HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

J. Biol. Chem., Vol. 277, Issue 16, 14077-14084, April 19, 2002

This Article Abstract Full Text (PDF) All Versions of this Article: 277/16/14077    most recent M200514200v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Magnusson, U. Articles by Mowbray, S. L. Search for Related Content PubMed PubMed Citation Articles by Magnusson, U. Articles by Mowbray, S. L. Social Bookmarking                      What's this?

Hinge-bending Motion of D-Allose-binding Protein from Escherichia coli

THREE OPEN CONFORMATIONS^*

Ulrika Magnusson, Barnali Neel Chaudhuri^§, Junsang Ko^¶, Chankyu Park^¶, T. Alwyn Jones, and Sherry L. Mowbray^**

From the  Department of Cell and Molecular Biology, Uppsala University, BMC, Box 596, Uppsala SE 751 24, Sweden, ^¶ National Creative Research Initiative Center for Behavioral Genetics, Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Yusong-Ku, Taejon 305-701, Korea, and the  Department of Molecular Biology, Swedish Agricultural University, BMC, Box 590, Uppsala SE 751 24, Sweden

Received for publication, January 17, 2002, and in revised form, January 30, 2002

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Conformational changes of periplasmic binding proteins are essential for their function in chemotaxis and transport. The allose-binding protein from Escherichia coli is, like other receptors in its family, composed of two / domains joined by a three-stranded hinge. In the previously determined structure of the closed, ligand-bound form (Chaudhuri, B. N., Ko, J., Park, C., Jones, T. A., and Mowbray, S. L. (1999) J. Mol. Biol. 286, 1519-1531), the ligand-binding site is buried between the two domains. We report here the structures of three distinct open, ligand-free forms of this receptor, one solved at 3.1-Å resolution and two others at 1.7-Å resolution. Together, these allow a description of the conformational changes associated with ligand binding. A few large, coupled torsional changes in the hinge strands are sufficient to generate the overall bending motion, with only minor disruption of the individual domains. Integral water molecules appear to act as structural "ball bearings" in this process. The conformational changes of the related ribose-binding protein follow a distinct pattern. The observed differences between the two proteins can be interpreted in the context of changes in sequence and in crystal packing and provide new insights into the nature of hinge bending motion in this class of periplasmic binding proteins.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Conformational changes play a vital role in the biological function of many proteins. The wide spectrum of conformational changes observed in crystal structures can be broadly classified as small amplitude shear motion, large amplitude hinge bending motion, or some combination of the two (1, 2). Shear motion generally represents the sliding movement of secondary structural elements on other parts of the tertiary structure, whereas hinge-bending motion is characterized by a few localized torsional rotations that combine to produce dramatic changes in the protein as a whole.

A classic example of the involvement of hinge-bending motion in protein function is found in the periplasmic receptors of the bacterial ABC transporter systems. Such systems use the energy of ATP to carry small ligands and ions across the cytoplasmic membranes of both prokaryotes and eukaryotes (3-5). A typical ABC system consists of an membrane-bound permease, an ATP-binding component, and, in most bacterial systems, a periplasmic receptor. Binding of a small molecule ligand to the periplasmic proteins favors their closure via large scale hinge-bending motions (6). These movements are required for productive interactions with the cognate membrane permeases and, in some cases, with membrane-bound chemotaxis receptors as well. The periplasmic sugar-binding proteins belong to a subfamily (pentose/hexose sugar receptors) of the larger family of periplasmic receptors (7). Crystal structures of several members of this subfamily have been reported in the closed, ligand-bound form, including allose-binding protein (ALBP¹ (8)), ribose-binding protein (RBP (9)), arabinose-binding protein (ABP (10)), and glucose-galactose-binding protein (GBP (11, 12)). Each consists of two similar Rossmann fold domains linked by a three-stranded hinge region. The binding site is located at the domain interface; extensive hydrogen bonding and hydrophobic interactions of the ligand with both domains of the protein stabilize the closed form. Although the periplasmic receptors can assume similar closed forms in the ligand-free state (e.g. Ref. 13), experimental data suggest that more open forms will predominate in the absence of ligand (6, 14-16).

