Leucine, Isoleucine, Valine-binding Protein from Escherichia coli STRUCTURE AT 3.0-A RESOLUTION AND LOCATION OF THE BINDING SITE*

The structure of the leucine, isoleucine, valine-bind-ing protein, an integral part of the high-affinity, branched-chain aliphatic amino acid transPo@ system in Escherichia coli, has been solved at 3.0-A resolution by x-ray crystallography. Five isomorphous heavy atom derivatives, including anomalous differences from a samarium derivative, were used. A model of the polypeptide chain backbone reveals two distinct, globular domains connected by three strands. Each domain consists of a &sheet core flanked by at least two helices on either side. Difference Fourier analyses of crystals soaked in L-leucine, L-isoleucine, or L-valine have located a single amino acid-binding site in the cleft formed by the two domains. Despite the lack of signif- icant sequence homology, the bilobate and secondary structure observed were similar to that found in the structures of L-arabinose- and D-galactose-binding proteins previously determined in our laboratory. and protein at resolution and the D-galactose-binding protein 3.0 been previously determined. A of

The structure of the leucine, isoleucine, valine-binding protein, an integral part of the high-affinity, branched-chain aliphatic amino acid transPo@ system in Escherichia coli, has been solved at 3.0-A resolution by x-ray crystallography. Five isomorphous heavy atom derivatives, including anomalous differences from a samarium derivative, were used. A model of the polypeptide chain backbone reveals two distinct, globular domains connected by three strands. Each domain consists of a &sheet core flanked by at least two helices on either side. Difference Fourier analyses of crystals soaked in L-leucine, L-isoleucine, or L-valine have located a single amino acid-binding site in the cleft formed by the two domains. Despite the lack of significant sequence homology, the bilobate and secondary structure observed were similar to that found in the structures of L-arabinose-and D-galactose-binding proteins previously determined in our laboratory.
Periplasmic binding proteins from Gram-negative bacteria are essential components in the high affinity, osmotic shocksensitive transport systems for carbohydrates, amino acids, and ions (1). Some of these proteins also function as receptors in chemotaxis (2). Of the five binding proteins crystallized in our laboratory (3), the remarkablyosimilar structures of the Larabinose-binding protein at 2.4-A resolution (4) and the Dgalactose-binding protein at 3.0 A (5) have been previously determined. A unified understanding of binding protein function requires the comparison of these sugar-binding structures with others differing in substrate specificity.
Osmotic shock-sensitive, high affinity transport of branched-chain aliphatic amino acids by Escherichia coli utilizes the leucine, isoleucine, valine-binding protein (6). The purified protein binds one molecule of substrate with a Kd of 2 x M (7,8). The amino acid sequence (9) indicates a molecular yeight of 36,770 for 344 residues.
The 3.0-A resolution structure of the leucine, isoleucine, valine-binding protein reported here adds an amino acid receptor to the family of transport protein structures determined in our laboratory. Furthermore, difference Fourier analysis to locate the amino acid-binding site in LIV-BP' is * This investigation was supported by Grants GM-26485 and GM-21371 from the National Institutes of Health and Grant C-581 from the Robert A. Welch Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. presented. Preliminary accounts of this work have been reported (10, 11). In order to crystallize the protein in a form originally obtained by Meador and Quiocho (14), it was necessary to remove endogenously bound amino acid (8) by a rapid, yet mild and effective procedure originally developed in this laboratory for L-arabinose-and D-galactose-binding proteins (15,16). To monitor the removal of bound substrate from LIV-BP, 5 p1 of stock [''C]leucine (100 pCi/ml) was added to 5 ml of LIV-BP (3.6 mg/ml) in 50 mM NaCl, 0.02% NaN3, 10 mM Tris-HC1, pH 7.6. The protein was dialyzed against 1 M guanidine HC1,lO mM Tris-HCl, pH 7.6, for 24 h a t 4 "C. Periodically, protein and dialysate samples were removed and counted to monitor the loss of ["Clleucine from the protein. The protein was exhaustively dialyzed against 10 mM Tris-HC1, pH 7.6, and chromatographed onto a 2 X 10 cm Bio-Gel P2 (100-200 mesh) column equilibrated with the same buffer. Less than 0.1% of [14C]leucine, initially added to the protein, remained after this treatment.
