Semiempirical AM1 and PM3 studies of the enzymatic mechanism of horse liver alcohol dehydrogenase

THEO CHEM Journal of Molecular Structure (Theochem) 364 (1996) 33-43

Semiempirical AM1 and PM3 studies of the enzymatic mechanism of horse liver alcohol dehydrogenase Sylvia A.M. Vanhommerig”>*, Robert J. Meierb, Lamoraal A.%. Sluyterman”, Emmo M. Meijera ‘Laboratory

ofOrganicChemistry, ‘DSM

Eindhoven University of Technology, P.O. Box 513,560U MB Eindhoven, The Netherlands Research, P.O. Box 18, 6160 MD Geleen, The Netherlands

Received 22 May 1995; accepted in final form 6 July 1995

Abstract Semiempirical (AM1 and PM3) calculations on active site models have been performed to study the mechanism of horse liver alcohol dehydrogenase (HLADH). The active site model used in the calculations consists of a Zn(I1) ion coordinated by derivatives of Cys 46, Cys 174, His 67 and an alkoxide/aldehyde, and also by derivatives of Ser 48 and NAD+/NADH. The theoretical calculations show drastic differences in ground state energy levels for model systems incorporating negatively charged cysteine residues compared with active site models based on neutral cysteine residues. The lower enzymatic activity of HLADH towards isopropanol can be rationalized using the active site models presented in this study. A negative charge on hydrogen being transferred in the transition state can be calculated, pointing to a hydride transfer mechanism. Probability calculations suggest that hydrogen tunnelling may occur. However, to draw

definite conclusions one should take into account the dynamics of the enzyme system. In agreement with literature data, water most probably does not act as a fifth ligand of zinc in the ternary complex. It is not clear from the calculations whether water is involved in the proton relay mechanism or not. Keywords:

Alcohol dehydrogenase; Enzymatic mechanism of HLADH; NAD+/NADH

1. Introduction Horse liver alcohol dehydrogenase (HLADH) is a NAD+/NADH dependent enzyme, stereoselectively catalysing the interconversions of various alcohols into their corresponding aldehydes and ketones [l]. In the past many kinetic studies have been carried out with HLADH using NAD+/NADH (analogues) and a great variety of

* Corresponding author.

coenzyme; Transition state modelling

substrates [2-51. We also examined the enzymatic activity of HLADH with various coenzymes and substrates [6] and performed several modelling studies on the ternary complex [7,8]. The generally accepted mechanism of this enzyme is binding of the coenzyme NADf first, followed by binding of the alcohol substrate [9,10]. Upon oxidation of the latter, the aldehyde/ketone product is released prior to the reduced coenzyme NADH [l 11. When a primary alcohol is used as a substrate, the rate-determining step is the dissociation of the coenzyme from the enzyme [12]. With the

0166-1280/96/$15.00 0 1996 Elsevier Science B.V. All rights reserved SSDI 0166-1280(95)04371-3

34

S.A.M.

Vanhommerig et al./Journal of Molecular Structure (Theochem)

secondary alcohol isopropanol, however, the hydride transfer is rate-limiting [ 131. Several three-dimensional (3-D) structures of HLADH have been determined, including an Xray structure of the complex of this enzyme with NADH and DMSO [14]. In this ternary complex, HLADH contains two Zn(I1) ions in each of its two subunits, and one zinc ion per subunit participates in the enzymatic reaction. The alcohol is expected to coordinate to zinc and to react as an alkoxide ion. The zinc ion is furthermore coordinated to Cys 46, His 67 and Cys 174 (see Fig. 1). As the

364 (1996) 33-43

resolution of the 3-D structure of the ternary complex is no more than 2.9 A [14], water molecules cannot be distinguished in this structure; it is therefore not clear whether or not a water molecule is coordinated to this Zn atom. Ramaswamy et al. [ 151 determined a ternary X-ray structure of HLADH with NADf and pentafluorobenzyl alcohol with a resolution of 2.1 A. In this structure the alcohol is indeed ligated to the zinc ion and, in addition, 12 buried water molecules were found in each subunit. Zinc-bound water is absent in the above mentioned structure. cys 174

Ser 48 7H3

-OH

-

S-

aldehyde

\-_o k

--_..

t,

=CH,

I

Fig. 1. (a) Schematic indicates the hydride

His 67

i-7

Cys 46

NADH

representation of the active site model with ethanol as substrate (system I), aldehyde transfer path. (b) Stick model representation of the aldehyde state with ethanol.

state. The dotted

arrow

S.A.M.

