Solution Structure of a Post-transition State Analog of the Phosphotransfer Reaction between the A and B Cytoplasmic Domains of the Mannitol Transporter IIMannitol of the Escherichia coli Phosphotransferase System*

The solution structure of the post-transition state complex between the isolated cytoplasmic A (IIAMtl) and phosphorylated B (phospho-IIBMtl) domains of the mannitol transporter of the Escherichia coli phosphotransferase system has been solved by NMR. The active site His-554 of IIAMtl was mutated to glutamine to block phosphoryl transfer activity, and the active site Cys-384 of IIBMtl (residues of IIBMtl are denoted in italic type) was substituted by serine to permit the formation of a stable phosphorylated form of IIBMtl. The two complementary interaction surfaces are predominantly hydrophobic, and two methionines on IIBMtl, Met-388 and Met-393, serve as anchors by interacting with two deep pockets on the surface of IIAMtl. With the exception of a salt bridge between the conserved Arg-538 of IIAMtl and the phosphoryl group of phospho-IIBMtl, electrostatic interactions between the two proteins are limited to the outer edges of the interface, are few in number, and appear to be weak. This accounts for the low affinity of the complex (Kd ∼ 3.7 mm), which is optimally tuned to the intact biological system in which the A and B domains are expressed as a single polypeptide connected by a flexible 21-residue linker. The phosphoryl transition state can readily be modeled with no change in protein-protein orientation and minimal perturbations in both the backbone immediately adjacent to His-554 and Cys-384 and the side chains in close proximity to the phosphoryl group. Comparison with the previously solved structure of the IIAMtl-HPr complex reveals how IIAMtl uses the same interaction surface to recognize two structurally unrelated proteins and explains the much higher affinity of IIAMtl for HPr than IIBMtl

The bacterial phosphoenolpyruvate:sugar phosphotransferase system (PTS) 2 comprises a fundamental signal transduction pathway whereby phosphotransfer via a series of biomolecular steps is coupled to sugar transport across the membrane (1)(2)(3)(4). The phosphoryl group originates on phosphoenolpyruvate, and the initial transfer steps, involving first enzyme I and subsequently HPr, are common to all components of the pathway. Thereafter, the phosphoryl group is transferred to sugar-specific enzymes II, of which there are four major classes, glucose (Glc), mannitol (Mtl), mannose (Man), and chitobiose (Chb), which share no sequence similarity with one another (2,3) and, with one exception (the B domains of enzymes II Mtl and II Chb ), no structural similarity with one another either (4 -15). Enzymes II generally comprise three domains, two cytoplasmic, A and B, and one transmembrane, C, which may or may not be covalently linked to one another (2)(3)(4). The A domain accepts the phosphoryl group from HPr and donates it to the B domain. Subsequently, the phosphoryl group is transferred from the B domain to the incoming sugar bound to the C domain. The complexes of the PTS are of considerable interest from the perspective of protein-protein interactions, since similar binding surfaces can recognize multiple, structurally different, targets. We have embarked on a long term structural study of the complexes of the PTS, and to date we have determined the solution NMR structures of the enzyme I-HPr complex (16), complexes of HPr with IIA Glc (17), IIA Mtl (18), and IIA Man (19), and the complex of IIA Glc with IIB Glc (19). In this paper, we present the solution structure of a post-transition state analog of the IIA Mtl -IIB Mtl complex. Specifically, we make use of the active site C384S mutant of IIB Mtl to generate a stably phosphorylated form of IIB Mtl (20,21) and the active site H554Q mutant of IIA Mtl to prevent phosphoryl transfer between IIA Mtl and IIB Mtl . (Throughout this work, residues of IIB Mtl are denoted in italic type.) The complex reveals the structural basis of specific recognition between IIA Mtl and IIB Mtl and the interactions involved in phosphoryl transfer.