Analysis of PUGNAc and NAG-thiazoline as transition state analogues for human O-GlcNAcase: Mechanistic and structural insights into inhibitor selectivity and transition state poise

Article abstract of DOI:10.1021/ja065697o

O-GlcNAcase catalyzes the cleavage of β-O-linked 2-acetamido-2-deoxy- β-D-glucopyranoside (O-GlcNAc) from serine and threonine residues of post-translationally modified proteins. Two potent inhibitors of this enzyme are O-(2-acetamido-2-deoxy-D-glucopyranosylidene)amino-N-phenylcarbamate (PUGNAc) and 1,2-dideoxy-2′-methyl-α-D-glucopyranoso[2,1-d]-Δ2′- thiazoline (NAG-thiazoline). Derivatives of these inhibitors differ in their selectivity for human O-GlcNAcase over the functionally related human lysosomal β-hexosamindases, with PUGNAc derivatives showing modest selectivities and NAG-thiazoline derivatives showing high selectivities. The molecular basis for this difference in selectivities is addressed as is how well these inhibitors mimic the O-GlcNAcase-stabilized transition state (TS). Using a series of substrates, ground state (GS) inhibitors, and transition state mimics having analogous structural variations, we describe linear free energy relationships of log(K_M/k_cat) versus log(K_I) for PUGNAc and NAG-thiazoline. These relationships suggest that PUGNAc is a poor transition state analogue, while NAG-thiazoline is revealed as a transition state mimic. Comparative X-ray crystallographic analyses of enzyme-inhibitor complexes reveal subtle molecular differences accounting for the differences in selectivities between these two inhibitors and illustrate key molecular interactions. Computational modeling of species along the reaction coordinate, as well as PUGNAc and NAG-thiazoline, provide insight into the features of NAG-thiazoline that resemble the transition state and reveal where PUGNAc fails to capture significant binding energy. These studies also point to late transition state poise for the O-GlcNAcase catalyzed reaction with significant nucleophilic participation and little involvement of the leaving group. The potency of NAG-thiazoline, its transition state mimicry, and its lack of traditional transition state-like design features suggest that potent rationally designed glycosidase inhibitors can be developed that exploit variation in transition state poise.

Full text of DOI:10.1021/ja065697o

Published on Web 12/23/2006
Analysis of PUGNAc and NAG-thiazoline as Transition State
Analogues for Human O-GlcNAcase: Mechanistic and
Structural Insights into Inhibitor Selectivity and
Transition State Poise
†
†
†
Garrett E. Whitworth, Matthew S. Macauley, Keith A. Stubbs,
‡
‡
‡
§
Rebecca J. Dennis, Edward J. Taylor, Gideon J. Davies, Ian R. Greig, and
,
†
David J. Vocadlo*
Contribution from the Department of Chemistry, Simon Fraser UniVersity, 8888 UniVersity
DriVe, Burnaby, British Columbia, Canada V5A 1S6, York Structural Biology Laboratory,
Department of Chemistry, UniVersity of York, York Y010 5YW, U.K., and Department of
Chemistry, UniVersity of Bath, Bath BA2 7AY, U.K.
Received August 7, 2006; E-mail: dvocadlo@sfu.ca
Abstract: O-GlcNAcase catalyzes the cleavage of â-O-linked 2-acetamido-2-deoxy-â-D-glucopyranoside
(O-GlcNAc) from serine and threonine residues of post-translationally modified proteins. Two potent inhibitors
of this enzyme are O-(2-acetamido-2-deoxy-D-glucopyranosylidene)amino-N-phenylcarbamate (PUGNAc)
and 1,2-dideoxy-2′-methyl-R-D-glucopyranoso[2,1-d]-∆2′-thiazoline (NAG-thiazoline). Derivatives of these
inhibitors differ in their selectivity for human O-GlcNAcase over the functionally related human lysosomal
â-hexosamindases, with PUGNAc derivatives showing modest selectivities and NAG-thiazoline derivatives
showing high selectivities. The molecular basis for this difference in selectivities is addressed as is how
well these inhibitors mimic the O-GlcNAcase-stabilized transition state (TS). Using a series of substrates,
ground state (GS) inhibitors, and transition state mimics having analogous structural variations, we describe
M I
linear free energy relationships of log(K /kcat) versus log(K ) for PUGNAc and NAG-thiazoline. These
relationships suggest that PUGNAc is a poor transition state analogue, while NAG-thiazoline is revealed
as a transition state mimic. Comparative X-ray crystallographic analyses of enzyme-inhibitor complexes
reveal subtle molecular differences accounting for the differences in selectivities between these two inhibitors
and illustrate key molecular interactions. Computational modeling of species along the reaction coordinate,
as well as PUGNAc and NAG-thiazoline, provide insight into the features of NAG-thiazoline that resemble
the transition state and reveal where PUGNAc fails to capture significant binding energy. These studies
also point to late transition state poise for the O-GlcNAcase catalyzed reaction with significant nucleophilic
participation and little involvement of the leaving group. The potency of NAG-thiazoline, its transition state
mimicry, and its lack of traditional transition state-like design features suggest that potent rationally designed
glycosidase inhibitors can be developed that exploit variation in transition state poise.
Introduction
using transition state (TS) mimicry as a guiding design
7
,8
principle. The expectation that stable molecules resembling
the enzyme catalyzed transition state, in either their charge or
geometry, will be potent competitive inhibitors stems from a
The development of rationally designed potent glycoside
hydrolase inhibitors has been a topic of continuing research for
years and has more recently enjoyed a surge of interest due to
1
2
1
,2
proposal, made first by Pauling, that enzymes catalyze
reactions by binding interactions realized preferentially with the
their increased uses as both research tools and as therapeutic
3
-5
agents.
Considerable efforts have been directed toward
12,13
preparing both potent6-8 and selective
9-11
transition state rather than with ground state (GS).
Such tight
glycosidase inhibitors
binding stabilizes these fleeting transition state structures and
†
Simon Fraser University.
University of York.
University of Bath.
(
7) Stutz, A. E. Iminosugars as Glycosidase Inhibitors, Nojirimycin and Beyond;
‡
John Wiley and Sons: Weinheim, 1999.
§
(8) Lillelund, V. H.; Jensen, H. H.; Liang, X.; Bols, M. Chem. ReV. 2002,
(
(
(
1) Tulp, A.; Barnhoorn, M.; Bause, E.; Ploegh, H. EMBO J. 1986, 5, 1783-
102, 515-553.
1
790.
(9) Macauley, M. S.; Whitworth, G. E.; Debowski, A. W.; Chin, D.; Vocadlo,
D. J. J. Biol. Chem. 2005, 280, 25313-25322.
(10) Siriwardena, A.; Strachan, H.; El-Daher, S.; Way, G.; Winchester, B.;
Glushka, J.; Moremen, K.; Boons, G. J. ChemBioChem 2005, 6, 845-
848.
2) Vosseller, K.; Wells, L.; Lane, M. D.; Hart, G. W. Proc. Natl. Acad. Sci.
U.S.A. 2002, 99, 5313-5318.
3) Butters, T. D.; Dwek, R. A.; Platt, F. M. AdV. Exp. Med. Biol 2003, 535,
2
19-226.
(
(
(
4) von Itzstein, M.; et al. Nature 1993, 363, 418-423.
(11) Ho, C. W.; Lin, Y. N.; Chang, C. F.; Li, S. T.; Wu, Y. T.; Wu, C. Y.; Liu,
S. W.; Li, Y. K.; Lin, C. H. Biochemistry 2006, 45, 5695-702.
(12) Pauling, L. Nature 1948, 161, 707-709.
