Solution Structure of the Phosphoryl Transfer Complex between the Cytoplasmic A Domain of the Mannitol Transporter IIMannitol and HPr of the Escherichia coliPhosphotransferase System*

The solution structure of the complex between the cytoplasmic A domain (IIAMtl) of the mannitol transporter IIMannitol and the histidine-containing phosphocarrier protein (HPr) of theEscherichia coli phosphotransferase system has been solved by NMR, including the use of conjoined rigid body/torsion angle dynamics, and residual dipolar couplings, coupled with cross-validation, to permit accurate orientation of the two proteins. A convex surface on HPr, formed by helices 1 and 2, interacts with a complementary concave depression on the surface of IIAMtlformed by helix 3, portions of helices 2 and 4, and β-strands 2 and 3. The majority of intermolecular contacts are hydrophobic, with a small number of electrostatic interactions at the periphery of the interface. The active site histidines, His-15 of HPr and His-65 of IIAMtl, are in close spatial proximity, and a pentacoordinate phosphoryl transition state can be readily accommodated with no change in protein-protein orientation and only minimal perturbations of the backbone immediately adjacent to the histidines. Comparison with two previously solved structures of complexes of HPr with partner proteins of the phosphotransferase system, the N-terminal domain of enzyme I (EIN) and enzyme IIAGlucose(IIAGlc), reveals a number of common features despite the fact that EIN, IIAGlc, and IIAMtl bear no structural resemblance to one another. Thus, entirely different underlying structural elements can form binding surfaces for HPr that are similar in terms of both shape and residue composition. These structural comparisons illustrate the roles of surface and residue complementarity, redundancy, incremental build-up of specificity and conformational side chain plasticity in the formation of transient specific protein-protein complexes in signal transduction pathways.

The bacterial phosphoenolpyruvate:sugar phosphotransferase system (1) is a classical example of a signal transduction pathway whereby the transfer of a phosphoryl group through a series of bimolecular protein-protein complexes is coupled with the transport of sugars across the membrane (2)(3)(4). The initial events of the cascade involve a common pathway: enzyme I, which is autophosphorylated by phosphoenolpyruvate at His-189 (in Escherichia coli), transfers the phosphoryl group to His-15 (in E. coli) of the histidine-containing phosphocarrier protein (HPr). 1 Subsequently, the phosphoryl group on HPr is transferred to a variety of sugar-specific carbohydrate transporters, known as enzymes II (5). The enzymes II are organized into several domains, some of which are covalently linked (5). There are two cytoplasmic domains, IIA and IIB; IIA accepts the phosphoryl group from HPr and then transfers it to IIB. The transmembrane domain IIC (and in some cases IID as well) catalyzes the translocation and phosphoryl transfer from IIB to the incoming sugar. There are five classes of enzymes II as follows: glucose-sucrose, mannitol-fructose, mannose-sorbose, lactose-cellobiose, and glucitol (3,5). The membranebound IIC domains comprise a variable number of transmembrane helices (5). The IIA and IIB domains for the different sugar classes bear no sequence or structural similarity to one another (6 -15).
The bacterial phosphoenolpyruvate:sugar phosphotransferase system provides a paradigm for understanding proteinprotein interactions and the factors governing their specificity. Thus, for example, HPr recognizes enzyme I and the various sugar-specific IIA domains, although all these target proteins are structurally dissimilar. We have recently solved the structures of the complexes between the N-terminal phosphoryl transfer domain of EI (EIN) and HPr (16) and between IIA Glucose (IIA Glc ) and HPr (17). In the present paper, we extend these studies to the solution structure determination of the complex between IIA Mannitol (IIA Mtl ) and HPr. (18). The expressed product from the recombinant plasmid (pLP1572) was identical in sequence to that from the clone described by Van Weeghel et al. (19).
Protein Expression-The recombinant plasmids, pSP100 for HPr expression (20) and pLP1572 for IIA Mtl expression, were introduced into E. coli GI698 for protein expression induced by tryptophan (21). Four-liter cultures were grown in minimal labeling medium (MLM) (22) supplemented with ampicillin (100 g/ml). For labeling with [ 13 C 6 ]glucose and/or 15 NH 4 Cl, the media contained 2.5 gm of [ 13 C 6 ]glucose/liter and/or 1 g of 15 NH 4 Cl/liter, respectively. Induction was continued overnight.