The structures of three ligand-free forms of RBP provided the first picture of the conformational changes in this subfamily of receptors (17). The two domains of each open RBP were shown to move as nearly rigid bodies at the hinge that joins them; the observed structures were opened by 43, 53, and 64° with respect to the closed receptor. Most structural changes involved only a few torsional changes in the hinge segments, although some minor repacking was observed where domain-domain interactions were lost in the opened receptors. Further, the three structures represented discrete points along a conformational trajectory, thus describing the motion that should apply to ligand capture as well as ligand release into the permease. In each open form, the two domains had a similar set of packing interactions that were not present in the closed form, and that would be expected to stabilize them in preference to other possible open conformations. Here, we describe three ligand-free, open forms of a second member of the subfamily, ALBP, solved at 1.7- and 3.1-Å resolution. These structures, together with that of the closed, ligand-bound form, supply specific information on the conformational changes of ALBP. As RBP and ALBP are the most closely related members of this family (amino acid sequence identity 35%), similarities and differences could readily be analyzed and extended to give a better understanding of the motions of the family as a whole.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Crystallization, X-ray Data Collection, and Processing-- Purified ALBP (18), concentrated to 5 mg/ml, was shipped frozen from Korea to Sweden and stored at 20 °C. Crystals were obtained by the hanging drop vapor diffusion method (19). Showers of small cubic crystals grew within 2-3 days in drops containing 2 µl of protein solution (5 mg/ml in 10 mM HEPES buffer, pH 7.8) and 2 µl of reservoir solution, equilibrated against 20-30% monomethyl polyethylene glycol 2000, 0.1 M Tris-HCl buffer (pH 9.0), and 0.01 M NiCl₂ at room temperature. These crystals grew slightly afterward, to a final size of ~50 µm³. Orthorhombic crystals (100 × 100 × 50 µm³) grew after several days using 30% monomethyl polyethylene glycol 4000, 0.1 M Tris-HCl, and 5 mM ZnSO₄. The presence of divalent cations substantially improved both types of crystals.

Crystals were soaked in mother liquor containing 15% glycerol for 2 min prior to freezing in liquid nitrogen. Diffraction data were collected from a single cubic crystal to 3.1-Å resolution using the synchrotron radiation source BW7B ( = 0.89 Å), at the EMBL outstation (DESY, Hamburg). Data to 1.7-Å resolution were collected from an orthorhombic crystal at ID14-EH4 at ESRF, Grenoble ( = 0.94 Å). Data were processed using DENZO (20), SCALA (21), and SCALEPACK (20); statistics are reported in Table I.

Structure of Cubic Form-- Initial attempts to solve the structure of the cubic form of ligand-free ALBP at 3.1-Å resolution, using either separate domains of the closed form of RBP (Ref. 9; Protein Data Bank code 2DRI) or the intact open forms of RBP (Ref. 17; Protein Data Bank code 1URP) as input models, were unsuccessful. When the ligand-bound, closed structure of ALBP had been solved and refined at 1.8-Å resolution (8), its domains were used as input models in AMoRe (22). Observed structure factors were sharpened by a B-factor of 10 Ų. An all atom model of domain 1 (residues 1-113 and 247-282) and a truncated, polyserine model of domain 2 (residues 114-246) provided solutions in rotation searches using data between 6- and 3.5-Å resolution. The fifth highest peak in the rotation search for domain 1 yielded a clear solution in the translation search, with a correlation coefficient of 0.38 and an R-factor of 45%. Phased translation function searches were then employed for domain 2, using data between 10 and 3.5 Å. The top solution, with a correlation coefficient of 0.56 and an R-factor of 39%, was the 38th peak of the original rotation search. After rigid body refinement of both domains, the correlation coefficient and R-factor were 0.63 and 36%, respectively. Clear electron density was observed at this stage (Fig. 1a), even for atoms missing from the initial model (e.g. residues 283-288 and most side chain atoms). This, together with the correct reassembly of the two domains into a contiguous protein, provided evidence that the replacement solutions were correct.

The initial 3.1-Å cubic model was subjected to several cycles of positional and B-factor refinement in CNS (23) and REFMAC (24), alternated with manual rebuilding in the program O (25). Where side chain conformation needed to be adjusted, the best fitting rotamer from the data base in O (26, 27) was used in the electron density. 10% of the data were used to monitor R-free (28). Harmonic restraints were applied to all atoms throughout the course of refinement, and small-molecule-derived parameters were applied in all geometrical restraints (29). Positional refinement with subsequent grouped B-factor refinement (one B-factor for each domain) was performed initially. The average B-factor for domain 2 was high. During the later stages of CNS refinement, B-factors were refined using three groups. Residues 139-147 and 168-193, which are the least ordered (although clearly present) part of the model, constituted one group. The rest of domain 2 and the complete domain 1 comprised the other two groups. Residues 174-176 were eliminated from the refinement; these residues were visible but ill defined. The final B-factor of domain 1 in the CNS refinement was 35.7 Å², that of the disordered region of domain 2 was 91.6 Å², and that for the rest of domain 2 was 63.5 Å². Superior results were obtained utilizing TLS refinement in REFMAC5 (30) prior to isotropic B-factor refinement. After this treatment, the average B-factors were reduced to 11.6 Å² for domain 1 and 12.2 Å² for domain 2, and the electron density for domain 2 was greatly improved. Final electron density in the hinge is shown in Fig. 1b. Electron density for residues 168-194 and 210-218 is still poorer than that for the rest of the molecule, although the main features are still quite clear. Residues 210-218 also show somewhat weaker electron density than the average, but again, their position is clear.

