Ribose and glucose-galactose receptors. Competitors in bacterial chemotaxis.

.J. Mol. Bid.

(199d) 227. 118

440

Ribose and Glucose-Galactose Competitors

Receptors

in Bacterial Chemotaxis

Sherry L. Mowbray Department of Molecular

Biology, Swedish University of Agricultural 24 Uppsala, Sweden

Scienms

S-757

(Received 17 January

1992; accepted 20 May

1992)

The periplasmic ribose and glucose-galactose receptors (binding proteins) of Gram-negative bacteria compete for a common inner membrane receptor in bacterial chemotaxis, as well as being the essential primary receptors for their respective membrane transport systems. The high-resolution structures of the periplasmic receptors for ribose (from Escherichia co&) and glucose or galactose (from both Salmonella typhimurium and E. coli) are compared here to outline some features that may be important in their dual functions. The overall structure of each protein consists of two similar domains, both of which are made up of two non-contiguous segments of amino acid chain. Each domain is composed of a core of p-sheet flanked on both sides with ol-helices. The two domains are related to each other by an almost perfect intramolecular axis of symmetry. The ribose receptor is smaller as a result of a number of deletions in its sequencerelative to the glucose-galactose receptor, mostly occurring in the loop regions; as a result, this protein is also more symmetrical. Many structural features, including some hydrophobic core interactions, a buried aspartate residue and several unusual turns, are conserved between the two proteins. The binding sites for ligand are in similar locations, and built along similar principles. although none of the specific interactions with the sugars is conserved. A comparison shows further that slightly different rotations relate the domains to each other in the three proteins, with the ribose receptor being the most closed, and the Salmonella glucose-galactose receptor the most open. The primary axis of relative rotation is almost perpendicular to that which describes the intramolecular symmetry in each case. These relative rotations of the domains are accompanied by the sliding of some helices as the structures adjust themselves to relieve strain. The hinges which are responsible for most of these relative domain rotations are very similar in the three proteins, consisting of a symmetrical arrangement of /?-strands and or-helicesand two conserved water molecules that are critical to the hydrogen bonding in the important interdomain region. A region of high sequenceand structural similarity between the ribose and glucose-galactose receptors is also located around the intramolecular symmetry axis, on the opposite side of the proteins from the hinge region. This region is that which is altered most by the relative rotations, and is the location of most of the known mutations which affect chemotaxis and transport in the ribose receptor. Keywords: receptors; X-ray crystallography; periplasmic-binding proteins; bacterial transport; bacterial chemotaxis

information about the features required for this function. Each of the two binding proteins is also The ribose and glucose-galactose receptors the primary receptor for a distinct membrane trans(binding proteins) of Gram-negative bacteria are port system (Galloway & Furlong, 1977; Anraku, 1968a; Boos, 1969). The processesof chemotaxis and unique in that they compete for a common membrane receptor in the chemotaxis system transport are otherwise independent, and mutants (Hazelbauer & Adler, 1971; Aksamit & Koshland, of normally bifunctional periplasmic receptors have 1974; Strange & Koshland, 1976). An analysis of the been found which lack one or the other function structural relationships between these two peri(e.g. see Ordal & Adler, 1973, 1974; Duplay & plasmic proteins should thus provide a good deal of Szmelcman, 1987; Duplay et al., 1987). The high418

1. Introduction

0022-2836/92/180418-23

$08.00/O

0

1992 Academic

Press Limited

Comparison

of the Ribose and Glucose-GalactoseReceptors

419

Table 1 Statistics for the various cmwdinate setsfrom the sourceslisted in Materials and Methods r.m.s. deviations Protein GBP-S-&I$ GBP-S-gal$ RBP-E$

GBP-E-glu§

Resolution (4

R-factorat

Bonds (4

Angles (“1

Dihedrals (7

Impropers (“1

24 1.7 1.7

161 197 187

6010 0.010 0.014

255 239 265

246 239 23.3

686 688 1.18

1.9

14%

Bond distances (4

Angle distances (4

Planar torsion angles ((0.‘)

0.024

0045

6.6

w h ere F is the structure factor amplitude. t &factor = Wc+,,l -IFc~,ell/Wod $ As reported by the X-PLOR analysis routines. 0 This protein was refined using the program PROLSQ (Konnert & Hendrickson, 1980). The statistics given are those reported by the authors in Brookhaven Protein Data Bank data set ZGBP. The statistics reported by X-PLOR are: bonds 6029 A, angles 423”, dihedrals 248”. improper bond angles 314”.

resolution X-ray structures of the glucosegalactose-binding protein from Escherichia coli (GBP-Et; Vyas et al., 1988) and Salmonella typhimurium (GBP-S; Mowbray et al., 1990; J.-Y. Zou & S. L. Mowbray, unpublished results) and of the ribose receptor of Escherichia coli (RBP-E; Mowbray & Cole, 1992) have been reported elsewhere. Here, I compare the structures of these three proteins and discuss the results in the context of their roles in chemotaxis and transport.

2. Materials

and Methods

The co-ordinates for GBP-S (2.4 A resolution structure, Mowbray et al., 1990: 1 A = 91 nm) and GBP-E (1.9 A resolution structure, Vyas et al., 1988, 1989) in complex with glucose are available from the Brookhaven Protein Data Bank (Bernstein et al., 1977) as data sets 3GBP and 2GBP, respectively. Those for RBP-E in complex with ribose (1.7 A resolution structure, Mowbray & Cole, 1992) and GBP-S in complex with galactose (1.7 A resolution structure, J.-Y. Zou & S. L. Mowbray, unpublished results) are currently in preparation for the Protein Data Bank as co-ordinate sets lDR1 and 4GBP, respectively. Some statistics for these refined structures are shown in Table 1. The electron density maps of RBP-E and GBP-S were extremely clear, except in a few places noted specifically below. For GBP-S, the naming conventions for solvents are those from the higher resolution structure. The program 0 (Jones et al., 1999) was used for all computer graphics work. The structural comparisons make use of the lsq feature of that program, which implements the least-squares alignment algorithm of T. A. Jones (personal communication). Initially a transformation was calculated relating the 2 structures in question, using the u-carbon atoms of pairs of segments t Abbreviations used: GBP, the glucose-galactose receptor (binding protein); RBP, the ribose receptor (binding protein). The suffix -E or -S indicates the protein is from Eschrichia coli or Salmonella typhimurium, respectively; r.m.s., root-mean-square.

which were designated to be equivalent after inspection of the structures. The transformation was then improved cyclically in chosen portions of the molecules, matching the longest fragments in which each pair of atoms agreed within an assigned cutoff (using the default of 3.8 A unless stated otherwise in the text). A minimum peptide length of 3 residues was allowed for each individual segment. The program cycles until convergence is reached, up to a total of 10 times. The matrices obtained from 0 which described the rotations were converted into polar angles using the program TRAP0 (from J. Deisenhofer of the Howard Hughes Medical Institute in Dallas, TX). The program DSSP (Kabsch t Sander, 1983) was used for the surface accessibility calculations and for the secondary structure and turn assignments, as well as for an estimate of hydrogen-bond strength based on geometry and backbone dihedral angles. Turns were defined as described by Chou & Fasman (1977). Surface accessibility representations for the graphics display were computed using the program of G. Vriend, obtained from the University of Groningen. Both this program and DSSP make use of the algorithm of Lee & Richards (1971), with a rolling sphere (set to 1.4 A in DSSP) to determine surface accessibility. The program X-PLOR (Briinger et al., 1987) was used for the estimates of residual energy and bond angle deviations in various portions of the structures. One method of measuring the differences in the angles relating the domains in each protein involved the defmition of two planes using 3 atoms from each domain. The plain of the 1st group (atoms 1 to 3) was defined by the vectors from atom 1 to atom 2, and from atom 1 to atom 3. The plane of the 2nd group (atoms 4 to 6) ww similarly defined. The normal to each plane was obtained as the cross product of the vectors defined by the 3 carbon atoms. The dot product of the 2 normals defined the cosine of the angle they subtend.

3. Results (a) Structures of GBP-S, GBP-E and RBP-E The general structure sof the ligand-bound forms of GBP-S, GBP-E and RBP-E are illustrated by

420

-

RBP-E in Figure 1. There are 309 resisues in each GUI’, and RBP has 271 residues. Each protein is a monomer with two domains of nearly equal size. The two domains are highly similar in general structure, i.e. a core of b-sheet flanked on both sides by a-helices. Each domain is similarly composed of two non-contiguous stretches of amino acid chain, with a total of three connections between the two domains; the topology of the connections is identical in GBP and RBP (Fig. 2). Sugar is bound in the cleft between the two domains. (h) Internal

symmetry

The two similar domains of RBP and GBP are related to each other by an axis of intramolecular symmetry, as shown in Figure 1. Domain 1 of RBP-E is defined by residues 1 to 103 and 236 to 263, and domain 2 by 104 to 235 and 265 to 271; residue 264 is between the two domains. When the first segment of domain 1 of RBP (residues l-103) was compared to the equivalent portion of domain 2 (residues 125-235), there were 82 out of 102 atoms matching, with an root-mean-square (r.m.s.) agreement of 1.80 8. The operation needed to rotate domain 2 onto domain 1 consisted of an approximately 178” rotation around an axis centered near residue 264. Most of the other segments of the domains also agreed well after the alignment, including the fit of helix E to helix J, and the /?-strand at 259-263 to that at 265-269. A /?-bulge is found only in domain 2 at residue 157; otherwise the sheets agreed extremely well. The poorest fits were of loop residues 41-42 to 163-164, and of helix C to helix H. The strand-like structure at 91-95 (where helix D has been deleted relative to GBP) was also not equivalent to helix I in domain 2. In the aligned sequences, 12 out of 82 residue pairs have the identical amino acid, giving a sequence conservation of 150/6. The first domain of GBP consists of residues 1 to 110 and 257 to 293, and the second domain of residues 111 to 256 and 295 to 309. Residue 294 is between the two domains; this residue is the st)ructural equivalent of residue 264 of RBP. Alignment of the two domains using residues 1 to 110 and 140 t’o 256 of GBP-S showed that this protein is much lesssymmetrical than is RBP. A tightened cutoff of 3.2 w was needed to prevent spurious alignments between the domains, and only 61 atoms could be brought into agreement with an r.m.s. difference of 1.76 A (chosen to be similar to that’ obtained for RBP above). The operation needed to rotate domain 2 onto domain 1 consisted of an approximately 174” rotation around an axis centered at residue 294. As was noted for RBP, the sheets showed the best agreement, with the exception of the B-bulge (residue 176) which was again found in only domain 2. The regions that agreed poorly in RBP generally agreed poorly in GBP as well, with several additional large regions of differences. The end of helix F did not match well to helix A due to a kink in the former, and helix E did not match well

Figure 1. The structure of REP-F:. illrrstraf ittg tltca c*ompositiott of each domain xs thta sum ot’ tur, ttott contiguous segments in the srquenct:. as well as t IIP symmetry of the arrangetnrnt of the 2 domains within t hc protein. The receptor is drawn as a ribbon t,ltat is cellarcoded according to residue number. beginning with red at residue I and rnding with blue at rrsidut, 271. Thea segments of the protein that include the hinge arr higtrlighted by filling the ribbon. and the location of t&--o water molecules that, BPP important in the hinge are shown as yellow spheres of radius 0.5 A. The axis of intratttolecrrlar symmet,ry would he perpendicular to the paper brtwrtqt these water molecules. to helix J. The agreement was, however, het,ter between helix 1) and helix I (compared to RBP), since both are helical st’ructures in GRP. In the aligned sequences of the domains. 8 out) of 61 residue pairs have the identical amino acids (13’$, identity). (c) Comparison of GBP-E and GHP-S Since GBP-S and GBPE will substitute for one another in viva (Miiller et al., 1982), information gleaned from both structures should be used in any comparison to RBP-E. Only 18 out of 309 residues differ between the sequences of GBP-S (BennerLuger & Boos, 1988) and GBP-E (Mahoney et ~2..

