or maltodextrin- binding protein, a primary receptor of bacterial

THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1991 by The American Society for Biochemistry and Molecular Biology, Inc.

VOl. 266, No. 8, Issue of March 15, pp. 5202-5219, 1991 Printed in U. S. A .

The 2.3-A Resolution Structure of the Maltose- or Maltodextrin- binding Protein, A Primary Receptor of Bacterial Active Transport and Chemotaxis*

(Received for publication, September 25, 1990)

John C. SpurlinoSglI, Guang-Ying LuSI), and Florante A. QuiochoS From the $Howard Hughes Medical Institute and the Department of Biochemistry and Structural Biology, Baylor College of Medicine, Houston, Texas 77030 and the §Department of Biochemistry, Rice University, Houston, Texas 77251

The three-dimensional structure of the maltose- or maltodextrin-binding protein (Mr = 40,622) with bound maltose bas been obtained by crystallographic analysis at 2.8-A resolution. nThe structure, which has been partially refined at 2.3 A, i! ellipsoidal with overall dimensions of 30 X 40 X 66 A and divided into two distinct globular domains by a deep groove. Although each domain is built from two peptide segments from the amino- and carboxyl-terminal halves, both domains exhibit similar supersecondary structure, consisting of a central @-pleated sheet flanked on both sides with two or three perallel a-helices. The groove, yhich has a depth of 18 A and a base of about 9 X 18 A, contains the maltodextrin-binding site. We have previously observed the same general features in the well-refined structures of six other periplasmic receptors with specificities for L-arabinose, D-gdaCtOSe/D- glucose, sulfate, phosphate, leucine/isoleucine/valine, and leucine.

The bound maltose is buried in the groove and almost completely inaccessible to the bulk solvent. The groove is heavily populated by polar and aromatic groups many of which are involved in extensive hydrogen- bonding and van der Waals interactions with the maltose. All the disaccharide hydroxyl groups, which form a peripheral polar surface approximately in the plane of the sugar rings, are tied in a total of 11 direct hydrogen bonds with six charged side chains, one Trp side chain, and one peptide backbone NH, and five indirect hydrogen bonds via water molecules. The maltose is wedged between four aromatic side chains. The resulting stacking of these aromatic residues on the faces of the glucosyl units provides a majority of the van der Waals contacts in the complex. The nonreducing glucosyl unit of the maltose is involved in approximately twice as many hydrogen bonds and van der Waals contacts as the glucosyl unit at the reducing end. The binding protein-maltose complex shows the best example of the extensive use of polar and aromatic residues in binding oligosaccharides.

The tertiary structure of the maltodextrin-binding protein, along with the results of genetic studies by a number of investigators, has also enabled us for the first time to map the different regions on the surface

of Health and the Welch Foundation. The costs of publication of this * This work was supported by grants from the National Institutes

article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “uduertisement” in accord- ance with 18 U.S.C. Section 1734 solely to indicate this fact.

7 To whom reprint requests should be addressed Howard Hughes Medical Institute, One Baylor Plaza, Houson, TX 77030.

I( Present address: Dept. of Biology, Beijing University, Beijing, China.

of the protein involved in the interactions with the membrane-bound protein components necessary for transport of and chemotaxis toward maltodextrins. These sites permit distinction of the “open cleft” (without bound sugar) and closed (with bound sugar) conformations of the binding protein by the chemotactic signal transducer with which the maltodextrin-binding protein interacts. While a similar mechanism of molecular recognition is operational for the transport system, the membrane transport components interact with sites on the binding protein that are distinct from the ones for chemotaxis.

In recent years the interests of our laboratory have centered on the tertiary structure and function of seven periplasmic binding proteins. Herein we describe the three-dimensional structures of the maltose- or maltodextrin-binding proteins (MBP)’ and its complexes with maltodextrins as determined by x-ray crystallography. We have previously determined and extensively refined the structures of the L-arabinose-binding proteins (ABP), the sulfate-binding protein (SBP), the phosphate-binding protein, the D-galactose/D-glucose-binding protein (GGBP), the leucine/isoleucine/valine-binding protein (LIVBP), and the leucine-specific binding proteins (LBP) (Quiocho and Vyas, 1984; Pflugrath and Quiocho 1985, 1988; Luecke and Quiocho, 1990; Vyas et al., 1987,1988; Sack et al., 1989a, 1989b; Quiocho, 1990).

These seven proteins represent about a third of the entire family of binding proteins found in the periplasmic space of Gram-negative bacteria. Although all binding proteins serve as primary high affinity receptors for the osmotic shock- sensitive active transport systems (for a recent review, Fur- long, 1987), only four (with specificities for D-ribose, D-galaC- tose/D-glucose, maltodextrins, and oligopeptides), also act as initial receptors for chemotaxis (Macnab, 1987). A recent report by Gilson et al. (1988) indicates the presence of similar proteins in a Gram-positive bacterium and a mycoplasma as an extracellular component anchored by means of an amino- terminal lipo-amino acid to the cytoplasmic membrane.

Binding proteins are monomeric with molecular masses ranging from 23,500 to 52,000 daltons, with most around 33,000. It is noteworthy that the two largest binding proteins have oligomeric ligands. MBP, the second largest, has a

The abbreviations used are: MBP, maltodextrin-binding protein; ABP, L-arabinose-binding protein; GGBP, D-galaCtOSe/D-glUCOSe- binding protein; LIVBP, leucine/isoleucine/valine-binding protein; LBP, leucine-binding protein; SBP, sulfate-binding protein; MIR, multiple isomorphous replacement; 6-iodomaltose, a-D-Glc-(1-*4)-6- iodo-6-deoxy-~-Glc; MES, 4-morpholineethanesulfonic acid.

5202

2.3-A Structure of the Periplasmic Maltodextrin Receptor 5203

relative molecular mass of 40,622 based on the amino acid sequence of 370 amino acids (Duplay et al., 1984). The oligo- peptide-binding protein is the largest ( M , = 52,000) and has recently been crystallized (Tolley et al., 1988). Binding proteins as a whole lack significant sequence similarity. An extraordinary feature is that while the binding proteins have a diverse set of ligands (monosaccharides, oligosaccharides, oxyanions, amino acids, oligopeptides, and vitamins), they bind their respective ligand with similar high affinities, Kd values around 5.0 X M (Miller et al., 1983; Furlong, 1987). Interestingly, the Kd values are very similar to the in uiuo K,,, values for active transport. Thus, the binding protein-dependent transport systems are very efficient, able to maintain an internal pool of lo4 to IO5 higher than the external pool.

Of the four sugar-binding proteins, the maltodextrin-binding protein has the most versatile sugar specificity. It can bind linear maltodextrins of two to at least seven ~ ~ ( 1 4 ) - linked glucosyl units and cyclic maltodextrins such as cyclo- maltohexaose and cyclomaltoheptaose. This versatile specificity is the reason that we prefer to use the name maltodextrin- binding protein rather than the original name maltose-binding protein. On the other hand, the other three sugar-binding proteins bind only monopyranosides; ABP binds L-arabinose, D-galactose, and D-fucose, GGBP binds D-galactose and D- glucose, and D-ribose-binding protein binds D-ribose (Miller et al., 1983; Furlong, 1987). Surprisingly, MBP does not bind D-glucose. The equilibrium dissociation constants ( Kd values of the linear and cyclic maltodextrins vary from 1.6 to 40.0 X

M (Miller et al., 1983; Quiocho, 1989),’ with maltotriose being the best ligand, binding about 20-fold more tightly than maltose. Kinetics of saccharide binding, studied by rapid- mixing stopped-flow technique, indicate a fast, second order process (Miller et al., 1983).’ Remarkably, despite the differences in the size and nature (linear and cyclic) of the maltodextrin ligands, the association rate constants fall within a very narrow range of 1-3 X IO7 M” s-’. Variations in the Kd values of the different MBP-maltodextrin complexes can be attributed primarily to the dissociation rate constants, which vary from 1 to 100 s-l (Miller et al., 1983).’ Furthermore, MBP, as well as ABP and GGBP, bind both the a- and p- anomers of the sugar ligands (Miller et al., 1983).

MBP serves as the initial high affinity receptor, but actual maltose transport across the cytoplasmic membrane further requires a membrane-bound hetero-complex consisting of the following: MalF (Mr = 57,000), a highly hydrophobic protein (Froshauer and Beckwith, 1984)); MalG (Mr = 32,200), an integral membrane protein (Dassa and Hofnung, 1985); and MalK (Mr = 40,700) (Gilson et al., 1982), a somewhat more hydrophilic protein which is considered to be peripherally bound at the inner surface of the cytoplasmic membrane (Shuman and Silhavy, 1981). MalK is a strong candidate for being involved in generating the energy required for active transport (Higgins et al., 1985). Two consensus regions for ATP binding which have been found in the MalK gene code (Higgins et al., 1986) are further indicative of the involvement of MalK. This sequence similarity also includes the P-glyco- protein associated with multidrug resistance in tumors and the product of the cystic fibrosis gene (Hyde et al., 1990). The organization of transport components is similar in the other of the binding protein-dependent active transport systems (Ames, 1986; Furlong, 1987). The entire maltose transport system is one of the best characterized, thus offering an excellent opportunity to understand the function and mech- anisms involved in specific high-affinity transport.

Deletion of binding proteins by genetic manipulation leads

* J. C. Spurlino and F. A. Quiocho, unpublished observations.

to abolition of the respective active transport (Ames, 1986; Furlong, 1987). Although pseudorevertants in the maltose system translocate maltose with only three membrane proteins and without a periplasmic binding protein, they do so very inefficiently since the K , of active transport is (-2 mM) which is several orders of magnitude greater than that of the wild-type system (0.001 mM) (Shuman, 1982). The mutant strains can achieve only about a 10-fold concentration of compounds, several orders of magnitude less than that achieved in the wild-type systems. The absence of binding protein almost completely abolishes the physiological role of these high affinity transport systems, which is to utilize nutrients present in the medium in low concentrations. Thus, contrary to the belief of others that binding proteins are only accessories (Hyde et al., 1990), binding proteins do play crucial roles in active transport and, hence, they should be considered as a ”supercharger” instead.

MBP is also the primary receptor for maltose taxis (Hazel- bauer, 1975). The Tar (Taxis to Aspartate and away from some Repellents) protein is the chemotactic signal transducer protein which forms a complex with MBP in order to initiate chemotaxis. The Tar protein is located in the cytoplasmic membrane as a homo-dimer (Falke et al., 1987).

We have previously proposed that the membrane-bound proteins recognize and bind to the liganded form of the binding proteins in preference to the unliganded one. This is a stringent requirement for chemotaxis to occur, since, if both forms were recognized equally, no differentiation between the presence and absence of ligand would be possible. The fact that there is an order of magnitude more binding protein than membrane-bound hetero-oligomer favors this preferential recognition for transport interactions as well. An implicit feature of this model is that the binding proteins undergo a ligand- induced conformational change, distinguishing the liganded form from the unliganded form. There is a large body of evidence from numerous sources (see “Discussion”) that sup- ports the existence of two conformations of binding proteins. This two-state model of a binding protein and the tight binding of ligand has been incorporated in a proposed mechanism of active transport (Quiocho, 1990).

While the structural analysis of the liganded form of the arabinose-binding protein and the galactose/glucose-binding protein has clearly revealed fundamental aspects of protein- sugar interactions, determining the MBP structure enables us to examine not only the mode of binding but also the conformations of both linear and cyclic oligosaccharide ligands, many of which are uncrystallizable alone. Structural determinations also lay the cornerstone for an understanding of the many genetic studies of the maltose-binding protein from several laboratories. Some of these studies have led to the identification of sites along the polypeptide chain of MBP which are involved in interactions with the cytoplasmic protein components of transport and chemotaxis. Others have identified regions which are dispensable and have no effect on binding, transport, or chemotaxis.