Thirteen water molecules were modeled in the electron density, four of them in the hinge region. While we would not normally place water molecules at this resolution, both electron density and local protein structure argued strongly for their presence here. Six of these were also present in the high resolution structure of the closed ligand-bound form. At the N terminus, there was strong positive electron density, suggesting the presence of two or more metal ions, although their position on or near the 3-fold axis made their exact placement difficult. Two Ni²⁺ ions were eventually modeled, with occupancies of 0.33 each, and obtained B-factors of 51.7 and 52.9 Å² in the refinement. At the same time, the R-factor dropped from 24.3 to 21.8, and R-free dropped from 28.8 to 27.2. We cannot show that this is the correct chemical model, but it appears to be more a accurate description than simply omitting the ions from the model and calculations.

Statistics relating to the final model are summarized in Table II. The Ramachandran plot had nine outliers. Two of these, residues Asp⁹¹ and Asp²²⁷, are outliers in the entire subfamily of periplasmic sugar-binding proteins. The rest were close to the allowed regions of a stringent boundary Ramachandran plot (31). Only nine side chains have rotamer side chain fit values of >1.5 Å, consistent with a well behaved structure (32, 33). Five of these are almost identical to the conformations found in the closed form, and others differ only slightly. Most residues in both domains have real space correlation coefficients higher than 0.7 (Ref. 34; computed using CNS). Coordinates and structure factors have been deposited with the Protein Data Bank (35, 36) with entry code 1GUB.

Structure of the Orthorhombic Form-- AMoRe was used to locate the two molecules in the asymmetric unit of the orthorhombic crystal, with data from 8- to 4-Å resolution. Initial attempts to solve the structure using the open form described above failed. An all atom model of domain 2 from the closed form of ALBP was then used for the rotation search. The highest peak gave a correlation coefficient of 0.19 and an R-factor of 53% when entered in the translation search. This solution, combined with the second best solution for domain 1 in the rotation search, yielded a correlation coefficient of 0.32 and an R-factor of 48% after rigid body fitting. These two solutions, which represented a contiguous protein, were then joined together to search for the second molecule. The original solution appeared as the top peak in rotation/translation searches. This molecule was fixed, and the second peak of the rotation search gave a good translation solution. A correlation coefficient of 0.44 and an R-factor of 44% were obtained after rigid body fitting, which allowed the two domains of each molecule to move independently.

Five cycles of simulated annealing and individual B-factor refinement of the initial 1.7-Å model, alternated with manual rebuilding, were carried out in CNS. A random set of 4% of the data was used in the R-free calculation. The final refinement of the structure was carried out in REFMAC5. Water molecules were added using wARP (37). Eight clear zinc ions were modeled into the electron density. These do not seem to have any functional role and probably serve to stabilize the crystal packing. Statistics relating to the final model are summarized in Table II. Final electron density in the hinge is shown in Fig. 1c. As for the 3.1-Å structure, Asp⁹¹ and Asp²²⁷ were outliers in the Ramachandran plot; the others were different in the various structures. Coordinates and structure factors have been deposited with the Protein Data Bank (35, 36) with entry code 1GUD.

Other Methods-- Coordinate sets used for the comparisons of the new ALBP structures were as follows: 2DRI (closed ligand-bound RBP structure (9)), 1URP and 1BA2 (open ligand-free RBP structures (17)), 1RPJ (closed ligand-bound ALBP structure (8)), 1GCA (closed ligand-bound GBP structure (12)) and 1ABE (closed ligand-bound ABP structure (10)), all available from the Protein Data Bank. Structures were compared using the programs O (34) and LSQMAN (38, 39). Rotations, translations, and axes describing domain movements were calculated using the program FIT.² Buried surface area was calculated with the algorithm of Lee and Richards (41), using a 1.4-Å probe. Figures were prepared with the programs O, Molray (42), and Molscript (43).