Comparison

of the Ribose and Glucose-Galactose

Receptors

421

Figure 2. The topology of RBP and GBP. and the location of some residues of interest as discussed in the text. The residues of the binding site are indicated by the filled letters, and other residues by open lettering. The metal ion in GBP (presumabl>- c-al(*ium) is indiratrd by M.

Figure 3. Stereo backbone showing the structure of GBP-S. The residues are colored to illustrate sequence changes between (>BP-S and GBP-E. with residues that are conserved shown in yellow. and residues that are different, shown in blue. Eve ry 20th residue is numbered.

Figure 4. The structures GBP-S (yellow) (gold), with domain 1 aligned to illustrate relative orientation of the domains in the 2 residues of the immediate hinge region (109-l 29% 295) are shown in cyan in GBP-E.

and GBP-E the different proteins. The 11, 2X-257,

1981; Scholle et al., 1987) giving an overall 94% sequence identity. When the backbones of the entire proteins are superimposed, the u-carbon atoms agree with an r.m.s. difference of @52 8. Only the positions of the C-terminal residues (309) differ by more than the default 3.8 a value for equivalent atoms. The amino acid changes are generally confined to surface residues which would not be expected to be important in the overall structure (Fig. 3); the fact that the receptors can replace each other in wivo also suggests that these residues are not involved in either transport or chemotaxis When the cc-carbon atoms of either domain of GBP-S were aligned separately to the equivalent residues of GBP-E, the agreement between those residues was significantly improved, indicating that the domains are related by a somewhat different rotation in the two proteins (Fig. 4). The r.m.s. differences for the separate alignments were @34 w for the 147 atoms of domain 1, and 0.21 A for the 159 atoms of domain 2; the improved agreement within each domain as measured by eye was dramatic. This level of agreement between the individual domains of the two structures correlates well with the mean of @33 a observed for the differences in five protein structures solved in different

Figure 5. Conformational changes in GBP-8 relative to GBP-E. A ribbon drawing was generated with the program Molscript (Kraulis, 1991). The relevant helices are lettered (solid type) according to the definitions in Fig. 2, and the directions of their movement are shown by arrows. The N and C termini are labeled in open type.

environments (Chothia & Lesk, 1986). When eyuivalent domains of each GBP are aligned, the other domains shows small differences in atomic position ( I 1 8) near the cleft, but larger changes (2 a or more) at the ends of the domain furthest from the cleft. To locate any more subtle changes in the two proteins, the set of atoms allowed in the comparisons was reduced to include only those most similar in the two structures. When the cutoff in the comparison was tightened to @3 8, and the first contiguous segment of each domain was used, the r.m.s. difference improved to 0 16 A (83 out of 110 atoms matching) for domain 1, and to @15 A (125 out of 146 atoms matching) for domain 2. This agreement is similar to that we have obtained from independent refinements of the same structure (see below for the structures of GBP-S with glucose and galactose) and agrees with an estimate of co-ordinate error in GBP-S (@18 A; Luzzati, 1952). Some atoms in each domain of the GBPs did not match within the 03 A cutoff. These changes primarily represent the sliding motions of helices along the direction of the helix axes in each case; the b-sheets are changed very little (see Fig. 5). In domain 1, helices A and J move by 05 to @7 !I and 0.35 A, respectively. In domain 2, helices T and K move by @5 A and 035 A. None of these changes

Comparison

of the Ribose and Glucose-Gakctose

could be attributed to the difference in the sequences of GBP-S and GBP-E. Those sequence differences are most’ly confined to surface residues, and even in the few places where interior residues are different (residues 28, 143, 163 and 250), the local structures aligned within the 03 A cutoff for backbone atoms. The extensive hydrogen bonding and hydrophobic interactions in the structural domains would be expected to result in a relatively rigid structure within each, as is in fact seen. Thus, most of the relative rotation should be due to changes in torsion angles in the three short segment’s of relatively unconstrained amino acid chain that connect bhe domains. Inspection of the individual structures suggests that the most important residues of this hinge region would be 109 to 112, 255 to 258 and 292 to 295. After superposition of the domains of each protein on their equivalents, these stretches of chain are clearly seen to be the pivot point for the rotation (see Fig. 4). The changes in main chain dihedral angles are primarily concentrated at the honds of 109-I 11, and 292-293. There is not much rot)ation at the residues around 256, as helix J moves slightly instead. The differences in hydrogenbond lengths and phi-psi angles in this hinge region are not, in fact’, significantly larger than those noted elsewhere in the two structures. The overall rotation effect is thus generated by the summation of a number of small but concerted changes in dihedral angles without gross structural differences in any one p1ac.e. The degree t’o which these proteins are more “open” or “closed“ relative to each other can be measured in several different ways. Since other minor conformational differences were noted, as well as the rotations, the measurement of the angle of relative twist was subject to the exact definition of the domains themselves and the point of rotation chosen: as well as the treatment of any translational components. In one method, the domains were each described as a plane defined by residues chosen from the first five strands of the sheets, since these appeared to vary least; sheet residues 38, 66 and 106 were used for domain 1, and 178, 210 and 252 for domain 2. One domain was aligned separately using the 0.3 A cutoff as described above, and the rotation measured as the difference in the angle between the planes of the other domains. By this measurement, GBP-S was determined to be about 2.7” more open than GBP-E when domain 1 was used for the alignment, and 3.5” more open when domain 2 was used in the alignment. In a simpler method, the distances were measured between several residues that represent the closest approach between the domains, on the lips of the cleft opposite the hinge. Residue 69 of GBP-E is 8.7 A from residue 151 and 8.1 A from residue 152, compared with 95 A and 8+6A for the same distances in GBP-S. An angle may be defined between residues 69 and 151 on the lips, and the hinge residue 294, which lies between the domains, roughly 22 A away. Using this measure, the relative angular change involved is estimated to be about 2”.

Receptors

423

Another simple method relates the 2 A difference at the ends of each non-aligned domain to their distance from the cleft (around 30 &A). giving an angular difference of 3%‘. By any of these measures then, a change in the inter-domain angle of roughly 3” is obtained: differences in the actual values in each case are largely due to different effective weighting of any translational components. All of these methods indicate that the rotations are around an axis roughly perpendicular to the axis of int’ramolecular symmet,ry (i.e. about an approximately horizontal axis in Fig. 1). 1 shall use GBP-S for most of t’he more detailed comparisons to RBP-E, as this struct,urr has been relined at higher resolution (1.7 A z~~sus 1.9 A resolut’ion), and the availability of the electron density maps allows a better estimate of which regions are best (and worst) determined in the structure. as well as any possible regions of conformational variability which are not reflected by the co-ordinates. (d)

Overnll

comparison of the GRP structurei;

RBP-I/:

a,nd

The superposition of RBP-E to t,he two GBP structures was performed initially using all a-carbon atoms in the alignments. When comparing the a-carbon atoms of RBP-E with those of GBP-S, there was an overall r.m.s. difference of 1.65 A (254 atoms matching). The same comparison with GBP-E gives an r.m.s. difference of 1.49 A (259 atoms matching). By this measure, there is an approximately 95% structural homology of the backbone atoms of RBP-E to those of the GBPs. The relationship of the complete structures of RBP-E and GBP-S is illustrated in Figure 6. RBP-E is smaller than the GBPs, with no insertions in the sequence relative to GBP, and deletions near residues 32, 100, 130, 170, 200, 280 and 302 (of the GBP sequence) giving rise to differences in the respective loops and the loss of helical segments D and K in RBP. An improved agreement between the st’ructures when the domains were considered separately indicated that there were differences in the relative domain closure of RBP-E compared to both GBPs. When the a-carbon atoms of the first contiguous segment of domain 1 of RBP-E were aligned separately to those of domain 1 of GBP-S or GBP-E (using residues 1 to 120 of RBP and 1 to 109 of GBP), the agreement was improved to 1.24 and 1.22, respectively, with 98 out of 102 atoms matching. The equivalent numbers for domain 2 (using residues 103 to 235 of RBP and 110 to 256 of GBP) and 1.25 and 1.21 (with 128 out of 133 atoms matching). To reduce the set of atoms used in the alignment to only those most similar in the structures, a cutoff of 1 A was used to compare the first contiguous segments of each; in this way, 54 out of 102 a-carbon atoms could be aligned in domain 1, compared to 78 out of 133 in domain 2, with r.m.s. differences of 656 A and 950 A: respectively.

S. I,. Mowbray

424

(a)

(b)

Figure 6. Ribbon

drawings of the backbone structures of (a) GBP-S and (b) RBP-E, aligned according to domain 1, prepared using the program Molscript (Kraulis, 1991). Regions in which the atoms differ by more than the default 3% A cutoff for the entire structure are colored darker to highlight them. These are residues 71, 94-97, 121-122, 146-151, 180 and 258 in RBP, and the non-equivalent residues 1,32, 73,96-104, 12%139,163-170,199-201, 279-288 and 301-309 in GBP. The metal ion of the calcium-binding site of GBP is shown (Ca) in the loop where it binds. Inspection of the domains after alignment with the tightened cutoff showed that the p-sheet structures of all three proteins were essentially identical.