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Purification of IIA Mtl (H554Q) and Phospho-IIB Mtl (C384S)-The region corresponding to the A domain (IIA Mtl ) of the mannitol transporter, residues 490 -637, was cloned as described previously (18). From this original construct, the active site histidine residue (His-554) was mutated to a glutamine to disable its phosphoryl transfer activity. The new construct was verified by DNA sequencing and then subcloned into a modified pET-32a vector (14) to form a thioredoxin fusion protein with a His 6 tag. After transformation with an expression vector, Escherichia coli strain BL21(DE3) (Novagen) was grown in either Luria Bertini or minimal media (with 15 NH 4 Cl and/or 13 C 6 -glucose as the sole nitrogen or carbon source, respectively), induced with 1 mM isopropyl-␤-D-thiogalactopyranoside at an A 600 of ϳ0.8, and harvested by centrifugation after 4 h of induction. After harvesting, the cell pellet was resuspended in 50 ml (per liter of culture) of 50 mM Tris, pH 7.4, 100 mM NaCl, 2 mM ␤-mercaptoethanol, 10 mM imidazole, and 1 mM phenylmethylsulfonyl fluoride. The suspension was lysed by three passages through a microfluidizer and centrifuged at 10,000 ϫ g for 20 min. The supernatant fraction was loaded onto a nickel-Sepharose column (ϳ20 ml; Amersham Biosciences), and the fusion protein was eluted with a 100-ml gradient of imidazole (25-500 mM). The fusion protein was then dialyzed against 20 mM Tris, pH 8.0, 200 mM NaCl, and 2 mM ␤-mercaptoethanol, and digested with thrombin (10 NIH units/mg of protein). Thrombin was removed by passage over a benzamidine-Sepharose column (1 ml; Amersham Biosciences), followed by the addition of 1 mM phenylmethylsulfonyl fluoride. The cleaved His 6 -thioredoxin was removed by loading the digested proteins over a nickel-Sepharose column. IIA Mtl (H554Q) was further purified by Sephadex-75 gel filtration column (Amersham Biosciences) equilibrated with 20 mM Tris, pH 7.4, 1 mM EDTA, and 0.01% (w/v) sodium azide.
The active site mutant of the B domain, IIB Mtl (C384S), of the mannitol transporter (residues 375-476) was cloned, expressed, phosphorylated, and purified as described previously (21).
With the exception of the titration experiments, NMR samples contained 3 mM IIA Mtl (H554Q) and 3 mM phospho-IIB Mtl (C384S) in 20 mM Tris-d 11 (pH 7.4), 0.01% (w/v) sodium azide, and either 90% H 2 O, 10% D 2 O or 99.996% D 2 O. Each sample contained complexes of one uniformly 15 N/ 13 C-labeled protein with its unlabeled partner. For titration experiments, a 0.6 mM concentration of either protein was titrated with the other protein using up to 1:12 molar ratios in 20 mM Tris, pH 7.4, and 0.01% (w/v) sodium azide.
NMR Spectroscopy-NMR spectra were collected at 30°C on Bruker DRX800, DMX750, DMX600, DRX600, and DMX500 spectrometers equipped with either an x,y,z-shielded gradient triple resonance probe or a z-shielded gradient triple resonance cryoprobe. Spectra were processed with the NMRPipe package (22) and analyzed using the program PIPP/CAPP/STAPP (23).
Side-chain torsion angle restraints were derived from 3 J NC␥ and 3 J CЈC␥ coupling constants measured using quantitative J correlation spectroscopy (27) in conjunction with short mixing time (30 ms) three-dimensional 13 C-separated NOE spectra recorded in H 2 O (26).
Axially stretched (28) neutral polyacrylamide gels (5% (w/v) polyacrylamide; 39:1 (w/w) acrylamide/bisacrylamide) were prepared as described previously (21). Residual dipolar couplings (RDC) were obtained by taking the difference in J couplings measured in aligned and isotropic (water) media. 1 J NH couplings were determined from a two-dimensional in-phase/antiphase 1 H-15 N HSQC spectrum. Measurements on the free proteins were carried out using a protein concentration of 0.5 mM; measurements on the complex were carried out on samples containing a 0.5 mM concentration of the 15 N-labeled partner and 2 mM of the unlabeled partner.
Structure Calculations-NOE-derived interproton distance restraints were classified into loose approximate distance ranges of 1.8 -2.7, 1.8 -3.5, 1.8 -5.0, and 1.8 -6.0 Å, corresponding to strong, medium, weak, and very weak NOE cross-peak intensities, respectively; an empirical correction of 0.5 Å was added to the upper distance bounds of distance restraints involving methyl groups to account for the higher apparent intensity of methyl resonances (29). NOEs involving nonstereospecifically assigned methyl, methylene, and aromatic protons were represented by a (͚r Ϫ6 ) Ϫ1 ⁄ 6 sum (26). The error range employed for the torsion angle restraints (represented by square well potentials) was Ϯ20°when a unique rotamer could be identified and Ϯ80°when the torsion angle could be narrowed down to two of the three rotamers (19).
Structures were calculated using conjoined rigid body/torsion anglesimulated annealing (30) with the program Xplor-NIH (31). The minimized target function comprises the experimental NMR restraints (NOE-derived interproton distances and torsion angles), a quartic van der Waals repulsion term for the nonbonded contacts (32), a multidimensional torsion angle data base potential of mean force (33), and a radius of gyration potential to ensure optimal packing (34).

