THE JOURNAL OF BIOLOGICAL CHEMISTRY

Prmted’in U.S.A. Vol. 256 No. 24 lasue of December 25, pp. 13213-13217.1981

L-Arabinose-binding Protein-Sugar Complex at 2.4 A Resolution STEREOCHEMISTRY AND EVIDENCE FOR A STRUCTURAL CHANGE*

(Received for publication, January 12, 1981)

Marcia E. Newcomer$, Gary L. Gillilandg, and Florante A. Quiochol From the Department of Biochemistry, Rice university, Houston, Texas 77001

The L-arabinose molecule (in the C1 pyranose chair conformation) has been fitted to the electron density corresponding to the bound sugar in the 2.4 A resolu- tion Fourier map of the L-arabinose-binding protein. The sugar molecule is buried in the cleft between the two lobes of the bilobate protein. All sugar hydroxyls are hydrogen-bonded to side chain residues: P-OH(1) to Lys-10 and Asp-90, OH(2) to Lys-10, OH(3) to Asn- 205 and Glu-14 (possibly via a water molecule), and OH(4) to Asn-232. Lys-10, Glu-14, and Asp-90 are asso- ciated with one domain while Asn-205 and Asn-232 are lodged in the other. Protein structural change accom- panying binding is indicated by the inaccessibility of the bound L-arabinose to the aqueous environment.

Periplasmic binding proteins are involved in the transport of a variety of low molecular weight compounds across the bacterial envelope and, for some, in bacterial chemotaxis toward or away from the same chemicals. One of these, the L-arabinose-binding protein mediates the high affinity uptake of L-arabinose in Escherichia coli (1). By a combination of low resolution crystallographic and chemical analyses, we have previously shown that the structure of the protein has been solved with bound L-arabinose (2, 3). The density cor- responding to the bound sugar molecule is located in the cleft between the two similar domains of the bilobate protein (3). However, the low resolution analysis has precluded a clear picture of the bound sugar molecule. More recently, the structure determination has been extended from 2.8 to 2.4 a resolution (4), thus paving the way for a detailed evaluation of the mode of binding of sugar ligands.

MATERIALS AND METHODS

We have previously studied the binding of D-galactose to ABP’ crystals by difference Fourier technique at 3.5 8, resolution (3). For the present study, a new set of complete data to 2.8 8, resolution was collected from crystals soaked in 10 mM D-galactose for 1 week. This

* This study was supported by Grant C-581 from the Robert A. Welch Foundation and United States Public Health Grant GM-21371. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. + Pre-doctoral Fellow of the Robert A. Welch Foundation. Present address, The Wallenberg Laboratory, University of Uppsala, Uppsala, Sweden.

5 Pre-doctoral Fellow of the Robert A. Welch Foundation. Present address, Laboratory of Molecular Biology, National Institute of Ar- thritis, Metabolism and Digestive Diseases, National Institutes of Health, Bethesda, Maryland 20205. 1 Recipient of a John Simon Guggenheim Memorial Foundation

Fellowship. To whom correspondence should be sent. ’ The abbreviation used is: ABP, L-arabinose-binding protein.

required a total of three crystals. The three data sets were correlated resulting in an Roverlap of 0.032 (based on intensity). The phases used for the difference Fourier calculation were those determined by multiple isomorphous replacement technique at 2.4 8, resolution (4).

Co-ordinates for th D-galactose molecule were obtained from the crystal structure determined by Sheldrick (5) and those for L-arabi- nose from the structure of DL-arabinose by Kim and Jeffries (6).

A molecular graphic system in the laboratories of Dr. E. F. Meyer and Dr. David R. Davies was utilized in the fitting of sugar models to the sugar density in Fourier maps. The 2.4 8, electron density of the L-arabinose-binding protein, computed with phases derived by mul- tiple isomorphous replacement technique (4), was used in this study.

RESULTS

The difference electron density map calculated at 2.8 with difference coefficient from crystals soaked in D-galactose and multiple isomorphous replacement phases (4) is shown in Fig. 1. This map c o n f i s the result of our initial study at low resolution which indicated that D-galaCtOSe binds to ABP crystals essentially in the same manner as L-arabinose does (3). The only difference density which is about 3x background (see figure) can be attributed to the additional -CH20H of the D-gdaCtOSe molecule.