A small, well formed crystal (<0.25 mm) could be grown larger by transferring it to a clean cover slip and rinsing it with 10% polyethylene glycol 6000, 50 mM NaC1, 0.02% NaN,, 5 mM Tris-HC1, 5 mM sodium citrate, pH 4.5, to dissolve micro-nucleation sites. After totally removing the wash solution, a drop consisting of 17 pl of processed protein and 13 p1 of precipitant was layered over the crystal and equilibrated uersus 18% polyethylene glycol 6000 as before. A single crystal as large as 1.5 mm routinely grew from a 30-p1 drop.
The polyhedral crystals of substrate-free LIV-BP, which diffract to at least 2-A resolution, belong to the space group P212121 with one molecule/asymmetric unit (14). Unit cell dimensions are shown in Table I.
* Based on intensity.
includes Friedel pairs. diffractometer with CuK, radiation, using a seven-step, 0.2" w-scan. The data were corrected for Lorentz and polarization effects, deterioration, and absorption by normal procedures. Data collection statistics for data used in the structure solution are presented in Table I.

RESULTS AND DISCUSSION
Harker sections from difference Patterson maps and threedimensional difference Fourier maps located the major heavy atom derivative sites for an initial multiple isomorphous replacement refinement at 4.1-A resolution (figure of merit = 0.76)." Sites for the samarium derivative were confirmed in anomalous and combined difference Patterson maps (17). A native data scale factor of 41.5 was determined by scaling the LIV-BP data to the native data of the L-arpbinose-binding protein, whose scale had been refined at 2.4-A resolution (4). The scale factor was later changed to 34.0, reducing the refined occupancy of the major samarium site to approximately 100%.
Extension of the data to 3.0-A resolution, followed by cycles of parameter refinement and phasing with the Sm anomalous differences (root mean square AF/F = 0.065 for the acentric reflections), gave a mean figure of mer$ of 0.68 for 6653 reflections. A summary of the final 3.0-A multiple isomorphous replacement phasing statistics is shown in Table 11.
"Best" Fourier maps, which were calculated from the 3.0-A resolution phases and contoured along the x and z crystallographic axes, revealed the intermolecular boundary (Fig. 1). The molecule clearly consists of two distinct, globular domains. Since the domains are tightly packed in the unit cell, each domain has close contacts with at least four neighboring, complementary domains from symmetry-related molecules.
Contoured Fourier maps were later examined on the new generation Evans & Sutherland PS 300 computer graphics system. An initial a-carbon trace made at 3. 2   by fitting a polyalanine structure to the 3.0-A map using FRODO, an implementation of Dr. T. A. Jones' program (19) for the PS 300.4 Fitting of the side chains for the known amino acid sequence of LIV-BP (9) is now in progress. The LIV-BP mol$cule is ellipsoidal, with overall dimensions 40 X 35 X 70 A, and consists of two globular domains connected by three polypeptide segments (Figs. 2 and 3a). Though lacking sequence similarity, each of the two domains has essentially the same secondary structure arrangement: a central, five-strand parallel @-sheet flanked by two major ahelices on either side running antiparallel to the sheet. The secondary structure is composed of 40% a-helix and 30% psheet arranged in the pap-folding units commonly seen in other proteins (20). The p-strands from the sheet in each domain run NH2 to COOH towards the wide cleft between the domains and exhibit the characteristic left-handed propeller twist: the NH, domain (Fig. 2, top) sheet swirls about 80" while the COOH domain (Fig. 2, bottom)  with a supersecondary fold equivalent to the NH2 domain. The polypeptide wanders into the solvent to create a large interdomain loop with a short helix. Parts of this loop are washed out in the map, indicative of a less rigid or disordered structure. The chain then returns to the NH2 domain with a long helix, curves to the bottom of the domain forming another short helix, and then meanders with a short two-strand, antiparallel, sheet-like structure before crossing back to the COOH-domain for the final 2 antiparallel @-strands.
Despite little sequence homology (21, 22), the secondary structure packing in each of the LIV-BP domains is very similar to that found in the corresponding domains of the bilobate ABP (Fig. 3b) and D-galactose-binding protein. Seventy-three per cent (100 C, s) of the NH,-terminal domain of ABP was found to be equivalent to the correspondiqg domain of LIV-BP (root mean square distance = 2.37 A) by the structure orientation technique of Rossmann and Argos (23) using the program OVRLAP by W. Bennett. Similarly, 112 a-carbons of the COOH-terminal domain of ABP (also 73%) were superimposed onto the corresponding atoms in the COOH-terminalodomain of LIV-BP with a root mean square distance of 2.81 A. Unusually high equivalence (80%) was also obtained in a comparison between the polypeptide backbone structures of ABP and D-galactose-binding protein (5).