Few of the large differences in the CBP structures were not directly explainable deletions in RBP relative to GBP. These

Table 2 Regions of RBP-E and GBP-S that are not structurally equivalent when aligned using a 1 A cutoff RBP

GBP

13-31

14-33

55-57 70-83

57-59 72-85

A. Domain 1

-

B. &main

about 100

231-254

258-265

104-120

111-127

133-153

127-137 150-172

176-182

195-203

2

Remarks Helix A slides by 1.5 to 2.5 A; related to difference in binding site interactions or domain closure; also deletion of 1 residue at GBP-32 Related to differences in helix C Helix C slides by 1.5 to 2 A and rotates about 25”; related to deletion at GBP-100 in nearby structure Deletion of 5 residues removes helix D from RBP Helix J slides by 1.5 to 2 A; related to deletion of 9 residues at GBP-280 or hinge movement Helix E slides by 2 A; related to deletion near GBP-130 or hinge movement Deletion removes calcium-binding loop in RBP Helix F slides by 1 A at beginning, end of helix kinked differently; related to deletion after GBP

207-208

228-229

226-227 271

247-248 301-309

170 and different

sugar

orientation

Related to deletion in RBP Related to sequence change; more local disorder in RBP Related

to deletion

of helix

K from

Deletion removes helix K from RBP

RBP

and RBP by the deletions,

Comparison


Receptors

425

Figure 7. Differences in domain closure between RBP-E (red) and GBP-S (gold). The a-carbon backbones of the entire proteins were aligned optimally. The atoms of the sugar molecules are shown with yellow carbon atoms for ribose, and gold carbon atoms for galactose; they are labeled at “equivalent” atoms in their rings (O-6 for GBP, H51 for RBP) to show differencesin the ring orientation. The 3 atoms in each protein that were used for the “closest approach” method of estimating domain rotation are labeled. of course, cause alterations in the loops directly involved, and the loss of helices D and K in RBP,

but they also result in some changes in nearby structures as well, mostly consisting of the relative movement of helices as was noted for the GBPs. The residues of RBP-E and GBP-S that differ by more than 1 L% after the alignment are described in Table 2; most will be discussedfurther below. The difference in angle between the domains of RBP

and

the

GBPs

was

estimated

as described

above for the GBPs. In this case, due to the larger differences between the structures compared, the measurements of the angles involved were more varied. In RBP-E, a plane was defined by sheet residues 36, 64 and 99 for domain 1, and 159, 189 and 231 for domain 2. Each of the domains of RBP and

GBP

was

aligned

using

the

1A

cutoff

about

4 A less than for the distances

between

RBP-S is partial (BuckenmeyerL Hermodson,1983).The sequence identity of RBP-S to RBP-E is 98 out of the 104 residues that have been determined in the protein sequence, for a homology of 94% in those regions. It is therefore presumed that, like the GBP structures, those of the RBPs will be highly similar. The sequenceof GBP-S shown here has histidine at residue 152 instead of alanine (Mowbray et al., 1990). Bars above the sequence indicate alignment of RBP-E and GBP-S within the 3.8 A cutoff for the whole molecules. Since these merely represent the locations of backbone atoms in similar positions, they are not equivalent in position to insertions and deletions in the protein-gene sequences. The residues that are known to be identical in all three (or four) proteins are marked with an asterisk (*). Structural elements (indicated as a for a-helix, 3 for 3-10 helix, b for b-strand and t for turn or bend) were assigned for the RBP-E and GBP-S structures using DSSP; each helix region is labeled according to the nomenclature of Fig. 2. The residues of the binding sites of RBP and GBP are indicated as r and g. respectively.

as

described above, then the difference in angle between the planes in the other domain was calculated. By this procedure, RBP-E is about 8” more closed than GBPS, and 7” more closed than GBP-E. The angular difference was also estimated using the distances between residues on the lips of the cleft, as for the GBPs. The distances from residue 67 to residues 134 and 135 are 5.1 A and 4.9 A, respectively,

Figure 8. Alignment of the sequences of RBP-E (Groarke et al., 1983), GBP-S (Benner-Luger & Boos, 1988) and GBP-E (Mahoney et al., 1981; Scholle et al., 1987) based on the structural results. The sequence of

the three equivalent residues (69, 151 and 152) in the GBPs (seeFig. 7). The domains of RBP-E are in fact sufficiently close that there are interdomain hydrogen bonds across the cleft, from atom 68-N to 134-0, from 68OG to 136-N, and from 91-N to 136OG. When an angle is defined using the equivalent residues and hinge residue 264, which lies between the domains about 23 d away, RBP-E is estimated to be about 11” more closed than GBP-S, and 9”

Figure 9. Stereo diagram of the backbone &u&n-e of RHP-E illustrating the conservation of residues in RBP GRP. Those residues found only in RRP are shown in whitfa. and those known to be conserved in all the RRP and ( sequences colored otherwise. Residues conserved for primarily st,ructural reasons are colored green, those in the bin site are clolored blur. and surface residues of possible functional significance are czolored red. Surface residues for w the reasons for conservation may be either structural or funrtional are shown in magenta. The view is approxim: down the intramolecular symmetry axis. Every 20th residue is numbered.

more than GBP-E. The fact that GBP-E is more similar t,o RBP-E in relative closure accounts for the better apparent agreement between those two structures when the whole molecules were compared. Again, the relative domain rotations are around an axis roughly perpendicular to the axis of intramolecular symmetry. When domain 1 is aligned, the difference in relative rotation of domain 2 is seen to begin at RBP 102-103 (GBP 109-110) and RBP 262 (GBP 292). Helix J moves as a unit, and so the associated hinge strand shows little relative rotation. When the domains 2 are aligned, twisting can be seen at GBP 257-258 relative to RBP 236-237, but the regions of GBP that are equivalent to those of RBP near residues 104 and 265 are different enough to make it difficult to precisely determine the point at which the rotation begins. (e) Sequence

alignment of RBP-E, RBP-S, GBP-E and GBP-S Based on the structural alignments, it is possible to generate accurate sequence alignments between

RBP-E (and the partial protein sequence of RBP-S) and the GBPs, as shown in Figure 8. The great structural similarity of GBP and RBP is thus seen to be based on a 24% sequence identity in the regions of the structures that are structurally equivalent. The types of amino acid residues that are conserved are distinctly non-random. Of the 62 residues known to be conserved in all four proteins, 11 are alanine, 10 are glycine, 17 are valine/isoleucine/leucine, 10 are aspartate or asparagine, and 14 are other residues. The locations of the residues which are known to be conserved are illustrated in Figure 9. These conserved residues fall into three general classes. The first class includes residues found in the sugar-binding sites. The second consists of residues that are conserved for primarily structural reasons. The third class is that of surface residues which may be significant in the function of these proteins in chemotaxis and transport. The second and third classes probably overlap. The nature of these conserved residues is discussed in more detail below.

Comparison

qf the Ribose and Glucose-GalactoseReceptors

427

(b)

Figure 10. Stereo diagramsof the binding sitesof (a) RBP-E and (b) GBP-S. The sameorientation is shownin the 2 cases, using the optimal alignment of the entire molecules. The residues are labeled with amino acid number and residue

type. (f) Binding sites All three struct’ures discussedhere contain bound sugar. Structures of GBP-S have been described both with bound glucose (Mowbray et al., 1990) and galactose (J.-Y. Zou & S. L. Mowbray, unpublished results). The crystals of these two forms of GBP were isomorphous, and the two structures show no appreciable differences (r.m.s. difference of 012 a for all cl-carbon atoms, with no changes in the interdomain angle or arrangement of residues within the binding site). The structure of GBP-E contains bound glucose (Vyas et al., 1988). In each of the GBPs, the sugar is bound primarily in the j?n-pyranose form, with the a-anomer being present at about 30% of the total. The structure of RBP-E was solved with the sugar bound exclusively as B-Dribopyranose (Mowbray & Cole, 1992), the most plentiful form of the sugar found in solution (Pigman & Anet, 1972). The residues of the binding sites are shown in the

context of the topology of the proteins in Figure 2, and stereo views are given in Figure 10. Zt may be seen that, while similarly positioned residues are used, the details of the binding site are substantially different in the two proteins. The sugar molecules are differently oriented, and the interactions they form with the proteins are completely different (see Figs 7 and 11). The same principle types of interactions are, however, used in both cases;both receptors make extensive use of hydrogen bonding, hydrophobic and steric factors in the binding of sugar. A complex web of hydrogen bonds involving residue side-chains from both domains provides the primary framework (Fig. 11). The capacity of the sugar hydroxyl oxygen atoms to accept multiple hydrogen bonds makes the ligand a net hydrogenbond acceptor in both proteins. Hydrophobic interactions with both faces of the sugars are provided by residues Phe15, Phel6 and Phe164 of RBP, and residues Phe16, Met17 and Trp183 of GBP. which

REP-rib

GBP-gal

Figure 11. Hydrogen-bonding patterns in the sugar-binding sites of RBP-E and GBP-S bound to Chose and galactosr. The carbon atoms of the sugar molecules are numbered. The direction of the dotted arrows indicates whether the involved groups are hydrogen-bond donors or acceptors, as indicated by the location of hydrogen atoms in t,hts refinement, as well as factors of distance and geometry. The binding site interactions of GBP-E (Vyas rt al.. 1988) with glucose are essentially identical with those shown here for GBP-S. In both proteins, residue Asp14 a& in alternat,ivr ways depending on whether glucose or galactose is bound. and residue Asp154 is used differently for the a ant1 fi-anomrrs of racah sugar. are drawn from equivalent positions in the two recept’ors. A water molecule forms a direct link to the sugar in the GBP binding sit’e (86 in GBP-S. 6313 in GBP-E), while two water molecules are involved more indirectly in the RBP-E binding site. The resulting highly tailored binding sites leave litt,le room for variatron in the type of ligand bound by each protein. RBP is able to accommodate only allose (in which the equatorial aliphatic hydrogen of ribose, H51, is replaced by a CH,OH group); that’ sugar has a measured K, of 3 x 10e4 M, compared with 3 x IV7 M for ribose (Aksamit & Koshland, 1972, 1974). Inspection of the structure suggests that, other sugars: including the cl-anomer of ribose itself, are excluded from binding due to sterie problems. GBP is able to accept both glucose and galactose with K, values of 2 x 10.. ’ M and 4 x lo-’ M, respectively, but other sugars are bound less tightly (Anraku 19683; Boos 1969; Zukin ef al.. 19776). The fl-anomer of each sugar is preferred (Richarme & Kepes, 1983; Miller et al., 1983). Only four residues, Phel5, Argl41, Asnl96 and Asp215 in RBP, the equivalents of Phe16, Arg158. Asn211 and Asp236 of GBP, are conserved out of the ten residues in RBP and 11 in GBP that interact directly with the ligand. While this is higher than the overall 24O/b sequence identity, it is heavily weighted by the fact that only certain residues will supply the needed hydrogen bonding and hydrophobic interactions. Since these “equivalent” residues form interactions with different atoms of the respective sugars, they are clearly conserved more because of their general character and loca-

tions on loops than because of specific roles in sugar binding. Asp89 of RBP and Asn91 of GRP are also similarly placed, but do not bind the samr sugar hydroxyl group. A closer functional relat’ionship exists bet’ween Asn 13 of RBP and Asp1 4 of G BP, as both are involved in binding the reducing end of the sugar. but one is a proton donor and tjhe ot,her an acceptor, and their conformations in t.hr binding site differ significantly. The other residues of thtb binding site come primarily from simila~rly located loops. but are not conserved. The residues of RKP which are equivalent to GBP His152 and Asp154 are not used in sugar binding. An additional point of interest’ is that t,hew is little symmetry in terms of the t’ype of interactions formed by binding site residues located in equivalent places in the two domains of either protein, although some are from symmetry-related loops. Only Asp89 and Asp215 (equivalent to Asn91 and Asp236 of GBP) are related in this manner. There is an additional level of conservation of tht binding site residues in their side-chain conformations. Where the same or similar residue is found at, a particular loop position, the side-chain conformations are almost identical. In all but one case (residue Asnl90 in RBP, equivalent to Asn211 of GBP), these residues represent as well a rotamer conformation which is the most frequent, or second most frequent residue conformation in the list of Ponder & Richards (1987). The similarity is probably largely the result of the very similar location of the sugar in the two proteins, but the consistent choice of the more frequent rotamers bespeaks

