Regulatory Interactions between a Bacterial Tyrosine Kinase and Its Cognate Phosphatase*

Background: Wzb phosphatase dephosphorylates Wzc tyrosine kinase in E. coli. Results: We have characterized the regulatory interactions between Wzb and Wzc. Conclusion: Wzb-catalyzed Wzc dephosphorylation relies on the increased local concentration of substrate due to mutual docking interactions. Significance: Structural elements that mediate the interactions between a bacterial tyrosine kinase and its cognate phosphatase have been ascertained for the very first time. The cyclic process of autophosphorylation of the C-terminal tyrosine cluster (YC) of a bacterial tyrosine kinase and its subsequent dephosphorylation following interactions with a counteracting tyrosine phosphatase regulates diverse physiological processes, including the biosynthesis and export of polysaccharides responsible for the formation of biofilms or virulence-determining capsules. We provide here the first detailed insight into this hitherto uncharacterized regulatory interaction at residue-specific resolution using Escherichia coli Wzc, a canonical bacterial tyrosine kinase, and its opposing tyrosine phosphatase, Wzb. The phosphatase Wzb utilizes a surface distal to the catalytic elements of the kinase, Wzc, to dock onto its catalytic domain (WzcCD). WzcCD binds in a largely YC-independent fashion near the Wzb catalytic site, inducing allosteric changes therein. YC dephosphorylation is proximity-mediated and reliant on the elevated concentration of phosphorylated YC near the Wzb active site resulting from WzcCD docking. Wzb principally recognizes the phosphate of its phosphotyrosine substrate and further stabilizes the tyrosine moiety through ring stacking interactions with a conserved active site tyrosine.

Tyrosine phosphorylation has been long realized to be a critical signaling mechanism in higher eukaryotes (1) that is responsible for almost every aspect of cell growth, differentiation, maturation, motility, and regulated cell death. On the other hand, although the existence of tyrosine phosphorylation in bacteria has been known for quite some time (2)(3)(4), its influence on bacterial physiology has been largely ignored. However, with mounting evidence of the role of tyrosine kinase signaling in the virulence of many pathogenic bacteria (5-7), its critical role in bacterial cell signaling is gaining prominence (8 -11).
The bacterial tyrosine kinase (BY-kinase) 4 family (10 -12) that is highly conserved (13) in both Gram-negative and Gram-positive species constitutes the largest family of protein-tyrosine kinases (PTKs) in bacteria. The central role of BY-kinase-mediated signal transduction in the synthesis and export of polysaccharides responsible for biofilm or capsule formation (9,10,12) is well established. In addition, BY-kinases also appear to participate in diverse cellular processes, including antibiotic resistance (14), lysogenization (15), and DNA metabolism (16), indicating that the extent of their influence on bacterial physiology could rival that seen for PTKs in life's critical processes in higher eukaryotes (17).
BY-kinases comprise a periplasmic/extracytoplasmic domain, two trans-membrane helices, and a cytoplasmic catalytic domain (CD) that is structurally distinct from those found in eukaryotic PTKs. Recently solved structures of the CDs of several BY-kinases, including Etk (18) and Wzc (19) (Fig. 1) from Escherichia coli and CapB from Staphylococcus aureus (20), confirmed their unique features highlighted by an absence of the two-lobed fold characteristic of the CDs of eukaryotic PTKs. Also striking was the close structural similarity of the BY-kinases to P-loop ATPases. BY-kinase CDs possess ATP-binding motifs reminiscent of the Walker motifs of P-loop ATPases instead of eukaryal kinase sequence motifs (21). These motifs include the following (Fig. 1): Walker A ((A/G)XXXXXGK(S/T); usually only the GK(S/T) element is conserved in BY-kinases); Walker B (DXXP, where is a hydro-phobic residue; the XXP sequence is a BY-kinase-specific extension of the Walker B motif); and an additional motif that is similar to a Walker B, called the Walker AЈ (DXDXR) (11,22).
Regulation of BY-kinases occurs through the intermolecular autophosphorylation and subsequent dephosphorylation of 3-7 tyrosine residues (the tyrosine cluster (YC)) located on the C-terminal tail (C-tail) (Fig. 1B) of the CD. Based on extensive biochemical and in vivo evidence (23), it is apparent that BY-kinases cannot be classified as being in "active" or "inactive" states, as for a majority of eukaryotic kinases, dependent simply on the phosphorylation state of key regulatory residues (24 -27). It is also clear that the overall level of YC phosphorylation rather than the quantitative phosphorylation state of any specific YC tyrosine is important for activity (28,29). Thus, the current model for BY-kinase activity suggests that activity is related to the cycling between a state where the YC is highly phosphorylated (YC high ) and a state where the level of YC phosphorylation is greatly reduced (YC low ); the presence of both of these states is essential in cellular physiology (10).