The 148 residue IIA Mtl domain, containing a methionine residue in front of alanine 491, was purified as follows. Washed cells from a 4-liter culture were suspended in 40 ml of 10 mM Tris-HCl (pH 7.5), 1 mM EDTA (Buffer A). The cell suspension was passed twice through a French press at 10,000 pounds/square inch. The ruptured cells were centrifuged at 100,000 ϫ g for 1 h. The supernatant fraction was applied to a DE52 anion-exchange column (1.5 ϫ 40 cm) equilibrated with 20 mM Tris-HCl (pH 8.5), 1 mM dithiothreitol (Buffer E). The column was washed with 100 ml of Buffer E supplemented with 100 mM NaCl (Buffer EE). Then a linear gradient (from 500 ml of Buffer EE to 500 ml of Buffer E ϩ 400 mM NaCl (Buffer F)) was run. Column fractions were screened by SDS-PAGE. Fractions enriched in IIA Mtl were pooled and concentrated by ultrafiltration through Filtron 3K membranes. Further purification was achieved by gel filtration chromatography on an AcA-44 Ultrogel column (2.6 ϫ 100 cm) run with 10 mM Tris-HCl (pH 7.5) ϩ 100 mM NaCl. Fractions enriched in IIA Mtl were pooled and concentrated as before. All protein samples used for the NMR studies were greater than 95% pure, as judged by both SDS-PAGE and two-dimensional 1 H-15 N correlation spectra.
Residual 1 D NH , 1 D C␣H␣ , and 1 D NCЈ dipolar couplings were obtained by taking the difference in the corresponding J splittings measured in magnetically oriented and isotropic (in water) media. The orienting liquid crystalline medium employed consisted of a 4 -5% C12E5 polyethylene glycol (PEG)/hexanol mixture with a surfactant to alcohol ratio of 0.96 (32). 1 D NH couplings were measured on IIA Mtl -HPr( 15 N) and IIA Mtl ( 15 N)-HPr complexes using two-dimensional in-phase/anti-phase { 15 N, 1 H} heteronuclear single quantum coherence experiments (33). 1 D C␣H␣ and 1 D NCЈ couplings were measured on IIA Mtl -HPr( 15 N/ 13 C) and IIA Mtl ( 15 N/ 13 C)-HPr complexes using three-dimensional (HCACO)NH and (TROSY)HNCO experiments (34), respectively. In all cases the concentration of labeled protein ranged from 0.6 to 1.0 mM, and the unlabeled protein was present in sufficient excess to ensure that the labeled protein was entirely in the bound state. The 1 D NH dipolar couplings were measured in 5% PEG/hexanol using a different stock of PEG/hexanol from the 1 D C␣H␣ and 1 D NCЈ dipolar couplings (which were measured in 4% PEG/hexanol). Consequently, two alignment tensors were employed: one for the 1 D NH couplings and a second for the 1 D C␣H␣ and 1 D NCЈ couplings. The magnitude of the alignment tensor (i.e. the values of axial component of the tensor, D a NH , and the rhombicity ) was obtained by singular value decomposition (35,36) using the x-ray coordinates of free IIA Mtl (1A3A (11)) and HPr (1POH (37)). The values of D a NH and , determined in this manner, are Ϫ13.3 and 0.43 Hz, respectively, for the 1 D NH set of dipolar couplings, and Ϫ9.4 and 0.58 Hz, respectively, for the 1 D C␣H␣ and 1 D NCЈ couplings. Note that small, unavoidable, variations in the concentration of PEG/hexanol among different samples only result in correspondingly small variations in the value of the axial component of the alignment tensor, D a NH . The dipolar couplings are appropriately normalized to reflect this.
Structure Calculations-Intermolecular NOE-derived interproton distance restraints were classified into two ranges, 1.8 -5.0 and 1.8 -6.0 Å (17). Intramolecular NOE-derived interproton distance restraints relating to the interfacial side chains were classified into four ranges as follows: 1.8 -2.7, 1.8 -3.5, 1.8 -5.0, and 1.8 -6.0 Å (38). An additional 0.5 Å was added to the upper bound for NOEs involving methyl groups, and distances involving methyl groups and non-stereospecifically assigned protons were represented by a (⌺r Ϫ6 ) Ϫ1/6 sum. 1 and 2 side chain torsion angle restraints were derived from analysis of heteronuclear 3 J couplings and NOE/(rotating frame Overhauser enhancement) experiments (26). Backbone / restraints for the four variable regions of IIA Mtl were derived from backbone chemical shift data using the program TALOS (39) in conjunction with knowledge of the ranges for these / angles observed in the four molecules of the crystal structure of free IIA Mtl .
Structures were calculated by conjoined rigid body/torsion angle dynamics (40,41) using the program Xplor-NIH (42). The target function minimized comprises the experimental NMR restraints (NOEderived interproton distances, torsion angles, and residual dipolar couplings (43)), a repulsive van der Waals potential for the non-bonded contacts (36), a torsion angle data base potential of mean force (44), and a radius of gyration restraint to ensure optimal interfacial packing (45).