P-DL-Arabinose crystallizes as a racemic mixture in which

I X 0

” -

0

8

FIG. 1. The 2.8 A difference Fourier map of ABP crystals soaked in 10 m~ D-gahCtOSe, 60% 2-methyl-2,4-pentanediol, 10 m~ maleate, pH 6.4. The origin is at the upper right and boundary: x = (0.1466 to 0.6127), y = (0.6101 to 0.9564), and z = (0.6588 to 0.7020). The peak of the highest difference density is indicated by +.

13213

13214 L -Ara- binding Protein-Sugar Complex at 2.4 A Resolution

the sugar is in a la2e3e4a chair conformation (where a and e refer to axial and equatorial positions for the hydroxyl groups, respectively) (6).

As is true for all monopyranosides, 8-L-arabinose can exist in either the C1 or 1C conformation; both conformers are illustrated in Fig. 2. However, the crystal structure of L- arabinose corresponds to the C1 conformation (6) and several

c 1 I 1c

I ‘ FIG. 2. C1 and 1C conformation of 8-L-arabinose.

FIG. 3. Two stereo views of the fit of the 8-L-arabinose model to the density of the sugar in a 2.4 A Four- ier map of !,-arabinose-binding pro- tein. The anomeric carbon is marked by a cube. Densities that extend beyond the sugar hydroxyls are from parts of side chain of amino acids (models not shown) liganded to the sugar. For instance the density contiguous to both OH(1) and OH(2) is from Lys-lO(NE) and to OH(4) from Asn-232 (ODI).

lines of evidence suggest that this is also the structure in solution (7-11). Furthermore, it is the C1 conformation that has the lower calculated free energy (10). The crystal structure of the D-galactose is also in the C1 conformation.

The C1 structure of the L-arabinose was thus fitted to the electron density of the bound sugar molecule in the 2.4 8, electron density map of the protein. A view of the fit with the hydroxyl at the anomeric sugar carbon in the orientation is shown in Fig. 3. It has been shown that whereas D-galaCtOSe is an excellent ligand, comparable to the natural substrate L- arabinose, 8-methyl-D-galactopyranoside is not (3). Although this observation does not indicate which anomer binds pref- erentially, it suggests that a free hydroxyl at the C-1 position is essential for binding of D-galactose.

The /?-anomer of L-arabinose can be fitted to the density at 2.4 8, resolution in two ways. Note that the two orientations illustrated in Fig. 2 (C1 and I) will have approximately the same electron density distribution (the rotation axis is perpen- dicular to the plane of the paper). However, the introduction of a substituent at the C-5 position (as in D-galactose) would distinguish between the two in a fit to the sugar density at 2.4 8, resolution. In fact, the identical location of the -CH2Br of 6-deoxy-6-bromo-~-galactose and of the -CH20H at C-5 of D-galactose obtained by difference Fourier analyses (3) (see also Fig. 1) is consistent with the orientation of the sugar shown in Fig. 3.

I

L -Ara-binding Protein-Sugar Complex at 2.4 A Resolution 13215

FIG. 4. Stereo drawing of the pro- posed complex between L-arabinose (heavy lines) and the L-arabinose- binding protein. Hydrogen bonds be- tween sugar and side chain residues are indicated by lines.

FIG. 5. A schematic diagram of the interaction of L-arabi- nose to ABP. A facsimile of this figure has been previously published in a paper by Argos et al. (27).

The fit of the L-arabinose molecule reveals that all sugar hydroxyls are within hydrogen-bonding distance (2.6-3.4 A) to amino acid residues. Stereo and schematic drawings of the proposed hydrogen bond scheme are presented in Figs. 4 and 5, respectively. As indicated, /I-OH( 1) is hydrogen-bonded to the carboxyl side chain of Asp-90. The e-NH3+ of Lys-10 is favorably positioned to interact with both OH(1) and OH(2) of the sugar. The OH(3) is linked to Asn-205 along with Glu- 14 possibly through a water molecule. The C-4 hydroxyl hydrogen bonds with Asn-232. Although the distance between the OH(3) and carboxylate oxygen of Glu-14 is about 3.8 A, the presence of continuous electron density between this residue and the bound sugar suggests the involvement of this residue. A water molecule is close to both Glu-14 and OH(3).