RESULTS AND DISCUSSION
Equilibrium Binding of IIA Mtl (H554Q) and Phospho-IIB Mtl (C384S)-The mannitol transporter consists of a single polypeptide comprising three independent domains (C, B, and A from N to C terminus), connected by long flexible linkers (39,40). The linker between the B and A domains is 21 residues and extends from residues 472-492 of the full-length protein (10,14). We studied the interaction of constructs comprising the isolated B (residues 371-474) and A domains (490 -637) by 1 H-15 N correlation spectroscopy. The combined results from two titration experiments, in which unlabeled phospho-IIB Mtl (C384S) was titrated into 15 N-labeled IIA Mtl (H554Q) and vice versa, are displayed in Fig. 1. Exchange between the free proteins and the complex is fast on the chemical shift scale, and binding is readily monitored by following the chemical shift perturbation of the 15 N-labeled partner upon the addition of the unlabeled second protein. The binding of the two proteins is weak, and nonlinear least squares optimization of all of the titration data simultaneously yields an equilibrium dissociation constant, K d , of 3.7 Ϯ 0.2 mM. Essentially identical results were obtained using the wild type proteins.  NOEs in a three-dimensional 13 C-separated (F 2 )/ 12 C-filtered (F 3 ) NOE experiment recorded in D 2 O are specifically observed from protons attached to 13 C (in the F 1 dimension) to protons attached to 12 C (in the F 3 dimension). In A, IIA Mtl (H554Q) is 15 N/ 13 C-labeled, and phospho-IIB Mtl (C384S) is unlabeled; in B, phospho-IIB Mtl (C384S) is 15 N/ 13 C-labeled, and IIA Mtl (H554Q) is unlabeled. The asterisks denote residual diagonal (autocorrelation) peaks arising from the labeled partner. (Note that 13 C-decoupling is not employed in the F 3 acquisition dimension, and hence residual autocorrelation peaks are split into two components separated by ϳ130 Hz, corresponding to the 1 J CH coupling; similarly, any incompletely suppressed cross-peaks arising from very high intensity intramolecular NOEs within the labeled partner will also be split by ϳ130 Hz and are therefore easily distinguishable from intermolecular NOEs (50).) Residues of phospho-IIB Mtl (C384S) are labeled in italic type.

TABLE 1 Structural statistics
The notation of the NMR structures is as follows: ͗SA͘ are the final 270 simulated annealing structures, (SA) r is the restrained regularized mean structure.

͗SA͘
(SA) r The intermolecular repulsion energy is given by the value of the quartic van der Waals repulsion term calculated with a force constant of 4 kcal⅐mol Ϫ1 ⅐Å Ϫ4 and a van der Waals radius scale factor of 0.78. The intermolecular Lennard-Jones van der Waals interaction energy is calculated using the CHARMM19/20 parameters and is not included in the target function used to calculate the structures. d Defined as the average root mean square difference between the final 270 simulated annealing structures and the mean coordinates. The value quoted for the complete backbone provides only a measure of the precision with which the orientation and translation of the two proteins in the complex have been determined and does not take into account the backbone accuracy of the x-ray/NMR coordinates of IIA Mtl and the NMR coordinates of phospho-IIB Mtl used for conjoined rigid body/torsion angle dynamics docking. The accuracy of the x-ray coordinates of IIA Mtl (including NMR refinement of the four variable regions comprising residues 540 -543, 555-567, 580 -585, and 593-599 on the basis of backbone RDCs) is likely to be around 0.3 Å, judging from the crystallographic resolution and R-factor (10) as well as the dipolar coupling R-factors (this paper). The accuracy of the restrained regularized mean coordinates of phospho-IIB Mtl can be estimated from the coordinate precision (0.3 Å) and the values of 14 -16% for cross-validated N-H dipolar coupling R-factors in two alignment media (phage pf1 and neutral polyacrylamide gel (21)), which suggest that the coordinate accuracy is comparable with a 1.5-2 Å resolution crystal structure (19,45). The percentage of residues in the most favored region of the Ramachandran map (51) is 92% for IIA Mtl and 91% for phospho-IIB Mtl .