Generation of the YC high state is achieved by intermolecular autophosphorylation facilitated by the oligomerization of the cytoplasmic CDs of the BY-kinases in which the YC of one subunit (substrate-acting) is inserted into the catalytic site of the neighboring subunit (enzyme-acting) in the oligomer, as in the case of eukaryotic PTKs (30). A highly conserved sequence motif (EXXRXXR) has been shown to be critical in facilitating the selfassociation of the CDs of the BY-kinases necessary to accomplish trans-phosphorylation. There are three specific lines of evidence for this. (i) This motif has been shown to play a central role in generating the octameric oligomerization state seen in the crystal structures of the CDs of Wzc (19) as well as S. aureus CapB (20), hinting at a general mode of oligomerization and trans-phosphorylation. (ii) Although the isolated CDs of BY-kinases are largely monomeric in solution (as will be demonstrated below), analytical gel filtration revealed the presence of a minor population of a high molecular weight species for the CD of a Wzc mutant (K540M) that cannot bind ATP and hence is incapable of generating a YC high state. This high molecular weight species was eliminated by mutating the conserved residues of the EXXRXXR motif to alanine (19). (iii) This same triple mutation reduced the autophosphorylation levels of Wzc (in the context of full-length protein) to around 10% of wild type (19), suggesting its importance for trans-phosphorylation. Although the tendency of the isolated BY-kinase CDs to oligomerize in solution appears to be low, this low self-affinity is likely to be enhanced severalfold in the context of the membranebound fully assembled signaling complex that involves additional interactions involving the periplasmic domains and perhaps the membrane segments (31). However, it is to be noted that isolated BY-kinase CDs have been shown to form higher order oligomers in vivo (29) using formaldehyde cross-linking. Thus, based on the current state of knowledge, the following hypothesis for BY-kinase regulation may be proposed. An association of the CDs is followed by trans-phosphorylation and generation of a YC high state. The CDs in the YC high state then dissociate to expose the active site (occluded in the oligomeric state) to downstream targets, such as UDP glucose dehydrogenase, a known BY-kinase target in both Gram-negative (32) and Gram-positive species (33). Regeneration of the YC low state of the CD occurs by the action of protein-tyrosine phosphatases (PTPs), and this reinitiates the cycle. There-fore, this cycling between YC high and YC low states and consequently between associated and dissociated states of the CD appears key for maintaining function (10). Thus, the critical players in sustaining this YC high /YC low cycle that occurs in the cytoplasm are the bacterial PTPs. Several of these PTPs belong to the low molecular weight protein-tyrosine phosphatase (LMW-PTP, also known as class II PTP) family.
Bacterial LMW-PTPs are wholly cytosolic and eukaryoticlike in overall structure (34 -37) (Fig. 2), although they lack several key features considered to be important for substrate recognition in eukaryotes, suggesting unique mechanisms of kinase docking and regulation. Functional BY-kinase/LMW-PTP pairs in bacteria include Wzc/Wzb (38) and Etk/Etp (E. coli) (39), Ptk/Ptp (Acinetobacter johnsonii) (40), and Yco6/ Yor5 (Klebsiella pneumoniae) (41), to name a few that have been characterized. Despite extensive genetic and biochemical evidence for the critical role of BY-kinase/LMW-PTP interactions in bacterial tyrosine phosphorylation-mediated signaling, no information is available on the structural elements on BYkinases that are important for their interactions with LMW-PTPs. Further, although structural work on bacterial PTPs, including the solution structure of E. coli Wzb (34), has allowed the generation of specific hypotheses on phosphotyrosine recognition by these enzymes, information on the specific nature of substrate recognition and dephosphorylation is lacking.
Toward our long term goal of characterizing BY-kinase/ LMW-PTP interactions that are key in maintaining the critical cycling between YC high and YC low states in terms of structure, dynamics, and mechanism, we have chosen the canonical BYkinase E. coli Wzc and its opposing LMW-PTP, Wzb, as a prototypical BY-kinase/LMW-PTP pair.
wzb and wzc belong to a gene cluster (cps) (42) that is responsible for the synthesis and export of the exopolysaccharide, colanic acid (23,43), an essential component of biofilms and part of a physiological response to osmotic shock, damage to the cellular envelope, desiccation, and similar environmental stress conditions. For example, in the pathogenic enterohemorrhagic E. coli (O157:H7), the causative agent of acute hemorrhagic diarrhea, colanic acid protects the bacterial cells from the acidic conditions of gastrointestinal fluid (44). The proper levels and size distribution of colanic acid are therefore essential for resistance and survival under stress, a process that is hindered in cells lacking either YC-phosphorylated Wzc or Wzc dephosphorylated specifically by Wzb (43).
Here, using a variety of biophysical and biochemical methods, including solution NMR spectroscopy, we define for the first time, the structural elements on the catalytic domain of Wzc (Wzc CD ) and on Wzb responsible for their mutual interactions. Our results also provide insight into the mechanism by which Wzb recognizes its substrate, the phosphorylated YC of Wzc CD .

Expression and Purification of Wzc CD and Wzb Constructs
Used for Biophysical Studies-wzc CD , wzc CD⌬C , and wzb were cloned into pET15b vectors, transformed into E. coli BL21 DE3 cells, and purified as described in detail previously (45) for Wzc CD (the various constructs utilized in this study are shown in Fig. 3). All proteins contained a thrombin-cleavable 21-resi-due N-terminal tag. This tag appeared to ensure the long term stability of the proteins for the NMR experiments and was therefore not cleaved. After purification, the Wzc CD and Wzc CD⌬C samples were incubated overnight with a 25-fold excess of ATP/MgCl 2 followed by exchange against NMR buffer (see below) to remove the excess ATP. The wzc CD,ERR/A and wzb C9S mutants were prepared from wzc CD and wzb, respectively, using the QuikChange kit (Stratagene), cloned into pET15b, expressed, and purified as the corresponding wild-type proteins.