5) Mendel, D. B.; et al. Antimicrob. Agents Chemother. 1998, 42, 640-646.
6) Vasella, A.; Davies, G. J.; Bohm, M. Curr. Opin. Chem. Biol. 2002, 6,
6
19-629.
(13) Jencks, W. P. K.; N. O.; Kennedy, E. P. Academic 1966, 273-298.
10.1021/ja065697o CCC: $37.00 © 2007 American Chemical Society
J. AM. CHEM. SOC. 2007, 129, 635-644
9
635

A R T I C L E S
Whitworth et al.
Figure 1. Catalytic mechanism used by O-GlcNAcase involves substrate-assisted catalysis. The first step of the reaction, cyclization, proceeds via attack
of the 2-acetamido carbonyl oxygen on the anomeric center to form a covalent bicyclic oxazoline intermediate. This step is facilitated by polarization of the
2
-acetamido moiety by an enzymic carboxyl group most likely acting as a general base catalyst. Departure of the aglycone is facilitated by general acid
catalysis provided by another carboxyl group in the enzyme active site. In the second step, ring opening, the oxazoline intermediate is broken open by the
general base-catalyzed attack of a water molecule on the anomeric center and general acid catalysis facilitating departure of the amide group. Both steps
occur with inversion of stereochemistry at the anomeric center such that the overall reaction proceeds with net retention of stereochemistry.
1
9
thereby results in the enormous rate accelerations of up to 10 -
fold seen for the most proficient enzymes.14 On the basis of
this rate acceleration the theoretical limit for the dissociation
constant of a perfect transition state analogue can be estimated
C-1 of the pyranose ring to mimic the geometric requirements
of the putative planar oxocarbenium ion-like transition state.
Second, nitrogen is often incorporated in place of C-1 or O-5
to mimic the relative positive charge development at these
centers within the transition state. Third, and less commonly,
conformationally constrained inhibitors are used that force the
pyranose ring to adopt a transition state-like conformation.6
Many inhibitors are defined as transition state analogues by
virtue of the principles used in their design and their superficial
resemblance to predicted transition state structures. In only a
very few cases have efforts been made to define whether these
molecules are genuinely transition state analogues, simply
serendipitous binders, or actually ground state analogues. Here
we investigate the binding of two known potent inhibitors to
human O-GlcNAcase; a family (for the classification of gly-
coside hydrolases see the CAZY database; http://afmb.cnrs-
mrs.fr/CAZY) 84 glycoside hydrolase catalyzing the cleavage
of O-linked 2-acetamido-2-deoxy-â-D-glucopyranoside (O-
-22
14
to be approximately 10 M. Of course, stable transition state
analogues necessarily vary from genuine transition state struc-
tures having partial bonds, and therefore it is impossible to
capture all the potentially available binding energy using a stable
molecule. Despite this limitation, very potent transition state
analogues have been found for several different classes of
enzymes.1
,23-25
5,16
For certain enzyme classes, extensive kinetic isotope effect
studies have been carried out that have enabled the definition
1
7,18
of geometric and electrostatic features of the transition state.
Such studies have led to the development of very tight binding
inhibitors that are now clinical drug candidates.19 For glycosi-
dases, however, despite considerable synthetic efforts the very
best inhibitors rationally designed as transition state analogues
typically have no lower than nanomolar affinities.6 Although
highly detailed kinetic isotope effect studies of the glycoside
hydrolases are lacking, a significant body of data has accrued
that provides limited insights into the general features of
transition states found for these enzymes.2
-8
GlcNAc) from serine and threonine residues of post-transla-
tionally modified nuclear and cytoplasmic proteins.2
6,27
Our
objective being to gain insight into the design of potent tran-
sition state analogues as well as to evaluate why O-(2-acet-
amido-2-deoxy-D-glucopyranosylidene)amino-N-phenylcarb-
0,21
Large and normal secondary R-D kinetic isotope effect studies
suggest that these transition states have an sp -hybridized
amate (PUGNAc, Figure 2, 1a-g) based derivatives bearing
2
28,29
bulky N-acyl groups show weak selectivity
for human
anomeric center with significant oxocarbenium ion-like char-
O-GlcNAcase over human lysosomal â-hexosaminidase, while
analogous derivatives of 1,2-dideoxy-2′-methyl-R-D-glucopy-
ranoso-[2,1-d]-∆2′-thiazoline (NAG-thiazoline, Figure 2,
acter.2
1,22
A corollary of such presumed planar oxocarbenium
ion-like transition states is that the pyranose ring is expected to
assume a conformation in which the C-2, C-1, C-5, and O-5
9
2a-g) show high selectivities.
20
atoms adopt a coplanar arrangement. Such limited kinetic
isotope effects, in conjunction with Brønsted analyses using
substrates having different leaving group abilities, have generally
suggested transition states in which the leaving group is largely
O-GlcNAcase uses a two-step catalytic mechanism involving
9,30
substrate-assisted catalysis
to form a transient oxazoline
31
intermediate (Figure 1) that is broken down to liberate the
free sugar hemiacetal and the protein. This overall catalytic
2
1
dissociated. Accordingly, the vast majority of rationally
designed glycosidase inhibitors exploit three principle design
(23) Tanaka, K. S. E.; Winters, G. C.; Batchelor, R. J.; Einstein, F. W. B.; Bennet,
features.6 First, an sp -hybridized center is often installed at
-8
2
A. J. J. Am. Chem. Soc. 2001, 123, 998-999.
(
24) Mosi, R.; Sham, H.; Uitdehaag, J. C. M.; Ruiterkamp, R.; Dijkstra, B. W.;
Withers, S. G. Biochemistry 1998, 37, 17192-17198.
(
(
(
14) Wolfenden, R.; Snider, M. J. Acc. Chem. Res. 2001, 34, 938-945.
15) Bartlett, P. A.; Marlowe, C. K. Biochemistry 1983, 22, 4618-4624.
16) Lewandowicz, A.; Tyler, P. C.; Evans, G. B.; Furneaux, R. H.; Schramm,
V. L. J Biol. Chem. 2003, 278, 31465-31468.
(25) Berland, C. R.; Sigurskjold, B. W.; Stoffer, B.; Frandsen, T. P.; Svensson,
B. Biochemistry 1995, 34, 10153-10161.
(26) Dong, D. L. Y.; Hart, G. W. J. Biol. Chem. 1994, 269, 19321-19330.
(27) Gao, Y.; Wells, L.; Comer, F. I.; Parker, G. J.; Hart, G. W. J. Biol. Chem.
2001, 276, 9838-45.
(
(
(
17) Schramm, V. L. Acc. Chem. Res. 2003, 36, 588-596.
18) Berti, P. J.; Tanaka, K. S. E. AdV. Phys. Org. Chem. 2002, 37, 239-314.
19) Gandhi, V.; Kilpatrick, J. M.; Plunkett, W.; Ayres, M.; Harman, L.; Du,
M.; Bantia, S.; Davisson, J.; Wierda, W. G.; Faderl, S.; Kantarjian, H.;
Thomas, D. Blood 2005, 106, 4253-4260.
(28) Stubbs, K. A.; Zhang, N.; Vocadlo, D. J. Org. Biomol. Chem. 2006, 4,
839-845.
(29) Kim, E. J.; Perreira, M.; Thomas, C. J.; Hanover, J. A. J. Am. Chem. Soc.
2006, 128, 4234-4235.
(20) Sinnott, M. L. Chem. ReV. 1990, 90, 1171-1202.
