Elusive transition state of alcohol dehydrogenase unveiled

Article abstract of DOI:10.1073/pnas.1000931107

For several decades the hydride transfer catalyzed by alcohol dehydrogenase has been difficult to understand. Here we add to the large corpus of anomalous and paradoxical data collected for this reaction by measuring a normal (>1) 2° kinetic isotope effect (KIE) for the reduction of benzaldehyde. Because the relevant equilibrium effect is inverse (<1), this KIE eludes the traditional interpretation of 2° KIEs. It does, however, enable the development of a comprehensive model for the tunneling ready state (TRS) of the reaction that fits into the general scheme of Marcus-like models of hydrogen tunneling. The TRS is the ensemble of states along the intricate reorganization coordinate, where H tunneling between the donor and acceptor occurs (the crossing point in Marcus theory). It is comparable to the effective transition state implied by ensemble-averaged variational transition state theory. Properties of the TRS are approximated as an average of the individual properties of the donor and acceptor states. The model is consistent with experimental findings that previously appeared contradictory; specifically, it resolves the long-standing ambiguity regarding the location of the TRS (aldehyde-like vs. alcohol-like). The new picture of the TRS for this reaction identifies the principal components of the collective reaction coordinate and the average structure of the saddle point along that coordinate.

Full text of DOI:10.1073/pnas.1000931107

Elusive transition state of alcohol
dehydrogenase unveiled
1
Daniel Roston and Amnon Kohen
Department of Chemistry, University of Iowa, Iowa City, IA 52242
Edited by Alan R. Fersht, Medical Research Council Centre for Protein Engineering, Cambridge, United Kingdom, and approved April 13, 2010 (received for
review January 22, 2010)
For several decades the hydride transfer catalyzed by alcohol de-
hydrogenase has been difficult to understand. Here we add to the
large corpus of anomalous and paradoxical data collected for this
reaction by measuring a normal (>1) 2° kinetic isotope effect (KIE)
for the reduction of benzaldehyde. Because the relevant equili-
brium effect is inverse (<1), this KIE eludes the traditional inter-
pretation of 2° KIEs. It does, however, enable the development
of a comprehensive model for the “tunneling ready state” (TRS)
of the reaction that fits into the general scheme of Marcus-like
models of hydrogen tunneling. The TRS is the ensemble of states
along the intricate reorganization coordinate, where H tunneling
between the donor and acceptor occurs (the crossing point in
Marcus theory). It is comparable to the effective transition state
implied by ensemble-averaged variational transition state theory.
Properties of the TRS are approximated as an average of the indi-
vidual properties of the donor and acceptor states. The model is
consistent with experimental findings that previously appeared
contradictory; specifically, it resolves the long-standing ambiguity
regarding the location of the TRS (aldehyde-like vs. alcohol-like).
The new picture of the TRS for this reaction identifies the principal
components of the collective reaction coordinate and the average
structure of the saddle point along that coordinate.
predicts a late TS for this slightly endothermic reaction (internal
Keq ¼ 0.15, i.e., ∼1 kcal∕mol) (9). Klinman also examined 2° ki-
netic isotope effects (KIEs) only to implicate an aldehyde-like TS
(8). 2° KIEs measure the ratio of reaction rates of substrates that
differ only in isotopic substitution of a bond that is not cleaved
during the reaction. They serve as tools for determining TS struc-
ture because they reflect changes in hybridization along the reac-
tion coordinate. The traditional formulation of 2° KIEs predicts
that a very early TS will give a KIE of unity and a very late TS will
give a KIE equal to the equilibrium isotope effect (EIE) (11). A
few years later, though, experiments on the homologous horse
liver ADH (hlADH) found that 2° KIEs on the nicotinamide ex-
ceeded the 2° EIEs. To explain these results, Cleland and co-
workers suggested that the reaction involves quantum mechanical
H tunneling and coupled motion between the 1° and 2° hydrogens
(
Fig. 1) (12). Subsequently, H tunneling appeared to be a com-
mon feature for H transfer in both enzymatic and nonenzymatic
reactions (1, 13).
Concurrent with these experimental findings, several theoreti-
cians developed a simplified empirical underpinning to the con-
cept of tunneling and coupled motion and predicted that
deviations from the semiclassical Swain–Schaad exponents
(
(
SSEs) (14) could serve as strong evidence for such behavior
15–17). Semiclassically, the SSE is
hydrogen tunneling ∣ enzyme kinetic ∣ secondary isotope effect ∣
Swain–Schaad
SSE ¼ lnðk ∕k Þ∕ lnðk ∕k Þ ¼ 3.3;
[1]
H
T
D
T
nzymes enhance the rates of chemical reactions by many or-
ders of magnitude, and extensive studies have uncovered
E
where k is the reaction rate with isotope i. Confirmation of this
i
many aspects of how they do so. The prevailing theory that en-
zymes stabilize the transition state (TS) relative to the reactants
explains many phenomena, and a great deal of contemporary
research focuses on how enzymes stabilize the TS. An unfortu-
nate hindrance to developing an understanding of how enzymes
stabilize the TS lies in the enigmatic nature of the TS. Many
enzymatic hydrogen (proton, hydrogen, or hydride) transfers,
for example, occur by quantum mechanical tunneling, where a
particle passes through an energy barrier because of its wave
properties (1–6).
prediction came quickly thereafter in a seminal paper by Klin-
man’s group that found a glaringly elevated mixed-labeling
SSE (mSSE) for 2° KIEs in yADH (18), where
H
H
H
T
D
D
D
T
mSSE ¼ lnðk ∕k Þ∕ lnðk ∕k Þ;
[2]
i
j
where k is the rate when i is the hydrogen isotope at the 1° posi-
tion and j is the hydrogen isotope at the 2° position. An elevated
mSSE later appeared in other ADHs (19, 20) as well as other
types of enzymatic hydrogen transfers (21).