Structures were visualized and analyzed with the programs VMD-XPLOR (46), RIBBONS (47), and GRASP (48). Reweighted atomic density maps were calculated from the ensemble of simulated annealing structures as described previously (49). Amino acid sequence searches and alignments were carried out using the program BLAST (50).
The coordinates and experimental restraints have been deposited in the RCSB Protein Data Bank (PDB accession code 1J6T).

RESULTS AND DISCUSSION
Structure Determination-The physiological form of enzyme II Mtl consists of the three domains, A-C, joined together in a single polypeptide chain by flexible linkers of variable length (51,52). The membrane-bound IIC Mtl domain is at the N terminus (residues 1-347), followed by the cytoplasmic IIB Mtl (residues 348 -489) with the IIA Mtl (residues 490 -637) domain at the C terminus (19,(53)(54)(55)(56). The isolated IIA Mtl domain is catalytically active and can accept a phosphoryl group from HPr and transfer it to both the isolated IIB Mtl domain as well as IICB Mtl (19,55). We have solved the structure of the IIA Mtl -HPr complex by multidimensional NMR. For consistency with previous structural work (11,57,58), residues 490 -637 of II Mtl are numbered 1-148 in the cloned, expressed IIA Mtl domain; hence, in this numbering scheme, the active site histidine (residue 554 in the full-length II Mtl sequence (59,60)) is located at position 65.
IIA Mtl and HPr are in fast exchange on the chemical shift scale. The equilibrium association constant derived from NMR titration studies by monitoring 1 H-15 N cross-peaks of 15 N-labeled IIA Mtl upon addition of unlabeled HPr is ϳ2 ϫ 10 4 M Ϫ1 . The lower limit for the dissociation rate constant, derived from the maximal observed 1 H N shift between free and bound states, is ϳ4500 s Ϫ1 .
A combination of isotopically ( 15 N and/or 13 C) labeled proteins was used to simplify the spectra for assignment purposes and to specifically observe intermolecular nuclear Overhauser effects (NOEs) (26). An example of the quality of the data is illustrated in Fig. 1 which shows a series of strips from threedimensional 13 C-separated/ 12 C-filtered NOE spectra in which NOEs are observed specifically from protons attached to 13 C on one protein to protons attached to 12 C on the other.
High resolution crystal structures of both HPr ((37) code 1POH) and IIA Mtl ((11) code 1A3A) are available and therefore permit the structure of the complex to be solved by conjoined rigid body/torsion angle dynamics (40,41), in a manner analogous to that employed for the HPr-IIA Glc complex (17). In this approach, it is necessary to either demonstrate that no significant conformational changes occur upon complex formation or to localize regions where such changes may occur. This can be rapidly assessed by the measurement of residual dipolar couplings, which provide highly sensitive long range orientational information (61)(62)(63)(64)(65).
For HPr, agreement between observed 1 D NH dipolar couplings measured on the complex dissolved in a dilute liquid crystalline medium of polyethylene glycol/hexanol (33) and those calculated from the x-ray structure of free HPr is excellent with a 1 D NH dipolar coupling R-factor, R dip NH , of 19.1% and a correlation coefficient of 0.96, consistent with the expected accuracy of a 2-Å resolution crystal structure.
The situation for IIA Mtl is more complex, because the crystal structure has four molecules (A-D) in the unit cell (11): molecules A and B, and molecules C and D are very similar with an overall backbone root mean square difference of ϳ0.3 Å, consistent with the expected coordinate errors for a 1.8-Å resolution structure. However, the backbone of one region (I, residues 51-54) is displaced by up to ϳ2 Å in molecule A relative to the other three molecules, and three regions (II, 66 -78; III, 91-96; and IV, 104 -110) exhibit maximal backbone displacements of up to 1.5-3 Å between the two pairs (A/B and C/D) of structures. Excluding these regions, the backbone root mean square difference between the four structures remains at ϳ0.3 Å. For the invariant core residues (i.e. all residues with the exception of those in the four variable regions), the average values of R dip NH and the correlation coefficients are 20.5 Ϯ 1.2% and 0.96 Ϯ 0.05, respectively, for the four molecules. For the four variable regions, on the other hand, R dip NH and the correlation coefficient have values of 32% and 0.92 for molecule A, 29% and 0.93 for molecule B, 28% and 0.94 for molecule C, and 22% and 0.96 for molecule D. Thus, the overall agreement between the measured and calculated 1 D NH dipolar couplings for the four variable regions in molecule D is essentially the same as that of the core residues, whereas the discrepancies for molecules A-C are significantly (ϳ50%) higher. The improvement for molecule D relative to the other three molecules is statistically significant, because it is larger than the intrinsic error, and implies that the structure of molecule D of IIA Mtl is closest to that in the IIA Mtl -HPr complex observed in solution.