(30) Cetinbas, N.; Macauley, M. S.; Stubbs, K. A.; Drapala, R.; Vocadlo, D. J.
Biochemistry 2006, 45, 3835-3844.
(
21) Davies, G. J.; Sinnott, M. L.; Withers, S. G. ComprehensiVe Biological
Catalysis; Sinnot, M. L., Ed.; 1998; Vol 1, pp 119-209.
(31) Macauley, M. S.; Stubbs, K. A.; Vocadlo, D. J. J. Am. Chem. Soc. 2005,
127, 17202-17203.
(22) Sinnott, M. L.; Souchard, I. J. Biochem. J. 1973, 133, 89-98.
636 J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007

Transition State Analogues for O-GlcNAcase
A R T I C L E S
Figure 2. Series of inhibitors and substrates used to study transition state analogy of PUGNAc and NAG-thiazoline.
mechanism is common to glycosidases from families 18, 20,
6, and 84, all of which have evolved two strategically
competitive inhibitor of both the family 20 â-hexosamin-
34,40
9
5
dases
and the family 84 O-GlcNAcases. This molecule has
positioned carboxyl residues located within the active site that
play key catalytic roles.32-35 Asp174 and Asp175 have been
identified as the two key catalytic residues of human O-
an obvious geometric resemblance to the oxazoline intermediate,
although it has been suggested it may bind tightly by virtue of
its resemblance to a geometrically related transition state.
Whether these two highly potent inhibitors of O-GlcNAcase
are truly transition state analogues, however, has yet to be
determined.
Using a combination of linear free energy analyses, compu-
tational modeling studies, and X-ray crystallographic analysis
of enzyme-inhibitor complexes, we evaluate whether these two
inhibitors bind as transition state analogues, ground state
analogues, or serendipitous binders. New insights into the
structure of the transition state for the O-GlcNAcase catalyzed
hydrolysis of glycosides are gained and, by analogy, into those
of related enzymes using such a substrate-assisted catalytic
mechanism.
30,36
GlcNAcase.
In the first step of the reaction, the cyclization
1
74
step, Asp directs and polarizes the 2-acetamido group to act
as a nucleophile and form the oxazoline intermediate. Asp
meanwhile acts as a general acid, encouraging departure of the
3
0
175
30
aglycone leaving group. In the second step, ring-opening,
174
Asp facilitates departure of the 2-acetamido group, while
Asp1 acts as a general base, promoting the attack of a molecule
75
30
of water to yield the â-hemiacetal product. The structure of
the two transition states flanking the oxazoline intermediate have
9,30
been investigated through substrate structure-function studies
and R-D kinetic isotope effects.31 These studies have suggested
an “exploded” SN1-like transition state in which both the leaving
group and nucleophilic carbonyl oxygen of the acetamido group
are distant from the anomeric center. Similar proposals have
been made for the mechanistically related family 20 â-hexos-
aminidases.37 The most common interpretation of the geometric
and electrostatic structure of such transition states for enzymes
using substrate-assisted catalysis is that the transition states are
Results and Discussion
The successful application of the principles of enzymology
to the design of transition state analogues can be gauged by the
increasing numbers of potent, rationally designed, transition
state analogues. For many inhibitors few criteria, other than their
potency, are used to define them as transition state analogues.
In many cases, the designation of transition state analogue may
be inadvertently misused to describe opportunistic inhibitors that
simply exploit favorable binding situations within enzyme active
sites. In recent years there has only been limited progress in
developing still more potent transition state analogues of
glycoside hydrolase inhibitors. Studies directed at clarifying
which inhibitors are transition state analogues, however, have
proven value in improving our understanding of transition state
structures as well as in the iterative design of improved transition
state analogues. We therefore have undertaken studies to define
the mode of binding of two well-characterized inhibitors of
human O-GlcNAcase: NAG-thiazoline and PUGNAc.
2
oxocarbenium ion-like with an sp -hybridized anomeric center
and delocalized positive charge between the endocyclic ring
oxygen and the anomeric center.2
0-22
The two best characterized inhibitors of O-GlcNAcase are
PUGNAc and NAG-thiazoline. PUGNAc, first prepared by
38
Vasella and co-workers, has been proposed to be a transition
3
9
2
state analogue by virtue of its sp -hybridized anomeric center
which is thought to have geometric resemblance to an oxocar-
benium ion-like transition state (Figure 1). NAG-thiazoline, first
prepared by Knapp and co-workers,40 is a highly potent
(
(
(
(
(
32) van Aalten, D. M.; Komander, D.; Synstad, B.; Gaseidnes, S.; Peter, M.
G.; Eijsink, V. G. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 8979-8984.
33) Tews, I.; Perrakis, A.; Oppenheim, A.; Dauter, Z.; Wilson, K. S.; Vorgias,
C. E. Nat. Struct. Biol. 1996, 3, 638-648.
Both NAG-thiazoline and PUGNAc are potent inhibitors of
O-GlcNAcase (Table 1) and bind nearly 240 000-fold more
tightly than the ground state inhibitor methyl 2-acetamido-2-
deoxy-â-D-glucopyranoside 4a (KI ) 11 mM, Table 1). Potent
inhibition alone, however, is not a good measure of transition
state analogy, and several other more rigorous approaches have
34) Mark, B. L.; Vocadlo, D. J.; Knapp, S.; Triggs-Raine, B. L.; Withers, S.
G.; James, M. N. J. Biol. Chem. 2001, 276, 10330-10337.
35) Markovic-Housley, Z.; Miglierini, G.; Soldatova, L.; Rizkallah, P. J.; Muller,
U.; Schirmer, T. Structure 2000, 8, 1025-1035.
36) Toleman, C.; Paterson, A. J.; Kudlow, J. E. Biochim. Biophys. Acta 2006,
1
760, 829-839.
(
37) Vocadlo, D. J.; Withers, S. G. Biochemistry 2005, 44, 12809-12818.
(
38) Beer, D.; Maloisel, J. L.; Rast, D. M.; Vasella, A. HelV. Chim. Acta 1990,
7
3, 1918-1922.
1
5,41-44
(
39) Rao, F. V.; Dorfmueller, H. C.; Villa, F.; Allwood, M.; Eggleston, I. M.;
van Aalten, D. M. EMBO J. 2006, 25, 1569-1578.
been advanced.
Arguably, the most rigorous approach
to establishing transition state analogy has been to use free
energy correlations between inhibitor binding and transition state
(
40) Knapp, S.; Vocadlo, D.; Gao, Z. N.; Kirk, B.; Lou, J. P.; Withers, S. G. J.
Am. Chem. Soc. 1996, 118, 6804-6805.
J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007 637

A R T I C L E S
Whitworth et al.