Yeast alcohol dehydrogenase (yADH) serves as an excellent
model system for enzyme-catalyzed H transfer because unlike
many other enzymes, the chemical step, oxidation of a primary
Some empirical models (22, 23) as well as quantum mechan-
ical/molecular mechanical simulation (QM/MM) studies (24–27)
have achieved moderate success in replicating the observed 2°
mSSEs. While not rigorously defined in those studies, the concept
of “tunneling and coupled motion” complicated the traditional
interpretation of 2° KIEs, so that the location of the TS need
not be proportional to the 2° KIEs. A more comprehensive em-
pirical model is proposed below.
þ
alcohol to an aldehyde by NAD , is rate-limiting with aromatic
substrates (7). Furthermore, the thermodynamics of the reaction
allow for examination of both the forward (alcohol to aldehyde)
and reverse (aldehyde to alcohol) reactions under similar
conditions (8). This type of nicotinamide-dependent redox reac-
tion appears ubiquitously in biology, and a detailed picture of the
TS of such reactions may facilitate many medical and industrial
applications.
Author contributions: D.R. and A.K. designed research; D.R. performed research; A.K.
contributed new reagents/analytic tools; D.R. and A.K. analyzed data; and D.R. and
A.K. wrote the paper.
Intriguingly, decades of experiments on the reaction catalyzed
by yADH have returned what appear to be contradictory results,
some suggesting an early TS, whereas others point to a late TS
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
(
7–10). In pioneering studies, Klinman measured linear free
energy relationships (LFERs) by using benzyl substrates in both
forward and reverse directions and found evidence to support an
alcohol-like TS (7, 8), despite the fact that Hammond’s postulate
1
To whom correspondence should be addressed. E-mail: amnon-kohen@uiowa.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/
doi:10.1073/pnas.1000931107/-/DCSupplemental.
9
572–9577 ∣ PNAS ∣ May 25, 2010 ∣ vol. 107 ∣ no. 21
www.pnas.org/cgi/doi/10.1073/pnas.1000931107

aldehyde (36). Our measurements differ from unity and appear
to be quite precise as a result of using only the aromatic-active
isozyme of yADH (10) and a broad range of time points.
The term tunneling and coupled motion has been applied to
ADHs and other systems that exhibit 2° KIEs outside the semi-
classical range (12, 28, 37, 38), as well as reactions that violate the
Rule of the Geometric Mean, as evidenced by inflated mSSEs
(
1, 18, 21). The term tunneling and coupled motion appears to
be used for both abnormalities in 2° KIEs. Below, we describe
the development of a model that interprets these abnormalities
without any explicit inclusion of “coupled motion” and show that
the two types of observations differ in their physical origins.
Fig. 1. Schematic TRS of the reverse reaction (aldehyde to alcohol) showing
°–2° coupling. The hydrogens are black, the carbons are gray, the oxygen is
1
red, and the three heavy atoms define the blue plane. Traditional models of
tunneling and coupled motion proposed that the reaction coordinate in-
volved motion of all three hydrogens (shown with arrows). The model
presented here parameterized the out-of-plane bending angles of the benzyl
substrate and nicotinamide cofactor, θs and θc, respectively, in order to obtain
a symmetric double-well potential along the H-transfer coordinate.
Modeling the TRS. The model developed here fits into the frame-
work of Marcus-like models of H tunneling in enzymes (1, 4,
30–32). A generalized functional form of these phenomeno-
logical models gives a rate constant (k) as
Here we report measurements of H/T and D/T 2° KIEs (cf.
Z
ðΔGþλÞ2
DAD0
DAD1
k ∕k and k ∕k ) on benzaldehyde in the reverse reaction that
k ¼ Ce−
4λRT
Fðm;DADÞ −EðDADÞ∕kT
e e
H
T
D
T
dDAD;
[3]
fall outside the range of the EIE to unity, a finding that mirrors 2°
KIEs on the nicotinamide cofactor (12). The KIEs reported here
completed the set of 2° KIEs for this reaction (both substrates in
both directions). This set guided our development of a more com-
plete empirical model that explains all the apparently contradic-
tory findings regarding the ADH reaction. This model implies a
where the first exponential is the traditional “Marcus term” for a
reaction with driving force ΔG and reorganization energy λ at
temperature T. The integral includes terms that adapt Marcus
theory to the situation of H tunneling. The first exponential inside
the integral, the “Franck–Condon term,” is a nuclear overlap in-
tegral, which describes the efficiency of tunneling through a given
barrier and is a function of the tunneling particle’s mass (m) and
the DAD. The second integrated exponential, the “Gating term,”
measures the relative contributions of the ensemble of DADs,
which are in dynamic equilibrium. This phenomenological model
was developed to describe the temperature dependency of 1°
KIEs, where most of the isotopic sensitivity is in the nuclear
overlap term. The mass of the 2° hydrogens, however, does
not significantly affect the integrated terms, and 2° KIEs are
mostly embedded in the Marcus term of this equation.