EXPERIMENTAL PROCEDURES

Materials-The D-maltose-binding protein used in these studies was isolated from the LA 3400 strain of Escherichia coli provided by Dr. Winfried Boos, University of Konstanz, Federal Republic of Germany. Maltose and maltotriose were obtained from Pfanstiehl Laboratories (Illinois). Cyclic maltodextrins were obtained from Sigma and all other linear maltodextrins (tetraose to heptaose) from Boehringer Mannheim. Dr. K. Omichi of Osaka University, College of Science, Osaka, Japan, kindly provided us with the following iodinated maltodextrins: a-D-Glc-(14)-6-iodo-6-deoxy-a-~-Glc (6- iodomaltose), 6-iodo-6-deoxy-a-D-Glc-(l4)-a-~-Glc, and a-6-iodo- 6-deoxy-~-G~c-(14)-a-D-G1cO(l-+4)-~-~-Glc. Polyethylene glycol

5204 2.3-k Structure of the Periplasmic Maltodextrin Receptor (8000) was obtained from Fluka AG. All other reagents and chemicals were of analytical reagent grade.

Protein Purification and Crystallization-The D-maltose-binding protein was isolated from the LA 3400 strain of E. coli grown on minimal media with maltose as the sole carbon source. The maltose- binding protein, with the rest of the periplasmic proteins, was released from the periplasm by the osmotic shock procedure of Neu and Heppel (1965). The shockate was filtered through 0.45-pm filters and then loaded onto a 6 X 30-cm DEAE-52 column equilibrated in 10 mM Tris-HC1, 0.02% sodium azide, pH 7.3. The column was eluted with a step gradient of 0.15, 0.30, 0.5 M KC1 in 10 mM Tris-HC1, 0.02% sodium azide, pH 7.3. The fractions containing MBP, determined using native acrylamide gels, were pooled and dialyzed against 10 mM Tris-HC1, 0.02% sodium azide, pH 7.2. Further purification was achieved by affinity chromatography using a column of cross-linked amylose and eluting bound MBP with 10 mM maltose (Ferenci and Klotz, 1978). Final purification was achieved using high performance liquid chromatography anion-exchange chromatography (Synchrom AX-300 column) and eluting with a 0.0-0.3 M KC1 gradient. Although the final purification step only removed a minute amount of contam- inant, the improvement in crystallizability was significant. The purified MBP was dialyzed against 10 mM sodium citrate, 0.02% sodium azide, pH 6.2, and concentrated to approximately 25 mg/ml in prep- aration for crystallization.

Crystals of MBP were obtained via the hanging-drop method. Twenty-five to 3O-pl drops (3.5 mg/ml MBP, 1 mM maltose, 16% (w/ v) polyethylene glycol 8000, 0.02% sodium azide, 10 mM sodium citrate, pH 6.2) suspended over 19% polyethylene glycol at 4°C produced small MBP crystals within 2 days. Small well-formed crystals could be grown larger when reseeded into drops containing 15% polyethylene glycol as the precipitating agent with all other constit- uents remaining the same. Crystals which grew as large as 1.5 mm, were harvested into a stabilizing solution of 25% polyethylene glycol 8000, 0.02% sodium azide, 1 mM maltose, 10 mM sodium citrate, pH 6.2.

Heavy Atom and Sugar Ligand Derivatives-Heavy atom and ligand replacement crystals were prepared by conventional soaking methods. The chosen compound was dissolved in 25% polyethylene glycol, 10 mM citrate, pH 6.2, at concentrations between 1 and 10 mM. The heavy metal solution was stirred overnight and the pH was checked and readjusted if necessary. As sodium azide can cause precipitation of heavy atom compounds, it was not included in the heavy atom solutions. It cases where the compound still formed a precipitate, MES was substituted for the citrate in the soak solution. Thirty-six different heavy metal compounds in a total of 50 different soak condition were tried during the derivative search.

In order to locate the sugar-binding site by difference Fourier technique, MBP crystals were soaked in solutions containing saccharide ligands and ligand analogs. The following compounds were tried a-~-Glc-(1~4)-6-iodo-6-deoxy-a-D-Glc, 6-iodo-6-deoxy-a-D-Glc- ( l -A)-a-~-Glc, a-6-iodo-6-deoxy-D-Glc-(14)-a-D-Glc-(1+4)-a-D- Glc, maltotriose, cyclic maltohexaose, and cyclic maltoheptaose.

Methods for Measuring, Processing, and Merging of Diffraction Data and for Heavy Atom Refinement and MIR Phasing-Intensity reflections were measured on a four-circle Syntex P21 diffractometer with CuKa nickel-filtered radiation from a sealed normal-focus x-ray tube operated at 50 kV and 35 mA. In the course of our structural studies of periplasmic binding proteins, which all diffract to better than 2 8, resolution, we have adapted a fairly standard set of procedures for obtaining high quality diffraction data, as well as for heavy atom refinement and MIR phasing. All these procedures have been described in detail in several papers, notably those by Gilliland and Quiocho (1981), Saper and Quiocho (1983), and Pflugrath and Quiocho (1988).

Electron Density Maps, Map Improvement, and Modeling-Elec- tron density maps derived solely by using MIR phases were calculated with “best” Fourier coefficients (figure-of-merit weight centroid phases) (Blow and Crick, 1959). Preliminary “mini-maps” at 4.1 A resolution were contoured on plastic sheets and were stacked for boundary determination and initial assignment of a-carbon coordinates. The C, trace from the mini-maps was used to obtain initial coordinates and all further fitting was done on a PS300 graphics unit.

Many methods exist to improve the quality of electron density maps utilizing an existing partial structure (Podjarny et al., 1987). A series of maps were used to fit the MBP structure, with each succes- sive map utilizing the previously fitted structure. As the completeness and quality of the structure improved, the contribution from MIR phasing was lessened. A combination of methods proved to be SUC-

cessful in improving phases for the MBP structure by density modi- fication. New phases were calculated from a map modified by impos- ing real space constrainst (positivity, connectivity, and constancy of the solvent regions) directly on the density3 and then combined with 3 A MIR phases to produce an improved map. An envelope was defined at a given radius from fitted atoms with regions outside the envelope truncated to solvent level. This envelope was used to obtain the phases which were combined with the original MIR phases to obtain a new map. The radius employed was varied from 5 8, at the beginning, when relatively few atoms had been fitted, to 2 8, at the final stages when most of the atoms had been placed. This procedure was iterated until convergence, signaled by lack of improvement in phases. The resulting maps were interpreted on the graphics display.

Subsequent maps were calculated usjng combined phases (aeomb)

from partial models and MIR/SIR (3 A/2.8 A) phases. In the final rounds of fitting the sequence, we used maps calculated with (2(Fot,al - IFcalcl) and (IFob.l - lFca~cl) coefficients and a,.lc phases obtained

2.3 8, resolutions. from initial rounds of restrained least squares refinement at 2.8 and

The published sequence of MBP (Duplay et al., 1984) was used to correlate the three-dimensional structure with the linear structure and assign residue numbers. Regions where several large, bulky groups occurred in close proximity in the sequence were used to align the sequence with the residues which had been fit to the electron density maps. The structure of ABP (Quiocho and Vyas, 1984) was also employed to explore structural similarities, as suggested by the two homologous stretches reported for ABP, GGBP, ribose-binding protein, and MBP (Duplay et al., 1984).

Computer Programs-The crystallographic program package PRO- TEIN (Steigemann, 1985 version) was used for data manipulation. All modeling was accomplished with the crystallographic fitting program PSFRODO (Pflugrath et al., 1984). ALB, a secondary structure prediction program (Finkelstein, 1975; Ptitsyn and Finkelstein, 1983), was used to predict the secondary structure of MBP. Restrained least squares refinement was done with PROLSQ (Hendrickson and Kon- nert, 1980) using the version with fast Fourier provided by B.C. Finzel, Upjohn Co.

Structural similarity of proteins was assessed by superpositioning their a-carbon structure. The stringency of the comparison was increased by using a progression rule to quality matches. OVRLAP, a program written by W. S. Bennett, MPI Moleculare Genetik, Berlin (using the method of Rossmann and Argos, 1977), was used to make this comparison. The solvent accessibility was determined by the method of Lee and Richards (1971), using the program ACCESS written by M. Handschumacher and F. M. Richards, Yale University.

RESULTS

Protein Purification, Crystallization, Diffraction Intensi- ties-In the course of our structural and biochemical studies, three preparations of purified MBP (-0.5 g each) were made. All migrated as a single band on polyacrylamide gel electro- phoresis, in the presence and absence of sodium dodecyl sulfate and exhibited Kd values of about 3 X M for maltose binding as measured by fluorescence titration technique (Miller et al., 1983) .

MBP crystals have the shape of a truncated diamond shield, resembling Superman’s emblem, with average dimensions of 1.0 X 1.0 X 0.4 mm. They belong to the space group C2 with unit cell dimensions of 105.8 X 68.6 X 57.8 8, (Table I). The asymmetric unit contains one protein molecule of 40,622 daltons. As expected from the use of maltose in the purification and crystallization steps, these crystals were proven to contain maltose (see “Sugar-binding site”). The form of these crystals is different from the one first reported by Quiocho et al. (1979), which had columnar shape and P21 space group. The original crystals were obtained from protein unexposed to any maltodextrin dyring purification or crystallization.

For the native 2.3-A diffraction data set (Table I), greater than 95% of the theoretical number of reflections (16,517 out of 16,991) were collected from 10 crystals. For all the individ-

N. K. Vyas, unpublished program after the methods of Shevitz et al., 1981 and Lunin et al., 1985.


TABLE I Maltose-binding protein crystal diffraction data statistics Derivative soak

Crystal No. of conditions No. of Unique crystals measurements reflections

Observed R,,,,." Resolution

Conc. Time A4 days % A

Native 7 42,106 16,517 95 5.6 2.3 5.4 2.8

Pb 2 1 12 9,036 7,873 98 7.3 3.0

4.2 3.2 6-Iodo-maltose 2 10 22 5,951 4,810 96 4.2 3.5 Maltotriose 1 10 14 2,982 2,584 96 3.6 4.1

Hg 3 1 13 10,671 9,183 97

Pt 1 1 8 2,713 2,429 90 5.5 4.1 DY 1 3 4 7 7,999 6,810 89

R,.,, = 100.- Z J I - I;[ where I , is observed intensity of a reflection and I is average intensity of the I , values.

Z I

TABLE I1 M I R and heavy atom phasing statistics at 3.0 A resolution

Definitions as per Blundell and Johnson (1976). rms, root mean square.

Resolution (A) 11.71 8.28 6.40 5.22 4.40 3.81 3.36 3.00 Total No. of reflections 95 243 423 672 891 1,229 1,673 2,103 7,329 Mean figure of merit ( m ) 0.84 0.83 0.82 0.80 0.81 0.62 0.46 0.34 0.57

6-Iodomaltose rms F H 89.76 87.65 86.80 84.55 82.35 80.93 0.00 0.00 84.02 rms residual 65.11 60.68 49.95 50.11 65.29 70.81 0.00 0.00 60.53 RcUai. 0.65 0.48 0.51 0.67 0.66 0.55 0.00 0.00 0.59

Pb rms F H 231.21 21.2.71 176.15 149.62 115.93 86.50 63.78 45.47 102.08 rrns residual 133.95 106.05 78.85 77.15 78.51 71.36 57.78 42.96 66.25 RCullie 0.77 0.60 0.63 0.57 0.72 0.77 0.83 0.74 0.69

Hg rms F H 221.07 180.58 152.89 119.03 89.07 61.89 42.55 26.40 82.37 rms residual 100.78 90.50 74.60 70.04 67.72 63.05 53.31 46.26 60.81 RCullis 0.24 0.45 0.37 0.57 0.66 0.66 0.82 0.88 0.56

Pt rms F H 344.19 283.82 242.48 219.95 223.51 233.44 0.00 0.00 240.55 rms residual 153.95 111.39 94.01 88.74 82.91 85.98 0.00 0.00 94.43 R c ~ I I ~ ~ 0.51 0.56 0.62 0.67 0.49 0.27 0.00 0.00 0.57

DYI rms F H 297.15 212.38 233.13 210.64 197.45 130.50 98.67 100.09 190.10 rms residual 124.72 106.78 98.85 95.67 101.11 88.34 63.89 57.23 98.27 RculIia 0.48 0.50 0.65 0.55 0.60 0.63 0.67 0.56 0.60

ual native crystal data R,,, (based on intensity) varied from 2.69 to 4.96. The agreement of a single crystal data set with the merged data set, R,,,,, varied from 6.12 to 8.92. The overall Rmerge is 5.6. Table I provides further evidence of the quality of the diffraction data for the native and heavy atom- derivative crystals and of the isomorphism of the derivatives.

