The N-acetyl neuraminyl oxecarbenium ion is an intermediate in the presence of anionic nucleophiles

Article abstract of DOI:10.1021/ja972503j

Solvolysis of CMP N-acetyl neuraminate (CMP-NeuAc) in 1.8 M acetate buffer at pH 5 containing 0.9 M azide results in the formation of both anomers of 2-deoxy-2-azido N-acetyl neuraminic acid in addition to N-acetyl neuraminic acid as determined by ¹H-NMR product analysis. A rate dependence on [azide] was observed with an apparent bimolecular rate constant of (2.1 ± 0.30) x 10^-3 M^-1 min^-1 which could only account for half of the azido- NeuAc formed. Comparison of rate, product ratio, and stereochemical data indicate that concurrent pathways for formation of N₃-NeuAc are operative, with 17% of product forming from reaction of azide and the tight ion pair. 12% via the solvent separated ion pair, and 6% from the free NeuAc oxocarbenium ion. From the corrected product ratio data, the lifetime of the oxocarbenium ion was estimated to be ≤ 3 x 10^-11 s. Solvolysis of CMP- NeuAc at pL. = 5.0 afforded an observed solvent deuterium isotope effect (SDIE) k(H₂O)/k(D₂O) = 0.45, consistent with specific acid catalysis of glycosidic bond cleavage. A SDIE of 0.66 for the apparent bimolecular azide trapping pathway was also observed. An apparent isotope effect of ~1.1 for trapping of the N-acetyl neuraminyl oxocarbenium ion by water was determined by product analysis of azide trapping in H₂O and D₂O. An ab initio transition state for attack of water on an N-acetyl neuraminyl oxocarbenium ion model was located which featured a hydrogen bond between the oxocarbenium ion carboxylate and water; proton transfer was not part of the reaction coordinate. It is proposed that the N-acetyl neuraminate carboxylate group stabilizes an intermediate oxocarbenium ion, but the barrier for capture by water is lowered by a transition state hydrogen bond.

Full text of DOI:10.1021/ja972503j

J. Am. Chem. Soc. 1998, 120, 1357-1362
1357
The N-Acetyl Neuraminyl Oxecarbenium Ion Is an Intermediate in the
Presence of Anionic Nucleophiles
Benjamin A. Horenstein* and Michael Bruner
Contribution from the Department of Chemistry, UniVersity of Florida, GainesVille, Florida 32611
ReceiVed July 24, 1997
Abstract: Solvolysis of CMP N-acetyl neuraminate (CMP-NeuAc) in 1.8 M acetate buffer at pH 5 containing
0.9 M azide results in the formation of both anomers of 2-deoxy-2-azido N-acetyl neuraminic acid in addition
1
to N-acetyl neuraminic acid as determined by H-NMR product analysis. A rate dependence on [azide] was
observed with an apparent bimolecular rate constant of (2.1 ( 0.3) × 10-3 M-l min-l which could only
account for half of the azido-NeuAc formed. Comparison of rate, product ratio, and stereochemical data
indicate that concurrent pathways for formation of N3-NeuAc are operative, with 17% of product forming
from reaction of azide and the tight ion pair, 12% via the solvent separated ion pair, and 6% from the free
NeuAc oxocarbenium ion. From the corrected product ratio data, the lifetime of the oxocarbenium ion was
estimated to be g3 × 10-11 s. Solvolysis of CMP-NeuAc at pL ) 5.0 afforded an observed solvent deuterium
isotope effect (SDIE) kH O/kD O ) 0.45, consistent with specific acid catalysis of glycosidic bond cleavage. A
2
2
SDIE of 0.66 for the apparent bimolecular azide trapping pathway was also observed. An apparent isotope
effect of ∼1.1 for trapping of the N-acetyl neuraminyl oxocarbenium ion by water was determined by product
analysis of azide trapping in H2O and D2O. An ab initio transition state for attack of water on an N-acetyl
neuraminyl oxocarbenium ion model was located which featured a hydrogen bond between the oxocarbenium
ion carboxylate and water; proton transfer was not part of the reaction coordinate. It is proposed that the
N-acetyl neuraminate carboxylate group stabilizes an intermediate oxocarbenium ion, but the barrier for capture
by water is lowered by a transition state hydrogen bond.
Introduction
the hydrolysis and synthesis of the glycosidic bond.4 The
Phillips hypothesis5 concerning the mechanism of the retaining
glycosylase hen egg white lysozyme (HEWL) included the
proposition that after cleavage of the glycosidic bond, an enzyme
bound oxocarbenium ion was formed which ion-paired with the
carboxylate of active site residue Asp 52. It was later found
that covalent adducts could be isolated for retaining glycosylases
which corresponded to collapse of the ion pair (or by direct
capture in the transition state),6 and the inability to trap a
glucosyl oxocarbenium ion in water led to the suggestion that
oxocarbenium ions are too unstable to exist with a significant
lifetime in the presence of anionic nucleophiles.7a,b
On the other hand, a contemporary crystallographic study of
lysozyme has suggested that HEWL Asp 52 is too distant to
form an acylal covalent intermediate without significant strain.8
Recently, azide trapping studies of mutants of the retaining
glycosylase â-galactosidase have been found to afford galactosyl
azide and galactose in ratios paralleling those obtained for
trapping of relatively stable carbenium ions in aqueous solution,
lending support for the existence of the galactosyl oxocarbenium
ion as an enzyme bound intermediate.9
CMP-NeuAc, 1, is the sugar-nucleotide1 donor substrate for
sialyltransferases, enzymes responsible for addition of N-acetyl
neuraminic acid (sialic acid) to oligosaccharide chains of
glycoproteins and glycolipids.2 The structure of 1 is unusual
for a sugar nucleotide because the leaving group is a mono-
phosphate, not a diphosphate, and because the glycon NeuAc
contains a carboxylate group directly attached to the anomeric
center. We are interested in the chemistry of 1 to help
understand the mechanistic features of sialyltransferases,3 and
because of its structure, it may represent a useful model for the
active site of enzymes which catalyze glycosyltransfer, as
described below.