The physical interpretation of Eq. 3 mirrors the VTST treat-
ment (39) that has been quite successful in modeling enzymatic
reactions that involve H tunneling (24, 25, 33). In these models,
the enzyme facilitates tunneling of the transferred particle by
creating energetic degeneracy between a donor and acceptor
state. When the two states are degenerate, the probability density
“
tunneling ready state” (TRS), which is a special case of the more
general TS (28). The TRS is the ensemble of states from which H
tunneling occurs and like the general case of TS is a saddle point
along a coordinate that represents the motions that bring the sys-
tem from the ground state to the TRS. The present approach uses
the framework of Marcus-like models of hydrogen tunneling (1,
4
, 28–32) and identifies some of the principle components of the
collective reaction coordinate. This empirically parameterized
model quantitatively replicates all of the 2° KIEs in the forward
and reverse directions, as well as their mSSEs, and the LFER
findings. The model also indicates an asynchronized hybridiza-
tion for the donor and acceptor carbons, which was first identified
for dihydrofolate reductase (DHFR) by QM/MM calculations
(
33). Finally, in accordance with the ensemble-averaged varia-
tional transition state theory (VTST) results from Truhlar,
Gao, and co-workers (24, 25), the model supports the recently
articulated hypothesis that the inflated mSSEs reported for many
ADHs since 1989 result from deflated 2° D/T KIEs because of the
shorter donor–acceptor distance (DAD) required for D transfer
(
1, 34). The present model mirrors some of the previous empirical
calculations on ADH and other systems (16, 17, 22, 23, 35) but
provides a more comprehensive molecular and electronic inter-
pretation. The interpretation of experiment by this method
agrees well with the results from QM/MM simulations (24, 25)
that also led to deviations from semiclassical 2° mSSEs in the for-
ward direction.
Results and Discussion
Measuring 2° H/T and D/T KIEs on Benzaldehyde Reduction. Tradition-
ally, models expected the magnitude of 2° KIEs to vary between
unity for a reactant-like TS and the EIE for a product-like TS (see
ref. 33 for a thorough discussion), but quantum mechanical
tunneling and 1°–2° coupled motion have given experimental
results that elude such simplistic interpretations. The 2° H/T
KIE measured here for the reduction of benzaldehyde by NADH
clearly escapes interpretation by semiclassical models. The value
we measured (as described in Materials and Methods) was
Fig. 2. Marcus-like model of a reaction with H tunneling. At the TRS (‡), the
reactant (Black) and product (Gray) surfaces are degenerate, which allows the
probability density of the hydrogen (Shaded) to spread from the donor well
to the acceptor well (i.e., quantum mechanical tunneling) at
a rate
1
.05 ꢀ :01, which is obviously outside the range of unity to the
dependent on the DAD and the particle’s mass. Once degeneracy is broken,
the hydride wave function can collapse into the acceptor well, giving a net
transfer (i.e., dissipative tunneling). The Marcus coordinate is a complex
amalgamation of many modes, some of the most important of which are
discussed in the text.
inverse EIE (0.75, ref. 36). An early attempt to measure this
KIE produced results within error of unity, which did not surprise
the authors because the KIE in the forward direction (alcohol to
aldehyde) was equal to the EIE, suggesting a TRS close to the
Roston and Kohen
PNAS
∣
May 25, 2010
∣
vol. 107
∣
no. 21
∣
9573

of the transferred particle (depending on its mass and the DAD)
passes from the donor well to the acceptor well (Fig. 2). Thus, the
TRS is the Marcus crossing point, at which H transfer from donor
to acceptor can occur. We approximated the TRS as a linear com-
bination of two states that are degenerate distortions of the heavy
atom framework and differ only as to whether the transferred
hydride is in the donor or acceptor well.
To find the structure of the heavy atom skeleton at the TRS, we
adapted Redington’s method to model the TRS of the proton
transfer in tropolone (40). The substrates were optimized to a
geometry with the transferred H at the midpoint on a straight line
between the donor and acceptor carbons by using a range of
DADs. Because these structures did not yield the degenerate
double-well potential necessary for tunneling, the hybridizations
of the donor and acceptor carbons were parameterized (see SI
Text) to give a symmetric potential energy surface for the motion
of the H between the donor and acceptor (Fig. 3). The TRS was
then approximated as a linear combination of the two states
where the transferred H is in the donor well or the acceptor well.
The geometric mean of the KIEs calculated for the donor and
acceptor states of the TRS at each DAD was compared with
experiment to determine the best fit. Parameterization of the
Zn-O bond length at the TRS also enhanced the fit to experimen-
tal KIEs and is discussed in SI Text. Fig. 4 shows the structure that
best matched experimental KIEs, Table 1 shows the values of the
KIEs, and Table 2 lists important geometrical values in the TRS
and their corresponding values in the ground state structures.
Whereas there might be other structural solutions that fit the
data, these are outside the conformational space we scanned
and might not be physically relevant. The solution we report
may not be the only one that reproduces some of the KIEs,
but it is physically meaningful and it meets a large number of
experimental restrictions that are quite sensitive to the geometric
parameters surveyed.
Fig. 4. TRS structure for H transfer found by the present methods, showing
all heavy atoms used in the model along with hydrogen atoms of particular
interest. The dividing surface of conformations that yield a TRS with two
degenerate wells (ΔED-A ¼ 0) is quite broad, so this structure represents
the weighted average (via Boltzmann distribution) of the full conformational
ensemble for which ΔED-A ¼ 0. The 1° hydrogen is shown in both the donor
and acceptor positions, but faded, because it has just half of its probability
density in each position. Important geometric values are listed in Table 2.