Structure Determination-MIR analysis was conducted in two stages initially at 4.1-A resolution and then finally at 3.0- A resolution. The phases for the 4.1-A studies were derived from three heavy atom derivatives, a three-site Hg derivative from a sodium mersalyl soaking experiment, a two-site Pb derivative from a Pb(NO& soak, a a three-site Pt derivative from a K2PtC14 soak. The major Hg site (site one), fixed at y = 0.25, was used to correlate the relative y coordinates of the other heavy atom sites. The mean figure of merit (( m ) ) of the 4.1-A MIR analysis was 0.63.

The 4.1-A MIR map, plotted at a scale of 0.12 cm/A onto plastic sheets, allowed clear delineation of the molecular boundary. It was also possible to obtain a preliminary backbone trace by marking the approximate positions of 243 a- carbon atoms. Since these positions were still short of the complete 370-residue sequence, the trace contained gaps. The

a-carbon coordinates were measured from these positions, transferred to a PSFRODO file for display on the graphics system and converte4 to a polyalanine model.

Following the 4.1 A MIR analysis, two additional derivatives were obtained a single-site 6-iodomaltose derivative and a seven-site DyI derivative from a dysprosium iodide soak. Instead of the single Hg site, the 6-iodomaltose site was used to fix the y coordinate of the other derivatives. Tht Hg data were extended to 2.8-A resolution and the Pb to 3.0 A. Several alternative cycles of le@. squares heavy atom refinement and phase calculation at 3 A ultimately yielded the phasing statistics and final heavy atom MIR parameters shown in Tables I1 and 111, respectively.

Starting with the 4.1-A initial polyalanine backbone structure superimposed on the 3-A MIR electron density map, the backbone trace was extended to eventually yield a 309-residue polyalanine model. This model, however, consisted of three separate long fragments of 79, 120, and 110 residues. The initial attempts to align the MBP sequence with these fragments took into consideration two observations. First, by comparing the sequence of MBP (370 residues) with sequences of the other sugar-binding proteins, arabinose-bind-

5206 2.3-A Structure of the Periplasmic Maltodextrin Receptor

TABLE I11 Heavv atom oarameters from the 3.0-A MIR Dhasine

Coordinates" ~~~ ~____ _______ ~~~

Derivative Site Temperature factor'

X Occupancy'

Y Z P1 P 2 83 Sr P b P6

6-Iodomaltose 1 0.2360 0.2500 0.6649 34.0 0.00155 -0.00122 0.00114 0.00208 -0.00221 0.00133

Pb 1 0.4886 0.1265 0.1618 42.5 0.00074 -0.00050 0.00031 0.00230 0.00460 0.00473 2 0.4392 0.4145 0.3306 42.1 0.00236 0.00546 0.00315 0.00400 0.00321 0.00475

Hg 1 0.4199 0.1861 0.4273 50.1 0,00118 -0.00405 0.00153 0.00372 -0.00277 0.00269 2 0.0615 0.2451 0.5497 48.6 0.00352 0.00192 -0.00237 0.00528 -0.00940 0.02169 3 0.1087 0.282 0.5814 12.1 0.00303 0.00339 0.01266 0.00605 0.01612 0.02122

Pt

DYI

1 0.4899 0.3362 0.9569 61.8 0.00131 0.00000 0.00182 0.00265 0.00000 0.00436 2 0.2905 0.2370 0.4209 27.4 0.00119 0.00000 0.00166 0.00242 0.00000 0.00398 3 0.4208 0.1836 0.4228 20.1 0.00136 0.00000 0.00189 0.00275 0.00000 0.00453

1 0.3535 0.4802 0.2915 37.9 0.00014 -0.00236 0.00081 0.00107 0.00083 0.00130 2 0.3423 0.4567 0.8460 33.5 0.00168 0.00099 0.00176 -0.00041 0.00014 0.00374 3 0.3701 0.1595 0.1020 43.7 0.00652 -0.00066 0.00209 -0.00138 0.00186 0.00428 4 0.1885 0.1722 0.5360 40.2 0.00261 -0.00616 -0.00425 0.00495 0.00757 0.00544 5 0.3785 0.4070 0.2949 39.9 0.00382 0.00124 0.00273 0.00268 -0.00010 0.00282 6 0.2934 0.4488 0.7620 40.6 0.01222 0.00041 0.00928 -0.00194 0.00063 0.00647 7 0.1662 0.2486 0.4977 44.4 0.03963 0.02564 0.05602 0.00463 0.03598 0.08176

a X, Y, Z are the fractional co-ordinates of the site. * Occupancies are in number of electrons.

Dl to 0 6 are the temperature factors according to the expression: T = exp(- plhz - @,hk - P3kl - &kZ - &hl - 0 6 1 7 .

ing protein (306 residues), galactose/glucose-binding protein (309 residues), and ribose-binding protein (271 residues), Du- play et al. (1984) observed that these four sugar-binding proteins have two regions with highly homologous sequences. The first consists of segments of residues 119-177 in MBP, 33-92 in ABP, 34-94 in GGBP, and 33-92 in the ribose- binding protein and the second consists of 288-343 (MBP), 199-255 (ABP), 205-256 (GGBP), and 184-235 (ribose-binding protein). Therefore, they concluded that MBP has an extra 86-residue extension at the amino- or N-terminal end and a slightly shorter carboxyl-terminal end.

The second observation was that the tertiary structure of ABP and GGBP are highly homologous despite only about 13% sequence identity between them (Quiocho and Vyas, 1984; Vyas et al., 1983, 1988, 1990). Least squares superpositioning of ABP and GGBP a-carbon backbone structures indicated about 80% a-carbon positions 0.f the two proteins superimposed (root mean square D = 2 A). This structural corroboration of the sequence alignment heavily influenced our expectations for the folding pattern of MBP.

Taking into consideration both the sequence and structural informations, the three fragments were aligned as follows. The short 79-residue fragment was assigned the N-terminal segment since it did not seem to exhibit the typical binding protein-folding pattern. This was followed by the 120-residue fragment, as its conformation roughly resembled that com- monly observed for about the first 100 residues of the other binding protein structures. The third fragment (100 residues) was connected last, forming the major part of the C-domain. It should be noted that the 79-residue fragment is proximal to the two larger fragments. The suggestion that the extra residues in MBP (as compared with the ABP or GGBP structures) are confined to the N-terminal segment could not be reconciled with any fittings tried.

The small 79-residue fragment was therefore shifted to the carboxyl-terminal end of the structure. In so doing, the N- terminal portion of MBP emulated the identical folding pattern observed for the same portion in all previously refined binding protein structures, not only the two sugar-binding

proteins but also the sulfate-binding protein and the Leu/Ile/ Val-binding protein (see discussion below). With this new arrangement of the three fragments, we could reconcile the residues criginally fitted with the sequence quite well.

The 3-A MIR phases, together wit) the Hg single isomorphous phases between 3 and 2.8 A, were combined with structure factors obtained from selected atomic mo(e1s obtained over the course of model fitting. The best 2.8-A phase combination had a mean figure of merit of 0.73. The side chains of only 23 of the total 370 residues could not be fitted to the maps calculated with combined phases. Most of these residues were found in regions of poor electron electron density, near the molecular surface and in loops. At this stage restrained least squares refinement was initiated, bringing the crystallographic R factor down from an initial value of 0.46 to 0.25. Moreover, using maps calculated with (21F01 - lFcl, a,) and ( IFo[ - IFcl, aJ, it was possible to fit the remaining 23 residues. The side chains of only 40 residues, mostly located near the solvent region, have incomplete electron densities. As native intensity data to 2.3-A resolution was available (Table I), initial rounds of PROLSQ refinement were done, achieving the same R factor as the one obtained at 2.8 A.

As one polypeptide stretch (residues 230-260) which was found to be out of register had to be shifted, the structure reported here varies slightly from the initial one reported (Spurlino, 1988; Martineau et al., 1990). The geometry of the partially refined 2.3-A structure has root mean square dcvia- tions from ideality of 03026 A for bond distances, 0.051 A for angle distances, 0.075 A for planar 1-4 distances, and 0.019 A for planes. The coordinates have been deposited in the Protein Data Bank.

Maltodextrin-binding Protein Structure-The structure of MBP ha! the shape of an ellipsoid with dimensions of 65 X 40 X 30 A. It is composed of two globular domains separated by a deep cleft or groove. Each domain is composed of segments from both the amino- or N-terminal and carboxyl- or C-terminal halves of the protein. Nevertheless, both domains have very similar arrangement of elements of secondary structure ("supersecondary structure"); each domain has a central

2.3-i Structure of the Periplasmic Maltodextrin Receptor 5207

FIG. 1. Stereo picture of the a- carbon trace of MBP. Every 10th residue is labeled, and N and C are the amino-terminal and carboxyl-terminal ends, respectively. The bound maltose is also shown in the center of the molecule.

FIG. 2. Ribbon-style drawing of MBP. Helices are presented as coils and strands are depicted as flat arrows. The @-strands are labeled with capital letters and helices with Roman numerals in the order of the polypeptide sequence. The site of the bound maltose is shown as two connected filled circles. This drawing was produced with a modified version of the RIBBON program (Priestle, 1988).

FIG. 3. Space-filling model of MBP. The orientation is the same as Fig. 1. The N-domain is colored blue. The C-domain is partitioned into two subdomains, the C1-domain, shown in yellow, and the C2-domain in gold. The limited accessible surface of the bound maltose (with gray carbons and red oxygens) is also clearly depicted.

core of a five-stranded @-sheet with two a-helices on one side of the sheet and three a-helices on the other. In spite of the integrated nature of the two domains, the domains are designated the N-domain and the C-domain as the former contains the N-terminal end and the latter the C-terminal end. The N-domain, which is composed of segments 1-109 and 272- 319, is smaller than the C-domain, which is made up of

segments 114-267 and 334-370. The C-domain could be further divided into two parts, a larger part called C1-domain and a smaller C2-domain (see below). Figs. 1 and 2 portray these general features of the MBP molecule at differing levels of abstraction and aid in an understanding of the following detailed description of the tertiary structure. (Note that @- strands are designated as capital letters in order of their occurrence along the polypeptide chain, and a-helices as Roman numerals in a like manner.) Maltose-binding protein is, in the main, an a/@ protein with 40% of the amino acids in a-helix, 20% in @-sheet. The remaining 40% is comprised of loops and coils.

The first section of the N-domain, consisting of the first 109 residues, has the secondary structure of ~Aa~j3BaII/3~aUI@D. The second 48-residue segment of the N-domain has the folding pattern @ L ~ I X ~ X @ M ( Y X I . The @-sheet is parallel with the exception of the fourth and the short sixth strand (L and M). The C-terminal ends of the parallel strands and the N- terminal ends of the helices terminate at the groove between the two domains. The entire six-stranded sheet is folded over on itself in a structure reminiscent of half a @-barrel (Figs. 1 and 2).

In the C-domain, the larger, 154-residue segment has a

and the much shorter second segment folds as three helices (OIXIICYXIIICXXIV). The central @-sheet is parallel, except strand L, with the C-terminal ends pointing into the groove (strands G and H are not part of the central sheet, see below).

The three peptide segments connecting the two domains consist approximately of residues 110-113,268-271, and 311- 315. The first two connecting segments (or cross-overs) from a short sheet-like structure. The first segment connects strand D in the N-domain to the C-domain strand E. The second segment bridges the C-domain strand K to the N-domain strand L. The last segment, in a helix to helix connection, bridges the N-domain to the C-domain (XI-XII). These cross- overs form the base of the groove. The existence of consecutive strands (e.g. 0~ and OH) and helices (e.g. XI-XIV) proves that MBP is not a pure a/@-type protein.

The last two closely associated helices (XI11 and XIV) and the loop between helices V and VI serves to extend the groove, thus creating a long tunnel between the two domains. A portion of the loop forms a short two-stranded antiparallel sheet (strands G and H) which lies beneath the helices. We have chosen to distinguish these elements (helices XI1 and XIV and strands G and H) as a third domain, thereby splitting the C-domain into two parts. The larger, C1-domain is equivalent to the standard binding protein C-domain. The smaller C2 domain is unique to MBP. Fig. 3 illustrates the domain divisions in MBP.

