J. Mol. Biol. (1990) 214, 7999802

COMMUNICATIONS

Insights into the Function of the Zinc hydroxide-Thrl99Glu106 Hydrogen Bonding Network in Carbonic Anhydrases Kenneth M. Merz Jr Department

of Chemistry and Department of Molecular The Pennsylvania State University University Park, PA 16802, U.S.A.

(Received 11 February

1990; accepted 13 April

and Cell Biology

1990)

The exact functional role of the zinc hydroxide (water)-Thr199-Glu106 hydrogen bond network in the carbonic anhydrases is unknown. However, from the results of molecular dynamics simulations (MD) we are able to better define its function. From computer graphics analysis and MD simulations on the zinc hydroxide form of human carbonic anhydrase II we find that this interaction forces the hydroxide hydrogen atom to be in a “down” position relative to the deep water-binding pocket. From previous work we have found that this pocket is a high-affinity binding site for COz. We also note that during the timescale of our simulation (126 ps) the hydrogen bonds between the hydroxide hydrogen atom and Thr199 and the one between Thr199 and Glu106 are not fluxional. We propose that the role of the zinc hydroxide (water)-Thrl99-Glu106 hydrogen bond network is to lock the hydrogen atom in the down position in order to expose the CO, molecule bound in the deep water pocket to a lone pair of the hydroxide oxygen atom. This would allow for the rapid reaction of the CO2 molecule around the zinc ion. Furthermore, if the hydroxide hydrogen atom were not locked in the down position the binding of CO, to the deep water pocket could be interfered with by the unrestrained hydroxide hydrogen atom (e.g. the N-Zn-O-H torsion could undergo rotational transitions that would partially block the deep water pocket). In summary, the roles we ascribe to this hydrogen bonding network are (1) to allow for facile access of CO, to the deep water pocket and (2) to allow for maximal exposure of a hydroxide oxygen lone pair to the COz carbon atom.

potential fifth co-ordination site that may or may not be occupied depending on the identity of the moiety occupying the fourth co-ordination site (see reviews and Kannan, 1980; Eriksson et al., 1989a,b). To the zinc-bound water (or hydroxide) molecule is hydrogen bonded the hydroxyl oxygen atom of Thr199, which is itself hydrogen bound, via its hydroxyl hydrogen atom, to GlulO6 (see reviews and Kannan, 1980; Eriksson et al., 1989a,b). The only known physiological role that this protein has is the conversion of COz into bicarbonate or vice versa. The zinc hydroxide (water)-Thr199-GlulO6 hydrogen bonding network in the active sites of human carbonic anhydrases (HCAs) has been the subject of much speculation concerning its function (Kannan, 1980; Eriksson et al., 1989a,b). This has been spurred on because the hydrogen bonding network involves the reactive center in the active site and because it is conserved in all of the carbonic anhydrase structures that have been solved to date (Kannan, 1980; Ericksson et al., 1989a,b). Further-

Human carbonic anhydrase II (HCAIIt) is a small monomeric protein consisting of 260 residues with a total molecular weight of about 30,000. The zinc ion in all known carbonic anhydrases is of catalytic importance. (Reviews of carbonic anhydrase chemistry are given by Bertini et al. (1981), Lindskog (1983), Lipscomb (1983), Packer & Sarkanen (1978), Silverman & Vincent (1983), Silverman & Lindskog (1988).) The co-ordination ,of the zinc ion is to three histidine residues, His94, 96 and 119 (see reviews and Kannan, 1980; Eriksson et al., 1989a,b). The fourth co-ordination site is occupied by a water molecule in the low-pH form of the enzyme (< 7), while in the high-pa form this site is occupied by a hydroxide ion (see reviews and Kannan, 1980; Eriksson et al., 1989aJ). There is a

t Abbreviations used: HCAII, human carbonic anhydrase II; HCAs, human carbonic anhydrases; MD, molecular dynamics; T.S., transition state; TIPSP, transferable intermolecular potential 3 point cliarge. 799 0022%2836/90/160799-04