Since the LIV-BP crystals used in the present structural analysis are devoid of bound leucine (see above), difference Fourier analysis was employed to locate the amino acid binding site. Diffraction data were collected from crystals soaked in solutions of 50 mM L-leucine, L-isoleucine, or L-valine (Table I). Each difference map showed a significant peak (-1.7 times the second highest peak) at the same location (Fig. 4). The proposed binding site is located in the cleft between the two domains but confined only to the wall of the NHn-terminal domain near two loops that connect @-sheets with helices (Fig. 3a). A model of the appropriate amino acid substrate was fit in the difference electron density. The identification of the amino acid residues involved in binding must await completion of the model fitting.
The structure determination of the leucine, isoleucine, valine-binding protein, together with the recent elucidation of a A resolution: brings to a total of four the binding protein structures thus far determined in our laboratory. An overall structural feature of this family has emerged all four receptors, including those specific for L-arabinose, D-galactose, and sulfate, are elongated by the presence of two lobes with similar secondary structure arrangement (4, 5 ) . Furthermore, the substrate binding sites of ABP (24), D-galactose-binding protein (5), sulfate-binding protein: and LIV-BP are located in a cleft formed between the two domains. Therefore, it is likely that these structural features will be preserved in other binding proteins. This will be particularly true for the leucinespecific binding protein, another protein involved in high affinity leucine transport, that shows at least 80% sequence homology with LIV-BP (12). Why do binding proteins have a bilobate structure? It appears to be important in the substrate-induced conformational change of the L-arabinose-binding protein Low angle x-ray scattering studies in solution show a 1-A decrease in the radius of gyration of ABP upon binding of L-arabinose (25). Such a significant change, indicative of a more compact, 5J. W. Pflugrath and F. A. Quiocho, in preparation. liganded structure, is consistent with the closing of a binding cleft by a hinge-like motion of the two domains. Only modest changes in the protein's internal energy are necessary to open and close the cleft by a flexible hinge mechanism (26). Local conformational changes, although not precluded, are insufficient to account for the difference in radius of gyration. Furthermore, the structure of ABP, which was solved with bound L-arabinose, displays a closed cleft with the sugar extensively liganded to residues from opposite walls of both domains and completely inaccessible to the solvent (27) (see Fig. 3b). The substrate is trapped within the cleft and the lobes must separate to allow the sugar to be translocated across the membrane.
The crystal structure of substrate-free LIV-BP has a very wide binding cleft, analogous to the proposed model for sugarfree ABP in solution. Upon binding of amino acid, the cleft could conceivably close by a relative twisting of the two domains around the three-strand hinge region. Nevertheless  (28).
Narrowing of the cleft may be hindered in LIV-BP because of the close packing of the molecules in the crystal (see Fig.  1). A similar reason was cited as the probable cause for the inactivity of citrate synthase in the crystalline state (31). Consistent with this, crystals grown from native, liganded LIV-BP and from substrate-free LIV-BP preincubated with amino acid are similar, but both are morphologically different from those used in this study and may reflect a substrateinduced conformational change.
In high affinity transport, periplasmic binding proteins interact with a membrane aggregate composed of at least three protein components (32)(33)(34)(35). TWO of these components are largely confined in the cytoplasmic membrane whereas the third is probably associated peripherally (32,33). Since it is present in only small amounts (32), the aggregate recognizes the liganded form of any one of a group of related binding proteins. Therefore, it is the binding protein that confers substrate specificity to the transport system by forming a tight complex with a specific substrate. A substrate-induced 3.0-A Structure of Leucine, Isoleucine, Valine-binding Protein conformational change, perhaps the closing of the binding protein cleft (see above), exposes and positions a recognition site that interacts with the membrane aggregate and initiates transport. A plausible complex formation between the membrane aggregate and the periplasmic receptor is suggested by the elongated and bilobate binding protein structures. If one assumes a "pore" in the membrane aggregate, the binding protein could span across the opening of the pore, each lobe binding to an exposed part of a transmembrane protein.
In such an arrangement, the binding site would be positioned directly over the pore, allowing efficient translocation of the substrate.