Miller et al. (2) have demonstrated that the single cysteine is in the vicinity of the binding site and most likely not directly involved in binding. Reaction of the thiol group with 2-chlo- romercuri-4-nitrophenol substantially diminished but did not completely abolish the binding affinity of the protein for L- arabinose. In the presence of the sugar substrate the sulfhy- dryl is protected from reaction. In addition, it was shown that the extent of activity lost is dependent on the size of the adducts; e.g. a modification of the thiol with CN retained most of the binding activity while the reaction with the mercurini- trophenol resulted in the least activity.

The location of the single cysteine, Cys-64, relative to the sugar molecule is shown in Fig. 4. The distance between the sulfur of Cys-64 and the closest L-arabinose hydroxyl (OH(1)) is too long for the formation of hydrogen bond. Moreover, such a bond is unlikely for the following reasons. 1) The hydrogen bond would deviate from linearity by more than 30’; this large a deviation was not observed for any of the

hydrogen bonds obtained for the sugar-protein complex. 2) /I-Methyl-D-galactose lacks only a free hydroxyl at the C-1 position, yet it does not bind to the protein. However, the protein modified with CN, and therefore lacking a free sulfhy- dryl group, still retains a substantial amount of activity ( K d

= 2 X M) (2). The effect of thiol modification on the activity could be explained by the proximity of Cys-64 to residues directly involved in the protein-sugar complex; Cys- 64 is particularly close to Asp-90 and Lys-10 (Fig. 4). Reaction of the thiol with sulfhydryl reagents could disrupt the orienta- tions of these residues and cause a diminution of binding activity, the extent being dependent on the size of the reagents.

DISCUSSION

Of the many protein structures known, only very few have been solved with substrates already bound, with the exception of cofactors. The structure of the L-arabinose-binding protein has been solved with bound L-arabinose (3). The binding protein consists of two similar lobes separated by a cleft (4, 12). Although low resolution studies indicate that this cleft contains the sugar-binding site (3), details of the nature of the sugar-protein interaction remain to be elucidated.

As a result of the recent successful fitting of the amino acid sequence (13) to the 2.4 A electron density map (4), a more detailed picture of the binding site, including the bound L- arabinose molecule has been obtained. The entire binding site pocket is lined with amino acids from both domains. One domain (the P domain) mainly contributes Lys-10, Pro-12, Glu-14, Trp-16, Phe-17, Cys-64, Asp-89, Asp-90, and Met-108, while the other domain (Q domain) shares Leu-145, Thy-147, Arg-151, Asn-177, Met-204, Asn-205, Ile-231, and Asn-232. Not all of these residues are, however, directly involved in binding.

The fitting of L-arabinose to the sugar density at 2.4 8, resolution allowed for the determination of a model for the sugar-protein complex consistent with the specificity of the protein. L-Arabinose, in the C1 chair conformation, is hydro- gen-bonded to residues from the opposite walls of the cleft. Because the residues involved directly in binding originate from the two domains and because of the lack of direct interaction between the opposite walls of the cleft (4), it appears that the binding of L-arabinose stabilizes a “closed” conformation.

A free anomeric “ O H is required for binding (3). Although a /I-L-arabinose model has been fitted, an a-hydroxyl at the anomeric carbon can be easily accommodated without steric hindrance. (In solution, the a and /I anomers exist in a ratio of 64:36 (11)). It is possible that both anomers of L-arabinose bind. Kinetic analysis of the binding of L-arabinose to ABP in solution, similar to that we have conducted on the D-galactose- binding protein (17), indicates that the a and /I forms of the

13216 L -Ara- binding Protein-Sugar Complex at 2.4 A Resolution

FIG. 6. Stereo drawing of D-galac- tose superimposed onto L-arabinose in the binding site of ABP. Gal is drawn with thin bonds and Ara with heaLy bonds.

FIG. 7. Space-filling model of the sugar-binding cleft of ABP showing parts of the two domains (gray and black) and the sugar molecule (white) buried in the cleft. The part of the sugar molecule which can be seen the most corresponds to the C-2 hy- droxyl.

-..

t

sugar bind with the same affinity.' The fitting of the p-form places the C-1 hydroxyl within hydrogen bonding distances to Asp-90 and Lys-10 (Fig. 4). On the other hand, the hydroxyl in the a-position could easily form a hydrogen bond with Thr- 147.:'

The mode of binding is consistent with results of substrate specificity studies which indicate that all four L-arabinose hydroxyls are required for binding to the protein (3, 14). The sugar analogues which do not bind to the protein are all known to be in the C1 chair conformation, with the exception of L-lyxose which is found to exist in both the C1 and 1C conformations (11). Therefore, the fact that these analogues are unable to bind to the protein is attributed to the lack of specific free hydroxyl(s) for H-bonding, rather than to an incompatible chair conformation.