Comparison


Figure 12. Stereo diagram showing the structure surrounding bonds to nearby atoms are shown asbroken lines. strategies which the proteins.

were part, of the overall

design of

(g) Other conserved features Most of the residues which are the same in RBP and GBP are conserved for reasons which are primarily structural. Not surprisingly, this category includes most of the conserved hydrophobic residues. Some are involved in the hydrophobic cores formed by the packing of the helices onto both sides of the sheets. A fairly large region of the hydrophobic core of domain 2 is preserved unchanged, and smaller portions of the hydrophobic cores are conserved elsewhere (see Fig. 9). Not all of the conserved hydrophobic core residues are actually found in the equivalent conformations, however, as will be discussed later. The class of residues conserved for structural reasons also includes t,he glycine and alanine residues that are required for correct helix-on-helix packing, and to allow room for binding site residues to be correctly placed. Other residues are conserved due to their role in supporting binding-site residues through hydrogen bonding. Some additional residues are conserved because of their effects on local conformation. These include three glycines, which are in lefthanded a-helical turns at the C-terminal ends of helices (Schellman, 1980), and three aspartate residues, which are involved in turns at the N-terminal ends of helices (Baker & Hubbard, 1984). Another case involves a buried aspartate residue at Asp191 of RBP, as illustrated in Figure 12. An equivalent structure exists in GBP, centered around Asp212. In each case, the aspartate is located next to a binding-site residue (Asnl90 of RBP, Asn211 of GBP), and its side-chain forms three hydrogen bonds to the main-chain amide nitrogen atoms of a nearby loop that is also involved in sugar binding (214-218 of RBP, 235-238 of GBP). That the aspartate residue is unprotonated is implied by the large number of acceptor hydrogen bonds and by the structural perturbations introduced as the protein tries to provide these hydrogen bonds. Residue 215 of RBP (235 of GBP) is outside the “allowed”

Receptors

429

the buried aspartate at, Asp191 of RBP. Hydrogen

region in a plot of backbone dihedral angles (Ramakrishnan & Ramachandran, 1965), is flagged as unusual in a search of the database of 32 highly refined protein structures in 0 (Jones et al., 1991), and shows residual strain (bond angle deviations of roughly 12”) after refinement. The burial of the charge is made possible only by the extensive hydrogen bonding involved; this residue will not be exposed even when the protein assumes a more open conformation. In addition to the main-chain amide nitrogen atoms, the hydroxyl groups of two nearby threonine side-chains (Thr217 and Thr232) in RBP, and one in GBP (Thr253, equivalent to RBP Thr232) provide hydrogen bonds t’o the carboxylate group. A glycine residue (RBP (~1~213, GBP Gly234) is conserved to allow room for the internal aspartate side-chain. Inspection of co-ordinate set, 1ABP (Gilliland & Quiocho, 1981) from the Protein Data Bank indicates that the similar arabinosebinding protein involved in transport in E. coli also has this buried aspartate. Residues Asp206, and Gly230 are present at the correct locations in the structure and Ser251 could function in the same way as Thr232 of RBP. There is also an usual multiple turn structure at residues 38 to 42 of RBP and 40 to 44 of GBP which is reflected in a strongly conserved segment of the sequences. The structural pattern observed for both proteins illustrated by RBP in Figure 13. The first turn is formed by the interaction of Asp38 with the main-chain amide group of residue 40. The second is a type III’ b-turn involving residues 39 to 42, leading into an Asx turn with residue Asn42 (Asp44 in GBP) at the beginning of helix B. This structural element includes two residues with phi values around +60” (3rd and 4th residues of the segment). The unit is also associated with a /?-bulge at the edge of the sheet of domain 1. This bulge is similar to the Gl bulge defined by Richardson (1981) but the presence of a type III’ turn rather than the glycine-containing type II turn results in the /l-turn twisting in the opposite direction relative to the sheet. This arrangement seems to be critical to the support of the residues in the first binding-site loop. There is a similar structure located in the equivalent place in domain 2. at residues 161 to 165 of RBP

Figure 13. Stereo diagram illustrating the multiple turn structure which is located at residues 38 to 42 in RBP-E. The hydrogen bondsthat are included in the turns and the P-bulgewith the next strand at residue9 are shownas broken lines.

(180 to 184 in GBP). The initial aspartate residue is replaced by ProI in RBP, but the local conformations of the loops themselves are almost identical to those of domain 1. The expression of this second loop in the structure is, however, slightly different in the two domains, with the /?-bulge being absent from domain 2, due to the different loop sizes at the other end of the relevant helix and the effects of nearby structure. These conformations result in the presentation of Gln40 and Asp163 of RBP, and Gln42 and Met 182 of GBP on the surface of the proteins. Some other structural features are preserved that are not directly reflected by the sequence conservation. Helix F is noticeably kinked in both proteins. Several hydrogen bonds within the helix (between residues 143 to 148 in RBP and 158 to 163 in GBP) are also detected as weak (energy < 1.5 kcal, with the normal being around 2.2 kcal: 1 cal = 4.184 J). This helix contains a binding-site residue (Arg141 in RBP, Arg158 in GBP), at a position just before the kink. The helices are tilted in a different direction and to a different extent in the two proteins (10” in RBP, 25” in GBP) due to a combination of several factors. The different orientation of the sugar molecule in the binding sites seemsto be related to the fact that the whole helix slides to a slightly different position within domain 2. In GBP, two threonine residues (159 and 160) and a buried water molecule (514) can hydrogen bond to provide stabilizing interactions with the main-chain oxygen atoms of the earlier portion of the helix, helping to keep the arginine aligned properly in the binding site. Two extra residues (171-172) in the loop at t’he end of the helix, however, allow more bending to occur in the second half. The last half of the helix F in RBP seemsto be tilted in the opposite direction due to the pull of its shorter loop. A similar kinked helix is found to include the equivalent binding-site arginine residue in the E. coli arabinose-binding protein (Gilliland & Quiocho, 1981) as well, and a Thr-Thr sequence is found at the same position as in GBP. There is a l-3 hydrogen-binding interaction in RBP at residues 88-90, and in GBP at the equiva-

lent residues 90 to 92; the central amino acid residue in both cases is an Asx residue involved in sugar binding. The COi to Ni+* distance in RBP is 2.9 A and 3.0 in GBP, and the strengths of the hydrogen bonds are estimated to be - 1.7 kcal and - 2.3 kcal, respectively. A full y-turn (Nemethy & Printz, 1972) is not completed in either case. The hydrogen bond between Ni to COi+2is not formed, since Ni is involved in the P-sheet. In GBP, an internal water molecule (S7) is also found inserted between Ni to in RBP, this space is filled by the side chain coi+2; of Pro65. A second B-bulge (Richardson et al., 1978) exists at residue 157 in RBP (176 in GBP), on the edge strand of the second domain. The interaction leaves 158-N (177-N in GBP) without a hydrogen-bonding partner. This feature is not found at the equivalent position in domain 1 of either protein. Its presence in domain 2 seemsto be t’he result of a one-residue insertion with little effect’ on the structure, except very locally, since the nearby sheet atoms still agree well with those of domain 1. Another pattern that becomes apparent in a detailed comparison is that’ of finding a high proportion of aspartate and asparagine residues in the context of structural oddities. In the caseof the l-3 hydrogen bonds, four are found in GBP and three in RBP; in five casesthe central residue is aspartate and the remaining two are asparagine. While this pattern is distinctly non-random, no obvious explanation can be offered concerning its origin, since the side-chains in no way force the conformations. With the exception of residues 89 of RBP and 91 of GBP, which were discussed above as Ramachandran “violations”, all have phi-psi values in the broad /?-region. A search of a number of highly refined protein structures (Baker & Hubbard, 1984) revealed that l-3 hydrogen bonding interactions are quite rare, but the authors made no note of any increased incidence of Asx residues associated with them. Asparagine residues (and more rarely lysine) are also found in left,handed a-helical turns (Schellman, 1980) at the end of the helices of RBP and GBP. Residues 78 to 83 in RBP (sequence ANQANI in RBP-F, ANQAKI in

( ‘omparison


Receptors

431

Figure 14. stereo diagram showingthe hinge region of RBP-E. Residues100 to 110, 232 to 242and 260to 268 of the hinge region are shown, and some surrounding features, including the buried residue at Aspl91, and the unusual peptide c-onformation ar residue 215. The backbone is colored according to residue number as described in Fig. I. The 2 water molecules of the hinge (860 and S61) are shown as red spheres with a radius of 0.25 A. The ribose molec~ule is shown in green. The main-chain-main-chain hydrogen bonds that are involved in secondary structure are shown as broken lines in line across the proteins, including the l-3 hydrogen yellow. Some important interactions that fortn a symmetrical bonding and close contacts as well asthe water interactions, are shownas dotted linesin cyan. This describesthe major bending plane in the molecule. Other important interactSions involving side-chains are shown by dotted lines in red.