Number of experimental NMR restraints
Such weak yet specific binding may at first appear surprising but is in fact optimally tuned to the intact biological system. The average endto-end distance of an unstructured 21-residue linker (given by ͌C n nl 2 , where n is the number of residues, C n is the characteristic ratio, which has a value of ϳ7 for n ϭ 21, and l is the average C␣-C␣ distance, which has a value of 3.8 Å) (41) would be expected to be ϳ46 Å. If one considers one of the domains to be fixed in space, then the effect of the linker is to constrain the second domain to a sphere of average volume 4.1 ϫ 10 Ϫ22 liters, which corresponds to an effective concentration of ϳ4 mM. Thus, in the intact protein one would expect the A domain to be bound to the B domain ϳ50% of the time, which is perfectly reasonable, given that the B domain must use approximately the same interaction surface, centered around the active site cysteine at position 384, to accept the phosphoryl group from the A domain and subsequently transfer the phsophoryl group onto the incoming sugar on the C domain.
It is also worth noting that the intact mannitol transporter is a dimer, with dimerization occurring solely through the transmembrane C domain (42). No homodimeric interactions between A domains or B domains are likely to occur, since the isolated A and B domains do not self-associate in free solution, even at high concentrations (in the millimolar range) (14,18). Clearly, phosphoryl transfer between the A and B domains can occur via either intra-or intersubunit interactions between the A and B domains. Mutagenesis and complementation experiments indicate that whereas phosphoryl transfer can occur via both routes, intrasubunit phosphoryl transfer is dominant (43). The above argument is consistent with the latter experimental observation, since the volume in which the A domain would be constrained relative to the B domain would be larger in the context of the intersubunit than the intrasubunit interaction, and hence the occupancy of the intersubunit IIA Mtl -IIB Mtl complex would be expected to be less than that of the intrasubunit complex.
Structure Determination-The solution NMR structures of IIB Mtl (14) and phospho-IIB Mtl (C384S) (21) have been determined previously, and the only structural differences between the phosphorylated and unphosphorylated forms are confined to the active site loop (resides 383-393) and involve a backbone atomic root mean square shift of only ϳ0.7 Å. Extensive use of residual dipolar couplings (RDC) was made in these structure determinations (14,21), and cross-validation using RDCs in multiple alignment media (21) indicate that the accuracy of the coordinates is high, with cross-validated dipolar coupling R-factors (44) of 14 -16% (in phage pf1 and neutral polyacrylamide gel), which is equivalent to values expected for 1.5-2 Å resolution crystal structures  (21,45). The crystal structure of IIA Mtl has been determined at 2.0 Å resolution (10). The unit cell in the crystal comprises a dimer of dimers (note that IIA Mtl does not self-associate in solution even at the high concentrations employed by NMR) (18). The structure of the four molecules of IIA Mtl in the unit cell are very similar, with an atomic root mean square difference of 0.3 Å, but four regions (residues 540 -543, 555-567, 580 -585, and 593-599), comprising loops, exhibit differences between the two pairs of molecules in the unit cell with maximal backbone atomic displacements in the range 1.5-3 Å. The structure of these loops in solution, both free and bound to HPr, coincides with that found in molecule D of the crystal structure as judged from RDC analysis, and these four regions were refined independently using RDCs in two alignment media in the solution structure determination of the IIA Mtl -HPr complex (18). RDCs measured in neutral polyacrylamide gel on free IIA Mtl (H554Q) are in excellent agreement with the coordinates of IIA Mtl in the IIA Mtl -HPr complex (18) with a dipolar coupling R-factor of 17%, a correlation coefficient of 0.98, and no deviations above average for the RDCs in the region of the mutation, indicating that the structure remains unaltered within coordinate errors (45,46). Since the chemical shift perturbations for both proteins upon formation of the IIA Mtl (H554Q)-phospho-IIB Mtl (C384S) complex are very small, the largest being less than 0.3 ppm in 1 H (for Asp-454 of IIB Mtl ) and 1 ppm in 15 N (for Leu-611 of IIA Mtl ), we conclude that the backbone of both IIA Mtl (H554Q) and phospho-IIB Mtl (C384S) remains essentially unchanged upon complex formation. Consequently, the structure determination was carried out using conjoined rigid body/torsion angle dynamics (28) based on the coordinates of free phospho-IIB Mtl (C384S) (21) (Protein Data Bank accession code 1VRV) and the coordinates of IIA Mtl in the IIA Mtl -HPr complex (18) (Protein Data Bank accession code 1J6T). The coordinates of the backbone and noninterfacial side chains were held fixed, and rigid body docking with full torsional degrees of freedom for the interfacial side chains was driven by interproton distance restraints derived from intermolecular NOE data coupled with torsion angle restraints derived from both heteronuclear coupling constant and short mixing time NOE data. Examples of the quality of the intermolecular NOE data obtained from three-dimensional 13 C-separated/ 12 C-filtered NOE experiments that provide exclusively intermolecular NOEs from protons attached to 13 C on the 13 Clabeled protein to protons attached to 12 C on the unlabeled protein, are shown in Fig. 2. A summary of the structural statistics is given in Table  1, a superposition of the backbone of the final 270 simulated annealing structures is shown in Fig. 3A, and an atomic density probability map representation of some interfacial side chains is depicted in Fig. 3B.