Because the variable regions I-IV of IIA Mtl compose several residues (specifically residues 52-54, 68, 92, 93, 96, and 109) at the periphery of the interface with HPr, the following strategy was employed. The starting coordinates comprised the free x-ray structures of E. coli HPr (37) and IIA Mtl (molecule D (11)), placed 30 -50 Å apart in random orientations. Structures of the complex were calculated from the experimental restraints by rigid body minimization (66), followed by conjoined rigid body/ torsion angle dynamics (40,41). The backbone coordinates and non-interfacial side chains (excluding the four variable regions of IIA Mtl ) were treated as rigid bodies throughout, with IIA Mtl held fixed, HPr allowed to rotate and translate, and the axis of the dipolar coupling alignment tensor free to rotate. The interfacial side chains of both IIA Mtl and HPr were given their full torsional degrees of freedom. The backbone and side chains of the four variable regions of IIA Mtl were also given torsional degrees of freedom; the / angles were restrained by squarewell potentials to the ranges covered by all four molecules in the crystal structure of free IIA Mtl , consistent with chemical shift data (39), and the side chain torsion angles of the noninterfacial residues were similarly restrained to the rotamers observed in the crystal structure, consistent with heteronuclear 3 J coupling measurements. In the case of region II, however, residues 64 -67 and residues 77-78 were given full torsional degrees of freedom, whereas residues 68 -76 were treated as a rigid body because the / angles for this region differed by less than 5°between all four molecules of the crystal structure. (Note residues 64 and 65 are included in region II to allow appropriate hinge movements to occur.) Within the confines of these torsion angle restraints, the conformations of the four variable regions of IIA Mtl in the complex are determined by the dipolar coupling restraints and intramolecular NOE-derived interproton distance restraints relating to these regions.
The IIA Mtl -HPr complex was solved on the basis of 872 experimental restraints, including 107 intermolecular NOE-derived interproton distance restraints and 528 residual dipolar couplings. A summary of the structural statistics is provided in Tables I and II The dipolar coupling R-factors for the complex, obtained with single alignment tensors for both proteins in the complex, are comparable with those of the free x-ray structures calculated using individual alignment tensors for each protein (Table I). Thus, one can conclude that the relative orientation of the two proteins is uniquely defined by the rotational and translational information afforded by the intermolecular NOE-derived interproton distance restraints and the orientational information afforded by the dipolar couplings (66).
To assess further the accuracy of the relative orientation of the two proteins in the complex, we carried out cross-validation using the dipolar couplings (67). This also probes the effects of errors in the coordinates. The 1 D NH , 1 D C␣H , and 1 D NCЈ dipolar couplings for each protein were randomly divided into two groups of equal size (A and B). In the first set of calculations, group A was used as the work set and B as the test set; the groups were reversed in the second set of calculations. Only the dipolar couplings in the work set are refined against. Compar- ison of the values of the working dipolar coupling R-factors, R dip (work), with the corresponding cross-validated ones, R dip (free), indicates that the values of R dip (work) and R dip (free) from the first and second set of calculations, respectively, are essentially identical, and similarly for R dip (free) and R dip (work) from the first and second set of calculations, respectively (Table   Table II) and the mean coordinates. The value quoted for the complete backbone atoms provides only a measure of the precision with which the relative orientation of the two proteins have been determined and does not take into account the accuracy of the x-ray coordinates of HPr (37) and IIA Mtl (11). The latter is expected to be ϳ0.3 Å, as judged from the crystallographic resolution and R-factors. The four variable regions of IIA Mtl comprise residues 51-54, 64 -78, 91-96, and 104 -110. (Note residues 64 and 65 are included in region II to allow appropriate hinge movements to occur since residues 68 -76 are treated as a rigid body.) f Defined as the average angular r.m.s. difference in the relative orientation of HPr and IIA Mtl between the mean coordinates derived from the simulated annealing structures calculated with all dipolar couplings and those derived from structures obtained in the first and second set of calculations with dipolar coupling cross-validation (Table II).