Sugar-binding Site-As MBP was purified and crystallized in the presence of maltose, we expected that the protein would

secondary arrangement of PEBE.~~@FBF(.VPGPH~VI@I~VII@J~VIIIPK


FIG. 4. Location of bound maltose. The density derived from a (21FJ - IFcI, a,) map is shown around the putative maltose location. The location of the maltose is designated by solid density contours. Also shown in the backbone trace in the area. This view is the same orientation as Fig. 1. The location and polarity of the bound maltose was confirmed by displacing maltose with maltotriose and 6-iodomaltose. The location of the iodine in the 6-iodomaltose was determined from a difference Fourier analysis (see "Experimental Procedures"). This site is shown as dotted density. The location of the third sugar in the maltotriose soak similarly analyzed is at the nonreducing end of the native maltose position. This site is shown as dashed density.

TABLE IV Protein sugar interactions

Residues that are within hydrogen-bonding distance (<3.5 A) and van der Waals contacts (<4 A) of the modeled maltose ligand. Note that one protein atom may have more than one bonding interaction or contact. The location of a residue within the protein is only mentioned the first occurrence of the residue. The residues listed here differ slightly from those found in Suurlino (1988). notably Tm-230. Tm-232. Glv-13. and As~-14.

~~ ~

Binding Polar contacts van der Waals contacts <4 A site Sugar atom Protein/water atom Location Location No. Residues

SI g l a-01 OD1 Asp14 NZ Lys'5 N-domain Loop PA+al Alaa, Glu"'

N-domain Loop P A - m l N-domain 12 Am", Asp14, Lys",

g l 0 2 NZ Lys'5 C-domain 6 G ~ u ' ~ ~ ,

OE2 Glu"' N-domain Cross-over 1 N + C 18 Wat

g l 0 3 g2 0 2 Maltose Wat

g l 06 Wat Wat

T y P , TrpZ3O

s2 g2 0 2 OD 1

gl 0 3 g2 0 3 NE1

OD1 NE

g2 0 4 NH1

g2 06 OE2 N

Asp65 N-domain Loop PC-WYIII N-domain 18 Trpfi2, Alafi3, Aspfi5,

Maltose C-domain 29 GlulS3, Pro15', Tyr15' N-domain Loop P c - w ~ ~ ~ Trp340

Asp65 N + C 47 Arg66 N-domain Loop &+allI Total 65 Arg+j6 Wat

TyP5 C-domain Loop P e a v

A r P

Trp62

~ 1 ~ 1 5 3 C-domain Loop @ e a v

have a bound maltose in the crystal structure. Moreover, as the structures of the liganded forms of ABP, GGBP, and SBP show ligand bound in the cleft between the two domains of each protein (Quiocho and Vyas, 1984; Vyas et al., 1988; Pflugrath and Quiocho, 1988), we further expected to find the maltose in a!imilar site. Indeed, an elongated electron density in the 2.8-A maps, which could not be attributed to the protein, could be seen at the center of the molecule, in the groove between the two domains (e.g. see Fig. 4). That this density belongs to a bound maltose was verified by two independent difference Fourier analyses using ( IFp-LI - I F P - ~ ~ ~ ( ) as coefficients (where P is MBP and L is a- ligand other than Mal) and phases obtained from the 3-A MIR analysis using only the Hg, Pt, and Pb derivatives. As the

native MBP (or P-Mal) structure is expected to contain bound maltose, a positive difference density peak would indicate that a given ligand was able to displace the bound disaccharide and the peak attributable to the moiety unique to that ligand. One of these ligands is a-~-Glc-(l+4)-6-iodo-6-deoxy-~-Glc or 6-iodomaltose which was also used as one of the heavy atom derivatives in the final 3-A resolution analysis (see above). It gave one very strong (>6 u) difference peak which overlapped with one end of the putative maltose density (Fig. 4). (The coordinates of this peak are essentially identical to the refined iodine position in the MIR analysis (Table III).) A similar analysis of a crystal soaked in maltotriose also revealed one significant difference density peak. Although this peak also overlapped the putative maltose density, it is

2.3-A Structure of the Periplusmic Maltodextrin Receptor 5209

FIG. 5. Hydrogen bonds between MBP and the bound maltose. A , a stereo picture of the hydrogen bonding interactions between MBP and maltose based on the initial PROLSQ refinement of MBP at 2.3 A. These interactions are also designated in Table V. Hydrogen bonds are shown as dashed lines. The hydrogen bonds involving water molecules (Table V) are not included since water positions have not been refined. Carbon atoms are depicted as single circles, oxygen atoms as double circles, and nitrogen atoms as triple circles. B, a sche- matic drawing of the proposed hydrogen bonding between maltose and MBP, dis- playing only the primary direct bonds between MBP and maltose.

A

15 15

8

located opposite to that of the 6-iodomaltose difference density. These results not only firmly establish the location of the sugar-binding site but also fix the polarity of the site. The 6-iodomaltose identifies the reducing end of the maltose while the maltotriose marks the third glucosyl unit located from the nonreducing end.

The other iodo-sugars and the cyclic sugars failed to displace the maltose in the crystal soaks as indicated by the absence of a difference Fourier peak.

The fitting of a model of the maltose to the sugar density in the 2.8-8, maps calculated with (21F01 - IFcl, a,), followefl by additional rounds of refinement using the native 2.3-A resolution data set (Table I) has enabled us to evaluate the mode of binding of the disaccharide. (The model was obtained from the crystal structure of a-maltose (Takusagawa and Jacobson, 1978).) A detailed and more accurate description of the atomic interaction between MBP and Taltose must await the completion of the refinement at 1.7-A resolution using data collected from one crystal on an area detector.

The deep maltodextrin-binding groove is bounded by the walls of t)e two domain? (about 18-8, high) and the base (about 9-A wide and 18-A long), which is mainly formed by the inter-domain connecting segments and helix XIII. The base of the binding groove has the shape of an arc. The bound maltose is buried in :he groove to the extent that only about 3.6% of the 545.5-A' accessible surface of the maltose is exposed to the solvent (Fig. 3).

Residues deployed from both walls are important in the binding of maltodextrins. The locations of those residues are

indicated in Table IV. Many of these residues are located in loops, those connecting strand A to helix I, strand C to helix 111, strand F to helix V (see Figs. 1 and 2 for reference). There are also binding site residues located in cross-over 1 (strand D-E) and in helix XIII.

At the present stage of analysis, the conformation of the bound maltose that has been modeled to the electron density appears to be very similar to that of the small molecule crystal structure of a-maltose (Takusagawa and Jacobson, 1978) (Fig. 4). (For brevity and convenience, we have in this paper identified the consecutive a-(l-4)-linked glucosyl units of any linear maltodextrin starting at the reducing end of the sugar, as gl, 82, 83.. .; for example see Fig. 5B). There is a twist of about -50" between the glucosyl units due to the rotations about two torsion angles, 4 (defined by (g2)05-(g2)Cl- (g1)04-(gl)C4) and + (defined by (g2)Cl-(gl)04-(gl)C4- (gl)C5). As a result (g2)02-H and (g1)03-H are within hydrogen-bonding distance.

The bound maltose is held in place by hydrogen bonds and van der Waals contacts (Table IV). The g2 or nonreducing glucosyl unit is more heavily involved than the gl unit. The binding of maltose results in the formation of 16 hydrogen bonds (Table IV), 11 directly with 8 residues of MBP (Fig. 5) and five likely by way of bound water molecules. All the sugar hydroxyls, but not the ring and glycosidic oxygens, participate in hydrogen-bonding (Table IV). While all these hydroxyl groups are located approximately in the plane of each glyco- pyranose ring, forming a peripheral surface, the hydrogen- bonding interactions are mainly concentrated along a ridge

5210


rn 1

FIG. 6. Stereo picture of the stacking interactions between the aromatic residues and maltose in the binding site as represented by van der Waals dot surfaces.

composed of these groups at positions 1 (a or /3 hydroxyl), 2 and 3 of the gl glucosyl unit, and 2,3, and 4 of g2 (Fig. 5). It is interesting that the (g2)04-H, the hydroxyl which would be involved in a glycosidic bond in longer dextrins, is involved in accepting a hydrogen bond solely from the protein (Arg6). This would facilitate the easy addition of an extra glucosyl unit in the longer maltodextrins. As the (g1)03-H is not engaged in direct hydrogen bonding with the protein and as the (g2)02-H donates a hydrogen bond to the protein, we conclude that the 0 3 hydroxyl group of the reducing sugar (gl) is a donor in the intra-sugar hydrogen bonding (Fig. 4). The small molecule structure of maltose clearly shows a similar intra-sugar hydrogen bond (Takusagawa and Jacob- son, 1978).

It is noteworthy that, with the exception of two NH groups (one from the peptide NH of T y r 1 5 5 and the other from the Trp6* side chain), all of the side chains that form direct hydrogen bonds with the maltose are charged, four carboxyl- ate, one guanidinium and one ammonium groups (Fig. 5 and Table IV). Thus a large proportion (9 out of 11) of the direct hydrogen bonds in the MBP-maltose complex is between neutral sugar-OH groups and charged residues.

Water molecules are often involved in protein-carbohydrate interactions, mediating hydrogen bonds between residues and sugar (Quiocho, 1986,1989). As ordered water molecules have not been included in the current structure analysis, they are not discussed in detail. However, there are clear indications of the presence of at least five water molecules near the bound maltose (Table IV). In the gl unit, there are two water molecules close to 06-H one to 02-H and one to 03-H. The fifth water is near 04 hydroxyl of the g2 unit. These water molecules appear to be further hydrogen bonded to other residues.

Many van der Waals contacts are also formed in the MBP-

maltose complex (Table IV). In these contacts there is extensive participation of aromatic side chains which include Trp6',

TrpZ3', and Trp340 (Table IV and Fig. 6). The interaction between the anomeric hydroxyl in the p position and TrpZ3O is intriguing; the hydroxyl is pointing at the six-membered ring (on the A-face side) of the indole side chain at a distance from the plane of the tryptophan ring of 2.8 A.

DISCUSSION

Binding-protein Structures-With the addition of MBP, we now have determined the structures of seven binding proteins, three with specificities for carbohydrates, two for amino acids, and two for tetrahedral oxyanions. Despite differences in their size, sequence (except between Leu/Ile/Val/-binding protein and leucine-specific binding protein, which have 80% sequence identity (Sack et al., 1989b)) and ligand specificities, the binding proteins share structural features that are essen- tial to understanding function (Sack et al., 1989a; Quiocho, 1990). First, the structures are prolate ellipsoids, consisting of two separate similarly folded globular domains. Although each domain is built from two noncontiguous polypeptide segments from the N-terminal and C-terminal half of the polypeptide chain, both domains exhibit similar supersecondary structure consisting of a central p-pleated sheet with two or three a-helices on each side. Second, the three different peptide segments that connect the domains are spatially close, although not sequentially, and provide a base (or boundary) for the deep cleft between the two domains and a flexible hinge as well. And third, the ligand-binding site is located in the cleft or groove between the two domains. Additionally, the first six secondary structural elements comprising about 100 residues in all six binding proteins have the identical topological arrangement of P A - ~ I - P B - ~ I I ~ c - o L I I I (Fig. 7).

Differences in the transitions or cross-overs from domain

2.3-A Structure of the P e r i p h m i c Maltodextrin Receptor 5211

YBP

SBP

C

ABP

QBP

C

@%e9 111

L w L w

FIG. 7 . Binding-protein topology. Topological packing diagrams for MBP, SBP, ABP, GGBP, and LIVBP. These diagrams are grouped according to their similarity in regard to folding pattern and cross-over transition style. Group I shown on the right and group I1 on the left (see “Results”).

to domain, are the primary differences in the folding topology of the binding protein structures. These differences have led us to divide the binding proteins into two groups. In group I, which is composed of ABP, GGBP, LIVBP, and LBP (Fig. 7), the transitions from domain to domain are from strand to helix for the first two cross-overs and strand to strand for the third cross-over. (Note that in LIVBP and LBP, an additional helix occurs in the second transition.) The sheet topology of both domains in this group of proteins is of the general form P2P1P3P4Pn, where in each domain the first four strands are from the larger polypeptide segment and Pn represents the last strand or strands which originate from the second, much shorter segment (Fig. 7). In ABP and GGBP n = 1 (strand K in the N-domain and strand L in the C-domain), whereas in LIVBP or LBP n = 2 (LK and MN).