Vanhommerig et al./Journal of Molecular Structure (Theochem)

364 (1996) 33-43

cys 174

Fig. 2. Active site model of the TS state as used by Ritter von Onciul and Clark [16]. (a) Arrows site model, system 1 (see Fig. 1). (b) Stick model representation of the TS.

In 1993, Ritter von Onciul and Clark reported a theoretical study of the catalytic oxidation of alcohols by LADH [16]. A model was used for the active site in which the catalytic zinc ion is surrounded by four ligands, a histidine, an alkoxide and two cysteine residues (see Fig. 2). However, they did not use the exact coordinates of the X-ray structure to build their model, and their optimized structures did not resemble the X-ray structure any more. Furthermore, the sulphur atoms of the cysteines contained hydrogen

indicate

the differences

from our active

atoms, resulting in an overall charge of the model of +2. However, it is far more likely that these cysteine residues coordinated to a zinc dication carry negative charges [ 14,17,18]. For example, with papain, it was shown that zinc is capable of removing a proton both from the thiol group of a cysteine and the imidazole group of a histidine in the active site [17]. Also, in several papers [14,18] dealing with HLADH ir has been indicated that the cysteines surrounding the catalytic zinc ion are negatively charged. Even more convincing is the

36

S.A.M.

Vanhommerig et alJJourna1

of Molecular Structure (Theochem] 364 (1996) 33-43

fact that the pK, of the zinc-bound water is shifted from 15.7 to 7.6 [18], suggesting a similar pK, shift for the cysteines (which normally have a pK, of ~9) causing complete ionization of both cysteines in the near-neutral pH range of the enzyme activity. The conclusion therefore seems justified that the zinc-coordinated cysteine residues in HLADH are deprotonated and present as thiolate anions. This would result in an overall charge of zero rather than + 2 for the model complexes as used by Ritter von Onciul and Clark [ 161. In the present paper, semiempirical calculations (AM 1 and PM3) are reported on active site models for the ternary complex of HLADH, using ethanol, 1-propanol and isopropanol as substrates. Our studies aim at improving the insight into the mechanism of hydrogen/hydride transfer in HLADH. The influence of negatively charged cysteine residues on the geometric and electronic features of the model has been examined. Finally, the effect of the presence of water molecules has been included in the current investigation.

2. Methods AM1 [19] and PM3 [20] calculations were performed with the semiempirical package MOPAC~.O on

COO-

H,N+

H,N+

Y s-

Cysteine

coo -

L NV

FH*

an Alliant FX28 16 computer. The crystallographic structure of the ternary complex of HLADHNADH-DMSO (2.9 A resolution) [14] was used to obtain a starting structure for the active site model. To limit the number of atoms involved in the calculations, derivatives of the amino acids and the coenzyme were used. Fig. 3 shows the original residues in the active site of HLADH and their derivatives as used in these calculations. The active site model consists of a zinc(H) ion, coordinated by derivatives of Cys 46, His 67, Cys 174 and alkoxide/aldehyde/ketone (at the position of DMSO), and also by derivatives of Ser 48 and NAD+/NADH. All residues (see Fig. 1) were separately optimized, and these optimized structures were inserted into the active site model at the same position as in the X-ray structure, using the “Molecular similarity --f Match atoms” option from the QUANTA program [22] to obtain the best fit for each derivative with its original counterpart. When performing this kind of calculations, there are two choices one can make: (1) optimize a!] atoms, or (2) apply constraints to maintain good agreement with the crystal structure used as starting structure and stay close to the realistic situation as in natural systems. The second option was chosen in this paper, and resulted in the carbon, sulphur and nitrogen atoms of the cysteines, serine,

CH2 NH

Histidine

k--i N-%/

Adenine

OH

Serine

NADH

W S-CH,

CH,-OH

NH

Fig. 3. Original compounds of the active site of HLADH (upper row) and their derivatives as used in the calculations (lower row).