Expression and Purification of Wzc CD Constructs Used for Enzymatic Assays-The DNA fragment encoding for wzc CD was PCR-amplified using either the wild-type E. coli (K12) DNA or the pQE30-Wzc cyto -ERR/A plasmid (19) and cloned into a pQE30 vector between the BamHI and HindIII restriction sites to generate wzc CD (or the corresponding mutants: wzc CD,ERR/A , wzc CD,F5 , and wzc CD,F5,ERR/A ). Wzc CD,F5 and Wzc CD,F5,ERR/A , both bearing N-terminal His 6 tags, were produced in E. coli Xl1-blue cells grown in LB medium supplemented with ampicillin and tetracycline. Overexpression was induced using 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside for 3 h at 37°C when the A 600 reached 0.4, followed by centrifugation for 10 min at 5000 ϫ g. The pellet was re-suspended in 5 ml of buffer A (50 mM Tris-HCl, 200 mM NaCl, 10% glycerol, and 10 mM imidazole at pH 7.5) containing lysozyme (1 mg/ liter), DNase I (6 mg/liter), and RNase A (6 mg/liter) by agitation for 15 min at 4°C. Cells were lysed by sonication and centrifuged for 30 min at 14,000 ϫ g at 4°C. The supernatant was then mixed with nickel-nitrilotriacetic acid-agarose matrix (Qiagen) and incubated for 30 min at 4°C under gentle agitation followed by three washes with buffer B (50 mM Tris-HCl, 200 mM NaCl, 10% glycerol, and 20 mM imidazole, pH 7.5). The proteins were eluted with buffer E (50 mM Tris-HCl, 200 mM NaCl, 10% glycerol, and 150 mM imidazole, pH 7.5). Eluted fractions were analyzed by SDS-PAGE, pooled, and dialyzed overnight against 1 liter of buffer D (50 mM Tris-HCl, 100 mM NaCl, 10% glycerol, 1 mM DTT, 1 mM MgCl 2 , pH 7.5) and stored at Ϫ20°C. Note that all Wzc CD (or mutants thereof) preparations produced as described above resulted in species with a heterogeneously phosphorylated YC (data not shown) in line with that observed for Etk (18) and CapB (20).
Expression and Purification of Wzb Constructs Used for Enzymatic Assays-The DNA fragment encoding wzb and the mutants wzb C9S , wzb L40A , wzb Y117A , and the double mutant wzb L40A/Y117A were PCR-amplified using either the pQE30-Wzb plasmid (38) or pQE30-Wzb L40A plasmid (created for this study; for the double mutant only) and cloned into a pQE30 vector between the BamHI and Acc65I restriction sites. These mutants bearing N-terminal His 6 tags were produced in E. coli Xl1-blue cells and grown in LB medium supplemented with ampicillin and tetracycline. The overexpression and purification protocol was identical to that described for Wzc CD (and variants) above.
Wzc CD C-tail Peptides-Peptides derived from the C-terminal tail of Wzc CD were custom-synthesized and HPLC-purified by Peptide2.0. These peptides included (i) WTpep ( 705 YQDYGYYEYEYKSDAK 720 ) (phosphorylatable tyrosine residues are shown in boldface type); (ii) MUTpep ( 705 YQDE-GEEEEEEKSDAK 720 , where the phosphorylatable tyrosine residues in WTpep were mutated to glutamate in order to mimic the charge state of a phosphorylated tyrosine residue; (iii) PHOSpep ( 712 EY*E 714 , where Y* represents 4-phosphomonomethylphenylalanine, a non-cleavable phosphotyrosine analog) (46). This peptide was chosen in order to test the affinity of Wzb for a singly phosphorylated species that would be a true mimic of phosphotyrosine and yet would not be cleaved by the enzymatic activity of Wzb.
Surface Plasmon Resonance Measurements-Analyses of the interactions between the different Wzc CD and Wzb constructs were performed using surface plasmon resonance (SPR) using a BIAcore X100 (GE Healthcare). All experiments were performed at 25°C. The buffers for the SPR measurements were filtered through 0.22-m filters and degassed extensively before use. CM5 (carboxymethylated dextran; GE Healthcare) sensor chips were covalently coated with each of the three ligands Wzc CD , Wzc CD⌬C , and Wzc CD,ERR/A (theoretical pI 8.9, 9.39, and 8.73, respectively). The ligands were dissolved to final concentrations of 21 g/ml in 10 mM sodium acetate (pH 5.3) and immobilized onto CM5 chips by amine coupling on flow cells 2 (FC2), yielding responses of 951.2, 1512.8, and 490.8 for Wzc CD , Wzc CD⌬C , and Wzc CD , ERR/A , respectively, after deactivation of the CM5 surface with ethanolamine-HCl. As control, flow cells 1 (FC1) were activated for amine coupling and deactivated under the same conditions as FC2 in the absence of ligand.