(4) Sinnott, M. L. Chem. ReV. 1990, 90, 1171-1202. Withers, S. G. Pure
Appl. Chem. 1995, 67, 1673-1682.
(5) Phillips, D. C. Proc. Royal. Acad. Sci. 1967, 57, 484-495.
(6) White, A.; Tull, D.; Johns, K.; Withers, S. G.; Rose, D. R. Nat. Struct.
Biol. 1996, 3, 149-154. Withers, S. G.; Street, I. P. J. Am. Chem. Soc.
1988, 110, 8551-8553.
A long standing question in the area of glycosyltransfer
centers on the nature of intermediates which may arise during
(1) Comb, D. G.; Watson, D. R.; Roseman, S. J. Biol. Chem. 1966, 241,
5637-5642.
(2) Harduin-Lepers, A.; Recchi, M.-A.; Delannoy, P. Glycobiology 1995,
5, 741-758. Varki, A. Glycobiology 1992, 2, 25-40. Reglero, A.; Rodriguez-
Aparicio, L. B.; Luengo, J. M. Int. J. Biochem. 1993, 11, 1517-1527. Hayes,
B. K.; Varki, A. J. Biol. Chem. 1993, 268, 16155-16169.
(3) Bruner, M.; Horenstein, B. A. Biochemistry 1998, 37, 289-297.
(7) (a) Amyes, T. L.; Jencks, W. P. J. Am. Chem. Soc. 1989, 111, 7888-
7900. (b) Banait, N. S.; Jencks, W. P. J. Am. Chem. Soc. 1991, 113, 7951-
7958.
(8) Strynadka, N. C. J.; James, M. N. G. J. Mol. Biol. 1991, 220, 401-
424.
(9) Richard, J. P.; Huber, R. E.; Heo, C.; Amyes, T. L.; Lin, S.
Biochemistry, 1996, 35, 12387-12401.
S0002-7863(97)02503-1 CCC: $15.00 © 1998 American Chemical Society
Published on Web 02/05/1998

1358 J. Am. Chem. Soc., Vol. 120, No. 7, 1998
Horenstein and Bruner
Figure 2. 1H-NMR product analysis for solvolysis of CMP-NeuAc in
the presence of 0.9 M total azide in H2O at pH 5.0. The region of the
spectrum shown corresponds to resonances for the C3 equatorial
hydrogens of the R/â anomers of N3-NeuAc and NeuAc. The reaction
conditions are described in the text.
Figure 1. Kinetics for acid-catalyzed solvolysis of CMP-NeuAc, 1,
with varied concentrations of NaN3 in H2O and D2O. The solvolysis
reactions were conducted at pL ) 5.04, 37 °C, in 1.8 M acetic acid Na
acetate buffer as described in the text. [azide] refers to [N3-] after
correction for pL. The squares (0) correspond to data for D2O; the
circles (o) correspond to data for H2O. The lines are the best fit to the
data by linear regression.
inverse solvent isotope effect of 0.45 ( 0.01 is expressed for
the hydrolytic reaction, and the ratio of the slopes show that an
inverse solvent isotope effect of 0.66 ( 0.12 exists for the azide
reaction. We considered the possibility that a salt effect was
operative in these experiments since the acetate buffer concen-
tration was fixed, while azide was varied. When the molarity
of the acetate buffer was varied along with azide so as to
maintain a constant ionic strength of 1.95, identical kinetics were
observed within experimental error. If a nonspecific salt effect
exists, it is small under the conditions employed in these
experiments.
The glycosyltransfer of CMP-NeuAc is salient to this issue
in that the NeuAc sugar contains an ionized carboxylate residue
adjacent to the anomeric carbon. The carboxylate could function
in a number of different capacities which are reflective of the
chemical reactivities found in glycosylase active sites. During
solvolysis, the carboxylate could stabilize an oxocarbenium ion
like transition state and/or oxocarbenium ion intermediate, act
as an intramolecular nucleophile, or mediate general acid-base
catalysis. The existence of an oxocarbenium ion derived from
N-acetyl neuraminic acid would provide an example supporting
the chemical possibility for existence of enzyme/glycosyl cation
ion pairs. Sinnott and colleagues have shown that solvolysis
of aryl glycosides of N-acetyl neuraminic acid can proceed with
nucleophilic participation of the carboxylate group in the
transition state, possibly resulting in an intermediate R-lactone.10
Recently it was shown that acid-catalyzed solvolysis of CMP-
NeuAc at pH 5 in water/methanol mixtures provides equal
amounts of R- and â-methyl glycoside products, consistent with
the formation of an oxocarbenium ion intermediate after
departure of CMP.11 In this work we present the results of
solvent deuterium isotope effect experiments, azide trapping
studies during solvolysis of CMP-NeuAc, and ab initio calcula-
tions in order to characterize the lifetime of the oxocarbenium
ion and the nature of acid/base catalysis over the reaction
coordinate.