RHP-S), corresponding to 80 to 85 in GBP (sequenceARGQNV) are such a case, where the lefthanded conformation is at residue 5 of each sequencxe.Another involving lysine includes RBP residues 149 to 154 with the sequence 1’AAHKF; the equivalent region of GBP is alt,ered by the insertion of two residues. This type of distribution of amino acids in (I-terminal regions of helices has been noted elsewhere (Schellman, 1980; Richardson & Richardson, 1988). (h) The hinge In RBP, the major bending regions of the hinge include residues 102-105, 234-237 and 262-266; the equivalent segments of GBP are 109-112, 255-258 and 292-296. These segments are situated around the axis of intramolecular symmetry. The local structure of this portion of RBP is shown in Figure 14. The details of the hinge region are very well conserved in GBP, in berms of the interactions of t’he protein structure and the water molecules involved, although not in terms of sequence. Only \‘a1263 (\:a1293 in GBP) is actually conserved, perhaps due t’o its favorable packing against helix J (which is also part of the larger structure of the hinge, and one of the helices which appears to “slide” with relative domain movements). In each protein. two symmetrically located water molecules are intimately involved in the transition of the hydrogen bonding from the p-sheet of one domain into that, of the other. The water molecule designated S60 in RBP hydrogen bonds wit)h 236-N and 264-O as well as 103.OG. This water molecule is

equivalent t’o S17 in GBP-S (S413 in (:BP-E) which hydrogen bonds t!o 257-N, 293-O and 294-C). In all three cases, this water is completely buried in the closed protein. A second water, 861, of RBP is involved in hydrogen bonding wit’h 104-N and 263. 0. The equivalent water: designat’ed S127 in GBP-S (S374 in GBP-E). is hydrogen bonded t)o 11I-N. 114. OEl and 293-O. The hinges also contain severa, 1-3 hydrogenbonding interactions. One exists in. the first hinge segment of each protein, at 103-105 of RBP-E and 110-I 12 of GBP. In each case, the COi to Nic2 distance is reasonable for the central hydrogen bond (3.1 A4 between 103-O and 105-N of RBP. 3.3 A between 110-O and 112-N of GBP-S). but the amide proton is not directly aimed at the carbonyl group (the so-called inverse y-conformation: Baker & Hubbard, 1984). The estimated strength of this hydrogen bond is - 1.2 kcal for GBP-S. - I.5 kcal for RBP. There is another l-3 hydrogen-bonding interaction in the hinge of GBP at residues 256 to 258 (sequence NIlA): with a distance of 2.8 ,%from 256-O t’o 258-h’ and an estimated strengh of - 1.8 kcal. This has no direct counterpart in RBP, but the equivalent QLP sequence has a close contact of 3.2 a between 235-O and 237-(11) at the same place. These interactions occur in symmetryrelated positions in the two domains (see Fig. 14), and appear to be forced by the path of the prot,ein from the P-strand into the helix, with the water interactions between. The central residue in each of the l--3 hydrogenbonding interactions in the hinge is an aspartate involved in a real or pseudo-Asx turn. The hydrogen

bonding distance for the side-chain interaction with t,he main-chain amide group is somewhat large for one in GBP (4.2 A from 111-O to 114-N) but, from the electron density it. is clear that the side-chain conformation is appropriate to an Asx turn. Perhaps the helix dipole (Ho1 et al., 1978) helps keep the side-chain properly oriented. As noted by Richardson $ Richardson (1988), these turns can involve interactions of the Asx with residue n+ 2 or r) +3, or bot.h. depending on the psi angle. This would seem to offer particular advantages in a region of conformational change, since it would nllm alternate ways of capping the helix, depending on the main-chain conformation. The psi angle of RBP 104 is 95”, that of GBP 111 is 104”, and that of GBP 257 is 85”, which are consistent with their current hydrogen bonding to residue n + 3 in each case. As mentioned above, helix J in RBP does not have an Asx turn, but it is already “capped” with a tightened turn in which Leu236-0 hydrogen bonds to Gln239-N and Tle240-N with a similar main-chain conformation (psi of residue 236 is 72”); it also has a proline at, residue 237. Proline residues are common in t,he hinge region (2 in RBP, 1 in (:BP). which opens up the possibilitv of ris-trans isomerizations in the conformational change. although all are in t,he bans conformation in the structures described here. The reason for the buried aspart.ate residue at residue 191 of RBP (conserved as Asp212 of GBP. as discussed above) also becomes apparent on studying the hinge. The “most unusual” aspect. of the structure resulting from t,hat aspartate is the conformation of the peptide at residue 215 of RBP (236 in GBP). A search of the protein database indicates t.his pept,ide would normally assume t,he opposite orientation in a similar structural context; this residue is also one of only two Ramachandran violations in each protein. That alternate peptide orientation, however, would disturb the hydrogen bonding between t’he b-strand segments on the hinge region at residue 235 in RBP (256 in GBP) and so would be highly disruptive of the hinge itself. The proteins thus force the needed peptide orientation at, this location by use of the buried charge. The involvement of t’his structural feature in both binding-site and hinge interactions explains the fact t,hat the surrounding regions of the two proteins show t,he highest sequence homology (Argos ef al.. 1981; Vyas e!t ~1.. 1991). It is significant. that the ot.her Kamachandran violation in both proteins (residue 89 of RBP. 91 of GBI’) is locat.ed at. the equivalent place in domain 1. and is preserved for the same reasons; this portion of the structure was discussed above as part of a l-3 hydrogen-bonding interaction in both proteins. The hinge segments including residues 262-263 of RHP and 256 of GBP were also noted to have large deviations in bond angles after refinement, (1 I” t,o 13”). It. should be noted that these deviations: and those noted above as being associated with the buried aspartatr residue. are the largest in the structures. and oc(*ur in regions for whic*h the electron

density is extremely clear, and the structures unambiguous. The density at Pro262 of RBP indicat,es that some additional deformation may be occurring here which was not allowed by t,he energy constraints applied in the refinement. These apparent regions of stain will he further investigated.

(i) Features

that are not conserved

or rise variable

A number of interesting features of RBP or GBP are not conserved in the other member of the pair. The metal-binding (presumably calcium) site found in both GBPs at residues 134 to 142 and 205 (Vyas et al., 1989; Mowbray et al., 1990) was not detected in the RBP-E structure, although the presence of metal at this or any other site was checked during the refinement using difference Fourier maps as had been useful for GBP-S. As the role of the metalbinding site is not clear in the GBP case, the significance (if any) of its absence from RBP is also not apparent. In RBP-E, the guanidinium group of -4rg141 of the ligand-binding site is almost parallel to the ring of Phe214 (at an angle of 17” and a distance of about 3.5 A). The interaction is based on an electrostatic attraction between the partial negative change near the aromat#ic ring (1,evitt 8r Perut’z, 1988) and the positively charged group of the arginine side-chain. This seems to be similar to cases described elsewhere in which aromatic rings act as hydrogen-bond acceptors (Burley & Petsko, 1986). It is, however. yuite distinct, as no proton is involved. This int.eraction. and that involvjng hydrogen honding to t,he side-cha.in of Asn 105. provide the correct orientation for the arginine in it,s role of sugar binding. Tn GBP. the hydrogen bonding of the side-chain of Arg158 to Tyr295 and Serll5 fulfill the same role. A l -~3hydrogen bond is found at the beginning of’ the first binding site loop of RBP-E, with a distance of 3.1 A from 13-N to 11-0, and an estimated strength of - 1.2 kcal. In GBP. t,his interact,ion does not occ’ur. as the equivalent proton (14-H) is involved in a p-turn with 11-G. The altered position of helix A in RBP makes the P-turn impossible. This intera(%ion is also associated with the hydrogen bond of 12-N t)o ThrlO-(X:1. As a result, of these interact,ions, the main chain of RBP has no h.vdrogen bond at 10-O. A buried water molecule (S36) is found in RBP with hydrogen bonds to the side-chains of Ser39, Glu46 and Asn64. This wat’er molecule is not, found in GBP. the interaction of the water and Glu46 being replaced by that of Gln45 and (Gln48. III GBI’. t’wo buried water molecules are found elsewhere. One (87) is located near residues 90 to 92. another (S14) is locat,ed in domain 2. hydrogen bonded to the side-chains of Oh1 19 and Asn210 and to the main c&hainat Arg158. Both of these were discussed above. Additional water molecules are trapped in the binding sites of both molecules when ligand was

Comparison


bound, rather than being truly buried. The positions of some but not all of these are conserved. The main-chain structure of RBP-E shows conformational variability in two places, as determined by disorder in the electron den&y maps. At. residue 113, the peptide is clearly seen to be distributed equally between two orientations in which the carbonyl oxygen atoms are 180” distinct from each other; the position of the other backbone atoms is essentially identical. The result is a mixture of type I and type II p-turns. As the third residue of the turn is Gly134, one would normally expect only the type II turn, as is in fact seen in this location in GBP (near residue Glyl51). This conformation is forced in GBP by the interaction with the side-chain of Glu156 with the main-chain amide nitrogen atom: at the expense of some small strain in the structure at 152-153. The difference may also be connected to the relative movement of helix F in RBP and GBP. The situation at t)he loop around RBP 207 is more complicated. The main-chain disorder here is sufficient t.o make the exact positions of residues 207-208 unclear. The structure of GBP-S shows primarily one of these possibilities (that with a type T turn wit,h an 0 to iv hydrogenbonding distance of 3.2 A) and much less disorder. The sequence of GBP is highly similar to that of RBP-E in this region, the primary difference being the replacement of Ser229 of GBP with Asp208 in R.BP. This may create more disorder in the lat.ter Jjrotein by pulling the loop closer to Lys206. In RBP, residues 27 and 50 were determined to be t*he sites of mutations that suppress defects in the leader sequence of t,he protein (Teschke et al., 1991). A link was established between the slowed folding of these mutant proteins and their enhanced transport compet,ence. The RBP-E structure shows that the locations of these mutations are correlat.ed with the presence of cavities in domain 1 of the wild-type protein. Mutation of the existing residue to one that’ is larger and more polar would, in each case, result in the cavity being filled, and some structural distortions. Based on this information, it was proposed that the formation of the first portion of this hydrophobic core was rate-limiting in the folding of RBP (Mowbray & Cole, 1992). The cavit.y near residue 50 of RBP is probably large enough to accept water, while GBP has no cavity near the equivalent residue 52, as determined from surface accessibility maps. The side-chain of GBP Leu63 instead moves to fill the hole left open by RBP Leu61. This leucine in turn is forced into it’s position in GKP by the altered conformation of the sidechain of Tle65 (compared to RBP Ile63); the effect originates with the replacement of t.he hydrogen bonding of Gln4.5 and Gln48 of GBP with Glu46 and a buried solvent (S36) in RBP. GBP does, however. have a fairly large cavity near residue 28, bordering residues 7. 64, 265, 266, 269; this is slightly differently placed than the cavit,y in RBP (a.bout) 2 A away in t)he aligned proteins). It would be predicted that suppressor mutations in GBP similar to those in RBP could occur at residues

Receptors

Va17 or Ala64 to make use of the cavity location.