The analytes, Wzb and Wzb C9S , were diluted in a buffer containing 50 mM phosphate, 137 mM NaCl, 2.7 mM KCl, 25 mM DTT, 5 mM EDTA, and 0.005% P20-surfactant at pH 7.4 (SPR buffer), to reach final concentrations ranging from 100 nM to 10 M. Each experiment consisted of three cycles where only the SPR buffer was passed over the CM5 sensor chip, followed by 11 cycles of different analyte concentrations (with one in duplicate). In each cycle, 50 l of SPR buffer was injected first to stabilize the base line, followed by injection of the analyte (0 -10 M) through the two flow cells at 30 l/min for 180 s. Responses detected on FC1 were considered to result from nonspecific binding and subtracted from the binding traces before detailed analyses. Dissociation of the bound analyte was achieved by passing SPR buffer over the two flow cells at 30 l/min for 600 s. After the dissociation phase, a solution containing 5 M NaCl and 4 M MgCl 2 and another containing 10 mM triethylamine (10 l/min for 60 s for each) were used to regenerate the CM5ligand surface. Each experiment was recorded in duplicate, and reproducible signals were detected for every ligand-analyte pair tested. FIGURE 1. BY-kinases. A, sequence alignment for BY-kinases. The conserved BY-kinase sequence motifs are indicated. The location of the ␣2 helix is shown by the horizontal black line, and the location of the EXXRXXR motif is shown by the yellow rectangle with the conserved residues (Glu, Arg, Arg) indicated by the asterisks. This alignment of BY-kinase sequences was used to generate Fig. 6C. B, structure of Wzc CD (Protein Data Bank code 3LA6) (19). Conserved structural elements are colored, and the side chains of conserved residues are shown in stick representations and labeled. The characteristic BY-kinase motifs (conserved residues are shown in boldface type) for Wzc CD are Walker A ( 533 GVSPSIGKT 541 ; red), Walker AЈ ( 558 VLLIDCDMR 566 ; orange), and Walker B ( 639 VLIDTPP 645 ; blue). Also shown are helix ␣2, containing the conserved 508 EAIRSLR 514 motif (yellow); the RK-cluster (Ser 473 -Leu 496 , magenta) (only Lys 477 and Lys 479 are seen in the crystal structure), and the C-tail (green; the YC tyrosines (Tyr 705 , Tyr 708 , Tyr 710 , Tyr 711 , Tyr 713 , and Tyr 715 ) are shown). The Wzc CD orientations are the same as in Fig. 6.
Transformation of experimental data were performed with the BIA-Evaluation software (GE Healthcare). No detailed kinetic analysis was performed for any of the sensorgrams, and equilibrium analysis to yield affinity values was performed for the Wzc CD⌬C (ligand)/Wzb C9S (analyte) pair (also for the Wzc CD /Wzb C9S pair at lower analyte concentrations) for which visual inspection of the sensorgrams confirmed 1:1 Langmuirtype binding. For these sensorgrams, the base line-subtracted ⌬RU values were related to the concentration of analyte (C analyte ) and fitted to Equation 1 to obtain a K d . The fits were performed using in-house software that utilized the ODRPACK subroutines (47).
The experimental methodology used for backbone resonance assignments of Wzc CD has been described in detail elsewhere (45). Backbone resonance assignments for Wzc CD⌬C were obtained using TROSY-based (in all cases) HNCO/ HN(CA)CO (512, 32, and 40 complex points with sweep widths of 13.35, 30, and 13 ppm in 1 H, 15 15 N, and 13 C dimensions, respectively) (50). No significant chemical shift changes were seen in the core residues of Wzc CD due to tail truncation (to generate Wzc CD⌬C ). Additionally, eight resonances corresponding to the 21-residue N-terminal tag (that includes 6 histidines) were also assigned, and these resonances showed no significant differences in their positions between Wzc CD and Wzc CD⌬C or for each species in the presence of Wzb. The corresponding tag in Wzb was also unaffected in the presence of Wzc CD or Wzc CD⌬C .
Nearly complete 15 N, 13 C, and 1 H resonance assignments for full-length Wzb are available from Lescop et al. (34). However, due to some differences under our conditions, the assignments of several of these resonances needed to be confirmed. Nonuniformly sampled versions of HNCO/HN(CA)CO (512, 32, and 50 complex points with sweep widths of 13, 32, and 13 ppm in 1 H, 15 N, and 13 C dimensions, respectively); HNCACB/ CBCA(CO)NH (512, 32, and 50 complex points with sweep widths of 13, 32, and 65 ppm in 1 H, 15 N, and 13 C dimensions respectively); and HNCA/HN(CO)CA (512, 32, and 50 complex points with sweep widths of 10, 32, and 30 ppm in 1 H, 15 N, and 13 C dimensions, respectively) were collected using a 30% sampling schedule generated by the sparse.py script (51). All data were collected at 700 MHz and 25°C and reconstructed using MDDGui software (52). 13 C, 1 H resonance assignments for the aromatic side chains of Wzb were obtained using (H␤)C␤(C␥C␦)H␦, (H␤)C␤(C␥C␦C⑀)H⑀ (512 and 40 complex points with sweep widths of 12 and 30 ppm in 1 H and 13 C dimensions, respectively) (53) and 13 C-edited NOESY-HSQC (512, 40, and 88 complex points with sweep widths of 12, 30, and 12 ppm in the 1 H, 13 C, and 1 H dimensions, respectively) experiments at 700 MHz and 25°C. A mixing time of 120 ms was used for the 13    beled Wzc CD at concentrations of 25, 50, 75, and 100 M, and 15 N, 1 H TROSY spectra were recorded for each titration point. All spectra were acquired at 900 MHz and 25°C using sweep widths of 13.35 ppm (512 complex points) and 30 ppm (128 complex points) in the 1 H and 15 N dimensions, respectively. Recycle delays of 1.5 s were used, and 16 transients were collected per t 1 point. Representative spectra corresponding to the various NMR-based titrations involving Wzc CD and Wzb are shown in Fig. 4. As can be seen, the spectra are of excellent quality.