The NMR experiments were carried out on samples comprising a 1:1 mixture of the proteins, each at a concentration of 3 mM. Under these conditions, only 30% of each protein is in the complex (i.e. the concentration of the complex in the sample is ϳ1 mM). This does not affect the observation of intermolecular NOEs. However, it does impede the use of RDCs to provide accurate and reliable information on the relative orientation of the two proteins in the complex. Since the observed RDCs in a fast exchanging system are a weighted average of the RDCs for the free and complexed protein, one could in principal back-calculate the RDCs for the pure complex on the basis of experimental RDCs measured on the free proteins and the mixture of the two proteins (19,47). However, this requires very accurate RDC measurements under nearly identical alignment conditions as well as accurate knowledge of the fraction of bound protein. Moreover, in a mixture of labeled and unlabeled partners, the fraction of the labeled partner (on which the RDCs are measured in the mixture) should typically exceed ϳ30% to ensure that the complex makes a significant contribution to the observed RDCs. Because, in this instance, the two proteins bind so weakly to one another, relatively high concentrations of the unlabeled partner (Ͼ2 mM) are required. Despite attempts using a variety of alignment media, we were not able to find experimental conditions that permitted reliable back-calculation of the RDCs due to either differential interaction of one of the partners with the alignment medium or differential perturbation of the alignment tensor by the unlabeled partner arising from molecular crowding at these protein concentrations.
Overall Structure of the Complex-Ribbon diagrams providing an overall view of the complex are shown in Fig. 4. The C␣-C␣ separation between the last ordered residue at the C terminus of phospho-IIB Mtl (residue 471) and the first ordered residue of IIA Mtl (residue 493) is 38 Å, a distance that can readily be accommodated by the flexible 21-residue linker. The interaction surface on IIA Mtl comprises 27 residues involving three segments of polypeptide chain: residues 538 -557 include the C-terminal end of helix ␣2, a small antiparallel ␤-sheet formed by strands ␤2 and ␤3 connected by a hairpin turn, and the active site residue at position 554; residues 480 -585 comprise a loop connecting strands ␤4 and ␤5; and residues 598 -612 encompass helix ␣3 (residues 600 -610). The interaction surface on IIB Mtl is made up of 18 residues located in four segments of polypeptide chain: residues 384-400 include the active site loop (residues 384 -389) and most of helix ␣1 (residues 390 -404); residue 414 is the C-terminal residue of strand ␤2; residues 430 -433 make up the first turn of helix ␣2; and residues 451-453 are located in the loop connecting strand ␤4 and helix ␣3. The residues comprising the interaction surface of IIB Mtl are highly conserved throughout Gram-positive and Gram-negative bacteria (48): 14 of 18 residues are conserved absolutely; two are subject to highly conservative changes and preserve the functional group (Arg-399 to lysine, His-430 to glutamine); and one residue, Gly-396, can be substituted conservatively by a serine. Only Lys-400 is subject to nonconservative substitutions from lysine in Gram-negative bacteria to asparagine or aspartate in Gram-positive bacteria. It is also worth noting that residues 382-394 comprising the complete active site loop and the first turn of helix ␣1 are conserved throughout.
The interaction surfaces of IIA Mtl and IIB Mtl are approximately circular in shape and comparable in size (28 ϫ 27 Å for IIA Mtl and 22 ϫ 27 Å for IIB Mtl ) and are composed of predominantly hydrophobic residues, with 60 -65% of the atoms being nonpolar. The interaction surface on IIA Mtl is concave, whereas that on IIB Mtl is largely convex, thereby providing overall complementarity of fit. The total accessible surface buried at the interface is 1575 Å 2 , of which 735 Å 2 originates from IIA Mtl and 840 Å 2 from IIB Mtl . The gap volume index (defined as the ratio of gap volume to interface accessible surface area) is 2.7, a value typical of optional complexes (i.e. heterocomplexes where the individual components of the complex can also exist as monomers) (49). The predominant intermolecular contacts between secondary structure elements involve a helix-helix interaction between helix ␣3 of IIA Mtl and helix ␣1 of IIB Mtl , oriented at an angle of ϳ60°.