Product Analysis of Azide Trapping. CMP-NeuAc (10-
15 mM) was solvolyzed at pL 5.04 ( 0.02 at 37 °C in 1.8 M
Na acetate buffers containing varied concentrations of NaN3
until HPLC indicated >98% conversion. No azide adducts were
detected at low azide concentration (0.005 M). 1H-NMR spectra
(Figure 2) of the reaction mixtures containing 0.9 M total azide
(actual [N3-] ) 0.63 M at pH 5.04) showed resonances for R-
and â-NeuAc equatorial C-3 hydrogens (R- δ2.72, dd; â- δ2.25,
dd), and resonances for R- and â-N3-NeuAc equatorial C-3
hydrogens were also observed (R- δ2.67, dd; â- δ2.29, dd).13
By integration of the equatorial hydrogens at C3, the relative
percentages of R- and â-NeuAc and R- and â-N3-NeuAc were
4, 61, 32, and 3, respectively. Of the original CMP-NeuAc,
35% was converted into N3-NeuAc. The results were similar
for solvolysis in D2O at pD ) 5.04 (Supporting Information).
The total azide employed was 0.9 M, but due to the pKa shiftl4
to 5.15 for hydrazoic acid in D2O, the actual N3- concentration
was 0.39 M. The relative percentages of R- and â-NeuAc and
R- and â-N3-NeuAc were 5, 68, 25, and 2, respectively.
Methanolysis of CMP-NeuAc in the Presence of Azide.
CMP-NeuAc was solvolyzed in 18% (v/v) aqueous methanol
at pH 5.0 in the presence of 0.9 M azide and in its absence.
After complete consumption of CMP-NeuAc, the reaction
mixtures were concentrated, worked up as described, and
subjected to a product analysis by 1H-NMR. In the absence of
azide, ∼1:1 R- and â-methyl ketosides of NeuAc were observed,
in agreement with previous results.11 An identical ratio of
methyl ketosides were formed in the presence of azide, in
addition to R- and â-N3-NeuAc and NeuAc in similar propor-
tions as described above for solvolysis in the absence of
methanol.
Results
Azide Rate Dependence for CMP-NeuAc Solvolysis in
H2O and D2O. The solvolysis CMP-NeuAc (500 µm) at pL
5.04 ( 0.02 at 37 °C in 1.8 M Na acetate buffers containing
varied concentrations of NaN3 was followed by HPLC to afford
apparent first-order rate constants. Figure 1 provides a plot of
the apparent first-order rate constants versus azide concentration
for reactions in H2O and D2O. The data show that the rate
exhibits a modest dependence on azide concentration. The
respective observed rate constants for solvolysis by water, kobs
and azide, kaz in H2O and D2O are Hkobs ) (5.2 ( 0.1) × 10-3
Ab Initio Transition State Structure for Capture of an
N-Acetyl Neuraminyl Oxocarbenium Ion Model by Water.
Previous work has established that a local minimum energy
conformation for the N-acetyl neuraminyl oxocarbenium ion
D
H
min-l, kobs ) (11.6 ( 0.1) × 10-3 min-l, kaz ) (2.1 ( 0.3)
× 10-3 M-l min-l, and kaz ) (3.2 ( 0.4) × 10-3 M-l min-l.
D
The ratios of the intercepts of the plot in Figure 1 show that an
(10) Ashwell, M.; Guo, X.; Sinnott, M. L. J. Am. Chem. Soc. 1992, 114,
10158-10166.
(11) Horenstein, B. A.; Bruner, M. J. Am. Chem. Soc. 1996, 118, 10371-
10379.

Enzyme Bound Glycosyl Oxocarbenium Ion Intermediates
J. Am. Chem. Soc., Vol. 120, No. 7, 1998 1359
Scheme 1
Figure 3. Ab initio (B3LYP/6-31G*) transition state structure for attack
of water on an N-acetyl neuraminyl oxocarbenium ion model. Water
oxygen O5 is attacking the anomeric carbon C1. Atom O2 is the
pyranosyl ring oxygen, C4, O8, and O9 are the carboxylate group atoms.
model featured a planar arrangement for the C-C-O-C atoms
comprising the oxocarbenium ion atoms with the carboxylate
group roughly perpendicular to this plane.13 A water molecule
was added to this structure which was then allowed to optimize
to a transition state using density functional theory (B3LYP/
6-31G*). The structure is presented in Figure 3, along with
key geometric parameters. This transition structure was char-
acterized by a single imaginary frequency of -192 cm-1. The
atomic displacements for the reaction coordinate featured
formation of the C-O glycosidic bond as the major contributor,
accompanied by pyrimidalization at the anomeric carbon, and
loosening of the oxocarbenium ion C-O bond. The “glyco-
sidic” C-O bond length is 2.07 Å, and the endocyclic C-O
bond is 1.30 Å. One water hydrogen atom is hydrogen bonded
to the carboxylate group with an O-O internuclear distance of
2.50 Å. The stable vibrational frequencies for the hydrogen
bond include an OH stretch of 2628 cm-1 and a complex mode
at 1808 cm-1 which couples the water HOH bend, and hydrogen
bonded OH stretch to the asymmetric CdO stretch of the
carboxylate.15 Transfer of the water hydrogen to the carboxylate
was not a component of the transition state.