2.8 Å, and the zero point energy of the hydride is above the bar-
rier even beyond that. This phenomenon echoes a model by
Hammes-Schiffer and co-workers that found no barrier to this
reaction at such short DADs but demonstrated how tunneling
can contribute to rates with a DAD in excess of 3.1 Å (26).
Furthermore, a model of the H transfer in soybean lipoxygenase
found that the “distance of most probable transfer” (the DAD at
the TRS) could be as long as 3.25 Å (41). To clarify, the DAD
refers to the distance between the two carbon atoms, but the
actual distance between the two minima of the double-well po-
tential at the TRS described here is just 0.8 Å, which is within
the de Broglie wavelength of H at room temperature (1.0 Å at
an energy of kBT). Nonetheless, to ensure the feasibility of
tunneling at 3.2 Å, we calculated the tunneling splitting for
the double-well potential in Fig. 3 and found a value of
0.19 kcal∕mol, which is similar to previous estimates of this value
using other levels of theory (26). The tunneling splitting is inver-
sely proportional to the rate of H tunneling at the TRS, and the
value we found indicates that once the system reaches the TRS,
all of the probability density of the hydride will pass from donor to
acceptor in 250 fs (42).
As one would expect, the TRS in this model has a somewhat
longer DAD (3.2 Å) than the distance of 2.6 Å cited in some pre-
vious ADH models (24, 27). That shorter distance referred to the
saddle point for the over-the-barrier process, and the authors of
those studies stressed that corrections for tunneling played a vital
role in replicating experimental KIEs. The DAD in our calcula-
tions, though, refers to that of the TRS, from which tunneling
occurs. In our model no barrier is present for DADs shorter than
Hybridization at the TRS. Parameterization of the donor and accep-
tor hybridization gave a particularly transparent view of the role
of rehybridization in the reaction. By using the parameterized
out-of-plane bending angle of the 2° hydrogens (Fig. 1), we
H
2
calculated hybridization (sp ) at the TRS ranging from sp to
3
sp where
Table 1. Computed and experimental 2° KIEs and EIEs
†
Forward*
Comp Exp
Reverse*
Comp Exp
Equilibrium
Comp
Exp
§
¶
§
§
§
¶
H2a (H/T)**
H2b (H/D)**
H2a (D/T)***
1.33§
1.33∥
1.05§
1.05
1.26§
1.33
0.89
∥
∥
1.06§
1.03
1.08
1.22§
0.97
1.24§
1.01
0.87
††
1.03
Experimental errors are omitted here for clarity but are available in SI
Text. The 2° hydrogens H2a and H2b are indicated in Fig. 4.
*
The forward direction is the oxidation of benzyl alcohol and the
reverse is the reduction of benzaldehyde.
†
Equilibria refer to the forward direction.
**H transfer.
Fig. 3. Least-squares quartic fit to 25 single point calculations at the
ꢁ
§
B3LYP∕6-31 þ G level along the H-transfer coordinate at the TRS. The
This work.
¶
vibrational wave functions of the ground state (Dashed Line) and first excited
state (Dashed-Dotted Line) of the tunneling hydride were calculated as
described in SI Text. This surface is equivalent to the 1D slice of the “hydride
coordinate” at the TRS in Fig. 2.
Ref. 36.
∥
Ref. 12.
***D transfer.
Ref. 18.
†
†
9
574
∣
www.pnas.org/cgi/doi/10.1073/pnas.1000931107
Roston and Kohen

Table 2. Properties of the TRS and ground states
replicate the observed 2° KIEs for D transfer nor does it give in-
flated mSSEs or the 1° KIEs. However, prompted by the recent
proposition of Klinman (1, 34), we tested whether a structure
with a shorter DAD would reproduce the observed behavior
for D transfer. Shortening the DAD by 0.2 Å did, in fact, give
the diminished value of 1.03 for the D/T 2° KIE with D transfer
in the forward direction, leading to an inflated mSSE in excellent
agreement with the experimental data (18). At the shorter DAD,
the effects of hybridization on energy (both total energy and
degeneracy between donor and acceptor) are nearly identical
to those at the longer DAD, so the hybridization at the TRS
is the same with the transfer of the heavier nucleus. Nonetheless,
noncovalent interactions between the substrates diminish the 2°
D/T KIEs and are the main source of the inflated mSSEs that
were long used to indicate tunneling and coupled motion in
the ADH reaction.
Reactants*
Products*
TRS
C7-C4**
C7-C4***
Zn-O
C7 hybridization
C4 hybridization
3.8 Å
3.8 Å
1.8 Å
3.8 Å
3.8 Å
2.3 Å
3.2 Å
3.0 Å
2.25 Å
3
2
2.76
sp
sp
sp
2
3
2.34
sp
sp
sp
*
Reactants and products refer to those of the forward
reaction.
*H transfer.
**D transfer.
*
*
1
80 − θ
H ¼ 2 þ 1
:
[4]
80 − θ0
Notably, the current model suggests that when a different
nucleus is transferred, the system goes through a different TRS.
Eq. 3 above predicts this type of behavior, because the contribu-
tion of the integral to the rate constant is greater for a heavier
nucleus at shorter DADs, and similar findings resulted from
ensemble-averaged variational transition state theory calcula-
tions of ADH (24). A recent VTST study of another system also
showed that an H transfer from NADH proceeds via large-
curvature tunneling, but the corresponding D transfer is domi-
nated by small-curvature tunneling (43). To further explore this
property, we calculated the H/D 1° KIE on the basis of the vibra-
tional zero point energy of H and D for the double-well potential
at a TRS of 3.2 and 3.0 Å, respectively, along with the difference
in energy between the two TRSs (see SI Text). This simple approx-
imation gave the value of 3.0 for the H/D 1° KIE in both direc-
tions, which agrees well with experimental measurements of this
value (7, 8). This supporting evidence for Klinman’s recent sug-
gestion of the source of the inflated 2° mSSE (1, 34) is likely to
assist the development of more general and less empirical models
that account for the effect of a shorter DAD on reactions with D
or T transfer in both enzymatic and nonenzymatic reactions.