MU-Glc NR
Table 1. Kinetic Data Determined Using Human O-GlcNAcase for the Three Series of Inhibitors and the Series of Substrates
PUGNR
selectivity (family (20/84))
GlcNR-THZ
GlcNR O-Me
R group
K
I
(µ
M)
K
I
(µM)
selectivity (family (20/84))
K
I(GS) (mM)
K
M
/kcat (mM min)
-
-
-
-
-
-
-
CH3
CH2CH3
0.046
1.2
2.5
40
220
9
1
1
11
6
g5
2
g6
0.07
0.12
0.25
1.5
57
1.6
5.7
1
270
1500
3100
100
11
11
21
0.13
0.15
0.19
1.8
1.8
5.0
(CH2)2CH3
(CH2)3CH3
(CH2)4CH3
CH(CH3)2
CH2CH(CH3)2
70
205
147
224
700
190
190
110
Scheme 1a
stabilization.15,44 Using this method to distinguish whether an
inhibitor is a transition state analogue, ground state analogue,
or simply a fortuitous binder involves systematic modification
of the inhibitor at one position and determining the resulting
changes in KI values. Analogous structural changes are also
made within the substrate, and kcat/Km values are determined,
since this parameter informs on the structure of the first
transition state reached along the reaction coordinate. The
inverse logarithm of these values provides free energy terms
that, when correlated to each other in a plot of log KI versus
log KM/kcat, may yield one of several patterns.1
5,44,45
The proper
interpretation of these linear free energy diagrams rests on
several practical assumptions:44 (1) the chemical step having
the transition state mimicked by the inhibitors must be rate
limiting for the reaction, (2) the intrinsic reactivity of the
substrates must not significantly change as a result of the
modifications, and (3) variation of the substituent should not
affect binding to the enzyme of the unmodified regions of either
the substrate or inhibitor. Once these requirements are satisfied
a strong correlation with a slope of 1 reveals the inhibitor is a
transition state analogue. This interpretation stems from the
expectation that changes made to a bona fide transition state
analogue should evoke equivalent changes in the energy of the
a
(
i) AcBr, heat; (iia) pyridine, MeOH; (iib)(a) acetone, 4-MU, Na 4-MU;
(
b) K2CO3, Et2O; (c) HCl; (iii) RCOCl, Et3N, DCM; (iv)(a) NaOMe, MeOH;
(b) 20:1 MeOH:acetic acid.
For O-GlcNAcases, we know that the N-acyl group of the
substrate is critically involved in catalysis and must undergo
conformational, electrostatic, and positional variation within the
active site of the enzyme on proceeding from the ground state
to the transition state.
this group would report clearly on the transition state structure
for these enzymes and that derivatives of PUGNAc (1) and
NAG-thiazoline (2), both previously synthesized within our
laboratory, would be useful probes.
within these inhibitors are small increases in the volume of the
alkyl group of the N-acyl moiety and they give rise to KI values
varying by over five orders of magnitude when assayed against
human O-GlcNAcase (Table 1). Such a wide variation in KI
values is beneficial for establishing a meaningful correlation.
With these molecules already in hand, we then prepared
9,31
We speculated that modifications of
46,47
actual transition state when introduced within the substrate.
Conversely, scattered plots having weak correlations suggest
the inhibitor is a poor transition state analogue or a fortuitous
9,28
The structural changes
7
binding inhibitor. Ground state mimics, on the other hand, are
best revealed by strong correlations having a slope of 1 between
log KI of the inhibitor and log KS values for binding of the
substrate (log KI for a ground state analogue or log KM for
enzymes with a rapid equilibrium substrate binding step may
also be used).
28
(
Scheme 1) the requisite series of substrates with analogous
It is noteworthy that in two closely related glycosidases from
family 13, correlations of log KI versus log KM/kcat have yielded
slopes other than 1 and these have been interpreted in entirely
modifications to the N-acyl group and measured the second-
order rate constants for human O-GlcNAcase catalyzed hy-
drolysis of these compounds (Table 1). Satisfying the first
requirement of the transition state analogy approach (vide supra),
we have previously shown, using a combination of kinetic
2
4,25
different ways.
These two studies, however, have been
carried out using enzymes in which mutations have been made
at several different residues within their active sites. It therefore
seems prudent to make systematic structural variations at only
one site. The best approach, first described by the Bartlett
31
isotope effects and a series of substrates bearing different
9,30
leaving groups,
that the second-order rate constant for
substrates bearing the 4-methylumbelliferyl leaving group
reflects the first chemical step. The second requirement of this
approach is fulfilled since the spontaneous hydrolysis rates of
these substrates are indistinguishable within error. Due to
limitations in the solubility of these substrates, however, we
were unable to obtain reliable KS or even KM values making
correlations with KS or KM values impossible. A series of methyl
glycosides (Figure 2, 4a-g) was therefore also prepared
(Scheme 1) as simple ground state substrates and were tested
as competitive inhibitors to obtain inhibition constants (Table
1) that are taken to be equivalent to true dissociation constants.
1
5
group, is to make a series of small structural perturbations to
the inhibitor at one site since these are most likely to report in
a sensitive manner on changes in the differences between the
actual transition state and the putative transition state mimic.
(
41) Rahil, J.; Pratt, R. F. Biochemistry 1994, 33, 116-125.
42) Dale, M. P.; Ensley, H. E.; Kern, K.; Sastry, K. A.; Byers, L. D.
Biochemistry 1985, 24, 3530-3539.
(
(43) Snider, M. J.; Wolfenden, R. Biochemistry 2001, 40, 11364-11371.
(44) Mader, M. M.; Bartlett, P. A. Chem. ReV. 1997, 97, 1281-1301.
(45) Bartlett, P. A.; Giangiordano, M. A. J. Org. Chem. 1996, 61, 3433-3438.
(46) Wolfenden, R. Nature 1969, 223, 704-705.
(
47) Lienhard, G. E. Science 1973, 180, 149-154.
638 J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007

Transition State Analogues for O-GlcNAcase
A R T I C L E S
logarithm of the inhibition constants (log KI(GS)) for the ground
state inhibitors showed no clear correlation (Figure 4C). These
two patterns are very similar to those observed by Bartlett and
15,48
co-workers,
which they used to show that phosphonamidates
are transition state analogues for the zinc peptidases. Together
these data strongly support the proposed mechanism involving
substrate-assisted catalysis from the acetamido group and re-
veal NAG-thiazoline as a good transition state mimic for
O-GlcNAcase catalyzed glycoside hydrolysis.
Using this approach, it is possible only to comment on the
transition state resemblance in regions where the inhibitor varies
in structure from the substrate. The specific structural features
probed by NAG-thiazoline encompass the entire thiazoline ring,
including the position of all associated atoms including C-1 and
C-2 of the pyranose ring. Notably, C-1 of NAG-thiazoline is
Figure 3. Inhibition of human O-GlcNAcase using ground state substrate
methyl 2-acetamido-2-deoxy-â-glucopyranoside. Inhibition of O-GlcNAcase
catalyzed hydrolysis of pNP-GlcNAc by the methyl glycoside (4a) shows
a pattern of competitive inhibition. The concentrations of 4a (mM) used
were 85.0 (4), 42.5 (9), 11.0 (0), 5.5 (b), and 0.0 (O).
3
2
an sp -hybridized carbon center rather than an sp -hybridized
center or basic nitrogen, features that are often assumed to be
essential features of transition state mimics for glycoside
hydrolases. The two best studied iminosugar inhibitors are the
isofagomine class, wherein C-1 is replaced with nitrogen and
the endocylic ring oxygen with a methylene unit, and the
nojirimycin class of inhibitors, wherein the endocylic ring
oxygen is replaced by a nitrogen atom. Both of these inhibitors
have been proposed to be transition state mimics although more
recent experimental data suggest that, at least for some enzymes,
this is not the case. Indeed, in a crystallographic complex of
GalNAc-isofagomine bound to a family 20 â-hexosamindase
that uses substrate-assisted catalysis, it was observed that the
protonated nitrogen engaged in an adventitious electrostatic
interaction with the general acid/base catalytic residue.49 For
nojirimycin, scattered linear free energy data was observed for
two related R-glucosidases, suggesting that this inhibitor is an
7
,25
adventitious binder to the two enzymes studied.
