A Hydrophobic Pocket in the Active Site of Glycolytic Aldolase Mediates Interactions with Wiskott-Aldrich Syndrome Protein*

Aldolase plays essential catalytic roles in glycolysis and gluconeogenesis. However, aldolase is a highly abundant protein that is remarkably promiscuous in its interactions with other cellular proteins. In particular, aldolase binds to highly acidic amino acid sequences, including the C terminus of the Wiskott-Aldrich syndrome protein, an actin nucleation-promoting factor. Here we report the crystal structure of tetrameric rabbit muscle aldolase in complex with a C-terminal peptide of Wiskott-Aldrich syndrome protein. Aldolase recognizes a short, four-residue DEWD motif (residues 498-501), which adopts a loose hairpin turn that folds around the central aromatic residue, enabling its tryptophan side chain to fit into a hydrophobic pocket in the active site of aldolase. The flanking acidic residues in this binding motif provide further interactions with conserved aldolase active site residues Arg-42 and Arg-303, aligning their side chains and forming the sides of the hydrophobic pocket. The binding of Wiskott-Aldrich syndrome protein to aldolase precludes intramolecular interactions of its C terminus with its active site and is competitive with substrate as well as with binding by actin and cortactin. Finally, based on this structure, a novel naphthol phosphate-based inhibitor of aldolase was identified, and its structure in complex with aldolase demonstrated mimicry of the Wiskott-Aldrich syndrome protein-aldolase interaction. The data support a model whereby aldolase exists in distinct forms that regulate glycolysis or actin dynamics.

Arp2/3 complex and induces conformational changes necessary for actin polymerization (15), but this domain also plays a role in the autoinhibition of WASP proteins (16), whereas the verprolin homology domain binds to actin monomers (17). The acidic domain of WASP family proteins is also required for binding to Arp2/3 (18), and it directs the interactions of WASP proteins with aldolase (19), suggesting that this interaction directs the actin cross-linking functions of aldolase.
The binding of aldolase with WASP proteins is competitively inhibited by its substrates and products, suggesting that the acidic domain of WASP interacts with the active site of aldolase. Here we report the crystal structure of rabbit muscle aldolase in complex with the 15 C-terminal residues of WASP. This structure reveals that the WASP interaction indeed disrupts the binding of substrates to the active site of aldolase and that WASP competes with the aldolase C terminus by interacting with the periphery of the active site. Finally, using the WASP-aldolase structure as a guide, we identified and solved the structure of aldolase in complex with a novel competitive inhibitor of the enzyme.
The results support a model where the regulation of actin dynamics is linked to the intermediary metabolism by aldolase.

EXPERIMENTAL PROCEDURES
Peptides and Inhibitors-WASP peptide (EDQAGDEDED-DEWDD) was synthesized by the Hartwell Center (St. Jude Children's Research Hospital, Memphis, TN). Naphthol AS-E phosphate (NASEP), C 17 H 13 NO 5 ClP, was purchased from Research Organic Inc.
Purification and Crystallization-Recombinant rabbit muscle aldolase was expressed and purified as described previously (20,21). Aldolase concentration was determined using an extinction coefficient of 0.91 (mg/ml) Ϫ1 at 280 nm (22). Enzymatic activity was monitored by spectrophotometry (23), and native aldolase was crystallized as described previously (24). The WASP-aldolase complex was co-crystallized by vapor diffusion from a 1:1 mixture of protein-peptide solution (0.084 mM aldolase tetramer and 1.5 mM WASP peptide in 20 mM Tris-HCl, pH 7.0) and precipitant solution (24% polyethylene glycol 550 monomethyl ether, 50 mM MgCl 2 , 0.1 M HEPES pH 7.5). This mixture was then equilibrated against a reservoir of precipitant solution.