Here θ is the out-of-plane bend at the TRS, and θ is that angle in
0
the relevant reduced form of each substrate. The calculated
2
.76
hybridizations at the TRS are sp
for the donor (C7 of the sub-
2
.34
strate) and sp
for the acceptor (C4 of the cofactor—Fig. 5).
The hybridizations of the donor and acceptor carbons, therefore,
are unsynchronized, as found for DHFR (33). Other models have
described anomalous rehybridization like this as 1°–2° coupled
motion (16, 17, 22, 23) but gave little insight into the physical
source of the coupling. The present picture provides a more
concrete description of the motion of the 2° hydrogens and
how it is necessary to give a symmetric double-well potential
at the TRS. The unsynchronized rehybridization reported here
and for DHFR (33) may be a general phenomenon in nicotin-
amide-dependent redox reactions and even in other types of
reactions involving H tunneling. We are currently examining this
question in other enzymes that involve both hydride and proton
transfers.
According to the current model, the physical origin of the in-
flated 2° H/T KIEs (relative to semiclassical predictions) is the
effect of tunneling of the 1° hydrogen (delocalized at the
TRS) on the 2° vibrational modes, as has been suspected since
the earliest observations of 2° KIEs outside the semiclassical
range (12, 18), and was first demonstrated for another ADH
by QM/MM calculations (24).
A Unified Picture of the Reaction Coordinate. In addition to replicat-
ing all of the KIEs measured for yADH, the model presented
here rationalizes the apparent paradox regarding the traditional
interpretations of 2° KIEs, LFERs, and the Hammond postulate.
The LFER experiments showed that substituents at the para po-
sition of the substrate have little effect on the oxidation of alco-
hols but substantially alter the rate of aldehyde reduction, which
typically indicates a TS that more closely resembles the alcohol
The Situation with D Transfer. Up to this point, we have addressed
only the (physiological) H-transfer scenario. Some of the most
inexplicable behavior of this reaction, however, particularly
inflated mSSEs, occurs when D is the transferred nucleus (1). Im-
portantly, the TRS structure we present for H transfer does not
(
7, 8, 10). For many years, this finding was seen as contradictory
to the data on 2° KIEs, which, when interpreted semiclassically,
suggest an aldehyde-like TS. To examine the present model’s con-
sistency with the studies on substituent effects, we explored how
two representative electron-donating and electron-withdrawing
para substituents (-OCH and -Br) affected the charge density
3
on the benzylic carbon in both directions. A recent analysis of
substituent effects on a similar system showed that Mulliken
charge (Q ) on the benzylic carbon, as calculated by the theory
M
used in this model, scales linearly with Hammett substituent con-
stants (44), suggesting that QM at the TRS vs. reactants (ΔQ )
M
could serve as a probe for reactivity. In accordance with experi-
mental rates, para substituents do not change ΔQM significantly
in the forward direction, but they give drastic changes in the re-
verse reaction (Fig. 6). That is not to say that QM on the benzylic
carbon does not change between the alkoxide and the TRS—only
that substituents do not affect that change. In the case of alde-
hyde reduction, though, substituents strongly modulate how
charge accumulates on the benzylic carbon during the reaction.
In terms of the Marcus-like model used as a framework for
these calculations, ΔQM appears to be a good probe for the re-
organization along the Marcus coordinate (the first exponential
Fig. 5. Rehybridization during the course of the reaction. The location of
the TRS is marked (‡) and the dashed line indicates a reasonable reaction
pathway from reactants to products that passes through the TRS. The solid
line indicates a reaction pathway with perfectly synchronized rehybridiza-
tion. The dotted line indicates the surface with perfectly symmetric rehybri-
dization.
Roston and Kohen
PNAS
∣
May 25, 2010
∣
vol. 107
∣
no. 21
∣
9575

Conclusions
For nearly four decades, data addressing the yADH catalyzed
reaction appeared contradictory and could not be rationalized
within a single model. New measurements of H/T and D/T 2°
KIEs for the yADH catalyzed reduction of benzaldehyde by
NADH deviated from the range predicted by semiclassical
theories. These data completed the set of 2° KIEs available
for this enzyme and allowed us to develop an empirical model
that is consistent with all the data. This model shed light on
the nature of the phenomenon described in the past as H tunnel-
ing and 1°–2° coupled motion (12, 15–18).
The picture of the reaction emphasizes the dual importance of
a short DAD as well as degenerate donor and acceptor wells at
the TRS. Such a picture opens the doorway to a better under-
standing of how enzymes stabilize the TRS. The current findings
are likely to spur both experimental and theoretical work to
better understand the TRS of yADH and other enzymatic and
nonenzymatic reactions. To test and consolidate this picture,
its predictions will be assessed experimentally. The present model
predicts, for example, that H/T 2° KIEs in the forward direction
will be greater with H transfer than with D transfer and that the 2°
SSE will conform to the semiclassical value so long as both the
H/Tand D/T KIEs are measured with the same nucleus at the 1°
position. Additionally, the findings presented here are likely to
inspire high-level theoretical studies of yADH, for which a crystal
structure is now available (Protein Data Bank ID code 2HCY), to
illuminate the question of how the substrates reach the TRS and
what role the enzyme plays in that process.