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Overall Structures-- The structures of three open, ligand-free forms of ALBP were obtained using molecular replacement, one refined at 3.1-Å resolution in a cubic space group, and two others found in the same asymmetric unit of an orthorhombic space group and refined to 1.7-Å resolution. Statistics describing the final structures are presented in Table II, and samples of electron density are shown in Fig. 1. Like the previously solved structure of closed, ligand-bound ALBP (8), those of the ligand-free protein feature two / (Rossmann fold) domains, each of which is composed of two distinct segments of polypeptide chain (Fig. 2, a and b). Strands B1-B5 and B11, together with -helices H1-H4 and H10, are found in domain 1 (residues 1-112 and 247-282). Domain 2 includes -helices H5-H9 and -strands B6-B10 and B12 (residues 113-246 and 283-288). Three connections result between the two domains (Fig. 3), which are designated here as connection I (residues 111-113), connection II (246-247), and connection III (281-284). The first two are - crossovers, which are more buried in the structure, while the third is a - crossover that lies diagonally across them.

View larger version (29K): [in this window] [in a new window]   Fig. 1.   Electron density maps. a, electron density (A-weighted map (40), contoured at 1), calculated using a polyserine model (molecular replacement solution, omitting residues 283-288 of domain 2) and the 3.1-Å cubic data, is shown in the region of connection III. The final 3.1-Å model in this region in shown for comparison. As expected, electron density was weaker for residues not included in the model (residues 283-288), although it was clearly present. b, the same region in the final refined 3.1-Å A-weighted map. c, electron density for the same part of the 1.7-Å orthorhombic structure for the O3 molecule.

View larger version (31K): [in this window] [in a new window]   Fig. 2.   ALBP conformational changes. Ribbon diagrams of closed, ligand-bound ALBP (a) and the open, ligand-free O3 form (b), as defined under "Results." The two structures are aligned to show the same view of domain 1 (at the bottom). Segments of O3 backbone structure that change by 0.7 Å or more on opening are shaded (residues 11-17, 38-44, 90-98, 130-132, 140-148, 163-164, 171-177, and 228-231).

View larger version (19K): [in this window] [in a new window]   Fig. 3.   Schematic diagram of the connectivity in the three-stranded hinge of ALBP. Two key water molecules (W1 and W2) are shown as circles, and their hydrogen-bonding interactions are shown as broken lines. The position of a third water (W3) that is only observed in ligand-bound ALBP is also indicated

Hinge Motion Opens Ligand-free ALBPs-- The structures of domain I and domain II are similar in all ligand-bound and ligand-free forms of ALBP. The ligand-free proteins are more open, as a result of largely rigid body rotations of the domains, using their three connections as a hinge. The 3.1-Å structure is opened by 37° compared with the closed form, while the two 1.7-Å forms are opened by 43 and 33° in the A and B molecules of the Protein Data Bank file (1GUD), respectively. For convenience in the following discussion, we will refer to the four conformational forms of ALBP as closed, O1 (33°), O2 (37°), and O3 (43°). The three very similar rotation axes that apply to the open forms of ALBP are shown in Figs. 4 and 5. The motions of O2 and O3 are essentially pure rotations about similar axes that pass close to two hinge residues: 113 (from connection I) and 246 (from connection II). The axis associated with O1 is parallel to the others but shifted by ~2.5 Å toward residue 246; this motion also includes an ~1-Å screw component along the rotation axis.

View larger version (16K): [in this window] [in a new window]   Fig. 4.   Rotations at the hinge. a, hinge regions in the O3 open (atomic colors) and closed (blue) forms of ALBP were superimposed using the C atoms of domain 1 (at the bottom). Hydrogen bonds are indicated as bubbled lines in green. Despite an overall 43° opening about the axis shown, the hydrogen bonding network is the same in both forms of ALBP. The apparent rotation axis appropriate to this form of ALBP is shown as a solid line. b, equivalent regions of closed (blue) and open (atomic colors) RBP are shown, aligned with the same view as for domain 1 of ALBP in a. During the 43° opening motion of RBP, a main-chain/main-chain hydrogen bond is gained, while W2 (indicated by blue bubbled lines for the closed form) is lost. The apparent rotation axis appropriate to this form of RBP is shown as a solid line.

View larger version (37K): [in this window] [in a new window]   Fig. 5.   Crystal packing for the O2 structure. One molecule is shown in dark gray, and the four symmetry-related molecules in the crystal packing that are closest to domain 2 are colored pale gray. The three observed axes of ALBP opening are indicated (marked O1, O2, and O3). The axes of libration for the two domains (indicated by L1 and L2, respectively), as suggested from TLS refinement, are also shown, with lengths proportional to the magnitude of the apparent motion.