The IIA Mtl -IIB Mtl Interface-A stereoview of the interface is shown in Fig. 5A, together with selected close-ups in Fig. 5, B and C; surface representations of the interfaces are given in Fig. 6A; and a schematic summary of the intermolecular contacts is provided in Fig. 7. A virtually complete, horseshoe-shaped ring of hydrophobic residues surrounds the active site phospho-Ser-384 of IIB Mtl (Fig. 6A, bottom panel). Several intermolecular interactions are noteworthy. The two methionines on IIB Mtl , Met-393 and Met-388, make extensive hydrophobic contacts to residues in two deep pockets located at the top and bottom halves, respectively, of the IIA Mtl interface (in the views shown in Figs. 5 and 6) that serve to anchor IIB Mtl onto IIA Mtl . The top pocket is formed by Leu-546, Gly-547, Glu-548, Ile-550, Ile-604, and Thr-608. The bottom pocket comprises Arg-538, the active site residue Gln-554, Val-557, His-600, and Ile-601. The residues in both pockets are either conserved or substituted conservatively in Gram-negative and positive bacteria (48). The close contacts between Met-338 of IIB Mtl and Arg-538 and Gln-554 of IIA Mtl serve to position the latter two residues such that the carboxyamide of Gln-554 is directed toward phospho-Ser-384, and the guanidino group of Arg-538 makes a direct salt bridge with the phosphate group of phospho-Ser-384. The critical role of Arg-538 is supported by its conservation throughout Gram-negative and Gram-positive bacteria, (48). The negative charge on Asp-385, located adjacent to phospho-Ser-384, is partially offset by a hydrogen bonding interaction with the hydroxyl group of Thr-542. There are also a number of weak complementary electrostatic interactions, asymmetrically situated along the outer edges of the interaction surfaces that facilitate the correct orientation of the two proteins. These include interactions between Lys-400 and Asp-612 and Asn-609, between Arg-399 and Thr-605 and Asn-609, between Asn-451 and Glu-582 and Asp-585, and between His-430 and Glu-581. These electrostatic interactions, however, are likely to be relatively weak, since the functional groups are separated by Ն4 Å.
The Phosphoryl Transition State-The phosphoryl transition state can be readily modeled by substituting Gln-554 for His and Ser-384 for Cys and minimizing the restrained regularized mean coordinates of the IIA Mtl -phospho-IIB Mtl complex subject to geometrical restraints for the His-phosphoryl-Cys transition state in conjunction with the experimental NMR restraints, only allowing the backbone of the active site and immediately adjacent residues (i.e. residues 383-385 of IIA Mtl and 553-555 of IIB Mtl ) and the side chains of the interfacial residues to move (15). The geometric restraints include N-P and S-P bond lengths, a planarity term to ensure that the phosphorus atom lies in the plane of the imidazole ring of His-554 and the sulfur atom of Cys-384, and bond angle terms to enforce trigonal bipyramidal geometry for the phosphoryl group (15). A dissociative transition state complex (N-P and S-P bond lengths given by the sum of the van der Waals radii of the atoms, 3.4 and 3.7 Å, respectively) can be formed with minimal atomic root mean square shifts of the backbone in the vicinity of the active site residues (Ͻ0.2 Å) and compensatory minor displacements of the side chains of Cys-384, Met-388, Arg-438, and His-554 to accommodate the change in the position of the phosphoryl group (Fig. 8). An associative transition state complex (N-P and S-P bond lengths of 2.4 and 2.8 Å) can also be formed but involves slightly larger backbone displacements (0.33 Å for residues 383-385 and 0.22 Å for residues 553-555). The transition state thus preserves all of the intermolecular interactions seen in the post-transition state analogue, including the neutralization of the phosphate group by the guanidino group of Arg-438. However, the hydrogen bonds to the phosphoryl group from the hydroxyl groups of Mechanism of Phosphoryl Transfer-In vivo phosphoryl transfer proceeds from IIA Mtl to IIB Mtl . A proposed mechanism for phosphoryl transfer is shown in Fig. 9. Modeling phospho-IIA Mtl on the basis of the crystal structure of IIA Mtl (10) suggests that the phosphoryl group, bonded to the N⑀2 atom of His-554, accepts three hydrogen bonds: two from the