Note that for all NMR-based titrations and spin relaxation experiments (described below), the last titration point contained an equimolar ratio of Wzc CD (or Wzc CD⌬C ) and Wzb for a total protein concentration of 200 M. Assuming a K d of ϳ2-5 M, we would expect ϳ80 -85% of the proteins to be in the complexed state. We did not use higher total protein concentration (i.e. saturating equivalents) to ensure sample stability during the measurements. However, no significant additional perturbations, compared with those at the equimolar ratio, were seen in the presence of saturating amounts of ligand at lower concentrations of the species being probed.
NMR Titrations to Determine Wzb/Phosphate Interactions-A series of 15 N, 1 H HSQC spectra were recorded using 100 M 13 C, 15 N-labeled Wzb, each with an incremental amount of potassium phosphate to obtain the following Wzb/phosphate ratios: 1:2, 1:3, 1:5, 1:10, 1:20, 1:40, 1:100, and 1:200. Titrations were performed in a buffer containing 20 mM BisTris, 50 mM NaCl, 25 mM DTT at pH 6.0 (BT buffer). The initial volume of 350 l increased to 363.5 l at the end of the addition, resulting in a dilution of less than 4%. All spectra were recorded at 600 MHz at 25°C using sweep widths of 13 ppm (512 complex points) and 32 ppm (128 complex points) for the 1  In all NMR-based titrations, chemical shift perturbations (⌬␦ i ) or signal attenuations (⌬I i ) for each residue at each titration point (indexed by i) were calculated using the following equations.
The 0 subscript indexes the 1 H/ 15 N chemical shifts (␦) or the intensities (I) for the reference state (no ligand present). Spin Relaxation Measurements-TROSY-based relaxation experiments to obtain spin-spin relaxation rates (R 2 ) (54) were recorded at 800 MHz and 25°C for Wzc CD , Wzc CD⌬C , and Wzb and for their equimolar mixtures. Samples typically were 100 M protein in NMR buffer, with the equimolar mixtures containing a 100 M concentration of each of the constituent proteins. Samples used (and relaxation delays) are provided in Table 1. Relaxation data were acquired with sweep widths of 13.35 ppm (512 complex points) and 30 ppm (128 complex points) in the 1 H and 15 N dimensions, respectively, for Wzc CD , Wzc CD⌬C , and their equimolar mixtures with Wzb and with sweep widths of 13 ppm (512 complex points) and 30 ppm (128 complex points) in the 1 H and 15 N dimensions, respectively, for Wzb and for its equimolar mixtures with Wzc CD and Wzc CD⌬C .
Enzymatic Assays-400 g of Wzc CD (or the corresponding mutant proteins: Wzc CD,ERR/A , Wzc CD,F5 , and Wzc CD,F5,ERR/A ) was incubated in the presence of 20 Ci of radioactive [␥-32 P]ATP (3000 Ci/mol) and 25 M non-radioactive ATP in a buffer containing 25 mM Tris-HCl, 5 mM MgCl 2 , 1 mM EDTA, 1 mM DTT at pH 7.5 for 30 min at 37°C. This was followed by extensive dialysis in the same buffer to remove radioactive ATP, followed by dephosphorylation using Wzb. For dephosphorylation, Wzc CD (or Wzc CD,ERR/A , or Wzc CD,F5 , or Wzc CD,F5,ERR/A ) and Wzb were mixed in a buffer containing 100 mM sodium citrate and 1 mM EDTA at pH 6.5 and incubated for a series of times (0, 1, 2, 5, 10, 20, and 30 min) at 37°C using several kinase/phosphatase molar ratios (5:1, 10:1, 20:1, 25:1, 30:1, and 35:1) with the quantity of Wzb being fixed at 2.85 pmol (in a reaction volume of 128 l corresponding to a concentration of 22.3 nM). Negative controls for the same times were performed in the absence of Wzb (data not shown). Samples were analyzed by SDS-PAGE and autoradiography. Bands corresponding to radioactive Wzc CD (or Wzc CD,ERR/A , or Wzc CD,F5 , or Wzc CD,F5,ERR/A ) were cut out from the gels and dissolved in 400 l of 30% hydrogen peroxide and 200 l of 60% perchloric acid for 4 h at 80°C. After the addition of 10 ml of scintillation liquid (ULTIMA GOLD TM XR), the radioactivity was measured in a scintillation counter (Tri Carb 2100 TR, Packard). The cpm values were then converted to pmol of phosphate remaining on Wzc CD,F5 (or Wzc CD,F5,ERR/A ). Michaelis-Menten parameters (V max and K m were estimated using Hanes-Woolf representations (Equation 4). Data analysis was carried out using in-house programs that utilize the ODRPACK (47) subroutines.