S.A.M.

Vanhommerig

et al./Journal

of Molecular

histidine and of the nicotinamide ring of NAD+/ NADH being fixed at their X-ray positions. The carboxamide side-chain of NAD+/NADH, all hydrogen atoms, the alkoxide and, if present, the water molecule were allowed to move freely. A GNORM of 0.1 was applied unless stated otherwise. For each model system three states were calculated: the ground state with NAD+ and alkoxide, the transition state and the ground state with NADH and aldehyde/ketone. They will be referred to as the alcohol state, the transition state (TS) and the aldehyde state respectively. The ground states were determined using the EF algorithm. Transition states were located by applying the SADDLE keyword [23]; some structures obtained with SADDLE were further refined using the TS keyword [24].

3. Results and discussion 3.1. The influence of the total charge of the model complexes

To allow a fair comparison between the results using geometrical constraints and a total charge Table 1 AMI and PM3 calculated heats of formation, keyword, unless stated otherwise Total charge

Geometries PM3 system 1 system 2 AM1 system 1 system lb system 2

Structure

(Theochem)

31

364 (1996) 33-43

equal to zero and those published by Ritter von Onciul and Clark, who started from a transition state model of the active site (see Fig. 2) without using geometrical constraints, we first followed their strategy. Within 1 kcal mol-’ differences we found the same energies for two of the three states (see Table 1). The results with the alcohol state could not be reproduced within better than 8 kcal mol-* , probably because the alcohol state had to be derived from the TS. We then compared two active site models with ethanol as substrate (as described under Methods), system 1, with S- groups and an overall charge of zero, and system 2, which contains SH groups and thus an overall charge of + 2. Both active site models consist of derivatives of NAD+/NADH, Ser 48, Cys 46, His 67 and Cys 174, as shown in Fig. 3. Furthermore, they contain ethoxide/acetaldehyde and a tetracoordinated zinc ion. For our active site model with negative charges on the cysteines and an overall charge of zero (system l), we find that the transition state of hydride transfer is 10.7 and 33.4 kcal mol-’ higher in energy than the alcohol state and 39.2 and 47.3 kcal mol-’ higher in energy than the aldehyde state, as calculated with

E,,, and AHr (in kcal molf’).

The transition

states are determined

using the

E act

AH,

SADDLE

AH,

alcohol state

transition state

aldehyde state

forward

backward

0 2

-71.2 +227.7

-37.8 +269.0

-85.1 +255.0

t33.4 141.3

+47.3 +14.0

-13.9 +27.3

0 0 2

-2.4 -2.4 +292.1

+8.3 +4.7 +319.5

-30.9 -30.9 +312.6

+10.7 +7.1 +27.4

+39.2 +35.6 f6.9

-28.5 -28.5 +20.5

t140.1 +139.2

+113.3 +113.3

+25.2 +32.4

+26.8 +25.9

-1.6 +6.5

based on X-ray structure

[14]

Geometries according to Ritter von Onciul & Clark [16] AM1 2 present results +114.9 results Ritter et al. [16] 2 +106.8 a AHr is defined as AH,’ (aldehyde state - AHrO (alcohol b The transition state was refined using the TS keyword.

state).

38

S.A.M.

Vanhommerig et al./Journal of Molecular Structure (Theochem)

364 (1996) 33-43

OCHCH3 L=Zn

/

: H-l._NAD +4 7

&!!