For the series of PUGNAc derivatives a poor correlation (r2
0.73, Figure 4B) with a slope of 1.07 ( 0.29 was observed
)
Figure 4. Comparison of transition state analogy free energy diagrams
for NAG-thiazoline and PUGNAc derivatives. (A, top left) Comparison of
PUGNAc series of inhibitors (1a-g) KI with methyl glycoside series
in a plot of log KI values of the inhibitor versus log KM/kcat
values for the series of glycoside substrates. As well, a poor
correlation is observed between log KI values and log KI(GS)
values for the ground state inhibitors (Figure 4A). Accordingly,
the scatter within both of these correlations indicates that
incremental changes of the N-acyl chain of the parent compound
are not viewed in a similar manner within the active site of the
enzyme as the incremental elongation of the N-acyl chain of
the substrate, indicating that PUGNAc is either a poor mimic
of the O-GlcNAcase-catalyzed transition state or simply a
fortuitous binder. Consistent with this interpretation, Bartlett
has described scatter of the data as an indication that the inhibitor
(
4a-g) KI. (B, top right) Comparison of PUGNAc series of inhibitors
1a-g) KI with substrate (3a-g) KM/kcat. (C, bottom left) Comparison of
(
thiazoline series of inhibitors (2a-g) KI with methyl glycoside series
(4a-g) KI. (D, bottom right) Comparison of thiazoline series of inhibitors
2a-g) KI with substrate (3a-g) KM/kcat.
(
Testing parent glycoside 4 as a representative member of these
ground state inhibitors revealed, as expected, a pattern of
competitive inhibition (Figure 3) and validated their use as
ground state analogues binding within the O-GlcNAcase active
site. Together, these four sets of compounds allowed us to
delineate the nature of the inhibition of human O-GlcNAcase
by NAG-thiazoline and PUGNAc.
44
may not closely resemble the transition state.
In the context of the traditional design principles used in
developing transition state analogue inhibitors of glycosidases,
the apparent weak transition state analogy of PUGNAc is
surprising. The trigonal anomeric center found within PUGNAc
is often considered to be a vitally important feature, and many
potent inhibitors contain such a moiety. It is worth noting that
NAG-thiazoline, which lacks all of the key design features of
transition state analogues, is an equally potent inhibitor as
PUGNAc. NAG-thiazoline may therefore mimic a transition
Correlations of log KI values for the series of NAG-thiazoline
derivatives (Figure 2, 2a-g) with the corresponding logarithm
of the inverse second-order rate constants KM/kcat for the
O-GlcNAcase catalyzed hydrolysis of a series of 4-methylum-
belliferyl glycoside substrates (Figure 2, 3a-g) reveals a strong
2
correlation (r ) 0.98) with a slope of 0.97 ( 0.06 (Figure 4D).
This correlation indicates that the incremental differences in
binding energy related to the elaboration of the thiazoline ring
with aliphatic chains of increasing length are nearly identical
to those perturbations observed for the transition state of the
enzyme catalyzed hydrolysis of the substrates bearing analogous
changes. Correlations of log KI values with the corresponding
(48) Phillips, M. A.; Kaplan, A. P.; Rutter, W. J.; Bartlett, P. A. Biochemistry
1
992, 31, 959-963.
(
49) Mark, B. L.; Vocadlo, D. J.; Zhao, D.; Knapp, S.; Withers, S. G.; James,
M. N. J Biol. Chem. 2001, 276, 42131-42137.
J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007 639

A R T I C L E S
Whitworth et al.
state that is significantly different than the canonical oxocar-
benium ion-like transition state that is widely perceived to be
found along the reaction coordinate of glycoside hydrolases.
Taft-like linear free energy analyses (log kcat/KM versus σ*)
of the O-GlcNAcase catalyzed hydrolysis of N-fluoroacetamido
glycosides have strongly implicated the acetamido group in
nucleophilic participation at the transition state of the rate
In view of this possibility we considered previous results in
9
,31
determining chemical step.
Furthermore, the linear free
this new context. The large secondary R-deuterium kinetic
31
energy analyses we describe here indicate that NAG-thiazoline
is a transition state analogue, an observation also consistent with
nucleophilic participation since the anomeric carbon and the
thiazoline sulfur atom are only 1.85 Å apart from each other.
However, whether NAG-thiazoline geometrically resembles the
transition state along the reaction coordinate, some point along
the reaction coordinate that is close to this saddle, or that the
enzyme catalyzed reaction is DN + AN with a short-lived, yet
discrete oxocarbenium ion, as has been suggested for â-galac-
isotope effects (RD-KIE) of 1.14 ( 0.02 measured previously
with human O-GlcNAcase suggests that the enzyme stabilized
transition states have oxocarbenium ion-like character with
2
significant sp -character at the anomeric center. Interpreting this
KIE as revealing a strict trigonal geometry at the anomeric center
in the transition state is complicated, however, as underscored
by studies of known SN2 reactions of the methoxymethyl model
system. Depending on the nature of the incoming nucleophile,
RD-KIE values traditionally held to be consistent with either
56
5
0
tosidase by Richard and co-workers, currently remains untested
SN1 (kH/kD ) 1.18) or SN2 (kH/kD ) 0.99) were observed.
in any glycoside hydrolase.
Further complicating matters is that equilibrium isotope binding
effects have been observed in several cases where deuterium is
present at the anomeric center.51 Accordingly, in the absence
of other supporting heavy atom isotope effects at the reaction
center, interpreting the magnitude of RD-KIE values at acetal
centers is difficult and cannot be used to discern the extent of
nucleophilic participation.52 Therefore, it seems equally pos-
sible that the RD-KIE values determined may reflect a non-
planar transition state having greater nucleophilic participation
and correspondingly less p-orbital character at the anomeric
center.
Assuming the transition state resembles NAG-thiazoline, as
suggested by the free energy analyses, one can approximate the
bond orders to both the leaving group and nucleophile. By
assuming the interatomic distance between C-1 and the carbonyl
oxygen in the transition state is 1.85 Å (the length of the C-S
bond length within NAG-thiazoline), a Pauling bond order of
0.26 to the nucleophile can be calculated. The Brønsted analyses
and the RD-KIE values previously measured for human O-
GlcNAcase suggest little involvement of the leaving group in
the transition state. On these bases and the crystallographic
analyses of the Bacteroides thetaiotaomicron O-GlcNAcase
homologue in complex with NAG-thiazoline, we estimate the
bond order to the leaving group is nearly 0 (0.01). In this
structure, a molecule of water sits approximately 2.8 Å above
the anomeric carbon in a position that is expected to be occupied
by the glycosidic oxygen of the substrate within the Michaelis
complex, and it is held in position by a hydrogen bond to the
Such differences in the extent of nucleophilic participation
and involvement of the leaving group have been well-
documented for the N-ribosyl hydrolases and have developed
with the support of X-ray crystallographic analyses of enzyme-
inhibitor complexes into the concept of transition state poise.