To test the activity of the mutant Wzb proteins, ␥-32 P-labeled Wzc CD (generated as described above) was incubated for 10 or 25 min in the absence or presence of either Wzb, Wzb C9S , Wzb L40A , Wzb Y117A , or Wzb L40A,Y117A . The mixtures were then analyzed by SDS-PAGE, and the extent of Wzc CD dephosphorylation was ascertained using direct film exposure and scintillation counting as described above.

RESULTS
In Vitro Interaction between Wzc CD and Wzb-As a first step toward detailed characterization of the interaction between Wzc CD and Wzb, we used SPR to estimate the affinity of their association. In all cases, we used Wzc CD (or mutants thereof; see Fig. 3 for the various Wzc and Wzb constructs used in this study) as a ligand affixed onto the sensor chip surface and Wzb (or mutants thereof; see Fig. 3) as an analyte. First, in order to test whether the C-tail of Wzc CD bearing the YC (the phosphorylated YC being the natural substrate of Wzb) is necessary for the Wzc CD /Wzb interaction, we used a Wzc CD construct missing the C-tail (Wzc CD⌬C ) as ligand. The sensorgram (Fig. 5A) for the Wzc CD⌬C /Wzb interaction was biphasic, especially at higher analyte concentrations, suggesting deviations from a simple 1:1 Langmuir-type binding, making it inappropriate for an equilibrium analysis. The curvature of the association stage of the sensorgram was not sufficient (fast association) for a meaningful kinetic analysis; therefore, such an analysis was not attempted for this (or for any other) case. However, when a catalytically inactive mutant of Wzb (Wzb C9S ; where Cys 9 , the catalytic cysteine nucleophile, is replaced by serine) (Fig. 3) was used as the analyte, a Langmuir-type behavior (Fig. 5B) was obtained (for the Wzc CD⌬C /Wzb C9S interaction), and an equilibrium analysis yielded a K d value of 5.1 Ϯ 0.2 M. Sensorgrams corresponding to Wzc CD /Wzb (Fig. 5C) and Wzc CD /Wzb C9S (Fig. 5D) interactions were highly biphasic, displaying a second slow process following the initial fast association. An equilibrium analysis (data not shown) of the Wzc CD /Wzb C9S sensorgram at lower analyte concentrations where the non-Langmuir behavior is less pronounced (for analyte concentration below 5.0 M) yielded a K d of 2.3 Ϯ 0.4 M, comparable with that obtained for the Wzc CD⌬C /Wzb C9S interaction. Overall, our analysis suggests that a region on Wzc CD distinct from the phosphorylated C-tail provides a majority of the binding energy for the Wzc CD /Wzb interaction.
Binding Site of Wzb on Wzc CD -Next, we attempted to ascertain specific residues on Wzc CD that interact with Wzb, using solution NMR techniques. Taking cue from the SPR results that showed that the YC of Wzc CD did not play a significant role in binding, we analyzed the spectral perturbations induced on a 1 H, 15 N TROSY spectrum of 15 N, 2 H-labeled Wzc CD⌬C in the presence of an equimolar ratio of full-length, uniformly 2 H-labeled Wzb. In addition to the uniform broadening of Wzc CD⌬C resonances due to complex formation, resulting in an increase in overall molecular mass ( 15 N R 2 for apo ϭ 26.4 Ϯ 5.9 s Ϫ1 , indicative of a monomer in solution; 15 N R 2 in the presence of Wzb at a 1:1 ratio ϭ 41.0 Ϯ 11.2 s Ϫ1 , indicating complex formation; data at 800 MHz with 100 M Wzc CD⌬C ), specific chemical shift changes and selective resonance attenuation were seen. The largest chemical shift changes were seen on the N terminus of ␣1 and the C terminus of ␣3. Some smaller but significant changes were also seen on ␣A and ␤7. However, an analysis of the selective signal attenuation (indicative of exchange on the intermediate time scale) was far more informative. Resonances corresponding to all of ␣2 and most of ␤A, ␤1, ␤6, and ␤7 were completely broadened out in the presence of Wzb. As shown in Fig. 6A, the residues that display chemical shift changes and selective attenuation form a continuous surface on the face opposite to that hosting the catalytic site of Wzc. In fact, no significant spectral perturbations (chemical shift changes or loss of signals) were observed at the active site. Note that 15 N, 1 H resonances other than those belonging to the Walker B motif are largely assigned for the active site (45).