433 in this

4. Discussion The X-ray structures show that the basic designs of RBP and GBP are very similar, as had been predicted earlier by sequence comparisons (Argos et al., 1981; Mowbray et al., 1990; Vyas et al., 1991). Each is a two-domain protein, thr domains consisting of a core of P-sheet decorated on both sides with cw-helices. The domains of each protein are related by an intramolecular axis of symmetry, wit.h RBP being the more symmetrical of the two. When the r-carbon atoms of the first domain of RBP were compared to those of the second domain, most were found to be located in equivalent places in the structure, while those of GBP agreed more poorly. A comparison of the overall structures of RBP and GBP indicates that the similarity goes further than t,he superficial level. The struct,ural agreement is in general much better between the backbone atoms in either domain of RBP and t,heir equivalents in the same domain of GBP (1.2 A) than between the two domains of either protein internally (1.8 A). This is consistent with the higher degree of sequence conservation between the struct’urally aligned regions of RBP and GBP (24’?&) than between the two domains of t,he same protein when t,hey are aligned (around 149,). The level of agreement observed in both cases is within the range of values found by Chothia B Lesk (1986) in their study relating sequence conservation and structural differences. RBP is smaller as a result of a number of deletions in its sequence relative to GBP. The deletions occur almost exclusively in the loop regions connect,ing units of secondary st,ructure. with the rxcept,ion of the removal of one helix in domain 1 and one at the C terminus in domain 2. RBP also has a larger relative number of ASX turns at the beginning of helices and left-handed r-helical turns at, t.he ends of heliaes; combined with the smaller loops, t)he result is a tighter structure with little extraneous material. The compactness of RBJ’ may explain the results of a search using environment.al strings, which det.ected similarity of its sequence and structure to the sequences of proteins of the lac repressor family (Bowie et al., 1991). but, a much lower degree of similarity of these prot’eins to GBP. The higher degree of intramolecular symmetry of RBP probably also results from this aspect of it’s struct,ure. A detailed comparison of the entire st.rurt,ures of RBP and GBP shows many common features. including it number of interactions in thr hydrophobic cores, a kinked helix, a buried aspartate residue, and several unusual turns. The larger 1ooJ)s of GBP seem to result in many of the conformational differences observed, and two are involved in the information of a calcium-binding site. ?r’o metal was found anywhere in t,he structure of RBT’, drspit,r explicit searches.

‘t’he residues of the sugar-binding site are drawn from loops at similar locations in the trio proteins. tnspe~~t~iott of the binding sites indicates that Lvhilr thta c*otntnott princ4plrs of hydrogen hottding. hytlroand steric* c~mst raint s il I? t)hobic* interac~tiotrs involvetl in both CYEW’S.the binding sit)es themselves ate substatttjially different. Although some binding site rf3idues are conserved. they seem to he present hecause of t)heir general ability io participate in hydrogen bonding or hydrophobic, itit~rrsctiotts. rather than due to any specific role itt sugar binding. sitrc*e they interac*t with different atoms of thtl respective sugSars. Sittcae ribose tnay be viewed as the (’ 3 epitner of gtuc~w (with the removal of a (‘H,OH group at (‘- 6). this almost tot,al redesign of t hr. binding site tnay at first seem somewhat of the structures surprising. A v Ioser inspection shows that the alternative epimer at (‘--3 cannot be ;rc,c.otnmotlat,ct1 in either protein without large c.ottformational changes. if the sugars are to be similarly oriented. The O-3 at’otrl of glucosr rvoultl c,ollide with Phe15 in RRP. and that of rihosch with TrI)lX3 in GHP. Since these non-polar residues provide criticaal h\-drophobic interac+ons with the sugar. the proteins are forced to solve the prohlrm in different ways. As a rrsult, ribose is rot ated iIt)OLlt 60 in the plane of the ring relative t,o galactosr. and sits slightly “flatt,er” in its binding site (ser Figs 7 and 10). A similar observation ma? be made caoncerning the mode of binding arabtnose to the iIrat)inose-l)indiIlg pro&in (xewcotner f4 (L1.. ISXln); it is distinct from both of the sugar-binding modes drsc*ribed here. This equivalent.-but -not-equivalent property of the binding sites should he noted 1)) those attempting to predict new st,ructures hased on c,xistitiy structural data. It mag not be possible to predic4 all asprc%s of the ttca struc+ures. and assumptions of st ricbt equivalence may lead to serious errors in the models resulting. 1)ata obtained by a nutnber of different, methods. including tluoreseence spect*roscopy (Boos et 01.. 1972: Miller pt ~1.. 1980). fluorescence rnergy transfer (Zukitt ut al.. 19770). measurements of Stokes radii (Zukin rt al.. 1979). small-angle X-ray scattering (Kewcomer ut al., 1981h) and nuclear magnetic resonance spectroscopy (Lucsk & Falke, 199ln,b), have suggested that’ bending motions in the hinge connecting the domains of GBP. RBP and related periplasmicb recept’ors arc linked with t’he binding of sugar. Solvent-accessibility calculations also indicate that ligand is almost c~otnpletely sequestered after binding to the receptors. arguing that, the proteins tnust be more open to accept or to discaharge t!he sugar. Theoretical studies have shown that a rigid angular rotation in the hinge of about’ 20” (as suggested by small-angle X-ray scattering experiments: Newcomer et al., 1981h) could take place without much strain (Mao et al.. 1982). X-rag struct’ural results show that two amino acid-binding proteins are able to assume a more open conformation (Sack et al.. 1989a.h). although no receptor structure has yet been described in bot’h the open and closed fortns.

The fact t.hat diEerertc,es are observed itt the rotations relating t)he two domains of K,HI’-E. (:UP-S and (:HP-I? thus has implications for the study of those ~onfortnational changes involved in the function of the protein. serving to highlight some regions of greatjest possiblr flexibility. R BP may be viewed as fu11y closed w&h Otrec direct hydrogen bonds forming across the (*left separatjiny t.he domains. The different orietttat~ion of the sugars in the binding site and other steric, c*ottstraitrts force (iKF’-E and (:IPP-S to be more open than It 131’ b>approximatrly 8” and 1W. respec4irely. The axis of rotation inrolrrd in Ihis domain opening is trt~;trl\ 1)erI)e~~dic~ular to that which relates the tlotttains through the intramolec~ular axis of sytnmet,ry. The difierencbes in relati\-tl orientation of’ the domains in the t,u’o (:Bf’s must havca irttother explanation. A detailed caomparisort of the st ru(‘tures c~ulti rela.te none of the tliffert~ric~es I)thtn-rett (:KP-S and GBP-E to the styuettc~t~ (‘tli~tl~~‘h het.wren tjhrtti. Thtb facat that these proteins (‘art replace ea(sh other it/ c>i~o (Miiller rt u/.. 19X:!) also argues that their solut,iott structures are assent iall! t’he same. The most probable sour(*e of the differ,t‘tt(‘e in ittterdomaitt angle then woukl setbitt t0 be the pa(akittg of t’hr protein tnolrc*ules into their rrspe(‘tive cryditl latticses. Large tlitYerenc*rs (u]) to 30” or more) in relatt,ive domain rotst.ions of bac+riophage TI tysozyme have heen attributed to c.rystal pacaking (Faber & Matthews. 1990). alt hottgh the open. unliganded prot)ein studied in that ww would be expected to be more flexible than Ott, c~losed. ligand-hound structures investigated iterr\. However. the differen~r in tarterg? involved is expec+tl to be small. even in the c:omplete opt~nitig of t>hr receptor as it ahangrs accompany the binding of sugar. ratlIter than a simple rigid movement, of the two domains. Alterations in the environments of resides on helices ?J and K, and possiblv others (but not, helix E). were noted in that case. in agreetnent with our results. Similar cha,rtgrs have noted in other proteins on

Comparison


Receptors

Figure 15. The a-carbon backbone structure of domain 1 of RBP-E and GBP-S, showing the mutation sites RBP Gly72 and GBP Gly74. RBP-E (bold line) and GBP-S (light line) were aligned to domain 1. A clear progression is seen from the excellent agreement at t)he multiple loop and residue 40 (at left) to the large differences arising from the deletion of helix 1) in RBP (at right). different exposure of the mutated residue in the two proteins. domain closure (Lesk & Chothia, 1984; Chothia & Lesk, 1985). Both of the liganded GBP structures presuma,bly represent forms found within the distribution of solution conformations. In bot)h RBP and GBP, most of the residues that provide binding-site interactions are “preformed”, that is they will be present in t’he correct conformation even in a more open receptor, even t’hough the binding site is found in separate halves. This is assisted by the fact’ that “common”, i.e. thermodynamically favorable, conformations are found for the residues of the binding sites. In each receptor, however, one binding site residue is located in the hinge (235 in RBP and 256 in GBP; see Figs I1 and 14). Small rotations in the hinge region could obviously have large effects on this residue’s relative position and availability for ligand binding. Its sidechain is. in fact, hydrogen bonded to groups in both domains. and so it is unclear to which domain this residue should be considered to belong. The fact that it is also located at the end of helix J. the hinge helix that showed the greatest relative movements in the three proteins, suggests that, this residue has a pivotal role in sugar binding and the movement within the hinge region. It, seems likely that’ sugar binds first to the preformed half-site of one domain, as was observed in the complex of the open form of t,he leucine-isoleucine-valine-binding protein with leucine (Saper & Quiocho, 1983: Sack et al.. 1989a). and that this hinge/binding-site residue is involved in bringing the second domain close enough (and in the correct orientation) for it to interact as well. The concept’ of symmetry comes in again in the study of the hinges which allow opening and closing of these proteins. Tn a macroscopic view. the most important elements of the hinges are seen to be located around the axis of intramolecular symmetry, as illustrated in Figure 1. The basic design involves