A reasonable and productive protonation site is the glycosidic
phosphate oxygen which would lead to the tight ion pair as the
first product. A partial reaction sequence is presented in Scheme
1. The classic specific acid catalyzed reaction would involve
rapid equilibrium protonation as shown in path A of the scheme.
It has been suggested19 that the pKa of the glycosidic phosphate
oxygen might be too low to afford a high enough concentration
of the CMP-NeuAc conjugate acid to be kinetically competent,
possibly requiring catalysis in the transition state as shown in
path B. At this time we cannot rule out either possibility. The
calculated equilibrium solvent isotope effect is ∼0.32 for
hydronium ion catalyzed generation of the tight ion pair. This
is somewhat more inverse than the experimentally observed
value of 0.45. Expression of the isotope effect may reflect
kinetic complexity as outlined in Scheme 2.22 If internal return
k-1 of the tight ion pair is similar to the net forward chemistry
of the ion pair k2, the size of the isotope effect will be sensitive
to the relative values of k-1 and k2, varying between the
equilibrium isotope effect for formation of the ion pair and the
kinetic isotope effect for cleavage of the glycosidic bond when
k-1 > k2 and k-1 < k2, respectively. Support that this scheme
may be operative follows from the observation that the solvent
Discussion
Solvent Isotope Effects for Solvolysis of CMP-NeuAc. The
solvolysis of CMP-NeuAc at pH 5 proceeds with an observed
inverse solvent deuterium kinetic isotope effect (SDIE) of 0.45,
in reasonable agreement with the experimental inverse SDIE
of 0.57 measured for solvolysis of R-D-glucose-l-phosphate.16
The result indicates specific acid catalysis and is consistent with
our earlier conclusion.l1 The results are distinct from those
recently reported for solvolysis of benzylic acetate containing
an ortho carboxylate group,17 and for aqueous decomposition
of CMP-KDO, a sugar-nucleotide similar to CMP-NeuAc by
virtue of a carboxylate at the anomeric carbon and a CMP
leaving group.18
(19) A referee suggested that the glycosidic phosphate oxygen could have
a pKa of -9 to -10. If this pKa is correct, the concentration of CMP-
NeuAc in reactive protonated form at pH 5 would be so low as to require
a rate constant of 1010-1011 s-l for glycosidic bond cleavage to maintain
overall kinetic competency. For a discussion of the timing of proton transfers
and solvent isotope effects, see: Schowen, R. L. Prog. Phys. Org. Chem.
1972, 9, 275-332.
(20) Alvarez, F. J.; Schowen, R. L. In Isotopes in Organic Chemistry;
Buncel, E., Lee, C. C., Eds.; Elsevier: Amsterdam, 1987; Vol. 7, Chapter
1. Kresge, A. J.; More O’Ferrall, R. A.; Powell, M. F. In Isotopes in Organic
Chemistry; Buncel, E., Lee, C. C., Eds.; Elsevier: Amsterdam, 1987; Vol.
7, Chapter 4.
(21) We point out that the rate law for solvolysis has two kinetically
equivalent expressions, the first involving spontaneous reaction of CMP-
NeuAc protonated (presumably) at the carboxylate group, and the second
for acid-catalyzed reaction with the carboxylate ionized.11 Our analysis of
the solvent isotope effects employed the second form of the rate law because
it is more chemically reasonable.
(12) Dabrowski, U.; Friebolin, H.; Brossmer, R.; Supp, M. Tetrahedron
Lett. 1979, 4637-4640.
(13) Horenstein, B. A. J. Am. Chem. Soc. 1997, 119, 1101-1107.
(14) Ritchie, C. D. Physical Organic Chemistry, 2nd ed.; Marcel
Dekker: New York, 1990; p 281.
(15) The Supporting Information includes coordinates for visualization
of these vibrational modes.
(16) Bunton, C. A.; Humeres, E. J. Org. Chem. 1969, 34, 572-576.
(17) Fife, T. H.; Bembi, R.; Natarajan, R. J. Am. Chem. Soc. 1996, 118,
12956-12963.
(18) Lin, C.-H.; Murray, B. W.; Ollman, I. R.; Wong, C.-H. Biochemistry
1997, 36, 780-785.
(22) The steady state assumption was applied to the tight ion pair, and
the expression shown in the scheme was obtained by dividing the rate
expression for reaction in H2O by that for reaction in D2O. The rate constant
k2 represents the net rate for reactions of the tight ion pair in the forward
direction.

1360 J. Am. Chem. Soc., Vol. 120, No. 7, 1998
Horenstein and Bruner
Scheme 2
Scheme 3
isotope effect is more normal for azide trapping than for
solvolysis (0.66 vs 0.45); the net rate constant k2 would be higher
with a contribution from azide capturing some of the ion pair,
with a resultant change in the observed isotope effect away from
the equilibrium isotope effect. Further study will be required
to resolve the structural features of specific acid catalysis in
this system and the possible effect of kinetic complexity on
observed kinetic isotope effects.