In order to obtain additional insight into the spectral perturbations, we analyzed a TROSY spectrum of Wzc CD⌬C in the presence of substoichiometric amounts of Wzb (1:0.25, the first point in our titration course, 1:1 being the last). In this case (i.e. for a Wzc CD⌬C /Wzb ratio of 1:0.25), although no significant chemical shift changes were seen (except for Ile 510 ; 0.19 ppm), resonances corresponding to almost all of ␣2 (Ala 506 , Ala 509 , Arg 511 , Ser 512 , Leu 513 , Arg 514 , Ala 520 , and Met 521 ) were broadened to below the noise level (average attenuation for ␣2 ϭ 0.80 Ϯ 0.23, compared with 0.35 Ϯ 0.29 for all resonances). Complete attenuation of some other resonances (Tyr 467 , Ile 470 , Leu 547 , Leu 663 , Ala 666 , Ile 694 , and Leu 695 ) was seen. Notably, the side chains of these residues lie in close spatial proximity to ␣2, suggesting the importance of ␣2 in Wzc/Wzb interactions. NMR titrations with Wzc CD produced similar results (Fig. 6B), again with ␣2 and spatially proximal regions displaying the largest specific spectral perturbations. ␣2 shows a degree of sequence conservation that is the highest outside the catalytic site (see Fig. 6C; also see Fig. 1A). Notably, ␣2 contains the conserved 508 EXXRXXR 514 that was discussed at length in the Introduction. It therefore appears that the Wzb docking site on Wzc CD overlaps with that shown to play a role in the oligomerization and consequent trans-phosphorylation of the latter.
As expected from the SPR results, minimal spectral perturbations were seen for the YC of Wzc CD . Note that the species probed by the NMR experiments (unlike the SPR results that involved the catalytically dead Wzb C9S mutant) corresponds to the dephosphorylated YC. Given the efficiency of Wzb, the dephosphorylation reaction was almost complete before any NMR data could be recorded (Ͼ90% dephosphorylated in 30 min at a 1:1 ratio; data not shown). Nevertheless, some attenuations were seen in the YC, including for Tyr 708 (0.52), Glu 714 (fully attenuated), and Tyr 715 (0.55). It is possible that this weak interaction of the dephosphorylated C-tail of Wzc CD with Wzb was the cause of the second slow association event seen in the SPR analysis of Wzc CD /Wzb interactions. However, as will be shown, the isolated dephosphorylated C-tail of Wzc CD does not bind Wzb in trans. The weak association seen here is due to its elevated local concentration as a result of the docking interaction of Wzb with the putative oligomerization surface of Wzc CD described above. This is also the likely mechanism for substrate (i.e. phospho-YC recognition), as discussed below.
Binding Site of Wzc CD on Wzb-As expected, an overall decrease in signal intensity was seen in the 15   near the catalytic site of Wzb (Fig. 7). The patterns of perturbations seen in the spectra of Wzb in the presence of either Wzc CD⌬C or Wzc CD were similar (Fig. 7, compare A and B). Overall, the magnitudes of the chemical shift perturbations were significantly smaller than those seen for Wzc CD⌬C or Wzc CD in the presence of Wzb, and as in those cases, an analysis of the selective attenuations was far more informative. A majority of resonances corresponding to the N terminus of Wzb, ␤1, ␣3, ␤3, and ␣4 were broadened to below the noise level. Most significantly, many resonances corresponding to the so-called P-loop that contains the CXXXXXR(T/S) ( 9 CVGNICRS 16 ) motif (Fig. 2), conserved in all Cys-based phosphatases (37), were significantly perturbed. However, a very interesting pattern emerged upon closer inspection of resonances corresponding to the P-loop of Wzb in the presence of Wzc CD⌬C (Fig. 7D). Despite the absence of the YC in the Wzc CD⌬C construct, significant spectral perturbations were noted at the catalytic site of Wzb. Residues that were broadened to below the noise level included the catalytic Cys 9 (that carries out the ini-tial nucleophilic attack on the phosphotyrosine moiety) and Arg 15 (that stabilizes the transition state). Given that these changes occur in the absence of substrate (i.e. the phosphorylated YC of Wzc), they probably indicate allosteric effects at the Wzb active site upon docking interactions with Wzc CD⌬C (this could also be the origin of the mildly biphasic behavior of the Wzc CD⌬C /Wzb sensorgram; see Fig. 5A). Additional perturbations, covering almost all of the P-loop (Fig. 7D), were seen in the presence of Wzc CD . These results suggest a scenario in which docking interactions occur between Wzc CD and Wzb in a YC-independent manner, raising the local concentration of the YC, which then samples the active site of Wzb. This is consistent with the results obtained from the Wzc CD side, as described above.