Opening of ALBP is the result of coupled rotations around main chain dihedrals in the three interdomain connections. Movements in the three hinge segments are largest at residues 111112, 245247, 283284 (Table III). The movements in connections I and II determine the direction of rotation, while those in connection III seem to be simply those required to accommodate changes in the other two segments. This is perhaps not surprising when the relatively exposed location of connection III is considered; it will not have the same constraints within the tertiary structure as the other two connections. Although it is clear that that the changes in  or  increase with the degree of opening, the relationship is not a linear one (the bonds that most closely follow a linear pattern are 112, 245, and 246). This serves as a reminder that the different parts of the protein must work together; conformational changes depend on a number of bonds rotating in a cooperative fashion.

View this table: [in this window] [in a new window]   Table III Changes in backbone torsion angles in the hinge connections of ALBP and RBP Each open form is identified by a label (e.g. O1), as well as the relative degree of opening in parentheses. The equivalent residues of the two proteins are aligned for each of the three connections. The  and  angles of ALBP model O2 are less dependable because of the lower resolution, but they show the similar patterns to O1 and O3. For RBP, the four wild-type structures are very similar, with / variations of only ~2 °; the values given here are those for the A molecule of the Protein Data Bank file (1URP).

Two water molecules are embedded in the hinge of all four forms of ALBP. One of these (W1 in Figs. 3 and 4a) hydrogen-bonds to the hydroxyl group of Thr¹¹² (from connection I) as well as to Asn²⁴⁸-N (near connection II) and Asp²⁸²-O (from connection III); it therefore links all three hinge segments near their borders with domain 1. The second water (W2) interacts with Asp¹¹³-N (from connection I) and Val²⁸¹-O (from connection III) while coordinating with another water molecule (W3) in the closed form. W3 also hydrogen-bonds to Ser²⁸³-O1 (from connection III) and Asp¹¹³-O (from connection I) in the closed protein but is lost from the open forms, as Ser²⁸³ moves further away. Both of the water molecules that remain lie on the domain 1 side of the axis and move with this domain; domain 2 actually pivots around W2 during opening.

Due to the nature of the packing and a high solvent content in the cubic crystal form (62%), motion of domain 2 of the O2 structure appears to be less constrained than for the O1 and O3 forms in the orthorhombic asymmetric unit. There was more disorder in those portions of the original electron density maps, and large improvements in both electron density maps and R-factors were obtained during TLS refinement of the cubic structure. Libration of the domains around the axes in Fig. 5 is suggested to occur. Libration for domain 2 is much larger than that for domain 1 and is centered around an axis perpendicular to the opening axes, reflecting motion within a large empty space in the crystal packing.³

Structure within the Domains-- Structural alignments of the individual domains of the various forms of ALBP are summarized in Table IV. Approximately 80% of the backbone atoms of each domain remain quite rigid during the conformational changes, with the -sheets providing the main sources of structural integrity. In addition to the hydrogen bonds and hydrophobic interactions in and around these sheets, the protein has other interactions that help maintain intradomain structure near the hinge during opening. Asp¹¹³ and Asn²⁴⁸, which cap the N termini of helices H5 and H10, respectively, are located symmetrically on either side of the hinge and retain their respective interactions in all structures. The side chain of Lys¹²⁰ moves appreciably to preserve a hydrogen bond with the main-chain oxygen atom of Ser²⁸³ as it moves during opening (2.9 Å in O1 and O3, 3.4 Å in O2, with good electron density in all cases). A hydrogen bond between Thr¹¹¹-OG1 and Asp⁹¹-O (2.8 Å) is maintained in all forms and reinforces the structure of domain 1 at that point. In domain 2, a series of hydrogen-bonding interactions linking Tyr¹⁹⁸-OH, Thr²²⁶-OG1, Gly²²⁵-O, and Thr²⁴⁵-N appear to have a similar effect. Only a few interactions are lost in the ALBP hinge on opening, including some involving a residue that also participates in ligand binding. In the allose-bound form, Gln²⁴⁷ forms hydrogen bonds with Asp⁹¹ and Thr¹¹² from domain 1, as well as with Arg¹⁵¹, Asp²²⁷, and Asn¹¹⁴ (indirectly via water) from domain 2. In the open structures, the interactions of Gln²⁴⁷ with domain 2 are lost, while those with domain 1 (including the water-mediated interaction) remain. The hydroxyl group of Thr¹¹¹ makes a hydrogen bond to Thr¹¹²-O in the closed and O2 structures (3.1-3.2 Å), which is not, however, seen in O1 or O3 (distance 3.6-3.9 Å).

View this table: [in this window] [in a new window]   Table IV Comparison of equivalent domains of ALBP All C atoms for each domain were initially superimposed; the r.m.s. difference using the two segments of each is reported in Å. The value in parentheses is the percentage of Cs that match when the alignment is improved using a 0.7-Å cut-off. The top-right half of the table gives the values for domain 1, while the bottom-left half (boldface type) refers to domain 2.