The transition state poise is defined by the bond orders between
the anomeric center and both the leaving group and nucleophile,
which can vary significantly as a function of the position of
C-1 along the reaction coordinate. In this context the glycoside
57
enzymic general base catalyst. These data, in combination with
other studies of the glycosidase mechanism, suggest that human
O-GlcNAcase possesses an oxocarbenium ion-like transition
state which is significantly stabilized by nucleophilic participa-
tion of the 2-acetamido group. This nucleophilic participa-
tion, in combination with the action of the catalytic residue
53
hydrolases, like the N-ribosyl hydrolases, appear to use an
electrophilic migration mechanism54 for which the reaction
coordinate diagram can be defined by the motion of the
anomeric carbon and where the total bond order between the
anomeric center, nucleophile, and leaving group is significantly
1
74
hydrogen bonding with the amide nitrogen (Asp in human
2
42
O-GlcNAcase and Asp
in the B. thetaiotomicron â-glu-
less than 1. Within crystallographically visualized Michaelis
complexes of â-D-glycoside hydrolases3
3,55
cosaminidase), delocalizes excess positive charge away from
the C-1-O-5 region of the pyranose and on into the oxazoline
ring. Such a dissociative transition state is consistent with the
the anomeric carbon
typically adopts a position aboVe the plane of the pyranose ring
with the exocyclic oxygen hydrogen bonding to the general acid
catalytic group. Within the newly formed intermediate, the
anomeric carbon is now positioned below the plane of the
pyranose ring bonded to the nucleophile. For O-GlcNAcase,
this nucleophile is the carbonyl oxygen of the acetamido group.
The transition state must accordingly resemble a structure sitting
somewhere along this pathway defined by these two end points
and not necessarily, as widely believed, equidistant between
them as would be found for an oxocarbenium ion.
9
,57
31
previously described Taft analyses, kinetic isotope effects,
and the free energy analyses we detail here. It is also reminiscent
of those determined empirically for N-ribosyl transferases using
a DNAN mechanism but most likely involving slightly greater
1
7,58
nucleophilic participation.
To gain greater insight into the basis by which NAG-
thiazoline mimics the transition state whereas PUGNAc does
not, we carried out crystallographic analyses of the selec-
tive O-GlcNAcase inhibitor, NButGT, complexed to a bac-
terial homologue of human O-GlcNAcase. This B. thetaioto-
micron â-glucosaminidase is an excellent model of human
(
(
(
50) Knier, B. L.; Jencks, W. P. J. Am. Chem. Soc. 1980, 102, 6789-6798.
51) Lewis, B. E.; Schramm, V. L. J. Am. Chem. Soc. 2003, 125, 4785-4798.
52) Huang, X. C.; Tanaka, K. S. E.; Bennet, A. J. J. Am. Chem. Soc. 1997,
1
19, 11147-11154.
(
(
53) Vocadlo, D. J.; Davies, G. J.; Laine, R.; Withers, S. G. Nature 2001, 412,
(56) Richard, J. P.; Huber, R. E.; Heo, C.; Amyes, T. L.; Lin, S. Biochemistry
1996, 35, 12387-12401.
8
35-838.
54) Fedorov, A.; Shi, W.; Kicska, G.; Fedorov, E.; Tyler, P. C.; Furneaux, R.
H.; Hanson, J. C.; Gainsford, G. J.; Larese, J. Z.; Schramm, V. L.; Almo,
S. C. Biochemistry 2001, 40, 853-860.
(57) Dennis, R. J.; Taylor, E. J.; Macauley, M. S.; Stubbs, K. A.; Turkenburg,
J. P.; Hart, S. J.; Black, G. N.; Vocadlo, D. J.; Davies, G. J. Nat. Struct.
Mol. Biol. 2006, 13, 365-371.
(55) Sulzenbacher, G.; Driguez, H.; Henrissat, B.; Schulein, M.; Davies, G. J.
(58) Scheuring, J.; Berti, P. J.; Schramm, V. L. Biochemistry 1998, 37, 2748-
2758.
Biochemistry 1996, 35, 15280-15287.
640 J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007

Transition State Analogues for O-GlcNAcase
A R T I C L E S
Table 2. Data Collection and Refinement Statistics for the
Structure Solution of B. thetaiotaomicron GH84 O-GlcNAcase with
NButGT
BtGH84 with N-butylthiazoline
Data Collection
space group
cell dimens
a, b, c (Å)
R, â, γ (deg)
wavelength
resolution (Å)
Rsym or Rmerge
I/σ(I)
C2
185.1, 51.7, 172.8
90.0, 100.1, 90.0
0.9930
40-2.25 (2.33-2.25)
0.086 (0.47)
18 (3)
a
completeness (%)
redundancy
99.9 (99.9)
4.5 (4.4)
Refinement
resolution (Å)
no. of reflecns
Rwork/Rfree
2.25
77 267
21/28
no. of atoms
protein
ligand
9500
32
waters
554
B-factors
protein
ligand/ion
water
30
19
30
RMS deviations
bond lengths (Å)
bond angles (deg)
Figure 5. Structural analyses of the binding of NButGT to BtGH84 and
its comparison to binding of PUGNAc to CfGH84. (A) Observed electron
density for the NButGT along with its flanking residues. The map shown
is a 2Fobs-Fcalc electron density map contoured at 1σ and is shown in
divergent (“walleyed”) stereo. (B) Structural overlap of the BtGH84 NButGT
complex (ligand, gray; protein, green) with the PUGNAc complex of the
Clostridium perfringens GH84 enzyme39 (yellow). The overlap was
optimized using the main-chain protein atoms only, in a radius of 15 Å of
the ligands, using the program QUANTA (Accelrys, San Diego, CA). This
figure was drawn using MOLSCRIPT63 and BOBSCRIPT.
0.023
1.99
a
Highest-resolution shell is shown in parentheses.
O-GlcNAcase since it too processes O-GlcNAc modified
proteins and also has striking sequence conservation when
compared to the human O-GlcNAcase. The structure of the
B. thetaiotaomicron GH84 O-GlcNAcase, in complex with
NButGT was solved at a resolution of 2.25 Å by molecular
replacement (Table 2). The structure crystallized in a C2 space
group (see methods) in which two molecules of BtGH84 lie in
the asymmetric unit in the same orientation separated by a
translation-only vector of 0, 0, 0.5. The BtGH84 structure can
be traced from residue 4 of the mature protein through to residue
64
It has been noted previously by us that derivatives of
PUGNAc having bulky N-acyl groups show weak selectivity
toward human O-GlcNAcase over human â-hexosaminidase,2
while the analogous NAG-thiazoline derivatives show high
8,29
9
selectivities. In an effort to better understand the molecular
basis for these differences in selectivity, we decided to compare
the binding modes of NButGT and PUGNAc by overlaying the
structures of these two inhibitors bound to the O-GlcNAcase
homologues (Figure 5B). The molecular basis for this discrep-
ancy in selectivity must stem from structural differences between
these molecules and how they interact with the active site of
human O-GlcNAcase.
There are several obvious structural differences between
NAG-thiazoline and PUGNAc. The first, and most obvious
difference is that PUGNAc bears a phenyl carbamate moiety
which is absent in NAG-thiazoline. This substituent extends out
from the active site and engages in several interactions that are
not found for NAG-thiazoline. The exocyclic oxime nitrogen
of PUGNAc hydrogen bonds with the general acid/base catalytic
5
89, with two short breaks in molecule B from 51 to 53 and
from 455 to 456. The C-terminal “CBM32-like” â-sheet domain,
which was partially ordered in the original native structure
solution, is completely disordered in this crystal form and could
not be modeled. We compared the structure of this complex
with a previously determined structure of NAG-thiazoline bound
to the same enzyme57 as well as to a complex of PUGNAc
bound within the active site of a Clostridium perfringens
homologue3
9,59
of O-GlcNAcase. Gratifyingly, NButGT and
NAG-thiazoline bind identically within the active site, and the
pyranose ring of the inhibitors within both complexes can be
overlaid exactly (data not shown). We find the only difference
between the two structures is where electron density corre-
sponding to the alkyl chain of NButGT extends into an active
site pocket (Figure 5A) that has been previously predicted to
accommodate such bulkier groups.9 The confirmation of
identical binding modes for the parent inhibitor and NButGT
is important in the context of the transition state analogy studies
described here since it satisfies the third implicit assumption of
the approach that analogous inhibitors bind in the same
orientation.