435

differences near the optimally according helix structure near The la,tt,er results in

three segments, each of which has elements from both domains. The first two are composed of a strand from domain 1 leading into helix E in the other domain, and the equivalent piece of structure originating in domain 2 (the strand leading into helix J). The third, consisting of the last strands of each domain, provides a bridge between t’he other two segments. The second half of the third segment provides an antiparallel, rather than parallel, interaction with the next strand in the sheet of domain 2, and so it is not precisely equivalent to the first half, to which it’ is, however, symmetrically and functionally very similar. Two water molecules are also located in symmetrical positions in the hinge, providing a continuous pattern of hydrogen bonds for the hinge segments as they cross from one domain to the other. These water molecules probably add a degree of flexibility that is useful in allowing bond rotations in the hinge region. They are located precisely a,t the positions that were noted to be involved in the relative domain movement, in the t,hree proteins studied here. A symmetrical pattern of close contacts (including l-3 hydrogen-bonding interactions) also exists in port,ions of the first and third hinge segment of each protein. Like the water molecules. these interactions may not be present in the fully open recept’ors, sincbe they are dictated by the type of p-into-r structure peculiar to the closed forms. In both RBP and GBP, some residual strain seems to be present in the hinge of the sugar-bound forms even after refinement, as suggested by the deviat,ions of some bond angles from more normal values. This residual sbrain might help keep the closed protein balanced at the point where it could easily open t,o discharge sugar, as might not otherwise be expected, based on the complex and favorable interactions formed with the sugar molecule in

the binding site. The bond rotations that accompany the opening of the protein would he expected to ease some of this strain. It, is also possible that relative sliding of the helices near t’he hinge on theil neighbors is a feature of the hinge-bending motion. as was in fact noted for helix ,J in t’he relative differences of RBP and the GBPs. Again, this would by consistent wit,h the nuclear magnet,ic resonance data (Luck $ Falke. 1991a) relating to a global conformational change in GBP on sugar binding. These “sliding springs“ could spread any strain from the hinge movements over a larger area, and so help dissipate it. Since RRP and GRP compete for a common membrane receptor in chemotaxis (Strange & Koshland. 1976), t’he produce of the try gene (Harayama et al., 1982), it, is obviously necessary to consider the structural results in this context. It was originally considered likely bot,h that’ GBP and RHP would bind to t’he same site on the Trg protein, and that the relevant Trg-binding sites on (:RP and RBP would be identical as well. Some genetic evidence exists (Park & Hazelbauer. 1986) that the former assumption was not. in fart,. quite correct, but) ra,ther that the binding proteins LIW similar. but not identical. sit’es on Trg. Biochemical data obtained with mutants (Hazelbauer & Adler. 197 1; Binnie et al., personal communication), as well as the struct~ural data presented here. make it clear that the second assumption was also not correct. Earlier alignments of the sequencesof these proteins had suggested some regions that were likely to be of interest in the context of chemotaxis. one of which included the region from GHT’ 599100 (Argos rt r&l.. 1981: Vyas et al.. 1991). A mutant of GBP-E had also been isolat,ed which was defect.ivcA in chemotaxis. but unimpaired in transport) (Hazelbauer 8 Adler, 1971; Ordal & Adler. 1974). This mutant was shown to involve a change of residue Gly74 of GBP to aspartic acid (Scholle et al.. 1987). The X-ray structure of this mutant GKP has been st’udied (Vyas et ~2.. 1988), and t)he changes shown to be confined to surface regions, as would be appropriate to the proposed Trg interaction site. The corresponding (fly72 to Asp72 mutant was therefore prepared in RBP (Binnie rf al., personal c*ommunication). The results of these hiochemieal studies showed t’hat the R’BP tnutant retained most of its c*hemotaxis functions, a,nd was somewhat defective in t’ransport. The availability of both structures makes this result quite easy to understand. It is possible t’o align domain 1 of RBP and GBP very accurately tjo inspect the nat’urr of this area in t’he two proteins (Fig. 15). This region is in fact one of those that is differently expressed within the tjwo structures due to t.he effect of neighboring regions (primarily the deletion of helix I> in RBP, as was shown earlier in Figs 6 and 8). Surface accessibility calculations show that Glyr72 of RBP has I ‘4’ surface area available, compared to 32 A* estimated for GBP. So t,he slightly impaired funct’ions in this RBP mutant c~mld be due more t’o distortions caused by the

introduction of the side-chain of aspart’ic acid at>this normally buried site than to any direct’ involvement in a protein-binding site in chemotaxis. Thus the interaction sites in RBP and GBP for Trg cannot he identical in character, although they may overlap; both could in fact use t’his particular portion of t)he surface. but they must do it in distinct ways. The structure within the individual domains can obviously be compared quite accurately. Roth domains, however, of RBP have been shown to be involved in the interaction with the respective proteins of chemotaxis and transport (R,. A. Binni. H. Zhang, S. I,. Mowbray &’ M. A. Hermodson. unpub lished results). Similar results have been obtained wit,h the related malt’ose and arahinosr receptors (Treptow & Shuman, 1988; Spurlino et nl., 1991; Hor, 1991). This makes the precise relationships bet’ween surfaces on the two domains particularly important. The surface of RRP-E is probably that, actually seen in solution. since the protein is full,y closed with several interdomain hydrogen bonds stabilizing the conformation. Because the domains of GBP-S and GBP-E differ somewhat in relative orientat.ion. the exact surface of GRP which would be available for proteirl-protein intItractions is not well definrd. Perhaps this “flexibility” of GBP helps explain why the response to liganded GBP is less than that> to liganded RBP (Hazelbauer & Adler. 1971): a lower concentration of GBP ma*v he found in t’he correct form to interact wit,h the mrmhrane receptor. Some equilibrium of alternate forms of th(b membrane receptor may also be involved. The residues known t)o be clonserved in the RBP and GBP sequenceswere shown in Figure 9. Ver! few have not already been fully explained by structural requirements and the functional features of the binding sites or hinge, although cllearly sornc’ residues may he (*onserved for bot)h structural and functional reasons. A pattern of (Gonsrrved surface residues is seen on the lips of the (Aleft, neat the intramolecular symmetry axis. Based on the mut,ant and structural rest&s described ahovcb. the region of conservation near residue Gly72 of RBP could be involved in both proteins. but in not)equivalent ways. This woulti in arly case serm t.o define

thr

far

edge

of

t’he

conserved

region.

These

residues definr an area approximat,ely 400 ‘I2 in size. Its location and extent. agrees very well wit,h t,he sites of most of t,he known chemotaxis mutants of RKP (Bitrnie et al., personal communication). and eoincidrs wit,h the single surface of the (:HI’s which is conspicuously free of amino acid suhst it ul ions (seeFig. 3). Very little is known about the structure of thr Trg prot,ein, hut, some implirationx can probably bc drawn from a comparison with t,he aspartatc receptor. which sequence comparisons havt shown to be homologous (Harayama rt al., 1982). The aspartatr receptor sends signals in response to aspartat.r as a result of its direct binding, and to maltose through its interaction with the liganded periplasmic maltose-binding protein. This protein has beert shown t,o be a dimer (Milligan 8.zKoshland.

Comparison


1988), but the available data suggest that both aspartate binding and signaling are carried out without, any co-operativity (Foster et al., 1985; Mowbray & Koshland, 1987). Evidence has been presented that signaling is carried out by monomer units independently (Milligan & Koshland, 1991). The recently reported crystal structure of the periplasmic ligand-binding domain of this receptor (Milburn et al., 1991) shows only a single aspartate bound per dimer, suggesting an alternate explanat,ion for the lack of co-operativity. The two monomer units are also sufficiently close for interaction of a single periplasmic protein with both membrane monomers simultaneously. This mechanism would explain the lack of co-operativity observed for the maltose response (Manson et al.. 1985). The proposed Trg-binding sites of RBP and GBP are raiher smaller- than the average size of the binding sites found protease-inhibitor and antibody-antigen complexes (1600 A2: Janin & (‘lothia, 1990) which is probably not surprising since those complexes are much tighter (Kd values of the order of lo-’ to 10-i’ M, rather than the 10e4 M expected here by analogy with the maltose-binding protein-aspartate receptor interaction; Manson et al.. 1985). Like the tighter-binding complexes. these surfaces of GBP and RBP are similar in character to the rest of the protein surface, with no apparent irregularities in the distribution of charged or apolar residues. This region of conserved residues has also been shown to be the site of the mutations of RBP that affect, transport (Binnie et al., personal communication); a similar region has been shown to be involved in the related ma.ltose and arabinose receptors (Treptow $ Shuman, 1988; Hor, 1991; D. G. Kehres & R. W. Hogg, unpublished results), as well. While the respective membrane transport systems are distinct, some features may be conserved, such as the utilization of similar surfaces of the periplasmic proteins. One remaining puzzle is the question of why it is the region surrounding the intramolecular symmetry axis that is the site of the chemotaxis and transport mutants. A related feature is the fact that these proteins are “closed” to the same degree, as defined by the nearly perfect alignment of the end strands of the two domains at the lips of the cleft in the same region; this seems to be the case in the closed structures of other binding proteins that are known (Gilliland & Quiocho, 1981; Pflugrath & Quiocho. 1985; Luecke & Quiocho, 1990; Spurlino et al., 1991). Possibly the main-chain amide and carbonyl groups of this region are used in binding the membrane proteins as well as the amino acid side-chains, although the main-chain involvement cannot be directly tested by mutant studies. The choice of this location is probably partly due to the fact that this surface would be that most drastically effect by the relative domain rotations of the type noted here, and so would be a good place for interaction sites that must be structurally distinct in the

Receptors

437

open and closed proteins. This region also has t’he advantage that it is the “watery” side of the cleft, i.e. the domains are already separated almost enough to allow passageof sugar out into the transport machinery. In each of the sugar-binding proteins, the sugar molecule lies quite flat in a binding site located roughly along the axis of intramolecular symmetry, despite an almost complete lack of other similarities in sugar orientation. Perhaps this assist’s the exit of t,he sugar molecule for t,ransport, much the same way that a coin must be correctly oriented to drop it through a slot. A minimal conformational change would thus be required to allow the depart,ure of t.he sugar molecule. perhaps while the binding proteins themselves remain in contact with the membrane proteins. The involvement of backbone atoms might be an advantage in the contact surface, as these would be more rigidly part of the structure than the side-chains, and could be better for linking any joint conformational changes of the periplasmic and membrane proteins. Effective sequestering of the sugar from the bulk solvent, during transport might result, as well. The rest of the answer may be found in the observation of Hor (1991) that the t’wo domains of maltose-binding protein make contact wit,h separate membrane proteins in transport (malF and malG). While it is not known if the single membrane component that seems to be involved in the ribose. galactose and arabinose transport systems is a present as a dimer. the cytoplasmic protein that is involved in energy coupling in each case is a highly conserved protein with internal symmetry. reminiscent of the binding proteins discussed here (Hogg et al.. 1991). This strongly suggeststhat this twofold relationship is an essential feature throughout these membrane transport systems. I would like to thank Maria Flocco and T. Alwyn Jones for several helpful discussions on methods of describing conformational changes, and Howard Shuman for communication of results prior to publication. This project was supported by a grant from the Swedish h’ational Research Counril (K-KU-9991300).

References Aksamit, R. & Koshland, D. E., .Jr (1972). binding protein of Salmonella typhimurium. Biophys.

Res.

Commun.

48,

A ribose Biochem.

1348-1353.

Aksamit. R. R. & Koshland. D. E., Jr (1974). Identification of the ribose binding protein as the in Salmonella receptor for ribose chemotaxis typhimurium. Biochemistry, 13, 4473-4478. Anraku, Y, (1968a). Transport of sugars and amino acids in bacteria. 1. Purification and specificity of the galactoseand leucine-binding proteins. ,J. Bid. (Them.

243,

3116-3122.

Anraku, Y. (196%). Transport of sugars and amino acids in bacteria. II. Properties of galactose- and leucinebinding proteins. J. Biol. Chem. 243, 3123-3128. Argos, P., Mahoney, W. C., Hermodson, M. A. & Hanei, M. (1981). St,ructural prediction of sugar-binding proteins functional in chemotaxis and t,ransport. J. Biol. Chem. 256, 435774361.

#.

438

L.