The sulfate-binding protein, a member of group 11, has sheet topologies and transitions different from group I. The sheet topology of both domains of SBP is of the general form P2P1P3&P4 (see Fig. 7 and compare with above). In this case Pn represents the initial strand from the first cross-over into each domain. Although in the C-domain the Pn strand is actually the first strand of the sheet it is considered here as a separate entity, since it is the first cross-over into the C- domain. The difference in sheet topology between SBP and the other binding proteins is attributed to the nature of the cross-overs between domains. In SBP the first two inter- domain connections are strand to strand and the third is helix to helix.

To which group does MBP belong? The maltodextrin- binding protein would be thought to most closely resemble the other sugar-binding proteins. Surprisingly, however, as can be seen in Fig. 7, MBP has sheet topology and cross-over connections virtually identical to SBP. There are additional elements of secondary structure as compared to SBP as should be expected in a protein that is more than 20% larger.

Further comparison was made by least squares superimpo- sitioning the a-carbon backbone coordinates of MBP with each of the other binding proteins. In this type of comparison, it is important to note that the structures of MBP, ABP, GGBP, and SBP are of the liganded, “closed cleft” forms, whereas the structures of LIVBP and LBP, being without ligands, are of the “open” forms. In the ligand-loaded structures, the two domains are close to each other and the ligand is essentially completely entrapped in the cleft between the two domains. On the other hand, in the structures of the unliganded forms of LIVBP and LBP, the two domains are widely separated, making the cleft very open and accessible (Sack et al., 1989a, 198913). Consequently, structural comparisons were done at two levels. First, each separate domain (N- or C-domain) of MBP was compared with the identical domain of ABP, GGBP, SBP, and LIVBP. Second, the entire backbone structure of MBP was superimposed with only ABP, GGBP, or SBP. (A whole protein comparison with LIVBP is of little significance.) Three general patterns emerge from these comparisons (Table V). The N-domains have the great- est structural homology. This is consistent with the previous observations that the first six secondary structural elements of all six binding protein structures have identical topology. Superpositioning of entire structures always produced an inferior alignment than achieved by superpositioning of like domains. This is attributed to slight differences, even among the structures of the proteins with bound ligand or closed forms, in the proximity of the two domains and of the relative orientations between the N- and C-domains. Finally, the highest structural homology is observed for the ABP and GGBP pair and the SBP and MBP pair. Indeed, the overlap of MBP with SBP was nearly as extensive as that previously found between GGBP and AB! which belong to group 1 (-70%, root mean square D = 2 A) (Vyas et al., 1983; 1990).

The superimposed regions of SBP and MBP include not only most of the elements of the secondary structure but also the hinge (cross-over) regions. Most of the 62 extra residues in MBP, as compared to SBP, can be accounted for by three extended loops and an additional helix at the carboxyl-terminal end. The loops occur between helix I11 and strand D (14 extra residues), helixes V and VI (14 extra residues, which include strands G and H), and strands K and L (11 extra residues). Helix XIV accounts for 20 additional residues.

The sheet topology of the N-domain of group 2-binding proteins reveals a unique and somewhat problematic folding pattern. Strand L, which follows the folding of the larger peptide segments in the N-domain and C-domain, intercalates between strands C and D in the N-domain sheet (Fig. 7). The mechanism of the insertion of strand L is probably similar to that originally proposed for SBP (Pflugrath and Quiocho, 1988).

Although MBP and SBP structures are highly homologous, the size of the cleft in MBP is much larger than the one in SBP. This is pot unexpected since the sulfate is significantly smaller (-3 A in diameter) than the ligands of MBP which not only include linear but also cyclic maltodextrins. Indeed, cyclic dextrins such as cyclic maltobeptaose, which has a diameter of 15.4 A and a height of 7.9 A (Saenger, 1983), bind with affinity similar to maltose (Miller et al., 1983). The size


TABLE V Comparison of the MBP structure with other binding-protein structures

Results of superpositioning the a-carbon structure of MBP to the a-carbon structures of other binding proteins. The comparisons were done with similar domains (e.g. N-domain versus N-domain) or whole structures. The minimization of the differences in the coordinates of a-carbons from two structures by the rigid body movement of one set of coordinates is a method to assess the similarity of structures. A residue pair is included in the equivalency calculation if the a-carbon positional difference is less than 3.0 A and part of a run of at least 3 residues. The N+C row is an additive total of the individual domain overlaps for easy comparison with the whole molecule alignment. rMS, root mean square.

Overlap molecule Percent No. possible rms D Agreement no. match

ABP complete molecule ABP N-domain ABP C-domain ABP N+C-domains

GGBP complete molecule GGBP N-domain GGBP C-domain GGBP N+C-domains

LIVBP N-domain LIVBP C-domain LIVBP N+C-domains

SBP complete molecule SBP N-domain SBP C-domain SBP N+C-domains

40 46 51 48

41 61 35 47

58 43 50

62 65 77 71

122 67 79

146

123 85 57

142

92 65

157

190 95

123 218

305 3.05 147 3.11 156 4.47 303

302 138 163 301

160 3.18 151 4.46 311

308 3.07 147 2.06 161 2.91 308

3.12 3.13 3.41

MBP N-domain versus C-domain 43 68 160 2.84

of the groove is large enough to easily accommodate the cyclic dextrins.

Protein-carbohydrate Interactions-One major and totally unexpected discovery resulting from our structure analysis of several binding proteins is that many ligands, although totally different (e.g. monosaccharides, oligosaccharides, oxydi- anions, and amino acids), are bound primarily by hydrogen bonds (Vyas et al., 1988, Pflugrath and Quiocho, 1985, 1988; Sack et al., 1989a; Luecke and Quiocho, 1990; Quiocho, 1989). This extraordinary feature is important for the function of these proteins (reviewed recently by Quiocho (1990)).

Despite considerable structural similarity between MBP and SBP the mode of binding of maltodextrins to MBP resembles the mode of binding of monosaccharides to ABP and GGBP. The major difference is that whereas the sulfate is hydrogen bonded mainly to main-chain NH groups, sugar ligands are hydrogen bonded almost entirely to polar side chains. Note that the sulfate bound to the sulfate-binding protein is not within van der Waals distance to a counter- charged residue or cation (Pflugrath and Quiocho, 1985,1988). The heavy reliance on the dipolar main-chain peptide units reflects the need not only to offset the hydration energy of the ionic sulfate ligand, which is much greater than that of uncharged saccharides, but also to provide a means to stabilize the isolated, uncompensated charges on the sulfate (Pflugrath and Quiocho, 1985, 1988; Quiocho et al., 1989). From the standpoint of diversity and specificity for carbohydrate ligands, the set of periplasmic proteins consisting of ABP, GGBP, and MBP is an excellent system to analyze and study using a variety of methods. The well-refined, high resolution structures of several complexes of the L-arabinose-binding protein and the D-galactose/D-glucose-binding protein with monopyranoside ligands have revealed many features of the atomic interactions between protein and carbohydrates, including the molecular basis for the unusual specificities exhibited by these proteins (including anomer and epimer rec-

ognitions) (Quiocho and Vyas, 1984; Quiocho, 1986; Vyas et al., 1988; Quiocho et al., 1989; Quiocho, 1990).4 This knowledge has been solidified and further enhanced by the results of the determination of the structure of MBP complexed with maltose.

Although the specificities of the three sugar-binding proteins are different, the binding sites are all located in the cleft between the two globular domains. Moreover, the topology of the polypeptide chain associated with these sites is very similar; loops between P-strands and helices in both domains are the main source of residues for ligand binding. Every ligand is almost completely engulfed in the cleft, with a loss of >96% of free accessible surface area of sugars, monosaccharides or disaccharide, in making contact with the three binding proteins (see Fig. 3). Consequently, every polar and nonpolar atom of the sugar ligand interacts extensively, via hydrogen bonds and van der Waals forces, with corresponding groups in the binding site cleft. These general features account, to a large extent, for the observation that the binding- protein carbohydrate complexes are some of the tightest.

In simple carbohydrates, the polar groups are made up almost entirely of peripheral hydroxyl moieties whereas nonpolar groups consist of C-H groups which protrude from each face of the sugar ring. The proportion between polar and nonpolar accessible surface areas (as defined by Lee and Richards, 1971) differs between monosaccharides and oligosaccharides. The polar group of atoms (oxygens) in monopyranosides (e.g. D-galaCtOSe, D-glucose, and L-arabinose), con- stitute a significantly larger percentage of accessible surface (-80%) than the nonpolar (C) atoms. On the other hand, in oligosaccharides the ratio of polar to apolar accessible surface area varies depending on the linkage- and orientations of the monomers. For example, of the 546-A* accessible area of free maltose (crystal structure conformation), polar groups con- stitute about 55% and apolar groups about 45%. As the sugars

N. K. Vyas and F. A. Quiocho, unpublished data.

2.3-A Structure of the Periplusmic Maltodextrin Receptor 5213

bound to the three periplasmic binding proteins are almost completely buried, both polar and apolar groups are heavily involved in binding. The locations (dictated by the stereochemistry of the carbohydrates) and accessibilities of each of these groups of atoms play a key role in molecular recognition and affinity of sugars.

The stereochemistry of the hydroxyl groups in monosaccharides is sufficient to determine the location of accessible polar, as well as apolar, groups. In oligosaccharides the location of polar patches is affected not only by the position of the hydroxyls, but also by the linkage of the sugars and the orientation of the monosaccharides with respect to each other. Maltodextrins ( a l 4 - l i n k e d glucose) present an asymmetric arrangement of hydrophobic and hydrophilic faces. The faces of the individual glucosyl units are located approximately -50" to each other with a slight twist (e.g. maltose; see Fig. 6). The A-faces of maltose form an outer hydrophobic arc, while the B-faces form a smaller inner arc which is punctuated by the glycosidic oxygen. The A-face of the sugar is defined as that for which the numbering of the carbon atoms is clockwise and the B-face counterclockwise. A much larger area of apolar C-H groups protrude from the A-faces than from the B-faces of maltose. The hydroxyls are concentrated at one edge of the oligosaccharide, much like the points on a crown. These features have important ramifications that are manifested in the binding pocket of MBP.

The polar groups of sugars, made up mostly of stereospecific and highly exposed hydroxyl groups, are naturally the ones involved in hydrogen-bonding interaction. Indeed, hydrogen bonds in the complexes of the binding proteins and saccharides are extensive and provide major contributions to the specificity and tight affinity of the binding sites. There is a total of 16 hydrogen bonds in the MBP-maltose complex (Table IV), 13 in the GGBP-glucose complex, and 10 or 11 in the ABP-arabinose complex (Quiocho and Vyas, 1984; Vyas et al., 1989; Quiocho, 1989).

In the complexes of ABP and GGBP with monopyranosides, the hydrogen bond potential of every polar group (hydroxyls and ring oxygen) of the sugars is satisfied. Moreover, all hydroxyl groups act simultaneously as hydrogen bond donors and acceptors of the type

(NH), + OH + 0

where NH and 0 are hydrogen donor and acceptor groups, respectively, and OH is a sugar hydroxyl, and n = 1 or 2. (There are some instances in which hydrogen bonding between binding protein and sugar is mediated by water, acting as both H bond donor and acceptor groups (Quiocho and Vyas, 1984; Quiocho et al., 1989).) Cooperative hydrogen bonding is also observed in the MBP-maltose complex (Table IV). All eight maltose hydroxyl groups form more than one hydrogen bond, including the intrasugar bond. Whereas six hydroxyls (01-H, 03-H, and 06-H of the gl unit, 02-H, 04-H, and 06-H of the g2 unit) participates both as hydrogen bond acceptors and donors (n = l ) , two hydroxyl (02-H of the gl unit and 03-H of the g2 unit) are fully coordinated (n = 2) including the C-0 bond.

Kinetic analysis of sugar binding to ABP, GGBP, and MBP originally indicated no preferential binding of either a- or p- anomeric form of the sugar ligands; both anomers are bound with similar affinity (Miller et al., 1983). Determining the high resolution structures of the complexes of ABP with arabinose, fucose, and galactose (Quiocho and Vyas, 1984; Quiocho et al., 1989) and complex of GGBP with glucose and galactose (Vyas et al., 1989)5 have revealed a common and

' N. K. Vyas, M. N. Vyas, and F. A. Quiocho, unpublished data.

simple mechanism for the recognition of both anomeric hydroxyls. In both proteins there is an oxygen atom of carbox- ylate side chain of an Asp residue which is precisely positioned to accept a hydrogen bond from either the a- or p-anomeric hydroxyl. The same mechanism exists in MBP (Fig. 4).