1
75
30
243
group (Asp within human O-GlcNAcase and Asp within
5
7
the B. thetaiotaomicron â-glucosaminidase ), and the phenyl
group of PUGNAc also engages in several hydrophobic and
stacking interactions with residues in the aglycon binding site.
It is interesting to note that N-acetylglucosamino-1,5-lactone
oxime (LOGNAc), which lacks the phenyl carbamate group of
PUGNAc, binds 30-fold more poorly than PUGNAc, providing
an indication of the importance of these serendipitous inter-
actions. A second structural difference between the two inhibi-
tors is that the hybridization at the anomeric center varies. This
,57
26
(
59) Ficko-Blean, E.; Boraston, A. B. Acta Crystallogr., Sect. F: Struct. Biol.
Cryst. Commun. 2005, 61, 834-836.
J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007 641

A R T I C L E S
Whitworth et al.
variation has an impact on the preferred conformation of the
4
pyranose ring; for PUGNAc the pyranose ring adopts a E
2
conformation enforced, in part, by the sp -hybridized anomeric
4
center, whereas NButGT sits in a C1 chair that is slightly
4
distorted toward the E conformation due to the rigidity of the
thiazoline ring. The structural consequence of these variations
is that the substituents of the pyranose ring, particularly the
3
-and 4-hydroxyl groups, adopt subtly different orientations.
Such apparently minor differences in the position of both ring
and substituent atoms can have pronounced effects on both van
der Waals interactions and hydrogen bonds. A third obvious
structural difference is that the N-acyl group of PUGNAc is
free, while NAG-thiazoline’s orientation is locked. This differ-
ence also results in important variations in the pKa of the
protonated nitrogen at the 2-position of the pyranose ring that
will be discussed in greater detail in context of the computational
studies (vide infra). The thiazoline ring is nestled between two
tryptophan residues which presumably serve to accurately
position the acetamido group of the substrate and stabilize the
positive charge that develops at the transition state during
formation of the oxazoline ring. Not only does the thiazoline
ring engage in stacking interactions with these residues, but it
also orients the alkyl substituent of the inhibitors. As seen in
the structural comparison, (Figure 5), the alkyl substituent of
NButGT is pointing directly down the active site cavity, whereas
the N-acetamido group of PUGNAc is pointed at a slightly
different angle. Consistent with this structural view, human
O-GlcNAcase is better able to cope with elongation of the
aliphatic chain of NAG-thiazoline derivatives as compared to
analogous derivatives of PUGNAc (Table 1). Because van der
Waals forces are powerfully distance-dependent, this apparently
small difference may account for the significant differences in
selectivity observed by us between the derivatives of these two
families of inhibitors.
Figure 6. Simplified reaction coordinate for human O-GlcNAcase-catalyzed
â-glucosaminide hydrolysis showing the MEPs for PUGNAc-analogue I
(
modeled with methyl rather than phenyl) and NAG-thiazoline II as
well as ground state III, transition state model IV, and oxazoline V.
NAG-thiazoline II is shown to share structural and electrostatic properties
of both transition state model V and oxazoline IV. PUGNAc-analogue I is
shown to share significant electrostatic properties with ground state
III, although it does display some transition state-like conformational
properties. MEPs (which all share a common scale) are shown projected
3
onto the 0.04 electrons/Å isodensity surface. Stick models of the calculated
These crystallographic analyses primarily furnish details of
the geometric features mediating interactions in the enzyme-
inhibitor complexes but do not provide great insight into
transition state structure nor the electrostatic features of the
inhibitors and reaction coordinate species. We therefore reasoned
that further insight into the potency, transition state analogy,
and selectivity of both NAG-thiazoline- and PUGNAc-derived
compounds toward human O-GlcNAcase could be gained from
a comparison of their molecular electrostatic potential surfaces
structures used to generate the MEPs are provided in the Supporting
Information.
reasonable qualitative, though not quantitatively accurate,
description of the reaction coordinate.
A comparison of our transition state model IV and the
oxazoline V demonstrates that C-O bond extension is associ-
ated with both increased planarity about the anomeric carbon
center as well as a contraction of the endocyclic C-1-O-5 bond.
The most striking difference between oxazoline V and transition
state model IV, however, is the build-up of positive charge,
not only at the anomeric carbon center but also on the anomeric
proton. This redistribution of charge is much more apparent than
other changes centered about, for example, the endocyclic ring
oxygen or within the oxazoline ring.
(MEPs) with that of the enzyme-bound intermediate or, better
still, the enzymic transition state. Work by the Schramm group
has shown that the similarity of transition states and inhibitors
6
0
may be probed qualitatively and quantitatively using MEPs.
Accordingly, we modeled both inhibitors, PUGNAc-analogue
I and NAG-Thiazoline II, as well as three species found along
the reaction coordinate for the cyclization step (Figure 6C,
bottom structures in color): an intact methyl glycoside ground
state III, the transition state IV we predict on the basis of the
TS analogy studies featuring a distorted oxazoline in which the
C-1 carbonyl oxygen bond distance was constrained to the C-S
bond distance within NAG-thiazoline,61 and last the oxazoline
intermediate V (Figure 6C, bottom structures in color). It is
important to note that these species are expected to provide a
To afford a careful evaluation of this difference, we verified
that this change in the MEP at the anomeric center was
independent of the protonation state of oxazoline V by calculat-
ing similar MEPs for the associated oxazolinium ion, thiazo-
(61) Using a different distance constraint for the model involving the distance
from the anomeric center to the 2′-carbon of the thiazoline ring leads to an
even tighter TS model with respect to the distance between the anomeric
center and the nucleophilic oxygen atom. This constraint allows the bond
between the 2′-carbon and the nucleophilic oxygen bond to flex with the
result that within the model it adopts a length greater than that found within
the model of the oxazoline intermediate; a result that is inconsistent with
the bond having an order between 1 and 2.
(
60) Bagdassarian, C. K.; Schramm, V. L.; Schwartz, S. D. J. Am. Chem. Soc.
1
996, 118, 8825-8836.
642 J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007

Transition State Analogues for O-GlcNAcase
A R T I C L E S
linium ion, and oxazolinium ion-based transition state model
potential observed in oxazoline V, thiazoline II, and the
transition state analogue IV. In this respect PUGNAc-analogue
I is significantly more substrate-like (structure III). The marked
changes in charge at this center along the reaction coor-
dinate are likely to be a significant target for selectiVe transition
state stabilization by human O-GlcNAcase, so any failure to
mimic transition state charge at this center is likely to hinder
the ability of PUGNAc and derivatives to act as transition state
analogues.
(in which the anomeric C-O bond length of the oxazolinium
ion was constrained to that found in the thiazolinium ion). These
MEPs may be found in the Supporting Information. A com-
parison of MEPs for the oxazolinium ion and the oxazolinium
ion-based transition state model showed similarly significant
changes in charge distribution on the anomeric carbon and
hydrogen atoms (with an increase in positive character in the
transition state model).