Baker, E. N. & Hubbard, R. E. (1984). Hydrogen bonding in globular proteins. Prog. Biophys. Mol. Biol. 44, 97-179. Henner-Luger, D. & Boos, W. (1988). The mglB sequence of Salmonella typhimurium LT2; promoter analysis by gene fusions and evidence of a divergently oriented gene coding for the mgl repressor. Mol. Cen. Genet. 214, 579-587. Bernstein. F. C., Koetzle, T. F., Williams, G. J. B., Meyer, E. T., Jr. Brice, M. D., Rodgers, J. R., Kennard, 0.. Shimanouchi, T. & Tasumi, M. (1977). The Protein Data Bank: a computer-based archival file for macromolecular structures. J. Mol. Biol. 112. 535~-542. Boos, W. (1969). The galactose binding protein and its relationhip to the /I-methylgalactoside permease from Escherichi coli. Eur. J. Biochem. 10, 66-73. Boos, W.. Gordon, A. S., Hall, R. E. & Price. H. 1). (I 972). Transport. properties of the galactose-binding of Escherichia co&: substrate-induced conformational change. J. Biol. Chem. 247, 917-924. Bowie, J. I:., Liithy. R. & Eisenberg, I). (1991). A method to identify protein sequences that fold into a known three-dimensional structure. Science, 253, 164-170. Briinger. A. T.. Kuriyan. J. &. Karplus, M. (1987). (:rystallographic R-factor refinement b.y molecular dynamics. Science. 235, 4588460. Buckenmeyer, G. K. & Hermodson, M. A. (1983). The amino acid sequence of n-ribose binding protein from Salmonella typhimurium STl. J. Biol. Chem. 258. 12957. Burley. 8. K. & Petsko, G. A. (1986). Amino-aromatic interactions in proteins. FEBS Letters, 203, 1399143. Chothia, C. & Lesk. A. M. (1985). Helix movements in proteins. Trends Biochem. Sci. 10. 116-118. Chothia. C. & Lesk. A. M. (1986). The relation between the divergence of sequence and structure in proteins. EMBO

J. 5, 823-826.

Chou. 1’. I’. & Fasman, G. D. (1977). B-Turns in prot.eins. J. Mol. Biol. 115, 135-175. Duplay, P. & Szmelcman. S. (1987). Silent and functional changes in the periplasmic maltose-binding protein of Escherichia coli K12. II. Chemotaxis towards maltose. J. Mol. Biol. 194, 675-678. Duplay. P., Szmelcman, S., Bedouelle, H. & Hofnung, M. (1987). Silent and functional changes in the perisplasmic maltose-binding protein of Escherichia cob K12. I. Transport of maltose. J. Mol. Biol. 194. 663-673. Faber. H. R. & Matthews, B. W. (1990). A mutant T4 lysozyme displays five different crystal conformations. Kature (London). 348, 263-268. Foster, D. L., Mowbray, S. L., Jap, B. K. & Koshland. I>. E.. . (1987). Sequence of the rngZB gene from Escherichia coli K-12: comparison of wild-t,ype and mutant. galactose chemoreceptors. 1Mol. Gen. Chemistry

study of protein.

4964-4970.

Mao, R.. Pear, M. R., McCammon. J. A. & Quiocho, F. A. Hinge-bending in L-arabinose-binding (1982). protein. J. Biol. Chem. 257, 1131-1133. Milburn. M. Prive, G. G., Milligan, D. L., Scott, W. G., Yeh. J., Jancarik, J., Koshland, D. E., ,Jr & Kim, S.-H. (1991). Three-dimensional structures of the ligand-binding domain of the bacterial aspartate receptor with and withbut a ligand. Science, 254, 1342-1347. Miller, D. M., Olson, J. S. & Quiocho, F. A. (1980). The mechanism of sugar binding to the periplasmic receptors for galactose chemotaxis and transport in Escherichia roli. J. Biol. Chem. 255, 2465-2471. Miller. D. M.. III. Olson. J. S., Pflugrath, ,J. W. & Quiocho, F. A. (1983). Rates of ligand binding to periplasmic proteins involved in bacterial transport and chemotaxis. J. Biol. Chem. 258, 13665-13672. Milligan, D. L. & Koshland, D. E.. Jr (1988). Site-directed cross-linking: establishing the dimeric structure of the aspartate receptor of bacterial chemotaxis. J. Biol.

Receptors

6268-6275.

Milligan, D. I,. & Koshland, D. E., Jr (1991). Intrasubunit signal transduction by the aspartate chemoreceptor. &%ence, 254. 1651-1654. Mowbray, S. I,. &Z Cole. L. B. (1992). The 1.7 A X-ray structure of t,he periplasmic ribose receptor of Escherichia coli. J. Mol. Biol. 225, 1555175. Mowbray. S. L. & Koshland, D. E., Jr (1987). Additive and independent responses in a single receptor: aspartate and maltose stimuli on the Tar protein. Cell, 50, 171 180. Mowbray. S. I,.. Smith. R. D. & Cole, L. B. (1990). Structure of the prriplasmic glucose/galactose receptor of Salwtonella typhimurium. Receptor. 1. 41-54. Miiller. N.. Heinr, H.-c:. & Boos, W. (1982). Cloning of nzglB. the structural gene for the galactose-binding protein of Salmonella typhimurium and Escherichia coli. Mol. Q’en. Genet. 185, 473-480. Nemethy. G. & Printz. M. P. (1972). The y-turn, a possible folded conformation of the polypeptide chain. Comparison with the /?-turn. Macromolecules. 5. 755-758.

Sewcomer, M. E.. Gilliland. G. L. & Quiocho, F. A. (1981a). L-arabinose-binding protein sugar complex at 2.4 A resolut’ion. J. Biol. Chem. 256, 13213-13217. Newcomer. M. E.. Lewis, B. A. & Quiocho, F. A. (1981b). The radius of gyration of L-arabinose-binding protein decreases upon binding of ligand. J. Biol. Chem. 256. 13218-13222. (:. &. Adler, J. (1973). Isolation and Ordal. complementation of mutants in galactose taxis and transport. J. Racteriol. 117. 5099516. Ordal, G. W. & Adler, J. (1974). Properties of mutants in galactose taxis and transport. J. Bacterial. 117. 5 17526. Park. (1. & Hazelbauer. G. L. (1986). Mutations speciiically affecting ligand interaction of the Trg ehemosensory transducer. J. Bacteriot. 167, 101-109. Pflugrath, *J. W. & Quiocho, F. A. (1985). Sulphate sequestered in the sulphate-binding protein of *Salmonella typhimurium is bound solely by hydrogen bonds. Nature (London), 314, 257-260. Pigman. W. & Amt. E. F. L. ,J. (1972). Mutarotations and

Genet.

208.

247-253.

Spurlino, J. C.. Lu, G.-Y. & Quiocho, F. A. (1991). The 2.3 A resolution structure of the maltoseor maltodextrin-binding protein, a primary receptor of bacterial active transport and chemotaxis .I. Biol. (‘hem. 266, 5202-5219. Strange. P. G. & Koshland, D. E., Jr (1976). Receptor interactions in a signaling system: competition between ribose receptor and galactose receptor in the cehemotaxis response. Proc. Nat. Arad. Sci., c!.S.A. 73, 762-766.

Teschke. C. M.. Kim. J., Song, T., Park, S.: Park, C. & Randall, I,. I,. (1991). Mutations that affect the folding of ribose-binding protein selected as suppressors of a defect in export in Escherichia coli. J. Riot Chem. 266. 11789-11796. Treptow, N. A. & Shuman, H. A. (1988). Allele-specific malE mutations that restore interactions between maltose-binding protein and the inner-membrane components of the maltose transport system. J. Mol. Riot

202,

809-822.

Vyas, M. N.. Jacobson, B. L. & Quiocho, F. A. (1989). The calcium-binding site in the galactose chemoreceptor protein: crystallographic and metal-binding studies. J. Biol.

Chem.

264,

20817-20821.

Vyas, N. K.. Vyas. M. N. & Quiocho, F. A. (1988). Sugar and signal-transducer binding sites of the Escherichia

coli galactose chemoreceptor protein. SciPncP. 242. 1290-1295. Vyas, iV. K., Spas. M. X. & Quiocho. F. A. (1991). (lomparison of the periplasmic receptors for L.arahinose, n-galactose/n-glucose, and o-rihose. J. Biol.

Chem.

266,

5226-5237.

Zukin. R. S.. Hartig, P. R. & Koshland, 1). E., tJr (1977~1). Use of a distant reporter group as evidence for a conformational change in a sensory receptor. I’rnc. :Va,t. Acad. Sci.. I:.S.A. 74, 1932-1936. Edited

Zukin. R. S., Strange. P. 0.. Heavey, L. R. & Koshland, I). E.. .Jr (19’77b). Properties of the galact,ose binding prot,ein of Salmonella typhimurium and E’ncherichia coli. Biochemistry, 16. 381-386. Zukin, R. S.. Hartig. P. R. & Koshland, D. E.. .Jr (1979). Effect on an induced conformational change on the physical properties of two chemotactic receptor molecules. Biochemistry. 18. 5599-5605.

by W. A. Hendrickson

Enemy attraction: bacterial agonists for leukocyte chemotaxis receptors.

Chamber for bacterial chemotaxis experiments.

Bacterial chemotaxis without gradient-sensing.

Statistical measures of bacterial motility and chemotaxis.

Stochastic gene expression: bacterial elites in chemotaxis.

Limits of feedback control in bacterial chemotaxis.

Universal response-adaptation relation in bacterial chemotaxis.

Change in membrane potential during bacterial chemotaxis.

Bacterial chemotaxis in a fixed attractant gradient.

A GPCR Handles Bacterial Sensing in Chemotaxis and Phagocytosis.

Identification of a methyl-accepting chemotaxis protein for the ribose and galactose chemoreceptors of Escherichia coli.

Coupled effects of chemotaxis and growth on traveling bacterial waves.

Control of tumbling in bacterial chemotaxis by divalent cation.

Methylation of a membrane protein involved in bacterial chemotaxis.

Bacterial chemotaxis to xenobiotic chemicals and naturally-occurring analogs.

Electron acceptor taxis and blue light effect on bacterial chemotaxis.

Adaptability of non-genetic diversity in bacterial chemotaxis.

Quantitation of the sensory response in bacterial chemotaxis.

The dynamics of protein phosphorylation in bacterial chemotaxis.

High Frequency and Diversity of Antimicrobial Activities Produced by Nasal Staphylococcus Strains against Bacterial Competitors.

Noise-Induced Increase of Sensitivity in Bacterial Chemotaxis.

Receptor interactions in a signalling system: competition between ribose receptor and galactose receptor in the chemotaxis response.

Direct imaging of intracellular signaling components that regulate bacterial chemotaxis.

Rest in underperforming elite competitors.