guanidino group of Arg-538 and a potential third from the imidazole ring of His-600 (N⑀2 atom) if the 1/ 2 conformation of the latter is changed from g Ϫ /g Ϫ (seen in the IIA Mtlphospho-IIB Mtl complex) to g ϩ /g Ϫ . (Note in the crystal structure of IIA Mtl both rotamers are observed (10).) The N␦1-H atom of His-554 accepts a hydrogen bond from the backbone of Val-452, thereby stabilizing the N␦1-H tautomeric state. The thiolate state of Cys-384 of IIB Mtl is stabilized by numerous hydrogen bonding interactions within the active site loop (14). Upon formation of the pretransition state phospho-IIA Mtl -IIB Mtl encounter complex, the hydrogen bond  between the imidazole ring of His-600 and the phosphoryl group is broken due to steric clash with IIB Mtl , which precludes the g ϩ /g Ϫ conformation. Nucleophilic attack at the phosphoryl group by the thiolate of Cys-384 results in the formation of a transition state in which the phosphoryl group accepts two hydrogen bonds from the backbone amide of Met-388 and Gly-389 of IIA Mtl and one hydrogen bond from the guanidino group of Arg-538 of IIA Mtl (Fig. 8). Resolution of the transition state to the post-transition state complex results in further electrostatic interactions in which the phosphoryl group accepts five hydrogen bonds from the active site loop of IIB Mtl (the backbone amides of Met-388, Gly-389, and Ser-391 and the hydroxyl groups of Ser-390 and Ser-391) and one from the guanidino group of Arg-538 of IIA Mtl and is partially neutralized by the positive helix dipole at the N terminus of helix ␣1 of IIB Mtl . Thus, the number of hydrogen bonding/electrostatic interactions involving the phosphoryl group in the post-transition state IIA Mtl -phospho-IIB Mtl complex is larger than in the pretransition state phospho-IIA Mtl -IIB Mtl complex. In addition, in the transition state, more hydrogen bonds to the phosphoryl group originate from IIB Mtl than IIA Mtl . Thus, whereas phosphoryl transfer between IIA Mtl and IIB Mtl is fully reversible in vitro, one would expect the flow of the phosphoryl group from IIA Mtl to IIB Mtl to be favored, in accord with the biological function of the PTS pathway.
Comparison of the Interactions of IIB Mtl and HPr with IIA Mtl -A comparison of the IIA Mtl (H554Q)-phospho-IIB Mtl (C384S) and IIA Mtl -HPr (18) complexes is provided in Figs. 6 and 7. Of the 27 residues of IIA Mtl that interact with IIB Mtl and the 25 that interact with HPr, 23 are shared by the two interfaces (Fig. 7). Thus, IIA Mtl uses essentially the same interaction surface to recognize two structurally different proteins, HPr and IIB Mtl . The shapes of the IIB Mtl and HPr interaction surfaces are similar but bear no similarity to one another in terms of the underlying backbone topology or the orientation of secondary structure elements relative to the IIA Mtl surface (Fig. 6). Indeed, helix ␣2 of IIA Mtl lies approximately orthogonal to helices 1 and 2 of HPr (cf. top panels of Fig.  6, A and B). There are, however, a number of key differences between the interaction surfaces of IIB Mtl and HPr that are noteworthy. First, the number of positively charged residues on the interaction surface of IIB Mtl is half that for HPr (2 compared with 4). Three of the positively charged residues on the HPr interaction surface are involved in electrostatic interactions with negatively charged residues on IIA Mtl , whereas for IIB Mtl , only Lys-400 interacts with a negatively charged residue (Asp-612) on IIA Mtl . This probably accounts for the much lower affinity (about 2 orders of magnitude) of the IIA Mtl -IIB Mtl complex relative to the IIA Mtl -HPr complex (18). Second, the interaction surface on IIB Mtl is significantly more hydrophobic than that on HPr (Fig. 6, A and B, bottom panels). Finally, the interaction surface on IIB Mtl includes a nega- tively charged residue (Asp-385), whereas there are no negative charges in the case of HPr. The latter is probably important in ensuring that direct interaction and phosphoryl transfer between IIB Mtl and enzyme I, bypassing HPr and IIA Mtl , does not occur; the presence of a negatively charged residue in close proximity to the active site residue in IIB Mtl would lead to unfavorable electrostatic interactions with the interaction surface on enzyme I that contacts HPr (16).