TABLE II Dipolar coupling cross-validation
The dipolar couplings, excluding those of the four variable regions of IIA Mtl , were randomly divided into two equal groups, A and B, each comprising 50% of the dipolar couplings. In the Set 1 calculations, group A was used as the work set and group B as the free set; in the Set 2 calculations, group B was used as the work set and group A as the free set. Thus, R dip (work) from Set 1 and R dip (free) from Set 2 comprise the same dipolar couplings; likewise for R dip (free) from Set 1 and R dip (work) from Set 2. II). Thus, the quality of the fit to the experimental dipolar coupling data is good, and the dipolar coupling data have not been overfitted (67). Comparison of the relative orientations of the two proteins in these two sets of calculations with that obtained using all the dipolar couplings indicates that the relative orientation of the two proteins is determined with a precision of Յ1.5°, which corresponds approximately to the expected uncertainty in the orientation of the dipolar coupling alignment tensor (68). In this regard it is worth noting that there is only a 1.9°difference in orientation between the alignment tensor used for the 1 D NH couplings obtained from one stock of PEG/hexanol and that used for the 1 D C␣H␣ and 1 D NCЈ couplings obtained from a different PEG/hexanol stock. The backbone r.m.s. difference between the mean coordinates derived from the first and second set of cross-validation calculations and those obtained with all dipolar couplings is Յ0.1 Å, which is comparable with the precision of the backbone coordinates.
As indicated above, one alignment tensor was employed for the 1 D NH dipolar couplings and another for the 1 D C␣H␣ and 1 D NCЈ couplings because different stocks of PEG/hexanol were used to measure these two sets of dipolar couplings. If the calculations are repeated using a single alignment tensor with an average value for the rhombicity and all the dipolar cou-plings normalized to a single D a NH value, the difference in relative protein-protein orientation is only 1.7°with a backbone atomic r.m.s. difference of 0.15 Å. This provides an upper limit estimate of the uncertainty in the relative protein-protein orientation introduced by inaccuracies in the value of the rhombicity.
It should also be noted that the difference in relative proteinprotein orientation for the structures calculated as described above with those calculated using molecules A or D of the crystal structure of IIA Mtl with their backbones held rigid differ by only 1.8 and 0.9°, respectively. This provides a measure of the uncertainty in the determination of the relative proteinprotein orientation introduced by inaccuracies in the crystal structure coordinates.
A final check on quality is afforded by calculations omitting all dipolar coupling restraints. The resulting values of R dip are on average only ϳ3% higher than those obtained when all the dipolar couplings are included in the structure calculation, the difference in relative orientation for the two proteins between the mean coordinates calculated with and without dipolar couplings is ϳ5°, and the corresponding backbone r.m.s. difference is ϳ0.5 Å.
Overall Description of the IIA Mtl -HPr Interface-Two views of the overall IIA Mtl -HPr complex, displayed as ribbon diagrams, are shown in Fig. 3 (throughout the text and figures, residues of HPr are denoted in italics). The IIA Mtl binding surface on HPr is convex and comprises two segments of the polypeptide chain (residues 12-27 and 46 -56) encompassing helices ␣1 (residues 16 -28) and ␣2 (residues 47-52), consistent with previous chemical shift mapping studies (69). HPr docks into a concave cleft on the surface of IIA Mtl formed by three distinct segments of polypeptide chain as follows: residues 49 -68 that include the C-terminal end of helix ␣2 (residues 41-52), an antiparallel ␤-sheet formed by ␤-strands 2 (residues 56 -58) and 3 (residues 61-62), and the active site histidine at position 65; residues 92, 93, and 96 located in the loop connecting ␤-strands 4 and 5; and residues 109 -128 that includes helix ␣3 (residue 111-121) and the N-terminal portion of helix ␣4 (residues 125-133). The interface is ϳ27 Å long and ϳ22 Å wide and comprises a total of 42 residues, 25 from IIA Mtl and 17 from HPr. The interfacial residues are distributed between ϳ60% non-polar residues and ϳ40% polar ones. The total accessible surface area buried upon complexation is ϳ1450 Å 2 of which 685 Å 2 originate from IIA Mtl and ϳ765 Å 2 from HPr.
A detailed view of the interface is shown in Fig. 4A together with a diagrammatic summary of the intermolecular contacts  g ϩ rotamer (conformer b). In the minor b conformer, the carboxylate of Glu-59 is within hydrogen bonding distance of the hydroxyl group of Ser-46 (Fig. 4A). In the major a conformer, Glu-59 can still potentially interact indirectly with Ser-46 via a water-bridged molecule (Fig. 4A).