Data Collection and Processing-A native aldolase crystal was soaked in a NASEP solution (mother liquor plus 3 mM NASEP) for 8 min. The crystal was cryoprotected by transfer through a cryobuffer solution (NASEP solution plus 20% glycerol) and flash-frozen in a stream of gaseous N 2 . Diffraction data were collected at beamline X8-C of the National Synchrotron Light Source (Brookhaven National Laboratory, Upton New York). A crystal of the WASP-aldolase complex was cryoprotected by transfer through a cryobuffer solution (mother liquor plus 11% polyethylene glycol 550 monomethyl ether) and flash-frozen in a stream of gaseous N 2 . Diffraction data were collected at beamline 22-BM at the Advanced Photon Source (Argonne National Laboratory, Argonne, IL). Data sets were processed with HKL2000 (25), and the results are summarized in Table 1.
Structure Determination and Refinement-The crystal soaked with NASEP inhibitor was isomorphous to the native aldolase (Protein Data Bank (PDB) code 1ZAH) (24) with monoclinic space group P2 1 . The WASP-aldolase complex crystal was also a P2 1 space group, and its structure was solved by molecular replacement using the structure of a native aldolase homotetramer (PDB code 1ZAH) as a starting model. Each asymmetric unit contains one aldolase homotetramer, which is the active form of the enzyme. All observed reflections having an I/(I) value of Ͼ1 were used in the refinement process. However, statistics and electron density maps were calculated to 2.05 Å, a resolution limit corresponding to the completeness of better than 70% in the highest resolution shell ( Table 1). The structures were subjected to iterative rounds of refinement with the Crystallography and NMR System (26) and model building using O (27). A portion of the aldolase C-terminal sequence between residues 346 and 357 rich in alanine and glycine amino acids was associated with weak electron density in both WASP-aldolase and NASEP-aldolase and was omitted from the refined model. Residues 358 -363 were clearly discernable in a simulated annealing electron density omit map.
Ligand modeling was based on interpretation of electron density shapes of an F o Ϫ F c -simulated annealing omit map and using PRODRG for topology and parameter generation (28). Binding by WASP peptide residues (498 -501) and NASEP were readily discernable and were associated with clearly defined electron densities in the active site. Difference electron density (F o Ϫ F c )-annealed omit maps calculated in the final round of refinement confirmed identical binding of ligands in all four subunits. Model statistics were evaluated with crystallography NMR software and PROCHECK (29) ( Table 1). The refined structures of the WASP-aldolase and the NASEP-aldolase complexes have, respectively, R cryst (R free ) of 0.152 (0.200) and 0.149 (0.197), with corresponding Luzzati atomic coordinate errors evaluated at 0.18 and 0.17 Å. Ramachandran analysis of structures with PROCHECK placed 91.8% (WASP-aldolase) and 92.5% (NASEP-aldolase) of nonglycine and nonproline residues in the most favorable region and the remainder in the allowed region, attesting to good model geometry in the structures.
The coordinates and structure factors of the WASP-aldolase complex and NASEP-soaked aldolase have been deposited in the Protein Data Bank under PDB codes 2OT0 and 2OT1, respectively. All figures in the present paper were prepared using the program PyMOL (30).
Structure Comparisons-Structure superposition was performed using PyMOL. To evaluate the overlap of binding loci between the aldolase substrate FBP and WASP and the conformational changes upon WASP peptide binding, the WASP-aldolase complex was superposed on the structure of aldolase covalently bound to FBP as a Schiff base intermediate (PDB code 1ZAI) (24) using C␣ atoms of residues 101-200. Residues common to both structures were selected on the basis of the lowest root mean square deviation upon superposition and comparable with the error in atomic coordinates of the model.
WASP Homologue Modeling-The C terminus region of T. gondii MIC2 (residues 765-768) and that of P. falciparum TRAP (residues 556 -559), both homologous to the WASP pep-tide, were threaded into the WASP peptide structure using PyMOL. Side chain conformations were selected from the PyMOL rotamer library that resulted in no steric clashes with aldolase active site residues.

RESULTS
The Structure of the WASP-Aldolase Complex-The interaction between the WASP C-terminal domain and glycolytic aldolase requires the penultimate tryptophan residue in the C terminus of WASP (19). This interaction is displaced by aldolase substrates, suggesting the aldolase active site also binds to WASP. However, precisely how aldolase interacts with WASP is unknown. To define this interaction, rabbit muscle aldolase was co-crystallized with a 15-mer WASP C-terminal peptide corresponding to the amino acid sequence EDQAGDEDEDDEWDD (residues 488 -502).