L-Zn PFCH3

=7 1 $2”=35

NAD+

6

___ I ______.______

-2.4

OCHCH3

AHf= -28 5 L=Zn/

i

HfN4D -309 alcohol state Scheme

transition state

I The energy differences in hydride-transfer

AM1 and PM3 respectively (see Scheme 1 and Table 1). The keyword TS resulted in AH,’ = +4.7 kcal mol-’ for the transition state of system 1. With the FORCE keyword the gradients and the number of imaginary frequencies of this calculated TS were checked. Although there are differences in the calculated activation energies with AM1 and PM3, the optimized geometries with AM1 and PM3 of the three states are almost identical. The energy differences found between the calculated AM 1 and PM3 structures are caused only by the differences in the applied method. Only with the optimized geometry of the alcohol state do the two methods yield different results (the amide side-chain of NAD is oriented differently). Qualitatively the values we calculated for system 2 (containing SH groups) resemble the ones obtained for the geometries of Ritter von Onciul and Clark [16] (see Table l), i.e. that the alcohol state is lower in energy than the aldehyde state with all these models. If we now compare our active site model system 1 (with S- groups) and system 2 (with SH groups), the most remarkable difference is the change in the sign of AHr with both AM1 and PM3 (see Table 1). Besides the differences in overall charge and in the side-chains of some amino acids (as illustrated with arrows in Fig. 2) we used geometrical constraints on some of the atoms, whereas Ritter von Onciul and Clark used none. These constraints were applied in order to maintain good agreement with the original X-ray structure. For example, the

aldehyde state calculated

with AM 1; pathway

of system

1

TS structure of Ritter vorr Onciul and Clark [16] shows for some atoms 5 A deviation from the original crystal structure. The accuracy of the atom posiiions in this crystal structure 6ADH [14] is 5 1 A. With our own model, using AM 1 and no or very few constraints, the optimized geometry resulted in ring opening of the pyridinium moiety, owing to the formation of a sigma bond between the ethoxide and the NAD+ ring. When we attempted to use PM3 and no constraints at all, the optimized structures deviated strongly from the X-ray structure. Use of these optimized PM3 structures as starting structures for subsequent AM1 calculations resulted in geometries with much lower energies, but very different from the original X-ray structures; far more so than one would expect taking into account the flexibility of the enzyme structure. Therefore geometrical constraints had to be included in our modelling studies (see Methods). 3.2. Mechanism and rate of hydride transfer in HLADH It is known that with the poor substrate isopropanol the hydride transfer step is rate limiting, in contrast to primary alcohols, where the dissociation of NADH from the binary enzyme complex is the rate-determining step [l l-131. To obtain more information about the mechanism of hydride transfer, the ground and transition states with three different substrates were calculated, using system 1

S.A.M.

Vanhommerig

et al./Journal

of Molecular

Table 2 AM1 calculated heats of formation, E,,, and AH, (in kcal mol-‘) coenzyme with three different substrates: EtO- = ethoxide, ProSubstrate

The hydrogen

E act

alcohol state

transition state

aldehyde state

forward

backward

-2.4 -13.1 -2.2

t4.1 -1.7 7.7

-30.9 -36.6 -36.2

+7.1 +l1.4 +9.9

+35.6 +34.9 f43.9

transfer

distances

are derived

from the corresponding

as the starting structure. In addition to ethanol, we also inserted 1-propanol and isopropanol in these active site models. The calculations show that with isopropanol the activation energy is a little higher than with ethanol (Table 2). This small increase in activation energy cannot explain the much lower k,,, found experimentally with isopropanol compared to ethanol (Table 3). If hydrogen or hydride tunnelling is also involved in this mechanism, then the hydrogen transfer distance (i.e. the distance that the hydrogen H* needs to cross from substrate to coenzyme) can affect the rate of tunnelling. The hydrogen transfer distance of 0.73 A in the case of isopropanol is significantly larger than the distances found with ethanol and l-propanol. However, the height of the barrier also plays an important role, and this barrier is lower with isopropanol than with 1-propanol. Some geometrical data are presented in Table 4. As is to be expected, the C4-Cl distances in the transition states are smaller than in the initial and final states. The angles Nl-CQ-Cl do not vary much from their mean values of 100” and 1XV’,respectively. Taking a closer look at the van der Waals interactions, we find that with system 1 (ethanol as Table 3 Relative turn substrates

k ca, k cat k cat

over numbers

of NAD+

(our results) (Park and Plapp) [22] (Dalziel and Dickenson)

a Unpublished

(Theochem)

results.