In the closed, liganded form, the sugar is completely buried at the domain interface in a site that suits it in both size and character. Indeed, most of the interactions that stabilize the closed form represent contacts between the protein and ligand. In the open, ligand-free ALBPs, the cleft is opened up to the solvent, so that the ligand can enter or exit freely. Some of the changes within the domains reflect relaxation of loops near the binding site when interactions to ligand or protein that are present in the closed form are lost (Fig. 2b).

In domain 1 of the closed form, two aromatic residues (Phe¹⁵ and Trp¹⁶), along with hydrogen-bonding residues Lys⁹, Asn¹³, Glu⁴², Asp⁹¹, and Gln²⁴⁷, form one-half of the ligand-binding site. The loop bearing residues 14-16 relaxes away from the binding site in all open forms. The greatest difference in the domain structure in the open forms is found in the B2-H2 loop of O1 and O3 (residues 39-45), where the entire loop moves further away from the cleft, and Glu⁴² points away from the binding site. The differences of the O1 and O3 structures could be partly an artifact of crystal packing, since the two molecules in the crystallographic asymmetric unit make contact at this point. However, these loops take on nearly identical conformations, implying that this is, in fact, a favorable structure. Changes at residues 91 and 247 are very small.

Trp¹⁷⁵, Ser¹⁴⁷, Arg¹⁵¹, Asn²⁰¹, and Asp²²⁷ of domain 2 form the other half of the ligand-binding site of ALBP. Changes in domain 2 on opening are generally smaller than those in domain 1 (Table IV). The tryptophan residue (from the B7-H7 loop) is stacked on top of the peptide plane of Gly¹⁴¹ (from the B6-H6 loop), as well as interacting with the nonpolar hydrogens of bound allose. The side chain ² angle of Trp¹⁷⁵ in the O2 structure seems to be rotated almost 180° with respect to the other structures, and the main-chain  values of 174 and 175 also appear to be relaxed from positive values; however, the electron density in this loop of O2 is somewhat weak. Few other significant changes are apparent in this domain; the remaining ligand-binding residues of domain 2 have good electron density consistent with conformations found in the ligand-bound form, conformations that are apparently preserved by supporting networks of hydrogen bonds.

Only a few specific inter-domain contacts exist in the closed form, localized at the lips of the binding cleft. In the closed form of ALBP, 68-O and the side chain of Glu⁹² of domain 1 make interdomain contacts with the B6-H6 loop (residues 145-147) from domain 2. This buried surface is completely exposed in the open forms, with the loss of five protein-protein interdomain hydrogen bonds. In the O2 form, Glu⁹² is reoriented slightly while maintaining a salt-link with Lys⁹ in the same domain; this interaction, however, is not seen in O1 and O3. The loop containing Asn¹⁴⁵ in domain 2 is altered in all open forms and becomes less ordered. The neighboring B7-H7 loop, which loses contacts with both the B6-H6 loop and the sugar, is disordered to an even greater extent, especially in O2, which appears to have a structure very different from the rest. In addition, domain motion causes the loss of a weak interaction between the ring of Phe⁴⁵ in domain 1 and Lys¹⁴² of domain 2.

Comparison with RBP and Related Binding Proteins-- Structures of RBP in both closed, ligand-bound and open, ligand-free forms allowed its conformational changes to be explored previously (17). The three ligand-free structures of RBP were calculated to be opened by 43, 50, and 64°; these forms are referred to as rO1, rO2, and rO3 here. (There are actually four copies of rO1 in the asymmetric unit of that crystal form, but they differ by 2°.) In RBP, the opening also reflects nearly pure hinge motions, which at first glance appears startlingly similar to that of ALBP (Fig. 6). However, a closer look shows that the direction of opening differs by 30-40° from that seen with ALBP (Fig. 4). The axis of rotation in RBP passes close to residues 103 and 234, which are equivalent to 112 and 246 of ALBP. The changes in main-chain torsion angles that open RBP are compared with those observed for ALBP in Table III.

View larger version (40K): [in this window] [in a new window]   Fig. 6.   RBP conformational changes. Ribbon diagrams of closed, ligand-bound RBP (a) and the open, ligand-free wild-type RBP (b) (43° of opening), as defined under "Results." The views are the same as those used for ALBP in Fig. 2, with respect to domain 1.