Fig. 6. The effect of para substituents on the forward and reverse reactions
catalyzed by yADH. In accordance with previous studies (7, 8) the experimen-
tal rates (Dotted Line) (10) showed that the forward reaction is unaffected by
þ
electronic changes (substituents with different values of
σ
)
but that
electron-withdrawing substituents greatly accelerate the reverse reaction.
Calculations (Dashed Line) of the reaction with a representative range of
para substituents (-OCH3,-H,-Br) showed similar trends in the change of
Mulliken charge (ΔQM) on the benzylic carbon between reactants and the
TRS. Substituents on benzyl alcohol (forward) do not affect the electronic
changes along the reaction coordinate, but substituents on benzaldehyde
(
reverse) severely alter the electronic changes that accompany the TRS.
term of Eq. 3). Reorganization energy indicates the extent to
which the electrostatic environment of the tunneling particle
must change in order to reach the TRS (45). The fact that elec-
tronic changes have little effect in the forward direction suggests
that the system is well preorganized for the forward reaction.
That is, by binding the substrates and forming the Zn-alkoxide
intermediate, the enzyme accomplishes much of the necessary
changes to reach the TRS. In the reverse direction, however, re-
organization after binding plays a crucial role and electronic
changes affect the extent to which the system must reorganize
to reach the TRS.
Materials and Methods
Detailed experimental and computational methods are available in SI Text. A
brief description is provided below.
Kinetic Experiments. The H/T and D/T 2° KIEs for the reduction of benzalde-
hyde were measured competitively in reaction conditions designed to mimic
3
those used in previous studies of this enzyme (18). [7- H]-benzaldehyde was
3
synthesized by reduction of benzoyl chloride by ½ Hꢂ-NaBH4 (48). [ring-
1
4
C]-benzaldehyde (as a trace for 1H or H) was commercially available
2
2
14
(
American Radiolabeled Chemicals). [7- H; ring- C]-benzaldehyde was pro-
duced by CN-catalyzed exchange of the aldehydic proton with D2O (49).
How can one reconcile an early TRS (for the forward reaction)
with the overall endothermicity of the reaction? The Hammond
postulate predicts a late TRS for an endothermic reaction. The
R-½4-2Hꢂ-NADH (NADD) was produced by ADH-catalyzed reduction of
þ
14
NAD by EtOH-d6 (50). To measure the H/T KIE, [ring- C]-benzaldehyde
3
was copurified with [7- H]-benzaldehyde and the substrate was consumed
by an excess of NADH in the presence of yADH. The D/T KIE used [7- H;
2
þ
equilibrium in solution favors the alcohol-NAD side of the re-
ring-14C]-benzaldehyde and NADD. The reactions were quenched at a range
of time points, and samples were analyzed as described in SI Text to measure
the depletion of tritium in the product as a function of fractional conversion,
to yield the KIE.
action (at catalytically relevant pH) (7), but the solution equili-
brium is not directly relevant to the rate-limiting hydride transfer
step under study. Measurements of internal (on the enzyme)
equilibrium gave a value of K ¼ 0.15 in favor of the alcohol-
eq
þ
NAD side, predicting a late TRS (9), but the internal equili-
Computational Modeling. KIEs and EIEs were calculated on the basis of
isotopic differences in vibrational frequencies between reactants and the
TRS according to the Bigeleisen equation (51) by using the program ISOEFF
brium is close to unity and very likely a combination of at least
two steps: the deprotonation of the alcohol and the hydride trans-
fer that follows (46). Indeed, QM/MM calculations for this reac-
tion found that the hydride transfer step is exothermic, despite
the fact that the overall reaction was endothermic (24, 27).
Our results indicating an alkoxide-like TRS for the hydride trans-
fer bolster the argument that this step truly is exothermic.
The overall course of the reaction appears to be the following:
After the Zn alkoxide forms, the Zn-O bond lengthens, accom-
panied by changes in the hybridization of the donor and acceptor
carbons to reach a symmetric double-well potential along the
H-transfer coordinate (Fig. 3). Once the system achieves this
TRS, rate-promoting vibrations of the enzyme (47) allow the sub-
strates to sample a range of DADs. When the DAD reaches 3.2 Å
0
7 (52). Vibrational frequencies were calculated by Gaussian 03 (53) at the
B3LYP level of theory. To model the effects of hydrogen tunneling on KIEs,
the TRS was treated as a simple linear combination of degenerate donor and
acceptor states:
1
1
ΨTRS ≈ pffi ffiffi Ψ þ pffiffi Ψ :
[5]
d
a
2
2
Vibrational frequencies at the TRS were approximated as the average
between the frequencies calculated for the individual donor and acceptor
states (see SI Text for details). To obtain fits to experimentally determined
KIEs, geometrical properties of the TRS were parameterized as described
in the text.
ACKNOWLEDGMENTS. We thank Bryce Plapp for assistance in expression and
purification of the enzyme and useful discussions. We thank Judith Klinman
and Chris Cheatum for useful discussions and insightful comments on the
manuscript. This work was supported by National Science Foundation Grant
CHE-0133117 and National Institutes of Health Grant R01 GM65368 (to A.K.).
D.R. is supported by a predoctoral fellowship from the National Institutes of
Health (T32 GM008365).
(
weighted average), the probability density of the hydride can
spread from the donor well to the acceptor well on a time scale
similar to the lifetime of the TRS. Once the degeneracy of the
TRS dissolves, the wave function of the tunneling hydride
collapses and it is trapped in the product state.