[23]

in the enzymatic

364 (1996) 33-43

39

and the hydrogen transfer distance (in A). NAD+/NADH = propoxide and iPrO_ = isopropoxide

AH,’

ethanol I-propanol isopropanol

Structure

reduction

TS structures

is used as

AH,

hydrogen-transfer distance=

-28.5 -23.5 -34.0

0.60 0.61 0.73

of model structures

resembling

Fig. 1.

substrate) in the aldehyde state the hydrogen H* to be transferred from C4 of NADH to Cl of the acetaldehyde is already close to Cl (see Fig. 4(a)). The C4-H* distance is extremely long (1.15 A versus a CH bond distance of 1.09 A obtained experimentally). In the transition state an increasing overlap occurs between H* and Cl; the distances between C4-H’ and H*-Cl are 1.65 and 1.24 A respectively. The charge on H’ in the TS state is -0.046, indicating a H- transfer reaction. In the alcohol state, there is no overlap between the C4 of NAD+ and the H* at Cl of the ethoxide (see Fig. 4(b)). As is to be expected, there is some asymmetry induced into the electron charges in the CH2 group of ethanol even in the alcohol state, being -0.0085, +0.0659 and +0.0313 units for H*, Cl and Hl respectively, favouring the H’ transfer. For the systems with 1-propanol or isopropanol as substrate similar hydrogen transfer patterns can be derived (not shown). 3.3. Hydrogen tunnelling effects So far we have neglected possible proton tunnelling effects and the influence of the protein

by horse

liver alcohol

ethanol

1-propanol

isopropanol

100% [8] 100% 100%

98%” 100% 119%

13% [8] 10%

dehydrogenase

with several

alcohols

as

40

S.A.M.

Table 4 AM 1 calculated

Vanhommerig et al./Journal of Molecular Structure

heats of formation,

EaCt and AH, (in kcal mol-‘)

Distances

AH,

alcohol state

transition state

aldehyde state

forward

backward

-74.5 -14.5

-59.5a -56.6

-96.1 -94.9

+25.0 f17.9

+36.6 +38.3

in system 6

Hl (WAT)-O(EtO-)

Hl (WAT)-O(Ser

3.10 4.81

4.10 3.75

2.19 4.71

2.49 3.98

364 (1996) 33-43

(in A)

E act

AH,’

system 5 system 6

and distances

(Theochem)

-21.6 -20.4

48)

GNORM 1.0

alcohol state aldehyde state GNORM 0.1

alcohol state aldehyde state

a The energy of the TS state after one cycle.

environment, which all tend to reduce the (apparent) activation energies of hydride transfer. Both the hydrogen tunnelling and the classical pathway depend on the height of the energy barrier. In addition, the tunnelling rate is affected by the hydrogen distance to cross. Depending on which side of the reaction one starts (the alcohol or the aldehyde state), the tunnelling effect is enhanced or decreased, depending on a negative or a positive AH,, respectively [25]. In the case of ethanol the barrier for the forward reaction (7.1 kcal mol-‘) is lower than the expected vibrational energy of the migrating hydrogen H’ (8.6 kcal mol-‘, equivalent to &_u = 3000 cm-‘) and the barrier can be taken 1-propanol and the classical way. With isopropanol the tunnelling probability can be calculated. Applying a barrier of 11.4 kcal mol-’ (see Table 2, forward reaction of I-propanol), a CH vibrational energy of 8.6 kcaj mol-’ and a hydride transfer distance of 0.61 A (derived from the corresponding TS structure), a probability of 5.1011 s-’ can be calculated [26] for the tunnelling of a hydrogen. For isopropanol the probability for hydrogen tunnelling is 8.10 11 s-l, which is a factor of 1.6 higher compared with 1-propanol. Since for the classical pathway a probability of 5 x lo5 s-* and 7 x lo6 s-l results for I-propanol and isopropanol respectively, it seems likely that tunnelling will occur in both systems. However, owing to the