Azide Trapping of the N-Acetyl Neuraminyl Oxocarbe-
nium Ion. A fascinating feature of CMP-NeuAc is the presence
of a carboxylate group adjacent to the anomeric center and how
it may influence glycosyltransfer. Kinetic isotope effects for
CMP-NeuAc solvolysis are most consistent with a very late
oxocarbenium ion-like transition state in which the carboxylate
group is coplanar with the oxocarbenium ion plane. Based on
the observation that methanol could be incorporated into NeuAc
during solvolysis of CMP-NeuAc to afford nearly equimolar
amounts of R- and â-glycoside, and by conformational analysis,
we propose that the transition state breaks down to an oxocar-
benium ion intermediate with the carboxylate group orthogonal
to the oxocarbenium ion plane.11,13
nium ion. In part this may be a manifestation of the carboxylate
group conformation: it may be near-planar in the ion pair due
to unfavorable electrostatics with the phosphate of neighboring
CMP, while the free oxocarbenium ion should favor a “twisted”
conformation.l3 Unfavorable electrostatic interactions between
the carboxylate and azide may exist for azide attack on the free
oxocarbenium ion; this is discussed further later. The results
show that the N-acetylneuraminyl oxocarbenium ion exists in
various intermediate forms, but most importantly, indicate that
a free oxocarbenium ion can exist in aqueous solution eVen in
the presence of a strong anionic nucleophile. The free
oxocarbenium ion is not especially long lived, and in the
following discussion we consider its lifetime.
We employed azide ion trapping experiments to obtain further
evidence for the nature of intermediates on the solvolytic
pathway.23a-c The data of Figure 1 show that solvolysis of
CMP-NeuAc has only a modest rate dependence on [azide],
reasonably via reaction of azide with the tight ion pair to afford
1
R-N3-NeuAc. However, H-NMR product analysis data show
that only half of N3-NeuAc formed during solvolysis can be
accounted for by kinetics. The balance of the N3-NeuAc is
assigned to arise from reaction of azide with the solvent
separated ion pair and the free oxocarbenium ion. The
conclusion that some reaction occurs with the free oxocarbenium
ion follows from the 1H-NMR analysis which identified â-azido
NeuAc, which most reasonably would derive from capture of
the free oxocarbenium ion. It would be difficult to justify
formation of â-azido NeuAc from the solvent separated ion pair
since the leaving group CMP is anionic and on the â-face.
Utilizing the NMR product ratios and the N3-NeuAc accounted
for by kinetics, we assign the relative proportions of NeuAc
and N3-NeuAc formed from the tight ion pair, solvent separated
ion pair, and free oxocarbenium ion, respectively, as 65:17:12:
6. Scheme 3 presents a mechanistic scheme which is consistent
with the results. Solvent ROH only attacks the free oxocarbe-
nium ion; the ion pairs are not captured by solvent. This point
is supported by the observation that methanolysis of CMP-
NeuAc in the presence and absence of azide gave identical ratios
of R- and â-methyl NeuAc glycoside products. If solvent
capture of ion-pairs were significant, addition of the much better
nucleophile azide would compete with solvent for formation
of the R-methyl glycoside, with an attendant change in the R/â
ratio of methyl glycosides, which was not observed. Interest-
ingly, azide appears selective over solvent for trapping ion-pairs,
but solvent is more selective than azide for the free oxocarbe-
Lifetime and Reactivity of the N-Acetyl Neuraminyl
Oxecarbenium Ion. The ratio of NeuAc to R/â N3-NeuAc
produced from the free oxocarbenium ion intermediate will be
equal to the ratio of rate constants for their formation, ks/ktrap
-
[N3-]. With knowledge of ktrap, ks can then be calculated; the
reciprocal of ks is equal to the oxocarbenium ion lifetime. The
product ratio NeuAc/R/â N3-NeuAc was 0.093, and the value
for ktrap is required to solve for ks. Values close to 5 × 109
M-1 s-l have been measured experimentally24 for reaction of a
carbenium ion and azide in aqueous solution. This value is
probably too high for trapping the N-acetyl neuraminyl oxo-
carbenium ion because (1) the N-acetyl neuraminyl oxocarbe-
nium has a net molecular charge of zero, not +1, and (2) the
preferred conformation of the carboxylate group of the N-acetyl
neuraminyl oxocarbenium ion should be orthogonal to the
oxocarbenium ion plane13 as in II. On both counts, ktrap for
(24) McClelland, R. A.; Banait, N.; Steenken, S. J. Am. Chem. Soc. 1986,
108, 7023-7027. McClelland, R. A.; Kanagasabapathy, V. M.; Steenken,
S. J. Am. Chem. Soc. 1988, 110, 69l3-69l4.
(25) The sensitivity of diffusion controlled rates to electrostatics is well-
known, see: Rice, S. A. In ComprehensiVe Chemical Kinetics; Bamford,
C. H., Tipper, C. F. H., Compton, R. G., Eds.; Elsevier: Amsterdam, 1985;
Vol. 25, Chapter 3.