There are only very minor differences between the hydrogen bonding interactions in the MBP-maltose complex and the ones observed in the monosaccharide-binding protein complexes. Not all of the polar groups (sugar ring and glycosidic oxygens) of the dissacharide are involved in hydrogen bonding with the protein. The use of a peptide unit ( T Y ~ ' ~ ~ ) in hydrogen bonding the maltose is seen for the first time.

Determining the structure of the MBP-maltose complex solidifies the previous results observed in the structures of ABP and GGBP indicating extensive use of polar residues with planar side chains in hydrogen bonding carbohydrates (Quiocho et al., 1989). What is now also clearly apparent is the dominant role of ionizable residues in sugar binding: 4 out 6 in ABP, 5 out of 8 in GGBP, 6 out of 8 in MBP. The extensive use of charged groups undoubtedly contributes significantly to the tight binding of sugars to these binding proteins (& = 3 to 30 X M).

The sugars make an unusually large number of van der Waals contacts upon binding, further contributing to the high affinity of the sugar sites in*the binding proteins. There are about 60 van der Waals (<4 A) contacts in the GGBP-glucose complex (or about five/sugar atom), 47 in the ABP-arabinose complex (or about 4/sugar atom), and 65 (a preliminary estimate) in the MBP-maltose complex (or about three/sugar atom). Many of these contacts result from hydrogen-bonding interactions, enabling more of the atoms of the polar residues to come within van der Waals distance of the sugar.

A new facet of protein-sugar interactions revealed in great detail in the structures of GGBP and ABP is the stacking of aromatic side chains against the pyranose ring, leading to formation of apolar van der Waals contacts (Quiocho and Vyas, 1984; Vyas et al., 1988; Quiocho, 1986, 1989). Similar, but more extensive, stacking interactions are also observed in the MBP-maltose complex (Fig. 6). The indole side chain of Trp6' not only forms a hydrogen bond, but is also roughly coplanar with the B-face of gl or reducing glucosyl unit and perpendicular to the B-face of g2 or nonreducing sugar. The A-face of Trp340 side chain is stacked against almost the entire A-face of g2 unit. The aromatic side chain of Tyr'55 is over the glycosidic bond and a portion of the A-face of the gl sugar. The interaction between the anomeric hydroxyl in the p orientation and TrpZ3O is intriguing; the hydroxyl is pointing at the six-membered ring (on the A-face side) of the indole side chpin at a distance from the plane of the tryptophan ring of 2.8 A. This interaction suggests a weak aromatic hydrogen bond with the reducing hydroxyl. Levitt and Perutz (1988) have previously suggested a similar type of hydrogen bond between amino groups and aromatics which is worth about 3 kcal/mol or half the strength of a normal hydrogen bond.

We have previously observed that in ABP, GGBP, SBP, and LIVBP one domain makes significantly more hydrogen- bonding interactions with ligand than the other (Sack et al., 1989a). This observation holds true in the case of the MBP- maltose complex; there are more H-bond interactions of the disaccharide with the N-domain than with the C-domain (Table IV). The majority of the van der Waals interactions and aromatic residues near the sugar are located in the C- domain.

There is an equilibrium between the open cleft and closed cleft forms of the binding protein. The ligand would bind to the open form, initially to one domain by virtue of the greater


FIG. 8. Aromatic residues in the oligosaccharide-binding groove of MBP. A stereo picture of MBP display- ing the large number of aromatic residues present in the protein. Many of these residues are in or close to the maltodextrin-binding site groove between the two domains. Aromatic residues play an important role in maltodextrin binding. The large number of aromatic residues seen in the cleft is unique to the MBP structure. The orientation of the picture is the same as Fig. 1.

number of H-bond interactions that will be formed. Cleft closure would enable the other domain to participate in binding and at the same completely entrap the ligand. Ligand binding stabilizes the closed form, and therefore the rate of dissociation of the ligand (which is seen to vary widely) will depend on the extent of the interactions formed in the closed complex (Miller et al., 1983). Evidence from structural (Sack et al., 1989a, 1989b; Quiocho, 1990), low angle x-ray scattering (Newcomer et al., 1981), kinetics of ligand binding (Miller et al., 1983),2 and theoretical studies (Ma0 et al., 1982) indicate the open and closed forms can be achieved by a bending motion about a hinge between the two domains.

Among the periplasmic sugar-binding proteins, MBP has significantly more aromatic residues (41) than ABP (26) or GGBP (22). Moreover, many of these residues (16, including the ones described above) are located in or near the maltodextrin-binding groove (Fig. 8). Undoubtedly many of these residues will be involved in binding longer linear maltodextrins, as well as cyclic dextrins. The wall of the C-domain facing the groove is particularly rich in aromatic residues (Fig. 9A). As can be seen in Fig. 9B, the bound maltose is cradled by three of these residues (Trp230, and Trp340). The relatively hydrophobic A-faces of the longer maltodextrins with a ribbon-like twist follow the aromatic residues on the face of the C-domain. (See above description of maltodextrins.)

Four of the 8 tryptophans present in MBP are locate9 in the cleft. Tryptophans 62, 230, and 340 are within 4 A of maltose, and TrpZ3* is separated from the sugar by TrpZ3O. The location of tryptophan residues relative to the bound ligand is of major interest since binding of oligosaccharides induces changes in the tryptophan fluorescence, and these changes have been conveniently used to measure equilibrium dissociation constants (&) (Szmelcman et al., 1976; Miller et al., 1983)' and the on ( k ' ) and off ( k ) kinetic rate constants of maltodextrin binding to MBP (Miller et al., 1983).'

Judging from the larger size of the ligand-binding site in MBP as compared to the smaller ABP- and GGBP-binding sites, there is indeed room for binding oligosaccharides longer than maltotriose (including the cyclic dextrins). The binding pocket for maltodextrins in MBP can be considered as a group of sites (SI, S2, etc.) for single glucose residues. X-ray analysis (see above) showed the existence of subsites S1 and S2, interacting with the reducing end and nonreducing end of maltose, respectively. Initial low resolution analysis of the complex of MBP with maltotriose (the best ligand) further revealed an additional subsite S3 which interacts with the g3 glucosyl unit. Subsites S3 and S4 are located near helix XIII. Although MBP binds linear maltodextrins of up to seven glucosyl units, it is doubtful, because of size limitations that the groove can have seven subsites.

The fact that cyclic maltodextrins bind to MBP indicates

a possible site in front of S1 (41) that can be occupied by a glucosyl unit of the cyclic sugar. The location of the putative -S1 site shows the presence of polar and aromatic residues. The fluorescent shifts induced by the binding of linear maltodextrins are opposite to those observed for binding of cyclic maltodextrins; linear sugars produced a slight red shift, while the cyclic ones resulted in a blue shift. The binding of cyclic sugars necessitate the presence of a glucose residue in front of site S1. This would result in Trp6' and TrpZ3O being in different environments when the cyclic sugars are bound, while Trp340 would remain in an equivalent environment (see Martineau et al., 1990 and below). But, thus far crystallographic studies of the binding of linear oligosaccharides do not show binding of a glucosyl unit at the 4 1 site. There may be enough differences in the relative configuration of the glucosyl unit in the cyclic dextrins as compared with the unit in the linear maltodextrins with a ribbon-like structure to disfavor the occupation of -S1 subsite by a linear dextrin.

It is possible that some of the glucosyl units of the longer linear maltodextrins are out in solution, making no contact with the protein. This is observed in the binding of the maltodextrins to the storage site of phosphorylase (reviewed in Johnson et al., 1988). It is also possible that longer dextrins will not necessarily have gl bound at S1 as observed for maltose and maltotriose (reported here) and maltotetraose.6 The photoaffinity labeling of MBP (Thieme et al., 1986) where the affinity label was attached to the reducing end of the sugar and successfully labeled MBP with up to 6 glucose residues attached to the affinity label also indicates that the reducing end of maltodextrins remains in contact with MBP, even for the larger dextrins. However, whether the contact site is the same for all oligosaccharides is unclear.

Determining the maltodextrin-binding protein structure brings to four proteins that bind oligosaccharide, including lysozyme, phosphorylase, and amylase, whose structures have been determined. Interestingly, only MBP is able to bind cyclic dextrins. In a review article by Johnson et al. (1988), it was noted that there is very limited similarity in the topology of the protein chains at the oligosaccharide recognition sites in lysozyme, phosphorylase (glycogen storage site), and amylase. The oligosaccharide-binding site in MBP also bears very little resemblance to any of the sites in the three enzymes. However, it, as well as the sites in ABP and GGBP, has many features which are similar to the catalytic site in phosphorylase. Like those of the three binding sugar-binding proteins (discussed above and described under "Results"), the catalytic site of phosphorylase is composed of residues located almost entirely in loops between p-strands and a-helices. This observation is not too surprising since phosphorylase is also an a/ p protein, and its catalytic site is also located in the cleft shared between two domains. Also, every sugar ( e g . glucose,

L. Rodseth and F. A. Quiocho, unpublished data.

I RG. 9. Stereo space-filling view of the C-domain highlighting aromatic residues in the binding

groove. A, the surfaces available for stacking with the sugar rings of maltodextrins are readily discernible in this representation showing the wall of C-domain that is facing the cleft with aromatic residues in green. B, the same picture as in A with the addition of the bound maltose (yellow). This shows the maltose being cradled by the aromatic residues resulting in the interaction with the A-faces of the glucose units of maltose. The extension of additional glucose can be envisioned as following the path (seen in A ) of aromatic residues on the face of the C- domain cleft.

heptinitol, etc.) that binds to the catalytic site of phosphorylase is almost completely buried and the hydrogen potential of almost every polar group is satisfied. Moreover, there is considerable similarity in the hydrogen-bonding pattern of the sugars bound in the three periplasmic sugar-binding proteins and in the catalytic site of phosphorylase. However, far less van der Waals contacts are formed on binding of glucose to the catalytic site of phosphorylase than on binding of monopyranoside ligands to ABP and GGBP. This is consistent with the fact that both ABP and GGBP (as well as MBP) have very high affiity for sugar ligands. Moreover, aromatic residues play a significant role in sugar binding to the three sugar-binding proteins, whereas none is involved in binding of monosaccharide in the catalytic site of phosphorylase. (Binding of oligosaccharide ligand in the catalytic site might involve a Tyr residue (Johnson et al., 1988).) Indeed, the oligosaccharide-binding site of MBP is unique in comparison to the sites in all other proteins that bind carbohydrates in its high content of aromatic residues. The site of MBP is also the only one known to bind very tightly both linear and cyclic maltodextrins.

The glycogen storage site of phosphorylase also binds longer

dextrins by filling sites at either end of the initially filled site for maltose. The fmst site added is at the nonreducing end of the disaccharide, the second at the reducing end of maltotriose (or maltose), and the third at the nonreducing end of maltotetraose (or maltotriose). The next site for binding succes- sively longer sugars in MBP is located at the nonreducing end of the shorter oligosaccharides. We have not seen the shift of the reducing sugar from site S1 for linear sugars, although data for maltohexaose and maltoheptaose are not yet available.

MBP does not bind D-glucose. The inability of D-glucose to bind in either subsites (S1 or S2), but especially in S2 where a glucosyl unit of maltose is involved in the formation of more hydrogen bonds and van der Waals contacts than in S1, may be due to the absence of hydrogen-bonding residues close to the glycosidic bond of maltose. The absence of these residues, especially hydrogen bond acceptors, would leave the donatable hydrogen of the C4 hydroxyl or C1 hydroxyl of D-glucose unpaired if the monosaccharide binds to either of the two subsites (Sl or S2). An unpaired protein in an inaccessible and low dielectric constant environment leads to instability, costing as much as 7 kcal/mol of energy (Jacobson and


Quiocho, 1988). It is also possible that glucose does not fully induce the closure of the cleft. Although the glycogen storage site of phosphorylase has four subsites for binding maltodextrins (maltose to maltoheptaose), the site does not bind D- glucose (Johnson et al., 1988). Interestingly, we note that, as in MBP, none of these subsites have a potential hydrogen bond accepting group near the glycosidic bonds.