Some degree of assessment of the relative importance of the various intermolecular interactions can be gauged by sequence comparisons of the interfacial residues of IIA Mtl and HPr among a variety of bacterial species. The species chosen comprise examples of Gram-negative (Yersinia pestis and Vibrio cholera) and Gram-positive (Bacillus anthracis, Bacillus subtilis, and Staphylococcus carnosus) bacteria, and the sequence alignments are shown in Fig. 4C. The overall sequence identities for the complete protein sequences range from 28 to 97% relative to the E. coli sequences (50). In the case of the Gramnegative bacteria, E. coli, Y. pestis, and V. cholera, sequence conservation among the interfacial residues is higher than that for the overall proteins. In addition, the few substitutions are highly conservative in terms of the intermolecular interactions in which they participate. Thus, for example, substitution of Sequence conservation between the E. coli proteins and those of the Gram-positive bacteria, B. anthracis, B. subtilis, and S. carnosus, is lower, ranging from 28 to 42% overall and 21 to 38% for the interfacial residues (Fig. 4C). These sequence comparisons highlight the critical importance of the intermolecular hydrophobic interactions involving the side chains of residues 16 (Thr or Ala), 17 (Arg), 47 (Leu or Ile), 48 (Phe or Met), and 51 (Gln or Met) of HPr and the side chains of residues 52 (Leu, Thr, or Val), 57 (Leu, Met, or Val), 112 (Ile or Leu), 115 (Ile or Leu), 119 (Thr, Ala, or Leu) and 122 (Leu, Phe, Ile, or Cys) of IIA Mtl . The intermolecular electrostatic interactions observed in the E. coli complex, however, are much less conserved, suggesting that these are not critical determinants of specificity. Thus, the salt bridge between Lys-49 of HPr and Asp-123 of IIA Mtl in the E. coli complex is abrogated by substitution of Lys-49 to an aliphatic side chain (Val or Gly) and Asp-123 to a neutral Ser. Salt bridges involving Lys-24/Lys-27 and Glu-92/Glu-93/Asp-96 in the E. coli complex can no longer occur because of substitution of the two lysines to neutral residues (Gln or Asn at position 24 and Ser at position 27), although water-bridged hydrogen bonding interactions are still possible. The probable water-mediated hydrogen bonding interaction between residue 12 (Asn, Thr or Ser) of HPr and 109 (Asn or Asp) of IIA Mtl does appear to be preserved except in the case of S. carnosus where Asn-109 is replaced by a Gly.
Although many features of the HPr and IIA Mtl interfaces are preserved between the Gram-negative and -positive bacteria, there is still considerable species specificity in terms of the interaction of HPr and IIA Mtl . Thus, phosphoryl transfer between E. coli (Gram-negative) HPr and S. carnosus (Grampositive) IIA Mtl is ϳ10-fold lower than that involving S. carnosus HPr (70). Because the disposition of intermolecular hydrophobic interactions is preserved, the reduction in the rate of the heterologous phosphoryl transfer is presumably due to a reduction in the surface complementarity for the inter-species interactions. Thus, for example, Phe-48 of HPr in the Gramnegative bacteria is substituted by Met in the Gram-positive bacteria (Fig. 4C) which will necessarily change the details of The Phosphoryl Transition State Intermediate-The phosphoryl group, originating on the N␦1 atom of His-15 of HPr (71,72), is transferred to the N⑀2 atom of His-65 of IIA Mtl (59) with inversion of the configuration at phosphorus (73). This indicates a transition state involving a pentacoordinate phosphoryl group in a trigonal bipyramidal geometry, with the donor and acceptor nitrogen atoms of the active site histidines in apical positions and the oxygen atoms of the phosphoryl group lying in the equatorial plane (74).
We modeled the transition state using essentially the same approach as that described previously (17) in the case of the IIA Glc -HPr complex; in particular, structures were recalculated introducing appropriate geometric restraints for the phosphoryl group in conjunction with all the experimental NMR restraints. Thus, the P-N distances were restrained to either 2 or 3 Å for associative (S N 2) and dissociative transition states, respectively; the phosphorus atom was restrained to lie in the plane of each imidazole ring. In addition, the backbone of the active site histidines and the adjacent residues on either side (i.e. residues 64 -66 of IIA Mtl and 14 -16 of HPr) were allowed torsional degrees of freedom with their and angles restrained by square-well potentials within Ϯ15 and Ϯ20°, respectively, of their values in the unphosphorylated complex. The results are displayed in Fig. 5A. The backbone r.m.s. difference between the mean coordinates of the structures calculated with and without the phosphoryl group is less than 0.1 Å, which is within the errors of the coordinates.
