional and TZP basis sets and
employing the ADF package [43]. The AMBER95 all-atom force
field [44] was applied to the MM region. The calculations of the
reaction pathways were performed along a predefined reaction
coordinate that characterizes the nucleophilic attack of the acceptor oxygen on the anomeric carbon C1 of the donor reaction.
The results showed that calculated reaction pathway is typical
of a concerted SN2-type mechanism. The predicted activation barrier of 19 kcal/mol correlated well with available experimental
data, while the protein contributes to reaction energetic by
decreasing the overall reaction barrier by 9 kcal/mol. The analyses
of calculated structures revealed a significant function of
two hydrogen bonds that contribute to the catalytic process. The
first is a strong hydrogen bond between the nucleophile oxygen
and the catalytic base oxygen and facilitates nucleophilic attack.

QM/MM Methods for Studying Enzymatic Reactions of Glycosyltransferases

497

Fig. 3 Schematic representation of partitioning of the entire complex into QM and

MM regions for GnT-I

The second is the interaction between HO6 from the acceptor and
-phosphate oxygen that stabilizes a negative charge on the
-pyrophosphate during the scission of the glycosidic linkage.
The calculated TS structure displays features (Fig. 4) characteristic of SN2 reaction. The nearly simultaneous nucleophilic addition
and dissociation steps are demonstrated by a distance of 1.912
and 2.542 for the forming and breaking glycosidic bond, respectively. The glucopyranose ring of GlcNAc undergoes significant
structural changes progressing from the ES to TS model. The
GlcNAc ring adopts the 4H3 conformation in the TS model with
almost planar arrangement around anomeric carbon, whereas in the
ES, the ring conformation is the 4C1 chair. This is accompanied by
a shortening of the ring oxygenanomeric carbon (C1O5) bond
length to 1.333 . This partial double-bond character of the C1
O5 bond is a result of delocalization of the O5 lone pairs into C1
and contributes to sp2 hybridization at the anomeric carbon. The
lone pairs delocalization also stabilizes the partial positive charge
on the anomeric carbon C1 and facilitates simultaneous interactions of C1 with the leaving group and attacking nucleophile. It is
noteworthy that cleavage of the C1O1 glycosidic bond is accompanied by the 17 rotation of the -phosphate oxygen. The calculated TS structure was supported by a normal-mode analysis of the
QM part of the model and, together with the calculated -deuterium
kinetics effect (KIE), supported an SN2-type mechanism.

498

Igor Tvaroka

Fig. 4 A geometrical representation of the transition state model obtained by

applying the MM(DFT)/MM method for GnT-I

Acknowledgements
This work was supported by the Scientific Grant Agency of the
Ministry of Education of Slovak Republic and Slovak Academy of
Sciences (grants VEGA-02/0176/09 and VEGA-02/0101/11).
References
1. Beyer TA, Sadler JE, Rearick JI, Paulson JC,
Hill RL (1981) Glycosyltransferases and their
use in assessing oligosaccharide structure and
structure-function relationships. Adv Enzymol
52:23175
2. Schachter H (1991) Enzymes associated with
glycosylation. Curr Opin Struct Biol 1:
755765
3. Kleene R, Berger EG (1993) The molecular
and cell biology of glycosyltransferases.
Biochim Biophys Acta 1154:283325
4. Montreuil J, Vliegenthart JFG, Schachter H
(1995) Glycoproteins. In: Neuberger A, van
Deenen LLM (eds) New comprehensive biochemistry, vol 29a. Elsevier, Amsterdam

5. Vocadlo DJ, Davies GJ (2008) Mechanistic

insights into glycosidase chemistry. Curr Opin
Chem Biol 12:539555
6. Tvaroka I (2005) Structural insights into the
catalytic mechanism and transition state of glycosyltransferases using ab initio molecular modeling. Trends Glycosci Glycotech 17:177190
7. Tvaroka I (2006) Molecular modeling of
retaining glycosyltransferases. In: Vliegenthart
JFG, Woods RJ (eds) NMR Spectroscopy and
computer modeling of carbohydrates, vol 930,
ACS Symposium Series. Washington, DC,
ACS, pp 285301
8. Wilson IBH, Breton CH, Imberty A, Tvaroka
I (2008) Molecular basis for the biosynthesis

QM/MM Methods for Studying Enzymatic Reactions of Glycosyltransferases

9.

10.

11.

12.

13.

14.

15.
16.

17.

18.

19.

20.