9
576
∣
www.pnas.org/cgi/doi/10.1073/pnas.1000931107
Roston and Kohen

1
2
3
4
5
6
7
. Nagel ZD, Klinman JP (2006) Tunneling and dynamics in enzymatic hydride transfer.
Chem Rev 106:3095–3118.
. Pu JZ, Gao JL, Truhlar DG (2006) Multidimensional tunneling, recrossing, and the
transmission coefficient for enzymatic reactions. Chem Rev 106:3140–3169.
. Hammes-Schiffer S (2006) Hydrogen tunneling and protein motion in enzyme reac-
tions. Acc Chem Res 39:93–100.
27. Cui Q, Elstner M, Karplus M (2002) A theoretical analysis of the proton and hydride
transfer in liver alcohol dehydrogenase (LADH). J Phys Chem B 106:2721–2740.
28. Pudney CR, Hay S, Sutcliffe MJ, Scrutton NS (2006) Alpha-secondary isotope effects as
probes of “tunneling-ready” configurations in enzymatic H-tunneling: Insight from
environmentally coupled tunneling models. J Am Chem Soc 128:14053–14058.
29. Kiefer PM, Hynes JTKohen A, Limbach H-H (2006) Isotope Effects in Chemistry and
Biology (Taylor & Francis, Boca Raton), pp 549–578.
. Kohen A (2006) Isotope Effects in Chemistry and Biology, eds A Kohen and H-H Limbach
(
Taylor & Francis, Boca Raton), pp 743–764.
3
0. Marcus RA (2007) H and other transfers in enzymes and in solution: Theory and
computations, a unified view. 2. Applications to experiment and computations. J Phys
Chem B 111:6643–6654.
. Masgrau L, et al. (2006) Atomic description of an enzyme reaction dominated by
proton tunneling. Science 312:237–241.
. Warshel A, et al. (2006) Electrostatic basis for enzyme catalysis. Chem Rev
3
3
3
1. Kuznetsov AM, Ulstrup J (1999) Proton and hydrogen atom tunnelling in hydrolytic
and redox enzyme catalysis. Can J Chem 77:1085–1096.
2. Knapp MJ, Klinman JP (2002) Environmentally coupled hydrogen tunneling—Linking
catalysis to dynamics. Eur J Biochem 269:3113–3121.
3. Pu JZ, et al. (2005) Nonperfect synchronization of reaction center rehybridization in
the transition state of the hydride transfer catalyzed by dihydrofolate reductase. J Am
Chem Soc 127:14879–14886.
1
06:3210–3235.
. Klinman JP (1972) Mechanism of enzyme-catalyzed reduced nicotinamide adenine
dinucleotide-dependent reductions—Substituent and isotope-effects in yeast
alcohol-dehydrogenase reaction. J Biol Chem 247:7977–7987.
. Klinman JP (1976) Isotope-effects and structure-reactivity correlations in yeast alcohol-
dehydrogenase reaction—Study of enzyme-catalyzed oxidation of aromatic alcohols.
Biochemistry 15:2018–2026.
8
9
3
3
3
4. Klinman JP (2006) Linking protein structure and dynamics to catalysis: the role of
hydrogen tunnelling. Philos T R Soc B 361:1323–1331.
. Dickinson FM, Dickenson CJ (1978) Estimation of rate and dissociation-constants invol-
ving ternary complexes in reactions catalyzed by yeast alcohol-dehydrogenase.
Biochem J 171:629–637.
5. Lin S, Saunders WH (1994) Tunneling in elimination-reactions—Structural effects on
the secondary beta-tritium isotope effect. J Am Chem Soc 116:6107–6110.
6. Welsh KM, Creighton DJ, Klinman JP (1980) Transition-state structure in the yeast
alcohol-dehydrogenase reaction—The magnitude of solvent and alpha-secondary
hydrogen isotope effects. Biochemistry 19:2005–2016.
1
0. Pal S, Park D-H, Plapp BV (2009) Activity of yeast alcohol dehydrogenases on benzyl
alcohols and benzaldehydes: Characterization of ADH1 from Saccharomyces carlsber-
gensis and transition state analysis. Chem Biol Interact 178:16–23.
1. Streitwieser A, Jagow RH, Fahey RC, Suzuki S (1958) Kinetic isotope effects in the
acetolyses of deuterated cyclopentyl tosylates. J Am Chem Soc 80:2326–2332.
2. Cook PF, Oppenheimer NJ, Cleland WW (1981) Secondary deuterium and N-15 isotope
effects in enzyme-catalyzed reactions—Chemical mechanism of liver alcohol-dehydro-
genase. Biochemistry 20:1817–1825.
1
3
7. Wilde TC, Blotny G, Pollack RM (2008) Experimental evidence for enzyme-enhanced
coupled motion/quantum mechanical hydrogen tunneling by ketosteroid isomerase. J
Am Chem Soc 130:6577–6585.
1
38. Karsten WE, Hwang CC, Cook PF (1999) Alpha-secondary tritium kinetic isotope effects
indicate hydrogen tunneling and coupled motion occur in the oxidation of L-malate
by NAD-malic enzyme. Biochemistry 38:4398–4402.
1
1
3. Truhlar DG Tunneling in enzymatic and nonenzymatic hydrogen transfer reactions. J
Phys Org Chem doi: 10.1002/poc.1676.