To accommodate the phosphoryl transition state, only small displacements (0.2 and 0.4 Å for the dissociative and associative transition states, respectively) in the backbone coordinates of the residues immediately adjacent to the active site histidines are required (Fig. 5A). This is mirrored by equally minor changes in backbone / angles. The only noteworthy difference appears to be in the / angles of Thr-16 at the beginning of the first helix of HPr; these change from Ϫ42/Ϫ67°in the unphosphorylated complex to the more helical values of Ϫ65/ Ϫ54°in the transition state complexes. The 1 and 2 rotamers of His-65 (g ϩ /g Ϫ ) of IIA Mtl and His-15 (g ϩ /g ϩ ) of HPr remain unaltered. The 1 and 2 angles of His 65 changes by less than 5 and ϳ30°, respectively; the 1 and 2 angles of His-15 change by 20 -30 and ϳ1°, respectively. In addition, the side chain of Arg-49 is slightly displaced (ϳ0.7-0.9 Å) by the presence of the phosphoryl group; this is achieved by very small changes (5-20°) in 1 / 2 / 3 / 4 side chain torsion angles that remain in the t/g ϩ /t/t rotamers. A detailed view of the active site illustrating interactions that stabilize the transition state is shown in Fig. 5B. The phosphoryl group sits in a hydrophobic cleft comprising Leu-52, Ile-112, and Ile-115 of IIA Mtl ; there are no negatively charged carboxylate groups in close proximity. The phosphoryl group is within hydrogen bonding distance of the guanidino group of Arg-49 of IIA Mtl and the hydroxyl group of Thr-16 and the backbone amides of Thr-16 and Arg-17 of HPr. The b conformer of His-111 does not interact with the phosphoryl group, but its N␦1 atom may be hydrogen-bonded to the carboxyamide of Asn-109; in the a conformer, the latter interaction is preserved but the N⑀2 atom is now in close proximity to both the phosphoryl group and the side chain carboxyamide group of Asn-12 of HPr. Although Arg-17 of HPr does not participate in any electrostatic interactions with IIA Mtl , it contributes to the pos- itively charged environment in the vicinity of the phosphoryl group.
Comparison with the IIA Glc -HPr and EIN-HPr Complexes-The interaction surfaces employed by HPr to recognize IIA Mtl and IIA Glc (17) and EIN (16) are very similar, overlap extensively, and consist of 17, 18, and 23 residues, respectively ( Fig.  6B, left, middle, and right panels). Of the 17 HPr residues that participate in the IIA Mtl -HPr complex, 15 are involved in the IIA Glc -HPr complex and 16 in the EIN-HPr complex (Fig. 4B). In the case of the IIA Mtl -HPr and IIA Glc -HPr (17) complexes, the overall accessible surface areas buried at the interface are essentially the same (ϳ1350 -1450 Å 2 ) and distributed similarly between HPr (ϳ45%) and the partner protein (ϳ55%). The surface area buried at the interface of the EIN-HPr complex (16) is about 40% larger and approximately equally divided between HPr and EIN.
The backbone scaffolds comprising the interaction surfaces on IIA Mtl , IIA Glc , and EIN are topologically different both in terms of structure and connectivity: IIA Mtl employs a mixture of helix and sheet (Fig. 7A); IIA Glc predominantly makes use of sheet (Fig. 7B); and EIN uses only helices (Fig. 7C). Despite these differences, some elements of commonality can be observed (Fig. 7D). Thus, strands ␤3 of IIA Mtl and ␤7 of IIA Glc , which bear the active site histidine, are aligned; helix ␣3 of IIA Mtl coincides with strands ␤5 and ␤6 of IIA Glc ; and helix ␣2 of IIA Mtl , helix ␣1 of IIA Glc and helix ␣2 of EIN partially overlap (Fig. 7D).
Surface representations of the interaction interfaces for the IIA Mtl -HPr, IIA Glc -HPr, and EIN-HPr complex are shown in Fig. 6 (left, middle, and right panels, respectively). The target surfaces that interact with HPr are broadly similar in all three cases comprising a central hydrophobic core located within a concave cleft surrounded by negatively charged residues (Fig.   6A). A number of electrostatic interactions are preserved in the three complexes. Thus, for example, Gln51 of HPr is buried in all three complexes (Fig. 6B) and adapts its conformation to form potential hydrogen bonds with Thr-116 of IIA Mtl , Ser-78 of IIA Glc , and Arg-126 of EIN (Fig. 6A). Likewise, Lys-49 of HPr forms potential salt bridges with Asp-123 of IIA Mtl , Glu-80 and Glu-86 of IIA Glc , and Glu-84 of EIN; Lys-24 and Lys-27 of HPr form potential salt bridges with Glu-92, Glu-93, and Asp-96 of IIA Mtl and Asp-144 of IIA Glc , and Glu-74 and Asp-82 of EIN. Ser-46 of HPr interacts with a negatively charged residue in both the complexes with IIA Mtl (Glu-59) and EIN (Asp-82), consistent with the finding that mutation of Ser-46 to Asp significantly reduces activity for phosphoryl transfer to these two enzymes (75).