21.
22.
23.
24.

of oligo- and polysaccharides. In: Fraser-Reid

BO, Tatsuta K, Thiem J, Cot GL, Flitsch S,
Ito Y, Kondo H, Nishimura S-I, Yu B (eds)
Glycoscience chemistry and chemical biology.
Springer, New York, NY, pp 22672323.
ISBN ISBN: 978-3-540-36154-1
Lairson LL, Henrissat B, Davies GJ, Withers
SG (2008) Glycosyltransferases: structures,
functions and mechanisms. Annu Rev Biochem
77:521555
Tvaroka I, Andr I, Carver JP (2000) Ab initio molecular orbital study of the catalytic
mechanism of glycosyltransferases. Description
of reaction pathways and determination of
transition states structures for inverting
N-acetylglucosaminyltransferases. J Am Chem
Soc 122:87628776
Tvaroka I, Andr I, Carver JP (2003) Catalytic
mechanism of the inverting N-acetylglucosaminyltransferase I: DFT quantum mechanical study of the reaction pathway and
determination of the transition state structure.
Glycobiology 13:559566
Andr I, Tvaroka I, Carver JP (2003) On
the reaction pathways and determination of
transition state structures for retaining
-galactosyltransferases. Carbohydr Res 338:
865877
Tvaroka I (2004) Molecular modeling insights
into the catalytic mechanism of the retaining
galactosyltransferase LgtC. Carbohydr Res
339:10071014
Lin H, Truhlar DG (2007) QM/MM: what
have we learned, where are we, and where do we
go from here? Theor Chem Acc 117:185199
Senn HM, Thiel W (2007) QM/MM studies of
enzymes. Curr Opin Chem Biol 11:182187
Senn HM, Thiel W (2007) QM/MM methods for biological systems. Top Curr Chem
268:173290
Hu H, Yang W (2008) Free energies of chemical
reactions in solution and in enzymes with ab initio quantum mechanics/molecular mechanics
methods. Annu Rev Phys Chem 59:573601
Senn HM, Thiel W (2009) QM/MM methods for biomolecular systems. Angew Chem
Int Ed 48:11981229
Tvaroka I (2011) QM/MM insight on enzymatic reactions of glycosyltransferases. Mini
Rev Org Chem 8:263269
Hu LH, Eliasson J, Heimdal J, Ryde U (2009)
Do quantum mechanical energies calculated
for small models of protein-active sites converge? J Phys Chem A 113:1179311800
http://www.Gaussian.com
http://www.scm.com
http://www.cse.clrc.ac.uk/qcg/games-uk
http://www.chem.ac.ru/Chemistry/Soft/
MOZYME.en.html

25.
26.
27.
28.
29.
30.

31.

32.
33.

34.

35.

36.

37.

38.

39.

40.

41.

42.
43.
44.

499

http://www.schrodinger.com
http://www.amber.scripps.edu
http://www.charmm.org
http://www.cse.clrc.ac.uk/qcg/chemshell
http://www.pdb.org
Modeller, A program for protein structure
modeling, http://salilab.org/modeller/modeller.html
Pettersen EF, Goddard TD, Huang CC,
Couch GS, Greenblatt DM, Meng EC, Ferrin
TE (2004) J Comput Chem 25:1605
(2009) Maestro version 9.0, New York, NY:
Schrdinger, LLC
Schrdinger Suite (2009) Protein preparation
wizard, Epik version 2.0, New York, NY:
Schrdinger, LLC
Morris GM, Lim-Wilby M (2008) Molecular
docking. In: Kukol A (ed) Methods in molecular biology, vol 443. Humana Press, Totowa,
NJ, pp 365382
Sumovski CV, Ochsenfeld C (2009) A convergence study of QM/MM isomerization energies with the selected size of the QM |region
for peptidic systems. J Phys Chem A
113:1173411741
Hu LH, Sderhjelm P, Ryde U (2011) On the
convergence of QM/MM energies. J Chem
Theor Comput 7:761777
Xu X, Alecu IM, Truhlar DG (2011) How well
can modern density functionals predict internuclear distances at transition state? J Chem
Theor Comput 7:16671676
Ka J, Tvaroka I (2009) Comparative DFT
study on the glycosidic bond in reactive species
of galactosyl diphosphates. Chem Papers
63:598607
Kozmon S, Tvaroka I (2006) Catalytic mechanism of glycosyltransferases: hybrid quantum
mechanical/molecular mechanical study of the
Inverting N-acetylglucosaminyltransferase I.
J Am Chem Soc 128:1692116927
Unligil UM, Zhou S, Yuwaraj S, Sarkar M,
Schachter H, Rini JM (2000) X-ray crystal
structure of rabbit N-acetylglucosaminyltransferase I: catalytic mechanism and a new
protein superfamily. EMBO J 19:52695280
Schachter H, Reck F, Paulsen H (2003) Use of
synthetic oligosaccharide substrate analogs to
map the active sites of N-acetylglucosaminyltransferases I and II. Methods Enzymol 363:
459475
(2009) Glide version 5.5, New York, NY:
Schrdinger, LLC
SCM (2005) ADF 2005.01, Amsterdam
Cornell WD, Cieplak P, Bayly CI, Gould IR,
Merz KM, Ferguson DM, Spellmeyer DC, Fox T,
Caldwell JW, Kollman PA (1995) J Am Chem
Soc 117(19):51795197