Functional Regions of MBP and Properties of Binding Pro- teins-There are four major regions in the tertiary structure of binding proteins that are functionally important for transport and chemotaxis: I, ligand-binding site region responsible for ligand binding; 11, inter-domain region composed of sites which are involved in bringing about precise alignment of the two domains especially in the liganded, closed-cleft conformation. This region includes the hinge segments connecting the two domains and the interface between the two domains; 111, membrane transport region consisting of an ensemble of surface sites which interacts with the components located in the cytoplasmic membrane involved in actual translocation of nutrients; IV, signal transducer region made up of a collec- tion of surface sites which interact with the transmembrane signal transducers to initiate chemotaxis. (The residues resid- ing in the four regions can be similarly grouped.) Whereas all binding proteins have regions 1-111, only those which are involved in chemotaxis also contain region IV. The maltodextrin-binding protein may have fifth region, studies have indicated that MBP interacts with the lumB protein located in the cell wall or outer membrane (Bavoil et al., 1980). However, the importance of such an interaction, which has to be unique to the MBP system, has been questioned (Brass et al., 1985).

Determining the three-dimensional structures of binding proteins has enabled us for the first time to locate and understand the functions of all four regions. The structures, especialiy of the ligand-free open form and liganded closed form (e.g. see above for MBP) are sufficient to reveal detailed features of regions I and I1 (summarized in Quiocho, 1990). On the other hand, locating regions I11 and IV must also take into consideration results of genetic studies. Fortunately, there is a large amount of genetic data available on the maltodextrin system.

An important feature of binding proteins is that they have at least two distinct conformations, an open unliganded structure in which the two domains are far apart and the ligand- binding site cleft is easily accessible and a closed liganded structure in which the two domains (especially at the lips of the cleft) are in contact and, thus, enclose the bound ligand. A motion about the hinge between the two domains is the best mechanism to bring about the two different conformations and to precisely juxtapose the residues in both domains for ligand binding (Newcomer et al., 1981; Sack et al., 1989a). The interdomain contacts (via hydrogen bonds, van der Waals contacts, and salt links) help stabilize the two domains precisely in the liganded closed form (Pflugrath and Quiocho, 1988; Sack et al., 1989a). The residues associated with region 11, which are found in the hinge segments connecting the two domains and at the interface between the two domains, indi- rectly affect ligand binding.

Interactions between region I11 or IV and the recognition sites on the other membrane components of the binding protein system must allow differentiation between the liganded and ligand-free forms of the binding protein to prevent nonproductive or wasteful interactions. This differentiation is in fact necessary since there is a large excess of a specific binding protein relative to corresponding membrane components. A simple and elegant mechanism to accomplish this necessitates the involvement of regions I11 (for transport

activity) or region IV for chemotactic activity), each region encompassing parts or sites located in both domains of the binding protein (Quiocho et al., 1977; Newcomer et al., 1981). Since ligand binding induces closure of the cleft, these sites of a given region would be at different distances from each other, depending on whether ligand was bound or not. There- fore, if both sites are required for a productive interaction, the unwanted, ligand-free binding protein could be easily discriminated against. The structure of the ligand-free form of the LIVBP or LBP protein is an example of what the unliganded form might look like, with the domains wider apart at the opening or top of the cleft (Sack et al., 1989a, 1989b). The structures of the ligadloaded ABP, GGBP, MBP, SBP, and phosphate-binding protein typify a liganded closed form.

There are a large number of MBP mutants with varying effects on the binding of maltodextrins and the in vivo transport of maltose and higher maltodextrins and chemotaxis towards maltose (Duplay et al., 1987; Treptow and Shuman, 1985, 1988; Manson and Kossmann, 1986; Kossmann et al., 1988). (See Table VI for a partial listing of the MBP mutants which have been characterized by these groups.) With the structure on hand, it is now possible to identify the location of these mutations, and thereby assign sites belonging to regions I11 and IV, in the structure of the wild-type MBP.

The sites for recognition by chemotactic and transport components need not be the same and in fact there is evidence from mutations in MBP as well as GGBP (Hazelbauer and Adler, 1971; Vyas et al., 1988) that they are distinct from one another. The mutants of MBP and GGBP that are transport competent but show no chemotactic response help to localize those sites associated with region IV. The best genetic experiment along this line indicates two distinctly separate sites in MBP which delineate region IV (Kossmann et al., 1988). One site revealed by two mutations (malE454 and malE461) which interrupt chemotaxis occurs in the N-domain on the surface at the loop between helix I1 and strand C. This site is about 20 A from the bound maltose. The second interaction site, mapped by mutant malE469, occurs at the same end of the cleft, but in the C-domain acd near the lip of the cleft in helix XIII. This site is about 10 A from the bound maltose. These mutation sites offer direct evidence that at least one recognition site in each domain of MBP is required for productive interaction of MBP with the Tar transducer, which is involved in triggering chemotaxis. The locations of these mutants are highlighted in Fig. 10. This finding does not rule out a tem- plate recognition scheme, where the recognitions site is an extended portion connecting both domains.

Countermutations were found in Tar to the MalE mutations at positions 53 and 55 that restored at a very low level (-10% of wild-type response) maltose chemotaxis (Kossmann et al., 1988). The mutations of MBP which they counter are from charged to neutral (D55N) or hydrogen bonding to hydrophobic (T53I). All the Tar mutations involved an Arg in the native Tar protein being mutated to a neutral species R73W, R64C, and R69C). As such the counter mutations can be seen as complementary surface changes to the MBP mutations which might allow the other chemotactic interaction site(s) to initiate, albeit poorly, chemotaxis.

MalE469 (T3451) as well as being nonchemotactic displayed interesting fluorescence properties, showing an increase in fluorescence upon binding maltose (Kossmann et al., 1988). MBP mutant WA158 (Martineau et al., 1990) was also im- paired in chemotaxis to maltose and showed no fluorescence shift upon binding maltose. The chemotactic disfunction of WA158 can be attributed to a regional deformation of the

2.3-A Structure of the Perblasmic Maltodextrin Receptor 5217 TABLE VI

MBP mutants The MBP mutants below have been divided into representative groups. The first group is suppressor mutants

that offer some evidence for the location of interactions with MalF or MalG, transport membrane components. . The second consists of point mutations that solely effect chemotaxis. The thud set does not transport maltohexaose.

The final set does not transport any maltodextrins. A more detailed listing of MBP mutants can be found in the references cited. wt, wild type.

KllI Response Chemotaxis Izd Mutant maltme todextrin' maltose (IOONM) maltose Location Ref.

100% 86% 95% 0 . 8 ~ ~ 0.8pM 0.5pM 0.5pM

% P M

wt 0.9pM 100% 100 641 10% 80 52' AlamGlu 642 100% 80 1.6b Ty?'OSer 623 40 3.2b Asp1% 632 <lo 11.5' Glyr3Asp 454 + 4 ThrS311e' 461 + 4 AspsAsnC 469 + 13 Th34snec 14oP 4 (370)+38 140c - 88 4 (370)+27 333 - 29 2.3 (297-312)+TDP 91 - 6 3 (364-370)+11

RN3d 301 - 107' 2.3 (212-220)+GS 346 - 268' 6 (207-216)+RIR 120 >lo0 (330-370)+TDP 183 >lo0 (340-357)+SDP 220 >lo0 (341-370)+YGSA

3.5

-

247

Treptow and Shuman, 1988 Treptow and Shuman, 1988 Treptow and Shuman, 1988 Treptow and Shuman, 1988 Manson and Kossmann, 1986 Manson and Kossmann, 1986 Manson and Kossmann, 1986 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987 Duplay et al., 1987

' Number indicates relative Kd for maltohexaose, + or - indicates growth on maltohexaose. ' Adjusted to achieve native Kd agreement (9.4 pM using dialysis). Location from Kossmann et al., 1988. RN3 is a strain containing a mutated MalF protein which blocks transport Duplay et al., 1987.

e As compared to mutant RN3 (Duplay and Szmelcman, 1987).

I

FIG. 10. Stereo space-fiiling representation of MBP showing sites for interacting with membrane transport (region 111) and chemotactic (region IV) components. Region III: residues 207-220, located in the C-domain (right side of picture), are shown in gold and are clearly distinct from the chemotactic sites. These residues represent the site of two deletion mutations (207-216 and 212-220) that do not transport maltodextrins. Residues 13 and 14, the sites of two suppressor mutations located in the N-domain, are also highlighted in gold. These residues present a proximal position for possible interactions with MalG. Region Iv: Residues 53, 55, and 345, shown in green, that are involved in interacting with the transmembrane signal transduced for chemotaxis. Point mutations of these residues produce MBP that is unable to initiate chemotaxis toward maltodextrins. Residues 53 and 55 are in the N-domain (left side of picture), while residue 345 is in the C-domain. They present clear evidence for recognition of "ligand loaded" binding proteins via a two-domain differential separation scheme.

surface features of MBP in the vicinity of residue 345 (helix XIII) caused by the loss of the bulky tryptophan side chain which makes many van der Waals contacts with helix XIII. There are 3 arginine residues clustered around residue 345, one of which (Ar$44) is involved in binding maltotriose. These arginines represent possible interaction sites which are near ThrM and present an easily recognized pattern. A structure feature of MBP which can clearly be seen in Fig. 10.

Several mutations have revealed sites belonging to region 111, which are involved in the interaction with the membrane components of maltodextrin transport. Again these sites are located in both domains. M a 3 0 1 (deletion of residues 212- 220 and insertion of Gly-Ser) and M a 3 4 6 (deletion of residues 207-216 and insertion of Arg-Ile-Arg) (Duplay et d, 1987) are two such mutant proteins. Purified MaLE301 protein binds maltose with a normal &, and the mutant strain


displays wild-type chemotactic response to maltose. The Kd and chemotactic response are slightly perturbed in MalE346. The region affected by these deletions is helix VI1 which is located in the C-domain (see Figs. 2, 3, and 10). The normal chemotactic response and Kd suggests that lack of transport is due to a faulty interaction with MalFG transport components. This could be a direct interaction with some portion of helix VI1 or with some portion of the C-domain ( ie . strand I) that is obscured by the rearrangement caused by the deletion of helix VII. A class I (binds maltose and maltohexaose and restores maltodextrin transport) suppressor mutant 642 (Y210S (Treptow and Shuman, 1988)) was also found in this region.

While a site that selectively affects transport has not been found in the N-domain, the region around helix I and strand B is a good possibility for its location. Two allele-specific MalE suppressor mutations 623 (D14Y) and 632 (G13D) were found in this region, evidence of possible interactions between MBP and the MalG protein (the mutant protein which they were selected on) (Treptow and Shuman, 1988). The locations of these 2 residues is also shown in Fig. 10.

Mutations of critical residues located in regions I and I1 will for the most part affect both chemotactic and transport processes. The mutation of Trp340 to an alanine results in a protein (WA340) unable to bind maltose ( K d > 1 mM) (Mar- tineau et al., 1990) and demonstrates the importance of Trp340. Other mutations implicate Trp340 as crucial for the binding of maltose and higher dextrins. Deletion of the two terminal helices (XI11 and XIV) of MBP results in protein unable to transport maltose (120, 183, 220 (Duplay et aL, 1987)). The Kd for maltose is greater than 100 PM for these mutants. Removal of Trp340 from the binding site could account for this loss of binding activity.

Other point mutations of tryptophan to alanine residues produced proteins with properties in agreement with the location of the tryptophan residues in the native structure. The loss of Trp6* (WA62, Martineau et al., 1990) is reflected in the 10-fold increase in Kd seen for this mutant. The effect of replacing TrpZ3O (5-fold increase in &) with an alanine (Martineau et al., 1990) is in agreement with its place in the binding site (see sugar binding above and Table IV).

Addition of residues to the C-terminal end of MBP resulted in mutants (MalE140p and MalE140c) that were dextrin minus, but otherwise normal. The extra residues introduced in these mutants could interfere with the binding of longer dextrins, which crystallographic evidence indicates would be bound with the additional sugar units near the terminal helices. Substitution of the last 6 residues of the C-terminal helix with 11 residues produced a protein (MalE91) able to transport maltose, but unable to transport longer dextrins or initiate chemotaxis. This mutation, as well as interfering with the binding of longer dextrins through effects on the terminal helices, is near residue 345, already implicated in interactions with the transducer. The nonchemotactic nature of this mutant is probably a result of interference with the interaction between MBP and the tar transducer near residue 345.