$03.00/O

0 1990 Academic Press Limited

more, the residues Thr199 and Glu106 are conserved in all of the carbonic anhydrases that have been sequenced to date (Tashian, 1989). On the basis of these observations, it is generally felt that this network is of functional significance. The hydrogen bonding network described here is to be differentiated from the “network” of water molecules that are in the active sites of carbonic anhydrases whose et al.! structures have been solved (Eriksson 1989a,b). The water molecule network has been implicated in proton transfer from zinc-bound water t,o His64 (Silverman & Vincent, 1983; Eriksson et al., 1989a,b), but this is not the subject of the present, research. The generally accepted zinc hydroxide mechanism for the hydration of CO, does not attribute any special function to the zinc hydroxide (water)-Thr199-GlulO6 network, beyond t*he fact that it is a hydrogen bonding network (Silverman &r Lindskog, 1988; Eriksson et al., 1989a,b). Kannan has suggested a catalytic mechanism in which this network is involved in an intramolecular proton relay that interconverts the zincwater and zinc hydroxide form of HCAs (Kannan, 1980). In order for this mechanism to work, though, the pK, of Glu106 would have to be anonymously high ( - 7.0 versus -4.O), which clouds the overall importance of this proton shuttle mechanism (Silverman &’ Vincent’, 1983). However, no one has been able to provide convincing evidence against this mechanism (e.g. perturbed Glu pK, values have been observed (Parsons & Rafter?;, 1972)), nor has anyone been able to define clearly the functional role for this hydrogen bonding network at the molecular level (see reviews and Kannan: 1980; Eriksson et al., 1989a,b). The location of CO, binding site(s) in HCAs have been the subject of much controversy (Led et al., 1987; Stein et al., 1977; 1982; Led & Keesgaard, Bertini et al., 1979; Yeagle et al., 1975; Riepe & Wang, 1968). The conclusion from the studies that have been done t*o identify the CO, binding site is that the meta,l ion to CO, carbon atom dist’ance is anywhere from 3.0 A (1 A=O*l nm) to about 10 A (Led et nl., 1982: Led & Neesgaard, 1987; Stein et al., 1977; Bertini et al.; 1979; Yeagle et al., 1975; Riepe & Wang, 1968). We have reported molecular dynamics (MD) simulations that have located two binding sites for CO, in the act’ive site of the zinchydroxide form of human HCATT (Merz, unpublished resultIs). Germane to the present discussion we have identified the so-called deep water pocket (see reviews and Kannan, 1980; Eriksson et al., 1989a,b) as the “high-affinity” or “reactrive” binding pocket, which we have identified as the A binding site (Merz, unpublished results). For a qualitative diagram of the location of active site residues and binding pockets in t,he HCATI active sit’e see Figure 1. The second binding site that we have observed may serve as a storage site and we have labeled it as t’he E site (Merz, unpublished results). The A site is surrounded by Va1121, 143, %07, Trp209, Leu198, Hisll9, Thrl99 and the residue consisting of ZnOH, while the E is formed

Thrl99-N-=--i

Thrl99-

.’ 061 a HOG I :

I

-1.9

s

E OEi\&/ CD GlulO6

Figure 1. A qualitative description of the import,artt, features of the HCAIB active. For Thr199 both the mainchain N-H and side-chain hydroxyl groups are shoam.

by His64, 94, 96, Ala65, Asn244, TyrT, Phe93 and ThROO. From our results on the location of the C’O, binding sites (Mere; unpublished results) a,nd ‘the data presented below, we are now able to give molecular level insights into the functional role of this network. The MD simulations were carried out using the MD module from the AMBER suite of programs. The AXBER (Weiner et al.. 1984, 1986) pot,ential funct’ion was used throughout. The three histidine residues around the zinc ion used the all-atom representation while the rest of the protein used the united-at*om representation. The starting st’ructure for the &ID simulation was first, minimized fully to remove any bad intermolecular contacts. This struc-kure was then solvated with a 22 A4 sphere of TTP3P (Jorgensen et al.: 1983) water molecules. which was active sit,e. centered at the zinc ion in the HCAII Any water molecules that come too close to a protein atom were removed. This resulted in about 300 wa,ter molecules solvating the active site of t,he protein. The water molecules were kept in this sphere with t,he use of harmonic rest’raining forces; which are applied to any water molecule t,hat strays out, of the sphere. The simulations were kept at a constant temperature (298 K) by coupling to a temperature bath (Berendsen et nl., 1984). All residues within 15 is of the zinc ion were permitt,ed to move during the course of the simulation, as were all water molecules, while residues lying outside this region were held fixed. The program SHAKE (\‘an Gunst,eren & Berendsen. 1977) was used to constrain

Communications

801

Table 1 Atom pair

L%tompairs? COZ Zn-C Zn-01 Zn-02 OGlLHZ HOGOEl HOGOE2 02-H 01-H

distances extracted from a 126 ps MD trajectory Min. (A) 2.52 3-32 2.82 376 1.57 1.55 1.84 2.06 3.40

Max. (L&S) 428 5.57 2.86 6.20 3.13 2.96 3.94 2.72 454

cc?>

(4

3.09 415 3.85 4.76 1.96 1.87 2.78 4.19 5.94

His94

Hmll9

1- The atom labels are given in Fig. 1.