The structure of the WASP peptide in complex with aldolase was solved to 2.05 Å resolution by molecular replacement using a native aldolase tetramer (PDB code 1ZAH) as the search model. Crystallographic refinement yielded a model with R cryst and R free values of 15.2 and 20.0%, respectively (Table  1). A difference (F o Ϫ F c )-simulated annealing omit map was used to build the WASP peptide in the active sites of tetrameric aldolase ( Fig. 1). Only WASP residues 498 -501 could be traced from the electron density and were included in the model. Residues 488 -497 and 502 appeared disordered and were not fitted into electron density. The modeled WASP peptide adopted the same conformation in all aldolase subunits and the N-to C-terminal orientation of the peptide was unambiguously assigned on the basis of the trace of residues 499 -501 in the electron density map. The relatively high average B-factor of the WASP peptides (60 Ϯ 7 Å 2 ) in comparison with the overall B-factor for the protein (23 Ϯ 10 Å 2 ) suggests a loose binding of WASP for the aldolase active site. To verify this weak binding, a crystal of the WASP-aldolase complex was soaked in the presence of a fresh 10 mM WASP peptide solution for 2 min, and the electron density map calculated from data collected on this crystal revealed no new additional FIGURE 1. WASP peptide bound to the rabbit muscle aldolase active site. The difference electron density shown was calculated from a 2.05-Å simulated annealing F o Ϫ F c omit map contoured at 3.0 level and encompasses Arg-42, Arg-303, and WASP. WASP peptide is depicted in orange. Only WASP residues 498 -501 were visible in the electron density map and included in the model. The green and magenta dashes illustrate hydrogen bonds and hydrophobic contacts, respectively.
where T is a test data set randomly selected from the observed reflections prior to refinement. Test data set was not used throughout refinement and contained 9% of the total unique reflections. e Analyzed by PROCHECK (29). features (data not shown). These data are consistent with a low ordered binding to aldolase by the WASP 15-mer peptide under these crystallization conditions. The binding site used by the WASP peptide delineates a novel hydrophobic pocket within the aldolase active site. This pocket is generated by the side chains of Arg-42 and Arg-303, which loosely sandwich (3.6 Ϯ 0.1-3.8 Ϯ 0.1 Å) the planar tryptophan side chain between them, and also by the Gln-306 side chain that contributes (3.8 Ϯ 0.1 Å) to the hydrophobic environment that surrounds the tryptophan side chain (Fig. 1,  magenta dashes). Moreover, the nitrogen atom of the tryptophan indole ring hydrogen bonds to the carboxylate group of the Glu-34 side chain. These two binding features, characteristic of a tryptophan side chain, account for the requirement of a tryptophan residue in the C-terminal domain of WASP (19). The WASP peptide contains a large number of acidic residues that are complementary to the essential basic character of the aldolase active site, which has a preponderance of lysine and arginine residues. Indeed, the carboxylate of the WASP Glu-499 side chain makes a salt bridge with aldolase Lys-146 N, WASP Asp-498 makes electrostatic interactions with aldolase Lys-41 N and Arg-42 NH 1 , and its carbonyl backbone makes a weak contact with the Arg-148 side chain. Arg-303 further stabilizes the peptide by interacting with WASP Trp-500 backbone carbonyl. However, the binding geometry (donor-acceptor angle) for these interactions is not optimal (31), as they do not fully respect the atom hybridization state, consistent with the low affinity of the peptide to aldolase active site. Hydrogen bonding involving other peptide moieties was not observed.
The NASEP-Aldolase Complex-To further probe the WASP binding site on aldolase, a native aldolase crystal was soaked with an inhibitor that mimics the tryptophan residue of WASP and which contains a negatively charged group. A family of naphthol phosphate-based inhibitors, differing in the substituents of the phenyl moiety (from Research Organic, Inc.), were tested on the basis of competitive inhibition of aldolase. The best inhibitor (K i Ϸ 0.1 mM), NASEP, containing a chlorine atom, was selected.