(23) (a) Young, P. R.; Jencks, W. P. J. Am. Chem. Soc. 1977, 99, 8238-
8248. (b) Ta-Shma, R.; Rappoport, Z. J. Am. Chem. Soc. 1983, 105, 6082-
6095. (c) Richard, J. P. Tetrahedron 1995, 51, 1535-1573.

Enzyme Bound Glycosyl Oxocarbenium Ion Intermediates
J. Am. Chem. Soc., Vol. 120, No. 7, 1998 1361
the data obtained in D2O, D Oks was estimated27 to be 3 × 1010
2
the N-acetyl neuraminyl oxocarbenium ion should be smaller
than for trapping simple carbenium ions because the electrostat-
ics are less favorable.25 To the best of our knowledge, more
appropriate experimental rate constants have not been reported,
so we employ the known value of 5 × 109 M-1 s-l as an upper
s-l, quite close to the value for H Oks giving an apparent solvent
2
isotope effect of 1.1. This result suggests28 that there is little
proton transfer during attack of water on the oxocarbenium ion.
We modeled the attack of water on the N-acetyl neuraminyl
oxocarbenium ion with ab initio calculations and located the
transition state structure shown in Figure 3. The attacking water
is in a hydrogen bond to the carboxylate group with the
hydrogen residing squarely on the water oxygen, and proton
motion within the hydrogen bond is not a part of the reaction
coordinate but is a stable vibration. Thus the transition state
hydrogen bond is very similar to a ground state hydrogen bond,
which would be expected to result in an isotope effect of
approximately unity,29 in agreement with the experimental
results. The hydrogen bonded array of water and carboxylate
can still be considered as a catalytic grouping on an electrostatic
basis.2lc,30 As glycosidic bond formation occurs, charge is
transferred to the attacking water. A hydrogen bond to the
negatively charged carboxylate would serve to stabilize the shift
in charge that occurs during progressive formation of the new
glycosidic C-O bond, lowering the barrier for reaction.
limit for ktrap
.
It has been pointed out that general base catalysis is not
anticipated to be significant for unstable carbenium ions.30b
Because of the proximate carboxylate in the N-acetyl neuraminyl
oxocarbenium ion, and its potential for facilitating addition of
water, we suggest that the separate consideration of oxocarbe-
nium ion stability and general base catalysis may not be possible
in the present system because these factors are likely to be
intertwined. A substantial intrinsic lifetime of the N-acetyl
neuraminyl oxocarbenium ion is predicted because it contains
a proximate stabilizing carboxylate group, but this is largely
offset because this same group is likely to facilitate water
addition to the oxocarbenium ion, shortening its lifetime.
With this consideration in mind, the calculated first-order rate
constant for water trapping of the cation, ks, is 3.4 × 1010 s-1
;
the lifetime (a lower limit) is 3 × 10-11 s, slightly briefer than
the earlier estimate.11 This lifetime is just sufficient26a,b to
account for the results of trapping by aqueous methanol in which
an approximately 1:1 ratio of methyl glycosides was observed
and proposed to arise by reaction of methanol with a diffusion-
ally equilibrated oxocarbenium ion. The N-acetyl neuraminyl
oxocarbenium ion lifetime of g3 × 10-11 s is at least 30 times
greater than the estimated lifetime7a for the glucosyl oxocar-
benium ion in pure water(∼10-12 s). It has been proposed that
a 2-deoxy sugar oxocarbenium ion will be more long lived than
the corresponding 2-hydroxy species by a factor of 4.7a So in
the comparison between the glucosyl oxocarbenium ion and that
of NeuAc, when the 2-deoxy factor is taken into account, the
relative stability conferred by the carboxylate group is 30/4,
which at 37 °C would be worth about -1.2 kcal/mol as a
minimum value.
The proposed interrelationship of N-acetyl neuraminyl oxo-
carbenium ion stability and intramolecular catalysis by the
carboxylate group may find parallel in enzyme-catalyzed
glycosyl transfer. Azide trapping studies9 on mutants of
â-galactosidase which had the catalytic general acid/base Glu
461 mutated to either glycine or glutamine showed a marked
ability to trap the galactosyl-enzyme intermediate to afford
azido- galactoside and galactose products; the wild type enzyme
does not incorporate azide. In part, removal of the Glu
carboxylate group favors reaction with anionic nucleophiles, but
also the apparent first-order rate constant for capture of the
galactosyl-enzyme complex by water was reduced: >74-fold
for E461G for reaction with water and 40 000-fold for E461G
for reaction with trifluoroethanol.
Water and Oxocarbenium Ion Carboxylate Group Inter-
action. Although the N-acetyl neuraminyl oxocarbenium ion
is more long lived than the glucosyl oxocarbenium ion, it is
surprising that it does not have a much longer lifetime than it
does. Even accounting for solvation, the immediately proximate
carboxylate group and oxocarbenium ion carbon should enjoy
a substantial stabilizing interaction that other unsubstituted
glycosyl cations do not. In comparison to a 2-deoxy pyranosyl
oxocarbenium ion, a 2-deoxy R-carboxylate substituted pyra-
nosyl oxocarbenium ion was estimated to be -17 kcal/mol more
stable by ab initio calculations which included an aqueous
solvation model.13 The implication is that the thermodynamic
stability is unrealized because of other factors. If the carboxylate
were to function as an intramolecular general base it would
shorten the lifetime by facilitation of oxocarbenium ion capture
by water (but not azide). This might be detectable by a solvent
isotope effect, but an experimental difficulty faced in establish-
ing this is that capture of the oxocarbenium ion by solvent is a
fast step, rendering it kinetically silent by conventional tech-
niques. We considered that competitive azide trapping might
provide a means of accessing this rapid step, where azide
(26) (a) The rate constant for trapping the free oxocarbenium ion by
water cannot be faster than the rate for diffusional separation of the tight
ion pair; an estimate26b of the diffusional separation rate constant is 1.6 ×
1010 s-1 for azide and a substituted benzyl carbenium ion. This is likely to
be a lower limit for CMP and the N-acetyl neuraminyl oxocarbenium ion
due to less favorable electrostatics. (b) Richard, J. P.; Jencks, W. P. J. Am.