Starting within domain 1 and proceeding into connection I, it can be seen that the bond rotations begin earlier in RBP than they do in ALBP. The reason for this appears to lie in the substitution of Ala¹⁰² in RBP with Thr¹¹¹ in ALBP. The hydrogen bond of the latter's side chain with Asp⁹¹-O seems to stabilize the structure at that point and force the rotation to begin at 111 instead of 111. This segment of ALBP then rotates around water W2 instead of forcing it out, as had been seen in RBP, where a new hydrogen bond results between 104-N and 263-O in all open structures. The other hinge water, W1, remains in place in open forms of both ALBP and RBP.

In connection II, it seems that changes of Phe¹⁸⁷ to Tyr¹⁹⁸ and Phe²¹⁴ to Thr²²⁶ and the new hydrogen bonds associated with these differences make nearby parts of domain 2 slightly more rigid in ALBP, compared with RBP. Changes in main-chain torsion angles thus begin later in ALBP, at 245. In both proteins, the changes in connection III seem to reflect the attempts of the entire strand (273-288 in ALBP) to adjust to movements in the other two connections; differences at bonds equivalent to ALBP's 281, 283, and 284 are most prominent in both proteins.

Changes within the domains of RBP on opening are much smaller than for ALBP.

ALBP and RBP represent the only members of this family of periplasmic binding proteins for which both open and closed structures are known. The closed forms of GBP (44, 45) and ABP (46) are, however, available. In closed GBP, both W1 and W2 are present, but Thr¹¹¹ of ALBP is replaced by Gly¹⁰⁹. In closed ABP, W1 is present, but W2 is already replaced by a direct hydrogen bond between 109-N and 283-O, and Thr¹¹¹ is replaced by Met¹⁰⁷. Such differences will surely affect the open conformations observed. The only conserved residue in the hinge of the four proteins is a valine equivalent to Val²⁸¹ of ALBP. It appears that this side chain's relatively small size allows it to fit in the structure at this point in both open and closed forms; its hydrophobic side chain places few demands on local tertiary structure, so it may provide a convenient bit of grease at this point of the hinge. Nearby Val¹¹⁰ is always valine or isoleucine, which may reflect a similar story.

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

ALBP was chosen for a detailed study of conformational changes because of its 35% sequence identity to RBP, the only other member of the subfamily for which ligand-associated conformational changes have been explored (17). It was thought that this relationship is close enough for similarities and differences in their behavior to be assessed with confidence but distant enough to provide an interesting comparison. The closed, ligand-bound proteins have an r.m.s. difference of 1.1 Å for the Cs of the individual domains (8). The opening motions of ALBP represent nearly pure hinge movements of its two domains with respect to each other. At first glance, its opening behavior is very similar to that of RBP (compare Figs. 2 and 4). The major torsional changes that open both proteins are restricted to a small hinge and result in an overall ~40° movement of the two domains as rather rigid bodies with respect to each other. We were surprised, however, to find that ALBP's hinge "bends" at different residues from RBP's and that the sense of the opening motion differs by ~40°. The same pattern is observed for all three open forms of ALBP and as it had been for all three open forms of RBP, so these patterns are protein-specific. We conclude that the open structures observed in our crystal structures are related to each other by low energy transitions within a restricted population that is characteristic of the protein involved. Particular conformers will be trapped in the various crystal packing environments.

It has previously been demonstrated that a related binding protein (GBP) has a great deal of flexibility in solution, with essentially any open conformation being attainable with some measurable frequency (47). Most of these are probably extremely uncommon, however, given that the construction of a hinge from three strands must place special constraints on the allowed motions. A rotation in any one of the connections must be accompanied by compatible changes in the other two, to minimize the disturbance of bonded and nonbonded interactions both locally and within nearby tertiary structure. The motions dictate both that the hinges be designed for flexibility and that the individual domains be designed for structural integrity. Although hydrogen-bonding interactions can be provided by solvent, the exposure of nonpolar side chains will disfavor any uncompensated changes that greatly disrupt the structure of the two domains. These factors should effectively limit the conformational space available for multistranded hinge bending, and that is indeed what we observe. This situation is quite different from the case of a single-stranded hinge like that found in T4 lysozyme, where almost any conformation appears possible (48).