The aldolase crystal soaked in a NASEP solution was isomorphous with the native crystal form, and its structure was solved to 2.05 Å resolution by molecular replacement using a native aldolase tetramer (PDB code 1ZAH) as the search model. The crystallographic refinement resulted in a model with R cryst and R free values of 14.9 and 19.7%, respectively ( Table 1). The difference (F o Ϫ F c )-simulated annealing omit map confirmed the presence of the inhibitor in the active sites of aldolase (Fig. 2). The binding mode was identical in all aldolase subunits.
The NASEP inhibitor utilizes the same hydrophobic pocket as the WASP peptide, again generated by the Arg-303 and Arg-42 side chains. Moreover, the NASEP phosphate group is stabilized by salt bridging one of the two alternate conformations displayed by Arg-303. The NASEP phosphate group also hydrogen bonds to the Gly-272 backbone carbonyl through a water molecule (not shown).
A Novel Binding Site in the Active Site of Aldolase-The most salient feature in the binding of both WASP peptide and NASEP inhibitor is their use of the same hydrophobic pocket (Fig. 3). To assess the minimum binding requirements, a native aldolase crystal was also soaked in a mother liquor solution containing 6 mM L-tryptophan for 7 min. However, electron density maps calculated from data collected to 2.0 Å resolution failed to reveal any distinguishing features (data not shown), underscoring the contribution of acidic residues of the ligand in binding. Thus, although a planar aromatic moiety is necessary for binding, it is not sufficient to ensure binding.
WASP Peptide Competes with the Substrate and the C Terminus of Aldolase-The effect of WASP peptide on substrate binding was evaluated by superposing the WASP-aldolase complex structure on the structure of aldolase covalently bound to its substrate FBP (PDB code 1ZAI) (Fig. 4). Active site adaptation is slight and requires only the reorganization of Arg-303 and Arg-42 side chains to create the hydrophobic pocket to accommodate the tryptophan side chain. The Arg-303 side chain stabilizes attachment by the FBP P1 phosphate and adopts a conformation that results in steric clashes with the WASP Trp-500 side chain and the WASP backbone. The Arg-42 side chain in the FBP-aldolase structure also collides with WASP Trp-500. Moreover, the WASP Glu-499 side chain clashes with FBP C4, C5, and C6 atoms. The binding of aldolase by WASP peptide and by FBP as well as by aldolase triose phosphate products are thus physically incompatible within the aldolase active site.
Furthermore, Fig. 4 shows that the WASP peptide clashes with the C terminus of aldolase, which can interact with the active site periphery vicinal to Arg-303. This conformation of the aldolase C terminus is also observed in the native enzyme (PDB code 1ZAH) (not shown). The C-terminal region mediates proton exchange at the level of the enamine intermediate in the aldolase catalytic mechanism (32), and the aldolase C-terminal residue Tyr-363 is essential for this proton transfer (33). The docking of the WASP peptide and the C-terminal region of aldolase to the active site appears mutually exclusive, as indicated by the conformation of the aldolase C termini in the crystal structure of the WASP-aldolase complex, which do not interact with the active site binding motif.
Modeling WASP C-terminal Homologues-To evaluate whether the interaction of WASP peptide with the active site of aldolase shared features with other peptides known to bind Schiff base-forming aldolases, the cytoplasmic C-terminal tail of the MIC2 protein from T. gondi (9) and that of the TRAP protein from P. falciparum (10) were modeled with PyMOL in the binding pocket using WASP coordinates as the template. Modeling consisted of threading the homologous MIC2 and TRAP C-terminal sequences onto the WASP peptide. Side chain rotamers could be selected from the PyMOL library that resulted in no structural clashes (Fig. 5). This strongly suggests a common binding site on aldolase for these different interacting partners, with each possessing a conserved tryptophan residue in their C-terminal domain and proximal acidic residues.