The four suppressor mutants (from Treptow and Shuman, 1985) (641, 642, 623, and 632) all map in the binding pocket of MBP. Mutant 641 could introduce severe steric hindrance thereby causing the 10-fold increase in Kd seen. Mutants 623 and 632 would also introduce steric hindrance due to the proximity of both to the bound sugar. Mutant 642 while not situated in the binding pocket is located directly above it. How these mutations restore dextrin transport is not readily apparent at this time. Further characterization and structural examination of these mutants could provide information on

the mechanism for transfer of maltodextrins from MBP to the MalFG complex.

Mutations which have no effect on MBP activity can also be informative about interactions between MBP and other protein components of transport and chemotaxis by deline- ating areas not involved in these interactions. MalE178, one such mutant protein, displays almost normal transport activity indicating that helix IV, which is at one end of the elongated structure, is not necessary for transport. Several epitopes have been successfully inserted at this deletion site without effect on the activity of MBP (Duplay et al., 1987). The exterior location of this deletion coupled with the extended loop connecting it to the rest of the structure allows the insertion of a wide variety of epitopes without any serious consequences for MBP activity of folding. The facile inser- tions of epitopes to MBP, the ease of purification of large quantity of the protein, and the fact that the MBP structure has been determined offer the unique opportunity of determining the tertiary structure of epitopes, especially those of proteins with unknown structures (Rodseth et al., 1990).

The structure of MBP further solidifies many of the common features of the structures, recognition of ligands, and function of the periplasmic receptors of active transport and chemotaxis (Pflugrath and Quiocho, 1988; Sack et al., 1989a; Quiocho, 1990). In view of the finding that, although the binding proteins vary in size (most have molecular weights of about 33,000) and sequence and have different ligands, they have common structural and functional features, it is realistic to assume that the membrane components also have common features, some of which might closely mimic those clearly established in the binding protein. For instance, the finding that the specificities, although diverse, and affinities of the binding proteins are achieved principally through hydrogen bonding and van der Waals interactions, it is possible that similar interactions could determine or modulate the affinity, as well as specificity, of the ligand-binding site(s) in the membrane-bound components.

Acknowledgments-For helpful discussions, assistance, samples, and more discussions, we thank Tim Reynolds, Lynn Rodseth, Jack Sack, and Nand Vyas.

REFERENCES

Ames, G. F.-L. (1986) Annu. Reu. Biochem. 5 5 , 397-425 Bavoil, P., Hofnung, M., and Nikaido, H. (1980) J. Bwl. Chem. 266,

Blow, D. M., and Crick, F. H. C. (1959) Acta Crystallogr. 14, 1195-

Blundell, T. L., and Johnson, L. N. (1976) Protein Crystallography,

Brass, J. M., Bauer, K., Ehmann, U., and Boos, W. (1985) J. Bacteriol.

Dassa, E., and Hofnung M. (1985) EMBO J. 4 , 2287-2293 Duplay, P., Bedouelle, H., Fowler, A., Zabin, I., Saurin, W., and

Hofnung, M. (1984) J. Biol. Chem. 259,10606-10613 Duplay, P., Szmelcman, S., Bedouelle, H., and Hofnung, M. (1987) J.

Falke, J. J., and Koshland, D. E., Jr. (1987) Science 237, 1596-1600 Ferenci, T., and Klotz, U. (1978) FEBS Lett. 9 4 , 213-217 Finkelstein, A. V. (1975) Doklady Akcad Nauk SSSR 223,744-747 Froshauer, S., and Beckwith, J. (1984) J. Bid. Chem. 259, 10896-

10903 Furlong, C. E. (1987) in Escherichia coli and Salmonella typhimirium:

Cellulur and Molecular Biology (Neidhardt, F. C., Ingraham, J. L., Low, K. B., Magasanik, B., Schaechter, M., and Umbarger, H. E., eds) pp. 768-796, American Society for Microbiology, Washington, D. C.

Gilliland, G. L., and Quiocho, F. A. (1981) J. Mol. Biol. 146, 341- 362

Gilson, E., Higgins, C. F., Hofnung, M., Ames, G. F-L., and Nikaido, H. (1982). J. Biol. Chem. 257,9915-9918

8366-8369

1202

Academic Press, New York

161,720-726

Mol. Bid. 194 , 663-673


Gilson, E., Alliong, G., Schmidt, T., Claverys, J-P., Dudler, R., and Hofnung, M. (1988) EMBO J. 7, 3971-3974

Hammes, G. G., and Schimmel, P. R. (1970) in The Enzymes (Boyer, P. D., ed) Vol. 11, pp. 67-115, Academic Press, New York

Hazelbauer, G. L. (1975) J. Bacteriol. 122,206-214 Hazelbauer, G. L., and Adler, J. A. (1971) Nature 230, 101 Hendrickson, W. A., and Konnert, J . H. (1980) in Computing in

Crystallography (Diamond, R., Ramaseshan, S., and Venkatesan, K., eds) pp. 13.01-13.23, Indian Academy of Sciences, Bangalore

Higgins, C . F., Hiles, I. D., Whalley, K., and Jamieson, D. J . (1985) EMBO J. 4,1033-1040

Higgins, C. F., Hiles, I. D., Salmond, G. P. C . , Gill, D. R., Downie, J. A., Evans, I. J., Holland, I. B., Gray, L., Buckel, S. D., Bell, A. W., and Hermodson, M. A. (1986) Nature 323,448-450

Hyde, S . C., Emsley, P., Hartshorn, M. J., Mimmack, M. M., Gileadi, U., Pearce, S. R., Gallagher, M. P., Gill, D. R., Hubbard, R. E., and Higgins, C. F. (1990) Nature 346 , 362-365

Jacobson, B. L., and Quiocho, F. A. (1988) J. Mol. Biol. 204, 783- 787

Johnson, L. N., Cheetham, J., McLaughlin, P. J., Acharya, K. R., Barford, D., and Phillips, D. C. (1988) Curr. Top. Microbiol. Zm- munol. 139, 81-134

Kossmann, M., Wolff, C., and Manson, M. D. (1988) J. Bacteriol.

Lee, B., and Richards, F. M. (1971) J. Mol. Biol. 5 5 , 379-400 Levitt, M., and Perutz, M. (1988) J. Mol. Biol. 201, 751-754 Luecke, H., and Quiocho, F. A. (1990) Nature 347,402-406 Lunin, V. Y., Urzhumtsev, A. G., Vernoslavia, E. A., Chirgardze, N.

A., Nevskaya, N. A,, and Fomenkova, N. P. (1985) Acta Crystallogr. A41,166-171

Macnab, R. (1987) in Escherichia coli and Salmonella typhimirium: Cellular and Molecular Biology (Neidhardt, F. C., Ingraham, J . L., Low, K. B., Magasanik, B., Schaechter, M., and Umbarger, H. E., eds) pp. 732-759, American Society for Microbiology, Washington, D. C.

170,4516-4521

Manson, M. D., and Kossmann, M. (1986) J. Bacteriol. 165 , 34-40 Mao, B., Pear, M. R., McCammon, J. A., and Quiocho, F. A. (1982)

Martineau, P., Szmelcman, S., Spurlino, J. C., Quiocho, F. A., and

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

Miller, D. M., 111, Olson, J. S., Pflugrath, J. W., and Quiocho, F. A.

Neu, H. C., and Heppel, L. A. (1965) J. Biol. Chem. 240 , 3685-3692 Newcomer, M. E., Lewis, B. A., and Quiocho, F. A. (1981) J. Biol.

Parkinson, J. S., Parker, S. R., Talbert, P. B., and Houts, S . E. (1983)

Pflugrath, J. W., and Quiocho, F. A. (1985) Nature 314,257-260 Pflugrath, J. W., and Quiocho, F. A. (1988) J. Mol. Biol. 200, 163-

J. Biol. Chem. 257 , 1131-1133

Hofnung, M. (1990) J. Mol. Biol. 214 , 337-352

Chem. 255,2465-2471

(1983) J. Biol. Chem. 2 5 8 , 13665-13672

Chem. 2 5 6 , 13218-13222

J. Bacteriol. 155 , 265-274

180

Podjamy, A. D., Bhat, T. N., and Zwick, M. (1987) Annu. Reu.

Priestle, J. P. (1988) J. Appl. Cryst. 2 1 , 572-576 Ptitsyn, 0. B., and Finkelstein, A. V. (1983) Biopolymers 22, 15-25 Quiocho, F. A. (1986) Annu. Reu. Biochem. 55, 287-315 Quiocho, F. A. (1989) Pure Appl. Chem. 6 1 , 1293-1306 Quiocho, F. A. (1990) Phil. Trans. R. SOC. Lond. B 326, 341-351 Quiocho, F. A,, and Vyas, N. K. (1984) Nature 310,381-386 Quiocho, F. A,, Gilliland, G. L., and Phillips, G. N., Jr. (1977) J. Biol.

Quiocho, F. A., Meador, W. E., and Pflugrath, J. W. (1979) J. Mol.

Quiocho, F. A,, Wilson, D. K., and Vyas, N. K. (1989) Nature 340 ,

Rodseth, L., Martineau, P., Duplay, P., Hofnung, M., and Quiocho,

Rossmann, M. G., and Argos, P. (1977) J. Mol. Biol. 109, 99-129 Sack, J. S., Saper, M. A., and Quiocho, F. A. (1989a) J. Mol. Biol.

Sack, J. S., Trakhanov, S. D., Tsigannik, I. H., and Quiocho, F. A.

Saenger, W. (1983) Biochem. SOC. Trans. 11, 136-139 Saper, M. A., and Quiocho, F. A. (1983) J. Biol. Chem. 258 , 11057-

Shevitz, R. W., Podjarny, A. D., Zwick, M., Hughes, J. J., and Sigler,

Shuman, H. A. (1982) J. Biol. Chem. 257,5455-5461 Shuman, H. A., and Silhavy, T. J. (1981) J. Biol. Chem. 256, 560-

Smoluchowski, M. V. (1918) 2. Phys. Chem. 9 2 , 129-168 Spurlino, J . C. (1988) The Three-dimensional Structure at the D-

Maltose-binding Protein from Escherichia Coli. Ph.D. dissertation, Rice University, Houston, TX

Stock, A. M., Mottonnen, J. M., Stock, J. B., and Schutt, C. E. (1989) Nature 337,745-749

Szmelcman, S., Schwartz, M., Silhavy, T. J., and Boos, W. (1976) Eur. J . Biochem. 65,13-19

Takusagawa, F., and Jacobson, R. (1978) Acta Cryst. B34 , 213-218 Thieme, R., Lay, H., Oser, A., Lehmann, J., Wrissenberg, S., and

Tolley, S. P., Derewenda, Z., Hyde, S. C., Higgins, C. F., and Wilk-

Treptow, N. A., and Shuman, H. A. (1985) J. Bacteriol. 163, 654-

Treptow, N. A., and Shuman, H. A. (1988) J. Mol. Biol. 202 , 809-

Vyas, N. K., Vyas, M. N., and Quiocho, F. A. (1983) Proc. Natl. Acad.

Vyas, N. K., Vyas, M. N., and Quiocho, F. A. (1987) Nature 327 ,

Vyas, N. K., Vyas, M. N., and Quiocho, F. A. (1988) Science 242 ,

Biophys. Chem. 16 , 351-373

Chem. 252,5142-5149

Biol. 133,181-184

404-407

F. (1990) J. Mol. Biol. 2 1 3 , 607-611

2 0 6 , 171-191

(1989b) J. Mol. Biol. 206, 193-207

11062

P. B. (1981) Acta Crystallogr. A37,669-677

562

Boos, W. (1986) Eur. J . Biochem. 160,83-91

inson, A. (1988) J. Mol. Biol. 204, 493-494

660

822

Acad. Sci. U. S. A. 80, 1792-1796

635-638

129n-1295 Vyas, M. N., Jacobson, B. L., and Quiocho, F. A. (1989) J. Biol. Chem.

~~.~

264,20817-20821 Pflugrath, J. W., Saper, M. A, and Quiocho, F. A. (1984) in Compu- tational Crystallography (Hall, S. R., and Ashida, T. eds) pp. 404- Vyas, N. K., Vyas, M. N., and Quiocho, F. A. (1991) J. Bid. Chem, 411, Clarendon Press, Oxford 266,5226-5237

or maltodextrin- binding protein, a primary receptor of bacterial

Documents