all bond lengths to their equilibrium values and a timest’ep of I.5 femtoseconds was used. The nonbonded pair list had a cut off of 10 A (simple truncation was used), was updated every 60 timesteps, and a constant dielectric of one was used throughout. The equilibration period was nine picoseconds and the simulation reported here was carried out for a total of 126 picoseconds. Every 100 timesteps energies and co-ordinates were saved for later study. The A binding pocket places the CO2 molecule in close proximity to the zinc ion and its associated ligands. Table 1 gives some of the pert’inent maximum, minimum and average distances taken from a 126 picosecond MD trajectory. All of the distances in Table 1 indicate that the CO, molecule remains site throughout the MD simulain the A binding tion. For example, the CO, carbon atom to the zinc-bound hydroxide oxygen (C-OZ) distance is on average only about 3 A long. From this MD trajectory we have also extracted information regarding the zinc hydroxide-Thr199 hydrogen bond as well as the Thrl99-Glu106 interaction. This is also given in Table 1. One immediately notes, for example, that the OGI-HZ distance indicates that this hydrogen bond is not broken during the course of the MD simulation. This result is also verified via visualization of the MD trajectories (MD movies). Thus these latter hydrogen bonds are very stable and the net result is to lock the hydroxide hydrogen atom in a down position relative to the A site (see Fig. I). Eriksson et al. (19896) have suggested that the main-chain N-H from Thrl99 hydrogen bonds with an oxygen atom from CO,. We find that’ the 02-H distance is on average 2.72 A, which in our opinion is too long for a strong hydrogen bond. This distance has a minimum value of about 2.0 A, which is more indicative of a hydrogen bond, but our results suggest t,hat this interaction is fluxional and at best weak. These results have several important implications for the activity of HCAs. In the absence of Thrl99 the N-Zn-O-H torsion would be free to undergo rotation transitions and the hydroxide hydrogen atom would be able to hydrogen bond with water molecules in the active site. The first effect this would have would be to partially block the deep

His96

I I I I I

Thrl99-

, OG,

\ HOG

\

\

\

\

\

Figure 2. A Newman projection of the HCAII active site in the region of the A binding site and the zinc ion with its associated ligands. The zinc ion is behind the OZ atom.

water pocket by either purely steric effects or by the possibilit,y that the hydroxide hydrogen atom could hydrogen bond with a water molecule in the A site (see Fig. 2). This would stabilize this water molecule and make it’ more difficult, for the CO, molecule to gain access and bind to the A site. Thus, the hydrogen bonding network by locking down the hydroxide hydrogen atom allows for the facile access of the CO2 molecule for the A site. The second adverse effect a freely rotating N-Zn-O-H bond would have on the activit,y of HCAs would be to interfere with the CO2 molecule reacting at the metal ion. By pulling the hydroxide hydrogen atom down away from the CO, molecule bound in the A site, a lone pair on the hydroxide oxygen atom is always exposed to the COz molecule. For a qualitative description of this see Figure 2. Quantum mechanical calculations (Merz et al., 1989; Pullman; 1981) indicate that t#he most’ favorable transition state (TX) is one where a hydroxide oxygen lone pair is pointing towards the CO, carbon and a CO2 oxygen atom is bound around the zinc ion trans to one histidine (His96 here) and staggered between the other two (His94 and His1 19). From Figure 2 we can see that) when the CO, molecule is bound to the A site it is in an orientation that’ allows it readily to attain the calculated transition

802

K.

M.

Merz

state structure (Merz et al., 1989, Pullman, 1981). Thus, the HCAs provide another example of enzyme-transition state complementarity (Fersht, 1985) in that the CO2 molecule is bound in a manner that is close in structure to the proposed T.S. for a CO, molecule reacting around a metal center. Another thing to note is that there is a favorable entropic gain by allowing the N-Zn-O-H bond to undergo internal rotation, but in order for the T.S. to be reached this internal degree of freedom would have to be hindered leading to an unfavorable entropic contribution to the free energy of activation for the hydration of CO*. However, by forming the hydrogen bond network this internal rotation is already restrained, which therefore does not cause an entropic penalty to the free energy of activation. There are two types of T.S. that have been proposed for the hydration reaction (Merz et al., 1989; Pullman, 1981; Liang & Lipscomb, 1987). One involves an inner-sphere reaction (CO, complexes with the zinc ion and is converted to bicarbonate (Merz et al., 1989; Pullman, 1981)), while the other goes via an outer-sphere reaction (hydroxide oxygen atom attacks the CO, carbon atom forming Zn-OH-C02, which undergoes a hydrogen shift to give zinc-bound bicarbonate (Liang & Lipscomb, 1987)). Given the fact that the A binding site is hydrophobic in character, it seems that the formation of a carboxylate ion in this pocket would be unfavorable. The main-chain N-H from Thr199 may stabilize the carboxylate, but from our results here the hydrogen bonding distance would be rather long. If the formation of the carboxylate and the hydrogen shift were a “concerted” set of processes the outer-sphere attack in our opinion would be more likely. In our discussion here the CO, molecule is not directly co-ordinated to the metal ion (Merz; unpublished results; Merz et al., 1989); however, some theoretical efforts have indica,ted that CO, may be co-ordinated directly t’o the metal ion prior to undergoing hydration (Pullman, 1981; Liang & Lipscomb, 1987). Either unto-ordinated or co-ordinated, though, most of the conclusions reached here are valid, with the one exception being that the entropic penalty would be paid during the co-ordination of the CO2 molecule for the latt,er. Finally, we note that our calculations do not address the issue of whether or not the proton relay mechanism of Kannan (Kannan, 1980) is a viable mechanistic alternative, but the roles ascribed here to the hydrogen bond network would also be operative in the Kannan mechanism. We thank the Center for Academic Computing at the Pennsylvania State University for generous allocations of Edited