DISCUSSION
Aldolase exhibits functional duality as a glycolytic enzyme that also interacts with diverse cellular and pathogenic partners. Here we defined the structure of the WASPaldolase complex, which identified a novel binding site on the aldolase surface. Further, this structure supports the concept that interactions of aldolase with the acidic domain of WASP may well regulate its functions in binding to and activating the Arp2/3 complex (34). Finally, this structure establishes a paradigm for the interactions of acidic peptides folded around a conserved tryptophan residue in a loose ␤-like turn, which here direct the interaction of WASP with the active site of aldolase.
Hydrophobic Interactions Mediate WASP Binding to Aldolase-Our structure demonstrates that binding of aldolase by the WASP peptide requires a conserved tryptophan residue, Trp-500. In turn, this residue loosely docks into a hydrophobic pocket on the aldolase surface that is formed by the aliphatic carbons of the side chains of basic residues in the active site of aldolase as well as by interactions with other conserved residues  . WASP competes with the aldolase substrate and C terminus in binding to aldolase active site vicinity. WASP peptide is depicted in orange. Superimposed in cyan is shown the covalent Schiff base intermediate formed between the substrate FBP and muscle aldolase in another crystal form (PDB code 1ZAI). Steric clashes occur between side chains of Trp-500 in WASP and both Arg-42 and Arg-303 of aldolase when FBP is bound in the active site. Additionally, the WASP Glu-499 side chain occupies the same position as the FBP C 4 , C 5 , and C 6 atoms. Attachment of triose phosphate products dihydroxyacetone phosphate and glyceraldehyde 3-phosphate would also be similarly compromised by WASP binding in the active site. WASP binding does not induce major conformational changes with respect to the native or FBP-bound enzyme forms. Only the side chains of Arg-42 and Arg-303 reorganize to facilitate WASP binding. WASP binding also conflicts with active site binding of the aldolase C-terminal region. In the WASP-aldolase complex, each C terminus residue Tyr-363 interacts identically with Arg-258 found in a subunit interface of an adjacent tetramer (not shown). In the crystal form of the native enzyme and FBP-bound aldolase, the C-terminal region in each subunit adopts different conformations, each capable of interacting with the active site, and in one subunit, C-terminal binding was clearly observed within the active site periphery. The interaction of the C-terminal region with the active site is consistent with a mechanistic role of Tyr-363 in mediating proton exchange during catalysis and in the crystalline state (M. St-Jean, and J. Sygusch, unpublished data). Binding of the C-terminal region at the active site periphery thus interferes with WASP binding in the crystal form of native aldolase and FBP-bound aldolase.
( Fig. 1). Although potentially electrostatic in nature, these interactions do not exhibit optimal binding geometry, and their number and strengths suggest a rather low affinity of WASP for aldolase. These rather modest hydrophobic interactions also explain the low affinity binding for the homologous TRAP 13-mer peptide with aldolase (10), whose modeled structure superimposes with that of WASP and aldolase (Fig. 5).
WASP Binding Competes with Active Site Ligands-The structure of the WASP-aldolase complex also demonstrates that binding of this enzyme substrate and products is mutually exclusive with WASP. Structural clashes also occur between WASP and the C-terminal region of aldolase (Fig. 4), which interacts in an intramolecular fashion with the active site of the enzyme. However, interference by the C-terminal regions at the periphery of the active site does not take place in the WASPaldolase complex structure, where the C-terminal regions for each subunit are distant from the active site and make crystal contacts with adjacent tetramers. Therefore, aldolase appears to exist in at least three, mutually exclusive distinct states, native versus substrate/product-or WASP-bound. In the cell, the intramolecular nature of the C terminus interaction with its active site would increase its effective concentration (35) and would thus be predicted to inhibit the binding of weak interacting competitors. WASP binding to aldolase would thus be pre-dicted to require a combination of low substrate concentrations, molecular crowding, and possibly sequestration of the aldolase C-terminal region, distal to the active site binding motif. Nonetheless, given the abundance of aldolase one would predict that such scenarios are frequent in the cell and facilitate aldolase-WASP interactions.