flexible and dynamic nature of the enzyme, the transfer distance will fluctuate considerably, as, therefore, will the tunnelling probability. In reality; therefore both the classical pathway and tunnelling are expected to occur. This might also explain the absence of measurable isotope effects in wild type HLADH [27]. 3.4. Water in the active site From X-ray studies of the apo conformation of HLADH (binary complex of HLADH with only the coenzyme bound) [28], it is known that there are many water molecules inside the enzyme and that one water is zinc-bound. Ritter von Onciul and Clark [ 161found it likely that the substrate replaces the zinc-bound water in binding to the zinc ion. Recent studies [15] on the ternary complex with pentafluorobenzyl alcohol as substrate indeed suggest that there is no water molecule present any longer as a fifth ligand of zinc. However, it cannot be ruled out that the water molecule may be too mobile to be recorded by X-ray crystallography. In order to check if the water molecule in the apo conformation remains zinc-bound during the calculations, we first performed a calculation of an active site model of the binary complex of HLADH [29] with one water molecule bound to zinc as a fourth ligand (thus before substrate

S.A.M.

Vanhommerig et al./Journal of Molecular Structure (Theochem)

364 (1996) 33-43

41

(4

(b) Fig. 4 The AM1 optimized geometries of the active site models with ethanol in van der Waals representation. (a) Aldehyde state; @I ho1 state.

bin ding has occurred). This active site model incl uded Zn(II), Cys 46, His 67, Cys 174 and H2( 3. After the AM1 calculations, during which the water molecule was allowed to move freely,

this water molecule remained coordinated to 1the zinc ion, resembling its original X-ray positicon. We then started from the optimized structures of system 1 (which contains alcohol and is based on

42

S.A.M.

Vanhommerig et al.lJournal of Molecular Structure (Theochem)

the ternary complex [14]), adding one additional water molecule close to the zinc ion (system 5), as found in the X-ray structure of the apo conformation of HLADH [29]. The results of the calculations show that the water molecule moves away from the zinc ion and remains at a 4 A distance from zinc, where it is not involved in the formation of hydrogen bonds, nor does it function as a fifth ligand of the zinc ion. This was also concluded from model-building experiments by Eklund et al. [29] and is in nice agreement with the X-ray studies on HLADH with NAD+ and pentafluorobenzyl alcohol [15]. It has been suggested [14] that the hydroxyl group of Ser 48 and the imidazole ring of His 51 are involved in a proton relay mechanism. It is unknown whether a water molecule is also involved in the deprotonation of the substrate. We studied the possibility that, after the formation of the alkoxide ion, a water molecule is involved in the Hbonding range alkoxide . . (H20) . Ser48. . His 51. In this model system (system 6), we again started from the optimized structure of system 1, and now the additional water molecule was placed in between the alkoxide and Ser 48, serving as a potential hydrogen bond acceptor and donor. AM1 calculations again show that the water molecule is moving away from the zinc ion. Minimizing the energy with a gradient norm of 0.1 kcal mol-‘, we find in the alcohol state that the water molecule takes a position in which it could be involved in the proton relay mechanism, because it is at H-bonding distance from EtO- and Ser 48 (see Table 4). A gradient norm of 1.0 kcal mol-t , however, causes geometries in which the water molecule does not form hydrogen bonds with the ethoxide and serine. (The energy differences between the states calculated with GNoRMm 1.0 and 0.1 are only 0.5 kcal mall’.) In conclusion, the water molecule could well be involved in the proton relay mechanism but, based on the present results, definite evidence is lacking.

4. Conclusions It is likely that the zinc-coordinated cysteine residues in HLADH are present as thiolate anions.

364 (1996) 33-43

Therefore active site models of HLADH should incorporate negatively charged cysteines. AM 1 and PM3 calculations show drastic differences in ground state energy levels for model systems incorporating negatively charged cysteine residues compared with active site models based on neutral cysteine residues. The present calculations do not provide an explanation as to why isopropanol is a poorer substrate than ethanol. This suggests that in the enzyme there are factors not accounted for in the simplified model used in our research. A negative charge on the hydrogen being transferred in the TS state can be calculated, pointing to a hydride transfer mechanism. Probability calculations suggest that hydrogen tunnelling may occur. However, to draw definite conclusions one should take into account the dynamics of the enzyme system. In the ternary model complex no water molecule is present as a fifth ligand of the zinc ion, in accordance with X-ray data.