The constraints on the design of an acceptable hinge are so great that this binding protein family has evolved special ways of dealing with the problem, one of which involves the creative use of water. In the closed ligand-bound forms of ABP (8), RBP (9), GBP (11, 12), and ABP (10), main-chain peptide groups in the hinge make hydrogen bonds with both protein groups and water molecules. Hinge waters are found in two similar locations in the closed forms; only ABP lacks the one that we have designated W2 (Fig. 3), having a direct hydrogen bonding between the two hinge segments instead. In all three known open forms of RBP (and in multiple observations of the 43° form), one of the waters (that equivalent to W2) is displaced and supplanted by a direct main chain/main chain hydrogen bond (17). The remaining water (W1) apparently acts as a molecular "ball bearing" in the hinge that can be rotated on quite freely during the domain rotations. This role is much more difficult to fulfill using protein side chains, since they are by nature more rigid. In the open forms of ALBP studied here, both of the hinge water molecules of the closed structure remain in place and adjust their interactions somewhat as the -strands of the hinge realign themselves. ALBP's opening motion thus proceeds along a path different from, although related to, that used by RBP (Fig. 4). It is also clear that residues near the hinge, such as Thr¹¹¹ of ALBP, can modulate the motion, either by stabilization of the relevant domain or by physically blocking certain routes. Differences in local sequence and variability in hinge waters lead us to suggest that the opening of GBP and ABP will represent additional variations on the themes we have so far observed.

Binding proteins of this family are also similar to two repressors, LacI and PurR (49, 50). However, the repressors function as dimers, so have additional, and different, constraints, and very different conformational changes are observed (51). No water is seen in the hinge regions of LacI and PurR, and the movements of the proteins when opening and closing are not limited to their hinges. The repressors open just enough to allow entry of ligand, each maintaining its dimer contacts, and translating its motions into proper placement of the attached DNA headpieces. Our overall conclusion from such studies is that we are not yet in the position where we can predict conformational changes with any accuracy, given only a single form of a protein. We can, however, confirm the conventional wisdom that main-chain segments unrestricted by local tertiary structure will be prime candidates for moving parts.

RBP generates new interdomain contacts near the hinge of its open forms, which bury ~200 Ų of its surface (17). A much larger surface has been observed for the open form of the (unrelated) maltose-binding protein (15). Such contacts could provide slight stabilization of the open forms and so help minimize the interference of ligand-free forms in transport and chemotaxis. However, we do not see a significant buried surface in the open ALBP structures reported here. Although the local protein structure of both domains is similar to that of RBP, very little surface is buried in the open forms, mostly at Ile²⁸⁴. In fact, we see no obvious means of stabilizing these particular open forms, beyond that of crystal packing. The energy of a crystal contact should compete favorably with low energy transitions that are occurring in solution. The situation may be fundamentally different for the maltose-binding protein, where the ligand-free form has a significant affinity for the membrane components of transport (52, 53). In general, any open form that would allow the free entry and exit of the ligand would be acceptable for the purposes of function. In ALBP, although not in RBP, there are additional conformational changes within the domains of the open forms that fall within regions believed to be involved in transport; these should make recognition of the open binding protein by the transport components even more difficult.

Ligand-free binding protein almost certainly opens and closes continuously in solution. When a sugar molecule binds to one domain of an open receptor, it will be trapped when the protein closes. The ligand must not be dislodged on closing, and the complete binding site must be assembled correctly from the two halves. (Some of the differences in the paths of motion observed for these proteins could actually have an origin in such phenomena; different ligands will place different demands.) The ligand-bound form of the binding protein will be correctly recognized by the cognate permease, and the bound sugar can be then be transferred through the membrane. A restricted path of reopening of the binding protein would have real physical significance, because the motions of the receptor and permease must be coupled during ligand transfer within the complex. Structural work with such complexes, where one partner is a membrane protein, will be extremely difficult. Studies of the individual periplasmic receptors, such as we report here, represent the only presently available way of gaining information about the functionally important motions that are an essential part of the transmembrane transport process.

	ACKNOWLEDGEMENTS

We thank Changhoon Kim and Gerard Kleywegt for helpful discussions, Alex Cameron for data collection and other assistance, and Maria Boström for efforts to produce better crystals for the open form of ALBP.

	FOOTNOTES

^* This work was supported by grants from the Swedish Natural Science Research Council (to S. L. M. and T. A. J.) and by the Creative Research Initiative Program (to C. P.).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

^§ Present address: UCLA-DOE Laboratory of Structural Biology and Molecular Medicine, Box 951570, University of California, Los Angeles, CA 90095-1570.

^** To whom correspondence should be addressed. E-mail: mowbray@xray.bmc.uu.se.

Published, JBC Papers in Press, February 1, 2002, DOI 10.1074/jbc.M200514200

² Program available on the World Wide Web at bioinfo1.mbfys.lu.se/~guoguang/fit.html.

³ Additional images showing conformational changes and differences between the open forms can be found on the World Wide Web at xray.bmc.uu.se/~mowbray.

	ABBREVIATIONS

The abbreviations used are: ALBP, allose-binding protein; ABP, arabinose-binding protein; GBP, glucose-galactose-binding protein; RBP, ribose-binding protein; r.m.s., root mean square.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

CiteULike

Complore

Connotea

Del.icio.us

Digg