An analysis of the structure and MEP of thiazoline II reveals
that it mimics some aspects of oxazoline V while in other aspects
it more closely resembles the transition state model IV. The
hexopyranose ring conformations of oxazoline V, thiazoline II,
and transition state model IV are all subtly different. The model
used for the reactivity of many glycosidases is one in which
electrophilic migration of the anomeric carbon center enforces
Another notable difference between NAG-thiazoline and
1
74
PUGNAc is that the polarizing residue (Asp
in human
3
0
242
O-GlcNAcase and Asp in the B. thetaiotaomicron â-glu-
57
cosaminidase ) is likely to engage in a much stronger interaction
with oxazoline IV and thiazoline II as compared to PUGNAc-
30
analogue I. As noted previously, the pKa of the nitrogen atom
will change during the cyclization step from around 15, for an
amide proton, to a pKa of about 5.5, the value found for
1
4
4
a conformational itinerary following the S3- H3- C1 path from
Michaelis complex to enzyme-bound intermediate (Figure 6A).
A comparison of our optimized structures indicates that, whereas
the pyranose ring of transition state model IV is best described
2
62
the 2-methyl-∆ -oxazolinium ion.
The pH-rate profile
(log(KM/kcat) vs pH) for the hydrolysis of p-nitrophenyl â-N-
acetylglucosaminide allows us to assign a macroscopic pKa value
4
174 30
174
as a H3 conformation and that of NAG-thiazoline as adopting
of approximately 5.0 to Asp
. The pKa of Asp , therefore,
4
a C1 conformation, oxazoline V is best described as adopting
much more closely matches those of the conjugate acids of
thiazoline II and oxazoline IV than that of PUGNAc-analogue
I. Because hydrogen bonds between residues with closely
matched pKa values are stronger than those between residues
whose pKa values differ greatly, we may reasonably expect
4
a H5 conformation. This conformation is a result of C-3-
C-2-C-1-O-5 coplanarity favored by formation of the five-
membered oxazoline ring which induces strain in the pyranose
ring. Although the precise conformational itinerary traveled
along the reaction coordinate is uncertain, it seems likely that
174
thiazoline-derived inhibitors to more tightly bind to Asp than
PUGNAc-derived inhibitors. From the perspective of Asp1
74
-
4
the less strained C1 conformation of thiazoline II lies confor-
binding, thiazoline-derived analogues will therefore be better
transition state analogues than PUGNAc-derived inhibitors,
which possess greater ground state character in this particular
interaction.
mationally closer to our transition state model.
Although NAG-thiazoline II bears a significant geometric
resemblance to transition state model IV, a comparison of the
MEPs for the three species under consideration reveal some
subtle differences. The larger van der Waals radius of sulfur
compared to oxygen is readily apparent when the MEPs are
visualized. The potential surrounding the sulfur atom is also
appreciably less negative than that surrounding the oxygen atom
of either oxazoline V or transition state model IV. Furthermore,
the potential distribution around the anomeric carbon and
hydrogen atoms of thiazoline II most closely resembles that of
oxazoline V when compared to that of transition state model
IV.
Conclusion
Using three approaches, including linear free energy correla-
tions, X-ray crystallographic analyses, and computational
modeling, we find that NAG-thiazoline is a TS analogue for
the human O-GlcNAcase catalyzed hydrolysis of â-glu-
cosaminides while PUGNAc is either a poor TS analogue or a
serendipitous binding inhibitor. These observations are surprising
in view of the conventional approaches used in the design of
TS analogues of glycoside hydrolases. Specifically, PUGNAc,
which incorporates a traditional transition state design feature,
The structural properties of thiazoline II and its MEP indicate
that it is likely to be a reasonable mimic of a species that lies
on the reaction coordinate between the region of the enzymic
transition state and the enzyme-bound intermediate. PUGNAc-
analogue I, in contrast, displays a number of markedly different
structural and electrostatic characteristics from oxazoline V and
the transition state model IV. The hexopyranose ring of
2
an sp -hybridized anomeric center, is a poor TS mimic. NAG-
3
thiazoline, which has an sp -hybridized anomeric center, is a
good TS analogue even though such a geometry is not
considered TS-like. It may appear counterintuitive that NAG-
thiazoline, given the tetrahedral geometry of its anomeric center,
should resemble the transition state for the hydrolysis of an
acetal since these transition states are widely perceived as
dissociative with significant oxocarbenium ion-like character.
It seems more obvious that NAG-thiazoline may resemble the
oxazoline structure found further along the reaction coordinate
and be more of a mimic of a high-energy oxazoline intermediate.
However, the C-S bond length of the thiazoline (1.86 Å) is
considerably longer than the C-O bond length (1.45 Å)
expected for the corresponding oxazoline. This longer bond
4
PUGNAc-analogue I adopts a E conformation and, as such, is
a marginally good mimic of the conformation of transition state
model IV. In a number of critical respects, however, PUGNAc-
analogue I is an unsatisfactory mimic of either the oxazoline V
or its preceding transition state. PUGNAc-analogue I fails to
capture any significant bonding between the N-acetamido moiety
and the anomeric carbon center: as indicated previously, this
interaction is a critical component of the reaction pathway at
the transition state.30 More critically, the MEP of PUGNAc-
analogue I indicates that there is significantly more negative
potential in the region of the anomeric carbon atom. This
negative potential is in striking contrast with the more positive
(
62) Porter, G. R.; Rydon, H. N.; Schofield, J. A. Nature 1958, 182, 927.
63) Kraulis, P. J. J. Appl. Crystallogr. 1991, 24, 946-950.
(
(64) Esnouf, R. M. J. Mol. Graphics Modell. 1997, 15, 132-134.
J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007 643

A R T I C L E S
Whitworth et al.
distance results in distortion of the pyranose ring of the
thiazoline and mimics a bond order of 0.26 between the
nucleophilic carbonyl oxygen and the anomeric center; a bond
order consistent with the significant nucleophilic participation
Acknowledgment. We thank the Western Canada research
grid (WestGrid) for a generous allocation of resources. D.J.V.
thanks the Natural Sciences and Engineering Research Council
of Canada (NSERC) for support of this research. D.J.V. is a
Michael Smith Foundation for Health Research (MSFHR)
Scholar and is supported as a Tier II Canada Research Chair in
Chemical Glycobiology. M.S.M. is supported as a Canada
Graduate Scholar (NSERC) and by a graduate student fellowship
from the MSFHR. G.J.D. thanks the Biotechnology and
Biological Sciences Research Council and the Royal Society
for funding. I.R.G. thanks the Royal Society (U.K.) for a
postdoctoral fellowship to Canada.
9
established previously. Crystallographic studies, computation-
ally determined structures, and molecular electrostatic potential
surfaces for NAG-thiazoline- and PUGNAc-based inhibitors of
human O-GlcNAcase support this view. On the basis of these
qualitative MEPs, thiazoline-derived O-GlcNAcase inhibitors
and related structures appear to be a better framework for
capturing a substantial fraction of the transition state binding
energy used by the enzyme when compared to PUGNAc-derived
inhibitors. These collective results, in combination with earlier
kinetics studies, also suggest late transition state poise for this
electrophilic migration mechanism with significant nucleophilic
participation and essentially no involvement of the leaving
groupsan observation consistent with studies of some of the
N-ribosyl hydrolases. This improved understanding of the
reaction coordinate adopted by human O-GlcNAcase, however
qualitative, should provide the basis for the development and
refinement of transition state analogues of human O-GlcNAcase
having optimized geometries.
Supporting Information Available: All experimental details,
including those for the synthesis of compounds 3a-g and 4a-g
as well as their structural proofs, enzyme assays, crystallographic
studies, and computational studies; optimized geometries for
all species described in the paper and MEPs for the protonated
oxazolinium, thiazolinium, and oxazolinium-based transition
state analogue; complete refs 4 and 5. This material is available
free of charge via the Internet at http://pubs.ac.org.
JA065697O
644 J. AM. CHEM. SOC.
9
VOL. 129, NO. 3, 2007