The model accounts for the observed change in tryptophan fluorescence resulting from the binding of sugar and, as dis- cussed previously, the sensitivity of the binding activity of the protein to thiol reagents.

Parsons and Hogg (15), as well as our laboratory,' have previously shown that the binding of L-arabinose to the pro- tein causes a blue shift and quenching of the tryptophan fluorescence and that prior chemical modification of the resi- due(s) by N-bromosuccinimide can partially abolish this change. These findings, coupled with the observation that chemical modification can be partially prevented by L-arabi- nose, suggest the presence of a t least 1 tryptophan residue in the binding site. The molecular model shows the presence of one tryptophan (residue 16) in the binding site, but this residue is not likely to be directly involved in binding (Figs. 4 and 5). However, it is interesting to note that the orientation of L-arabinose in the binding site is such that its most hydro- phobic side, created by the -CH?- at the 5 position and C- 4 with an axial hydroxyl, is toward Trp-16 and Phe-17. ' D. M. Miller, 11. J . S. Olson, J. W. Pflugrath, and F. A. Quiocho,

manuscript in preparation. .' Unpublished results.

Ligand-induced fluorescence changes have been observed in several other binding proteins, e.g. D-galactose-binding protein (16, 17), maltose-binding protein (18), and histidine- binding protein.' We have recently exploited this property to study the kinetics of substrate binding to these proteins (Ref- erence 17).'

Sugars with different substituents at the C-5 position (rel- ative to L-arabinose) are able to bind to the L-arabinose- binding protein, some with tight affinity, i.e. D-gahCtOSe ( K O = 4 X 10" M), D-fucose (2 X IO-' M), D-galaCtUrOniC acid (2 X M) (3). (For comparison, the Kc/ of the Ara. ABP complex is 3 X lo-' M (3).) However, large substituents, such as in the case of 6-deoxy-6-iodo-~-galactose, prevent binding. Results obtained from difference Fourier analysis of the bind- ing of D-galactose and the electron-rich analogue 6-deoxy-6- bromo-D-galactose not only led to the location of the sugar- binding site (3) but proved to be instrumental in resolving the ambiguity in fitting a P-L-arabinose molecule to the 2.4 A electron density map (see above). Apart from the additional -CH20H, the binding of D-galaCtOSe is very similar to L- arabinose (Fig. 6). The finding that the "CHcOH substituent in D-galactose can be accomodated without steric interference with the protein accounts very nicely for the apparent non- essentiality of this substituent. However, a " C H J would probably not be easily accommodated, consistent with the fact that 6-deoxy-6-iodo-~-galactose does not bind to the protein (3).

Even before the fitting of the L-arabinose molecule could be accomplished, it became immediately apparent that the sugar is not only imbedded in the cleft between the two domains of the bilobate protein, but largely inaccessible to the solvent. This observation is clearly borne out by the space- filing model of the binding site with the bound sugar (Fig. 7). (This figure shows the most exposed view of the sugar mole- cule possible.) An accessibility calculation by the method of Lee and Richards (19), using a probe of radius 1.4 A further confirms this observation; only the C-2 hydroxyl of the sugar

L -Am-binding Protein-Sugar Complex ut 2.4 A Resolution 13217

is exposed, but only by 0.5 A.2 These findings clearly indicate that the binding site cleft must be more open in the unliganded protein. A likely way of accomplishing this structural change is suggested by results of small angle x-ray scattering experi- ment in solution (20). The removal of bound L-arabinose from ABP is accompanied by a 1 A increase in the radius of gyration of the protein. This increase has been interpreted in terms of an opening of the cleft which could result from one lobe rotating relative to the other lobe about a hinge deep in the base of the cleft.

Recent genetic and chemical evidence indicate that specific interactions between binding proteins and membrane-bound components are essential for transport (21-23) and chemotac- tic behavior (24, 25). This finding suggests that membrane receptors recognize and bind to the liganded form of the binding proteins in preference to the unliganded form. Con- sequently, this interaction is dependent upon a substrate- induced conformational change on part of binding proteins. Results presented above provide conclusive evidence for a structural change in the L-arabinose-binding protein.