Specific Inhibition of the Aldolase-WASP Interaction-Guided by its similarity to the tryptophan residue of WASP and its ability to competitively inhibit aldolase, we identified a naphthol phosphate-based inhibitor coined NASEP. Our structure of the NASEP-aldolase complex underscored the specificity and the novel features of this binding locus with respect to the active site. Indeed, here the aromatic moiety of the inhibitor competes for the same hydrophobic pocket as the WASP peptide and utilizes the same mode of binding (Fig. 3). Tryptophan, however, does not bind at similar concentrations, indicating that interaction by the NASEP phosphate group with conserved Arg-303 is required for attachment (Fig. 2). Thus, this inhibitor has the potential to interfere with any WASP-type interactions made by aldolase and would thus be predicted to disable proper regulation of actin dynamics, a notion which clearly needs to be put to the test.
Binding Modes of the Aldolase Partners-Other cellular partners known to bind to aldolase share a common acidic motif with a central tryptophan, including cortactin (DDWET) (19), Bloom syndrome protein (DDDWED) (19), and sorting nexin 9 (DDWDEDWDG) (36), suggesting they all bind to aldolase in a fashion similar to that described here for WASP. Further, our modeling of the interactions of WASP homologues TRAP and MIC2 with aldolase predicts a similar structure and mode of binding. Similar to its interactions with WASP, the binding of aldolase to F-and G-actin also requires Arg-42 and Arg-148 (37)(38)(39)(40). However, in contrast to WASP binding, actin binding to aldolase is not disrupted but merely weakened (ϳ10-fold) by FBP (41). Nevertheless, the fact that binding by aldolase reactants diminishes both actin and WASP binding to aldolase (19,40) indicates overlapping features in their binding to aldolase.
Role of the Aldolase-WASP Interaction in Actin Dynamics-Aldolase accounts for up to 7% of soluble protein within the cell (42) and its broad cytoplasmic distribution (5) may promote the recruitment of WASP and its many other cytoskeletal partners. Indeed, aldolase localizes to active ruffles in the leading edge of motile cells (43) along with N-WASP (44,45). Molecular crowding of aldolase in such a scenario would inhibit actin dynamics by favoring sequestration of actin monomers, cortactin, and WASP by aldolase. This is consistent with viscometric findings, where equimolar concentrations of aldolase inhibits actin polymerization (41) and reverses actin gelling (40). However, FBP blocks these effects, indicating that FBP destabilizes aldolase complexes with actin or WASP homologues (9,10). Thus, the release of aldolase by substrate suggests that, in the cell, activation of glycolysis would simultaneously liberate WASP to activate Arp2/3 and increase the pool of free actin monomers, facilitating actin nucleation.
The interactions of aldolase with WASP and its homologues are likely to occur at the onset of actin polymerization (19,46), when glucose levels are low and glycolysis is not fully active. Residues 498 -501 corresponding to the sequence DEWD of the WASP peptide were used as a template to model the bound conformation of homologous peptides known to interact with homologous aldolases and also implicating a tryptophan residue. Shown in gray and yellow are aldolase and WASP residues 498 -501, respectively. Side chains of WASP residues were substituted using PyMOL software with side chains of the amino acids of homologous sequences DMWM (cyan) and NEWN (magenta), corresponding respectively to the cytoplasmic C-terminal tail of the MIC2 protein from T. gondi and that of the TRAP protein from P. falciparum. Each substitution was made using side chain conformations available in the PyMOL rotamer library and was devoid of steric clashes. The green dashes illustrate hydrogen bonds.
WASP-aldolase interactions would then transiently inhibit cortical F-actin polymerization by preventing the association of WASP with the Arp2/3 complex, which also requires the WASP acidic domain. In turn, as FBP levels rise and become sufficient to sustain ATP production, FBP binding would then displace WASP, leaving it free to nucleate actin polymerization. Such a model is concordant with the dependence of actin polymerization on ATP supplied by glycolysis (47), as uncoupling glycolysis from ATP generation results in actin polymerization and the accumulation of dispersed F-actin aggregates (48 -51). In this regard it is also noteworthy that FBP is a protective agent in experimental ischemia, where it appears to exert its action by inhibiting F-actin filament aggregation (52). Fluctuating levels of FBP thus may act as a switch that controls, in both space and time, the role of aldolase as a regulator of both intermediary cell metabolism and actin dynamics.