Chem. Soc. 1984, 106, 1373-1383.
(27) A value of 4.2 × 109 M-1 s-l for trapping by azide in D2O was
used to account for the 1.2-fold higher viscosity of D2O versus H2O at 37
°C. Kirshenbaum, I. Physical Properties of HeaVy Water; McGraw Hill:
New York, 1951.
(28) A near unity solvent isotope effect could be the result of fortuitous
cancellation of nonunity contributions and must therefore be considered
cautiously.
(29) Schowen, K. B.; Schowen, R. L. Methods Enzymol. 1982, 87, 551-
606.
(30) (a) Swain, C. G.; Kuhn, D. A.; Schowen, R. L. J. Am. Chem. Soc.
1965, 87, 1553-1561. (b) Richard, J. P.; Jencks, W. P. J. Am. Chem. Soc.
1984, 106, 1396-1401.
(31) Zapata, G.; Vann, W. F.; Aaronson, W.; Lewis, M. S.; Moos, M. J.
Biol. Chem. 1989, 264, 14769-14774.
trapping in H2O versus D2O could provide H Oks/D Oks. From
2
2

1362 J. Am. Chem. Soc., Vol. 120, No. 7, 1998
Horenstein and Bruner
were passed through a Pasteur pipette containing Amberlite IR120-H+
resin to remove cations, then concentrated in Vacuo, and exchanged
against D2O three times for NMR analysis. The desalting step was
necessary as it was observed that if this step were neglected, the solution
was quite alkaline after concentration, resulting in exchange of the
â-hydrogens in NeuAc in D2O, complicating the analysis. 1H-NMR
analysis (Figure 2) was used to determine the ratios of R- and â-N3-
NeuAc and of NeuAc based on integration of the respective C3
equatorial hydrogens in the following way. The resonances for R- and
â-N3-NeuAc were assigned by the reported chemical shifts for these
compounds.12 The upfield half of the R-NeuAc resonance (doublet of
doublets, dd) was buried under the R-N3-NeuAc dd, while the upfield
half of the â-N3-NeuAc dd was too close to the â-NeuAc dd to
satisfactorily integrate these latter two resonances. The area for
R-NeuAc was taken to be twice the area of the resolved downfield
half of its dd; the area corresponding to the resolved half of the
R-NeuAc was then subtracted from the area of the overlapping R-N3-
NeuAc and R-NeuAc multiplets to provide the corrected integration
for R-N3NeuAc. The corrected integral areas for â-N3-NeuAc and
â-NeuAc were determined in the same way. The equilibrium ratio of
R- and â-NeuAc was calculated to serve as an internal check of the
method: we calculate 6:94 and 7:93 R/â ratios after solvolysis in H2O
and D2O, respectively; the literature reports 5:95 and 8:92 equilibrium
ratios for R/â NeuAc in aqueous solution,34 indicating that the method
is reliable.
Calculations. Ab initio calculations were performed using Gaussian
94,35 revision C.3 on a Silicon Graphics Indigo XZ workstation and an
IBM RS 6000 SP. Calculations employed the 6-31G* basis set and
density functional theory (DFT) with the Becke 3 parameter exchange
functional and the Lee-Yang-Parr correlation functional.36,37 The
transition state structure for attack of water on the R-carboxylate
substituted pyranosyl oxocarbenium ion13 was optimized to tight
criterion (Figure 3). A frequency calculation on the transition state
structure afforded a single imaginary frequency which corresponded
to atomic motion on the desired reaction coordinate. Cartesian
coordinates for the transition state structure and selected sets of
Cartesian displacements for vibrational modes of interest are provided
in the Supporting Information.
Conclusions
The key features of the solvolytic pathway of CMP-NeuAc
at pH 5 are (1) glycosidic bond cleavage is specific acid
catalyzed; (2) formation of N3-NeuAc from CMP-NeuAc occurs
from ion pairs and the free oxocarbenium ion; (3) the free
oxocarbenium ion has a lifetime of g3 × 10-11 s; and (4)
capture of the oxocarbenium ion by water does not appear to
utilize intramolecular general base catalysis, but the rate may
be facilitated by favorable electrostatic interaction between the
attacking water and carboxylate group. The N-acetyl neuraminyl
oxocarbenium ion exists as an intermediate in the presence of
azide and its own carboxylate group, providing an example of
the chemical feasibility of the intermediacy of this oxocarbenium
ion in sialyltransferase and neuraminidase active sites. How-
ever, the lifetime of the intermediate should be short in the
presence of catalytic functionality that facilitates transfer to the
glycosyl acceptor.