Further crystallographic and chemical studies are being undertaken in order to understand in more detail the confor- mational rearrangement and its role in transport. Recently we have been able to crystallize “sugar-free” ABP using a method different from that utilized for the Ara-ABP4; sugar-free ABP crystals are obtained from polyethylene glycol,5 whereas Ara- ABP was obtained from 2-methyl-2,4-pentanediol (26). Struc- tural analysis of the sugarless protein is presently underway. At the same time we are refining the native structure at 1.85 A resolution.

Acknowledgments-We are deeply grateful to Drs. E. F. Meyer and D. R. Davies for use of the computer graphics in their respective laboratory. We thank Dr. R. Feldmann for the computer graphics- generated space-filling model.

The purified ABP was rid of bound L-arabinose by the procedure outlined in Reference 20 and subsequently chromatographed onto a Bio-Gel P-2 column. Further details of this mild, effective, and general procedure for binding proteins will be published elsewhere.

S. Bowen and F. A. Quiocho, unpublished data.

REFERENCES 1. Brown, C . E., and Hogg, R. W. (1972) J . Bacteriol. 111,606-613 2. Miller, D. M., 111, Newcomer, M. E., and Quiocho, F. A. (1979) J.

3. Newcomer, M. E., Miller, D. M., 111, and Quiocho, F. A. (1979) J.

4. Gilliland, G. L., and Quiocho, F. A. (1981) J. Mol. Biol. 146, 341-

5. Sheldrick, B. (1976) Acta Crystallogr. B32, 1016-1020 6. Kim, S. H., and Jeffries, G. A. (1967) Acta Crystallogr. 22, 537-

7 . Lemieux, R. U., and Martin, J. C. (1970) Carbohydr. Res. 13,139-

8. Whiffens, D. H. (1956) Chem. Znd. (Lond.) 964-968 9. Reeves, R. E. (1949) J. Am. Chem. Soc. 71, 215-217

10. Reeves, R. E. (1950) J . Am. Chem. SOC. 72, 1499-1506 11. Angyl, S. J. (1972) in The Carbohydrates (Pigman, W., and

Horton, D., ed) Vol. IA, pp. 195-213, Academic Press, New York

12. Quiocho, F. A., Gilliland, G. L., and Phillips, G. N., Jr . (1977) J. Biol. Chem. 252,5142-5149

13. Hogg, R. W., and Hermodson, M. A. (1977) J. Biol. Chem. 252, 5135-5141

14. Hogg, R. W., and Englesberg, E. (1969) J. Bacteriol. 100,423-432 15. Parsons, R. G., and Hogg, R. W. (1974) J. Biol. Chem. 249,3602-

16. Boos, W., Gordon, A. S., Hall, R. E., and Price, 0. H. (1972) J.

17. Miller, D. M., 111, Olson, J. S., and Quiocho, F. A. (1980) J . Biol.

18. Szmeleman, S., Schwartz, M., Silhavy, T. J., and Boos, W. (1976)

19. Lee, B., and Richards, F. M. (1971) J. Mol. Biol. 55,379-400 20. Newcomer, M. E., Lewis, B., and Quiocho, F. A. (1981) J. Biol.

21. Ordal, G. W., and Adler, J. (1974) J. Bucteriol. 177, 517-526 22. Ames, G. F-L., and Spudich, E. N. (1976) Proc. Natl. Acad. Sci.

23. Ames, G. F-L., and Nikaido, K. (1978) Proc. Natl. Acad. Sci. U.

24. Strange, P. G., and Koshland, D. E., Jr.(1976) Proc. Natl. Acad.

25. Springer, M. S., Goy, M. F., and Adler, J. (1979) Nature 280,279-

26. Quiocho, F. A., Phillips, G. N., Jr., Parsons, R. G., and Hogg, R.

27. Argos, P., Mahoney, W. C., Hermodson, M. A., and Hanei, M.

Biol. Chem. 254, 7521-7528

Biol. Chem. 254, 7529-7533

362

545

161

3607

Biol. Chem. 247,917-924

Chem. 255,2465-2471

Eur. J. Biochem. 65,13-19

Chem. 256, 13218-13222

U, S. A. 73, 1877-1881

S. A . 75,5447-5451

Sci. U. S. A. 73, 762-766

284

W. (1974) J. Mol. Biol. 86, 491-493

(1981) J. Biol. Chem. 256,4357-4361