Acknowledgement We thank A. Ritter von Onciul and T. Clark for useful discussions and for sending us their optimized transition state structure.

References [l] C.-I. Brand&n, H. JGrnvall, H. Eklund and B. Furugren, in P.D. Boyer (Ed.), The Enzymes, Vol. 11, 3rd edn., Academic Press, New York, 1975, p. 103. [2] H. Eklund, J.-P. Samama, L. Wall&, C-I. Brand&r, A. Akeson and T.A. Jones, J. Mol. Biol., 146 (1981) 561. [3] H. Eklund, J.-P. Samama and L. Wallen, Biochemistry, 21 (1982) 4858. [4] E.S. Cedergren-Zeppezauer, J.-P. Samama and H. Eklund, Biochemistry, 21 (1982) 4895. [5] H. Eklund, J.-P. Samama, B. Plapp and C.-I. Branden, J. Biol. Chem., 257 (1982) 14349. [6] N.A. Beijer, H.M. Buck, L.A.& Sluyterman and E.M. Meijer, Biochim. Biophys. Acta, 1039 (1990) 227. [7] S.A.M. Vanhommerig, R.J. Meier, L.A.A% Sluyterman and E.M. Meijer, J. Mol. Struct. (Theochem), 304 (1994) 53. [8] S.A.M. Vanhommerig, L.A.R. Sluyterman and E.M. Meijer, Biochim. Biophys. Acta, in press.

S.A.M.

Vanhommerig er al./Journal of Molecular Structure (Theochem)

[9] H. Theorell and B. Chance, Acta Chem. Stand., 1127. [IO] CC. Wratten and W.W. Cleland, Biochemistry,

5 (1951) 2 (1963)

935. [1 l] J. Kvassman and G. Pettersson, Eur. J. Biochem., 103 (1980) 557. [12] K. Dalziel and F.M. Dickenson, Biochem. J., 100 (1966) 34. [13] D.-H. Park and B.V. Plapp, J. Biol. Chem., 267 (1992) 5527. 1141 H. Eklund, J.-P. Samama and T.A. Jones, Biochemistry, 23 (1984) 5982. [ 151 S. Ramaswamy, H. Eklund and B.V. Plapp, Biochemistry, 33 (1994) 5230. [16] A. Ritter von Onciul and T. Clark, J. Comput. Chem., 14 (1993) 392. [17] L.A.fE Sluyterman and J. Wijdenes, Eur. J. Biochem., 71 (1976) 383. [18] J. Kvassman, A. Larsson and G. Pettersson, Eur. J. Biothem., 114 (1981) 555.

364 (1996) 33-43

43

[19] M.J.S. Dewar, E.G. Zoebisch, E.F. Healy and J.J.P. Stewart, J. Am. Chem. Sot., 157 (1985) 3902. [20] J.J.P. Stewart, J. Comput. Chem., 10 (1989) 209 and 221. [21] ~op~c6.0, Program No. 455 from Quantum Chemistry Program Exchange (QCPE), Indiana University, Bloomington, IN. [22] QUANTA/CHARM is a molecular modelling software package of Molecular Simulations Inc., Waltham, MA, USA. [23] M.J.S. Dewar, E.F. Healy and J.J.P. Stewart, J. Chem. Sot Faraday Trans. 2, 3 (1984) 227. [24] J. Baker, J. Comput. Chem., 7 (1986) 385. [25] M.D. Harmony, Chem. Sot. Rev., 1 (1972) 211. [26] P.W. Atkins, Physical Chemistry, 2nd edn., Oxford University Press, Oxford, 1982, pp. 408-409. [27] B.J. Bahnson, D.-H. Park, K. Kim, B.V. Plapp and J.P. Klinman, Biochemistry, 32 (1993) 5503. [28] H. Eklund, J.-P. Samama, L. Wallen, C-I. Branden, A.A. Akeson and T.A. Jones, J. Mol. Biol., 146 (1981) 561. [29] H. Eklund, B.V. Plapp, J.-P. Samama and C.-I. Branden, J. Biol. Chem., 257 (1982) 14349.