4. Swain CG, Stivers EC, Reuwer JF, Schaad LJ (1958) Use of hydrogen isotope effects to
identify the attacking nucleophile in the enolization of ketones catalyzed by acetic
acid. J Am Chem Soc 80:5885–5893.
39. Truhlar DG, Garrett BC (1984) Variational transition-state theory. Annu Rev Phys Chem
35:159–189.
40. Redington RL (2000) H atom and heavy atom tunneling processes in tropolone. J Chem
Phys 113:2319–2335.
1
5. Saunders WH (1985) Calculations of isotope effects in elimination-reactions—New
experimental criteria for tunneling in slow proton transfers.
07:164–169.
6. Huskey WP, Schowen RL (1983) Reaction-coordinate tunneling in hydride-transfer
reactions. J Am Chem Soc 105:5704–5706.
7. Huskey WP (1991) Origin of apparent Swain-Schaad deviations in criteria for tunnel-
ing. J Phys Org Chem 4:361–366.
8. Cha Y, Murray CJ, Klinman JP (1989) Hydrogen tunneling in enzyme-reactions. Science
J
Am Chem Soc
41. Meyer MP, Klinman JP (2005) Modeling temperature dependent kinetic isotope effects
for hydrogen transfer in a series of soybean lipoxygenase mutants: The effect of
anharmonicity upon transfer distance. Chem Phys 319:283–296.
1
1
1
1
1
4
4
2. Cohen-Tannoudji C, Diu B, Laloë F (1977) Quantum Mechanics (Wiley, New York).
3. Pang JY, Hay S, Scrutton NS, Sutcliffe MJ (2008) Deep tunneling dominates the
biologically important hydride transfer reaction from NADH to FMN in morphinone
reductase. J Am Chem Soc 130:7092–7097.
2
43:1325–1330.
4
4. Essa AH (2007) Hammett MSP Taft DSP analysis of substituent effects on Mulliken
9. Bahnson BJ, Park DH, Kim K, Plapp BV, Klinman JP (1993) Unmasking of hydrogen
tunneling in the horse liver alcohol-dehydrogenase reaction by site-directed muta-
genesis. Biochemistry 32:5503–5507.
0. Kohen A, Cannio R, Bartolucci S, Klinman JP (1999) Enzyme dynamics and hydrogen
tunnelling in a thermophilic alcohol dehydrogenase. Nature 399:496–499.
1. Alston WC, Kanska M, Murray CJ (1996) Secondary WT and D/T isotope effects in en-
zymatic enolization reactions. Coupled motion and tunneling in the triosephosphate
isomerase reaction. Biochemistry 35:12873–12881.
2. Rucker J, Klinman JP (1999) Computational study of tunneling and coupled motion in
alcohol dehydrogenase-catalyzed reactions: Implication for measured hydrogen and
carbon isotope effects. J Am Chem Soc 121:1997–2006.
3. Kohen A, Jensen JH (2002) Boundary conditions for the Swain-Schaad relationship as a
criterion for hydrogen tunneling. J Am Chem Soc 124:3858–3864.
charges of 1-(arylmethylene)-1H-cyclopropanaphthalene. Int
07:1574–1577.
J Quantum Chem
1
4
4
5. Villa J, Warshel A (2001) Energetics and dynamics of enzymatic reactions. J Phys Chem
B 105:7887–7907.
6. Shearer GL, Kim KY, Lee KM, Wang CK, Plapp BV (1993) Alternative pathways and
reactions of benzyl alcohol and benzaldehyde with horse liver alcohol dehydrogenase.
Biochemistry 32:11186–11194.
2
2
4
4
4
7. Antoniou D, Schwartz SD (2001) Internal enzyme motions as a source of catalytic
2
activity: Rate-promoting vibrations and hydrogen tunneling.
05:5553–5558.
J Phys Chem B
1
8. Entwistle ID, Boehm P, Johnstone RAW, Telford RP (1980) Metal-assisted reactions. 8.
Selectivity in the reaction of organic halides with tetrahydroborate and the reduction
of acyl halides to aldehydes. J Chem Soc Perk T 1:27–30.
2
2
2
4. Alhambra C, Corchado JC, Sanchez ML, Gao JL, Truhlar DG (2000) Quantum dynamics
of hydride transfer in enzyme catalysis. J Am Chem Soc 122:8197–8203.
5. Alhambra C, et al. (2001) Canonical variational theory for enzyme kinetics with the
protein mean force and multidimensional quantum mechanical tunneling
dynamics. Theory and application to liver alcohol dehydrogenase. J Phys Chem B
9. Chancellor T, Quill M, Bergbreiter DE, Newcomb
M (1978) Formyl-D aromatic-
aldehydes. J Org Chem 43:1245–1246.
50. Rafter GW, Colowick SP (1957) Enzymatic preparation of Dpnh and Tpnh. Method
Enzymol 3:887–890.
1
05:11326–11340.
51. Bigeleisen J (1949) The relative reaction velocities of isotopic molecules. J Chem Phys
17:675–678.
52. Anisinov V, Paneth P (2007) ISOEFF 07. (Technical University of Lodz, Lodz, Poland).
53. Frisch MJ, et al. (2004) GAUSSIAN 03. (Gaussian, Inc, Wallingford, CT) Revision E.01.
2
6. Webb SP, Agarwal PK, Hammes-Schiffer S (2000) Combining electronic structure
methods with the calculation of hydrogen vibrational wavefunctions: Application
to hydride transfer in liver alcohol dehydrogenase. J Phys Chem B 104:8884–8894.
Roston and Kohen
PNAS
∣
May 25, 2010
∣
vol. 107
∣
no. 21
∣
9577