Experimental Section
Materials. Buffers and reagents were purchased from Sigma and
Fisher. Plasmid pWV200B harboring the expression construct31 for
E. coli CMP-NeuAc synthase was a gift from Dr. W. F. Vann at the
National Institutes of Health. CMP-NeuAc synthase32 and CMP-
NeuAc11 were prepared as previously described.
Instrumental. 1H-NMR spectra were measured at ambient tem-
perature on a Varian Gemini300 spectrometer operating at 300 MHz.
The acquisition pulse angle was 30°. Spectra were obtained in 99.9%
D D2O referenced to the HDO peak (4.80 ppm). A preacquisition
saturation of the HDO line was used for solvent suppression. HPLC
was performed on a Rainin HPXL gradient unit interfaced to a
Macintosh personal computer. A Rainin Dynamax UV-1 detector was
employed to monitor separations at 260 nm.
Measurement of Apparent First-Order Rate Constants for
Solvolysis. Solvolysis reactions were conducted at 37° C in 0.5 mL
polypropylene microfuge tubes. Reaction mixtures were 500 µM in
CMPNeuAc, 1.8 M CD3COONa buffer, pL ) 5.04 ( 0.02. Reactions
were initiated by addition of the appropriate volume of buffer to a
solution of CMP-NeuAc in deionized water and allowed to proceed
for 3-4 half lives. The course of solvolysis was followed by HPLC
(MonoQ HR10/10, 85 mM NH4HCO3, 15% methanol, pH 7.8, 2 mL/
min, A260) whereby integration of the unreacted CMP-NeuAc (13.6
min) versus the product CMP (18.4 min) allowed calculation of the
percent remaining CMP-NeuAc using eq 1. The extinction coefficients
of CMP-NeuAc and CMP were determined to be identical within
experimental error. The time points and corresponding progress of
reaction data were fit to eq 2 using MacCurveFit to obtain the best fit
for the apparent first-order rate constant for solvolysis, kobs; plots of ln
%CMPNeuAc versus time showed excellent linearity over 3-4 half lives.
Acknowledgment. Support of this work by the National
Science Foundation (CAREER award MCB-9501866), the
University of Florida Division of Sponsored Research, and the
Northeast Regional Data Center at the University of Florida is
gratefully acknowledged. We thank Dr. W. F. Vann of the NIH
for the gift of an E. coli clone containing a CMP-NeuAc
synthase expression plasmid. We wish to thank a referee for
several insightful comments.
Supporting Information Available: 1H-NMR spectral data
for product analysis of azide trapping in D2O. 1H-NMR spectral
data for methanol incorporation in the absence and presence of
azide. Cartesian coordinates for selected normal vibrational
modes of the calculated transition state structure (4 pages).
See any current masthead page for ordering and Internet access
instructions.
% CMPNeuAc ) ACMPNeuAc/(ACMP + ACMPNeuAc) * 100 (1)
ln %CMPNeuAc ) kobs*t + %CMPNeuAc0
(2)
Solvolyses of CMP-NeuAc for Product Analysis. Reactions (1.0
mL) were conducted in 1.5 mL polypropylene microcentrifuge tubes
and consisted of 10 mM CMP-NeuAc, 1.8 M CD3COONa pL 5.0 in
the presence of the appropriate amount of NaN3 and/or methanol. For
reactions in D2O, 0.4 was added to the observed pH meter readings
obtained with a combination glass electrode (Orion).33 The reactions
were maintained at 37 °C and monitored by HPLC until CMP-NeuAc
consumption was greater than 98% (14 h, pH 5.0). Reaction mixtures
JA972503J
(34) FrieboIin, H.; Supp, M.; Brossmer, R.; Keilich, G.; Ziegler, D.
Angew. Chem., Int. Ed. Engl. 1980, 19, 208-209.
(35) Gaussian 94, revision C.3; Frisch, M. J.; Trucks, G. W.; Schlegel,
H. B.; Gill, P. M. W.; Johnson, B. G.; Robb, M. A.; Cheeseman, J. R.;
Keith, T.; Petersson, G. A.; Montgomery, J. A.; Raghavachari, K.;
Al-Laham, M. A.; Zakrzewski, V. G.; Ortiz, J. V.; Foresman, J. B.;
Cioslowski, J.; Stefanov, B. B.; Nanayakkara, A.; Challacombe, M.; Peng,
C. Y.; Ayala, P. Y.; Chen, W.; Wong, M. W.; Andres, J. L.; Replogle, E.
S.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Binkley, J. S.; Defrees, D. J.;
Baker, J.; Stewart, J. P.; Head-Gordon, M.; Gonzalez, C.; Pople, J. A.
Gaussian, Inc.: Pittsburgh, PA, 1995.
(32) Liu, J. L.; Shen, G.-J.; Ichikawa, Y.; Rutan, J. F.; Zapata, G.; Vann,
W. F.; Wong, C.-H. J. Am. Chem. Soc. 1992, 114, 3901-3910.
(33) Schowen, K. B. J. In Transition States of Biochemical Processes;
Gandour, R. D., Schowen, R. L., Eds.; Plenum Press: New York, 1978;
Chapter 6.
(36) Becke, A. D. J. Chem. Phys. 1993, 98, 5648-5652.
(37) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785-789.