Comparison of the IIA Mtl (H554Q)-phospho-IIB Mtl (C384S) and IIA Glc -IIB Glc Complexes-Although the A and B components of enzymes II Glc and II Mtl bear no sequence, secondary structure or topological similarity to one another (2,3,5,6,14,15), the two complexes share a number of features in common as well as some significant differences. In both cases, phosphoryl transfer occurs from a histidine on the A domain to a cysteine on the B domain (2,3). Unlike II Mtl , the A and B domains of II Glc are expressed as separate polypeptides (2,3). The general surface features of the binding interfaces on IIA Glc and IIA Mtl for their target proteins (HPr and the corresponding B domains) are broadly similar in terms of size and shape, each comprising a central hydrophobic region surrounding the active site residue and an outer ring of charged residues (this work) (15). However, the IIA Glc interface has a much larger preponderance of negatively charged residues (8 versus 4 for IIA Mt ; cf. Ref. 15). The active site loops of IIB Glc and IIB Mtl display similarities in so far that the thiolate state of the active site cysteine is stabilized by hydrogen bonding interactions with backbone amide protons located in the active site loop, reminiscent of the active site loop of eukaryotic protein tyrosine phosphatases (14,15). However, the structural correspondence with the latter is far more extensive for IIB Mtl (14,21) than IIB Glc (15). The affinity of IIA Glc for IIB Glc (15) is about 2 orders of magnitude higher than that of IIA Mtl for IIB Mtl . From a functional perspective, this is fully consistent with the fact that IIA Glc and IIB Glc are expressed as individual proteins (2, 3), whereas IIA Mtl and IIB Mtl are expressed as a single polypeptide chain connected by a long flexible linker (3). Examination of the two interfaces suggests that the difference in affinities can be largely attributed to electrostatic interactions. In the case of the IIA Glc -IIB Glc complex, there are six complementary interactions between positively and negatively charged residues (15), whereas there is only a single such interaction for the IIA Mtl -IIB Mtl complex (Fig. 6A).
Conclusion-The structure determination of the IIA Mtl (H554Q)phospho-IIB Mtl (C384S) complex completes the cytoplasmic complexes of the mannitol branch of the PTS. The structure explains how IIA Mtl uses the same interface to recognize both its upstream, HPr (18), and downstream, IIB Mtl (this paper), interaction partners, although structurally HPr and IIB Mtl are completely dissimilar, and provides a rationalization for the very different affinities of the two complexes. Thus, surface complementarity is largely achieved through hydrophobic interactions, while electrostatic interactions serve to modulate affinity. The orientation of two proteins within a protein-protein complex is clearly determined by the cumulative effect of numerous interactions, each of which makes a small contribution to the overall interaction energy. However, the structures suggest that the two complexes may use different initial determinants to drive the docking of the proteins in the correct configuration. For the IIA Mtl -HPr complex, it seems likely that electrostatic interactions between oppositely charged residues, distributed asymmetrically at judicious locations along the outer edge of the interaction surface, may play a key role in the initial docking event. Thus, a cluster of two positive charges at one edge of the HPr interaction surface (Lys-24 and Lys-27) complement a cluster of two negative charges on IIA Mtl (Glu-591 and Asp-585, respectively), whereas a single charge on the opposite edge of the HPr interaction surface, Lys-49, matches a single charge (Asp-612) on IIA Mtl (Figs. 6B and 7). For the IIA Mtl (H554Q)-phospho-IIB Mtl (C384S) complex, on the other hand, the electrostatic interactions are clearly very weak, and interactions between complementary, localized hydrophobic features of the interaction surfaces may be more important, with the two conserved methionines of IIB Mtl serving as hooks that latch on to the two deep hydrophobic pockets on the surface of IIA Mtl (Figs. 5, B and C, and 6A). Methionine is ideally suited to such a role, since its unbranched hydrophobic side chain can sample extensive configurational space, thereby permitting optimization of its interactions with a target surface. Indeed, extensive use of methionines is employed by calmodulin to recognize many different interaction partners (50). Thus, like other complexes of the PTS, the IIA Mtl (H554Q)-phospho-IIB Mtl (C384S) complex illustrates the versatility of a protein interaction network in which each protein recognizes its upstream and downstream partner using the same interaction surface. Key features of these interactions are complementarity of shape and residue type that can be achieved using a wide array of underlying structural elements; surface side-chain conformational plasticity, particularly involving long side chains (both hydrophobic and charged), to optimize intermolecular interactions; asymmetric distribution of complementary, intermolecular electrostatic interactions to aid in guiding correct docking; and finally extensive redundancy, thereby ensuring that the overall interaction energy is not dominated by any single interaction but by a multitude of interactions, each contributing only a small proportion of the total interaction energy. The last is important, since, as clearly illustrated by the structures of various protein-protein complexes of this PTS (this paper) (15)(16)(17)(18)(19), not every charged residue located in the binding surface of a particular PTS protein need be involved in electrostatic interactions with complementary residues of its diverse interaction partners. Whereas the presence of inherent redundancy does not impact specificity, it will generally result in relatively weak protein-protein interactions. In the case of the PTS, such weak interactions, which are tuned to the microto millimolar range, are critical to function that requires rapid dissociation of transient protein-protein complexes to permit efficient phosphoryl transfer down the reaction cascade.