by P.

nilr

IBM3090-600S computer time and thank for helpful discussions and suggestions.

10.

Silverman

References Berendsen, H. J. C., Potsma: J. P. M., van Gunsteren, W. F., DiNola, A. D. & Haak, J. R. (1984). J. C&m. Phys. 81; 3684-3690.

Bertini,

I., Borghi,

E. $ Luchinat,

C. (1979). J. dmer.

Chem. Xoc. 101, 706997071.

Bertini;

I., Luchinat,

Bonding

Eriksson.

(Berlin),

C. & Scozzafava, A. (1981). Struct. 48, 45-91.

A. E., Jones, A. T. & Liljas,

Proteins,

$.

(1989a).

4, 274-282.

Eriksson, A. E., Kylsten, I’. M., Jones; T. A. $ Liljas, A. (19898). Proteins, 4, 283-293. Fersht, A. R. (1985). Enzyme Structure and Mechanism. W. H. Freeman & Co., New York. Jorgensen, W. L., Chandrasekhar. J.; Madura, J., Impey. R. W. & Klein, M. L. (1983). J. Chem. Phys. 79. 92G935.

Kannan, K. K. (1980). In Biomoleeular Structure, Conformation, Function & Evolution (Srinivassan, R.. Pergamon Press, Oxford. ed.), vol. 1, pp. 165-181, Led, J. J. & Neesgaard; E. (1987). Biochemistry, 26, 1833192. Led, J. J., Neesgaard, E. & Johansen, J. T. (1982). FEBS Letters, 147, 74-80. Liang, J.-Y. & Lipscomb, W. N. (1987). Biochemistry, 26. 5293-5301. Lindskog, S. (1983). In Zinc Enzymes (Spiro, T. G.; ed.), pp. 78-119, John Wiley & Sons, New York. Lipscomb, W. N. (1983). Anmu. Rev. Bioch.em. 52, 17-34. Merz, K. M., Jr, Hoffmann, R. & Dewar: M. cJ. S. (1989). J. Amer. Chem. Sot. 111; 563CS-5649. Parsons, S. M. & Raftery, M. A. (1972). Biochemistry. 11, 1630-1633. Packer, Y. & Sarkanen, S. (1978). Advan. Enzymol. 47, 1499274. Pullman, A. (1981). Ann. X. Y. Acad. Sci., 367, 340-X%. Riepe, M. E. &, Wang, J. H. (1968). J. Biol. Chem. 243. 2779-2787.

Silverman, D. N. & Lindskog, S. (1988). Act. Chem. Res. 21, 30-36. Silverman, D. N. & Vincent. S. H. (1983). CRC Crit. Rev. Biochem.

14. 207-255.

Stein, P. J.; Merrill,

S. P. & Henkens.

R. W. (1977).

Chem. Sot. 99. 3194-3196. R. E. (1989). Bioessays, 10, 186-192.

J. Amer.

Tashiaq Van Gunsteren, W. F. & Berendsen, H. J. C. (1977). Xol. Phys. 34, 1311-1327. Weiner, S. J., Kollman, P. A., Case, D. A.. Singh; U. C., Ghio: C., Alagona, Cr.; Profeta, S., Jr $ Weiner, P. (1984). J, Amer. Chem. Sot. 106. 765-784. Weiner, S. J.: Kollman, P. A., Nguyen, D. T. 8: Case. D. A. (1986). J. Gomput. Chem. 7, 230-251. Yeagle, P. L., LochMiiller, 6. H. & Henkens, R. W. (1975). Proe. IVa,t. Acad. Sci.. U.S.A. 72, 454-458. Wriyht