In addition to the similarities noted above, there are also a number of noteworthy differences between the three complexes. Although the negatively charged residues are almost uniformly distributed around the periphery of the interaction surfaces on IIA Glc and EIN, they are clustered in essentially one region of IIA Mtl (along the top rim of the interaction site in the view shown in Fig. 6A). From a structural perspective, this has consequences with regard to the role of the conserved Arg-17 of HPr. Arg-17 is critical for phosphoryl transfer between EIN and HPr and between HPr and IIA Glc (76,77), and its function is to neutralize a pair of negatively charged residues that are located in close spatial proximity to the active site histidines: Glu-67 and Glu-68 of EIN (16), and Asp-38 and Asp-94 of IIA Glc (17) (Fig. 6, A and B). In the case of IIA Mtl , these negatively charged residues are replaced by one positively charged residue, Arg-49, and two hydrophobic residues, Leu-52 and Val-68 (Fig. 6A, left panel). Thus, in the IIA Mtl -HPr complex, the neutralizing role of the guanidino group of Arg-17 is no longer required, and the aliphatic portion of the Arg-17 side chain simply packs against the methyl groups of Leu-52 and Thr-53 (Fig. 4A). The guanidino group of Arg-49 of IIA Mtl , on the other hand, forms a potential hydrogen bond with Thr-16 in the unphosphorylated complex (Fig. 4A) and interacts with the phosphoryl group in the transition state (Fig. 5). This suggests that mutation of Arg-17 to a neutral residue would have little impact on phosphoryl transfer between HPr and IIA Mtl . This is borne out experimentally where mutation of Arg-17 to His, Ser, Cys, or Gly has no effect on V max and only increases K m by a factor of ϳ2 (76). Mutation of Arg-17 to a negatively charged residue (Glu), on the other hand, increases the value of K m 100-fold, while leaving V max unaltered (76); this is presumably due to either the introduction of a negative charge in close proximity to the phosphoryl group or possibly to a potential electrostatic contact with Arg-49, thereby hindering the interaction of Arg-49 with the phosphoryl group. Mutation of Arg-49 of IIA Mtl , however, would be predicted to abrogate phosphoryl transfer between HPr and IIA Mtl . This prediction is supported by the observation that Arg-49 appears to be absolutely conserved among IIA Mtl s from different bacterial species (Fig. 4C).
The location of the active site histidines of IIA Mtl , IIA Glc , and EIN in the three complexes is also of interest. In the case of EIN (16), the active site His-189 is located at the very edge of the interaction surface, and adjacent residues in the sequence of EIN are not in contact with HPr (Fig. 6A, right panel). In contrast, in the case of both IIA Mtl and IIA Glc , the active site histidines, His-65 and His-90, respectively, are located just off-center of the interaction surface (Fig. 6A, left and middle panels, respectively). In all three complexes, the phosphoryl group in the transition state intermediate is located between the N␦1 of His-15 of HPr and the N⑀2 atom of the active site histidine of the partner protein, and the 1 / 2 angles of His-15 are in the same rotameric states (g ϩ /g ϩ ). Consequently, the side chain conformation of the active site histidine of the partner protein must accommodate its different location relative to His-15 of HPr in the three complexes. This is achieved by different combinations of rotameric states for 1 and 2 : g ϩ /g ϩ for His-65 of IIA Mtl , t/g Ϫ for His-90 of IIA Glc , and t/g ϩ for His-189 of EIN.
An example of conformational side chain plasticity is afforded by Phe-48 of HPr. Phe-48 participates in crucial intermolecular hydrophobic interactions in all three complexes (Fig.  6) and adapts its side chain rotamer conformation to optimize these; in the IIA Mtl and IIA Glc complex, the 1 angle of Phe-48 is in the t rotamer, whereas in the EIN complex it is in the g Ϫ rotamer.
Concluding Remarks-The structure of the E. coli IIA Mtl -HPr complex reveals that specificity is achieved through a large set of hydrophobic interactions, supplemented by a few electrostatic/hydrogen bonding interactions around the periphery of the interface. Comparison of the IIA Mtl -HPr (this paper), IIA Glc -HPr (17), and EIN-HPr (16) complexes show how HPr can use essentially the same interaction surface to recognize a plethora of structurally unrelated proteins. Three key factors are involved as follows: shape and residue complementarity of the interaction surfaces, even though the underlying backbone structural elements of the target surfaces are dissimilar; large interaction surfaces that can support redundancy of interactions and that minimize the effect of any single interaction (that is specificity is built up by a large series of interactions, each of which makes only a small contribution to the overall interaction energy); and side chain conformational plasticity to optimize intermolecular contacts and surface complementarity. These features are likely to be characteristic of many proteinprotein interactions in signal transduction pathways that involve the rapid formation and dissociation of temporally transient, yet specific protein-protein complexes.