Substrate Recognition by Wzb-Next, in order to ascertain the key determinants of substrate recognition by Wzb, we monitored changes in 15   Residues that display chemical shift changes greater than the mean ϩ 2 S.D. (0.036 ppm for Wzc CD⌬C ; 0.044 ppm for Wzc CD ) are colored green; residues that are broadened to below the noise level are colored orange. C, sequence conservation (also see Fig. 1A) in the CDs of BY-kinases plotted on the Wzc CD surface using a cyan (0% conservation) to magenta (100% conservation) gradient. The EXXRXXR (black), Walker A (green), Walker AЈ (yellow), and Walker B (red) motifs are indicated by the solid ovals. residues that are broadened to below the noise level are colored orange. C, sequence conservation (also see Fig. 2A) in LMW-PTPs plotted on the Wzb surface using a cyan (0% conservation) to magenta (100% conservation) gradient. D, close-up of the Wzb active site in the presence of either Wzc CD⌬C (left) or Wzc CD (right). Significant perturbations at the active site in the absence of substrate (i.e. the C-tail in the case of Wzc CD⌬C ) indicate allosteric effects upon docking interactions. The P-loop and D-loop containing key catalytic site residues are labeled. Also labeled is the so-called W-loop that, in the case of Wzb, does not contain a tryptophan. Selected residues that depict significant spectral perturbations are shown in stick representations and labeled. the presence of an increasing concentration of a 16-residue peptide (WTpep) corresponding to the dephosphorylated C-terminal tail (Tyr 705 -Lys 720 ). Lack of any significant spectral perturbations (data not shown) confirmed the extremely low affinity in trans of the dephosphorylated C-tail of Wzc CD for Wzb. Additionally, no significant spectral perturbations were seen in the presence of another 16-residue peptide (MUTpep; WTpep carrying a Y708E/Y710E/Y711E/Y713E/Y715E mutation to mimic the charge state of the fully phosphorylated YC). In vivo the YC has been found to be multiply phosphorylated (20,23,28), and this highly phosphorylated YC is the natural substrate of Wzb. It therefore follows that Wzb is capable of organizing several negative charges near its active site. Hence, that Wzb does not bind MUTpep is due to the fact that glutamate, although an excellent mimic for phosphoserine or phosphothreonine, is an extremely poor mimic for phosphotyrosine (55). However, significant spectral perturbations were seen in the presence of phosphate. A large number of chemical shift changes (Ͼ0.04 ppm; Fig. 8A) were seen emanating from the active site of Wzb in the presence of a 5-fold excess of phosphate. Interestingly, resonances corresponding to Cys 9 , Val 10 , Glu 20 , Leu 40 , and Tyr 117 only appear in the presence of a large excess of phosphate (200-fold excess under our experimental conditions), whereas additional resonances corresponding to residues Gly 11 , Ile 13 , Cys 14 , and Arg 15 can only be seen in phosphate buffer. The appearance of additional resonances in the presence of phosphate has been noted by Lescop et al. (34) (Note that the titrations involving Wzc CD , Wzc CD⌬C , and Wzb, described earlier were carried out in phosphate buffer; this was to allow the analysis of spectral perturbations for several resonances at the Wzb active site that are exchange-broadened and not observable in the absence of phosphate, as described above; see "Experimental Procedures"). It therefore appears that the phosphate moiety plays a significant role in substrate recognition by Wzb. It is to be noted that Wzb has no activity toward phosphoserine or phosphothreonine, 5 although selectivity toward phosphotyrosine is probably obtained in part by the depth of the catalytic cavity of Wzb (phosphoserine and phosphothreonine side chains are not long enough to reach the catalytic elements), and additional stabilization of the large hydrophobic tyrosine moiety near the active site is required for stable processing following phosphate recognition.
In order to ascertain the origin of the selectivity of Wzb toward phosphotyrosine, one would ideally use a peptide phosphorylated on a single tyrosine. However, such a phosphotyrosine peptide, being a natural substrate for Wzb, would be efficiently dephosphorylated before any NMR experiments could be conducted. Therefore, we monitored the spectral perturbations in Wzb in the presence of a tripeptide, EY*E (PHOSpep, where Y* represents 4-phosphomonomethylphenylalanine). The choice of this peptide incorporating an unnatural amino acid was guided by the fact that it truly mimics a phosphotyrosine (in a way that glutamate does not) in that it contains both a phosphate group and an aromatic ring. Additionally, being non-cleavable, it is stable under the enzymatic activity of Wzb. Despite the fact that the YC has been found to be multiply phosphorylated in vivo, as discussed earlier, the overall size of the Wzb catalytic cavity ensures that it cannot accommodate more than one phosphotyrosine at any given time, a condition that is effectively simulated by PHOSpep.
Unlike in the case of phosphate, where a large number of chemical shift changes were seen, suggesting chemical exchange on the fast time scale (as expected from a K d value of 630 Ϯ 118 M obtained by fitting the chemical shift changes against phosphate concentration), PHOSpep generated significant changes in intensity without any major chemical shift changes. In addition, several peaks showed doubling, with one peak reducing in intensity and the other increasing in intensity with increasing PHOSpep concentration, indicative of slow exchange on the chemical shift time scale (Fig. 8B). Fitting the changes in intensity with PHOSpep concentration yields a K d of 261 Ϯ 65 M, suggesting an affinity that is comparable with that seen for phosphate alone. This reinforces the fact the majority of the binding affinity of Wzc CD for Wzb is provided by docking interactions that do not involve the YC. A point to note is that the slow exchange regime seen here, although unusual for the weak affinity, is however not unprecedented (56,57). This apparent anomaly could be due to an unusual k on or the result of a slow conformational change following the binding event.
Here we analyze these effects only qualitatively in the context of YC recognition; a more detailed analysis, not relevant in the present context, is left for future studies. Overall, the spectral perturbations seen in Wzb in the presence of PHOSpep are similar to those seen with phosphate alone. However, PHOSpep induces a larger degree of spectral perturbations, in relative terms, on helix ␣5 of Wzb compared with phosphate (Fig. 8, compare A and C). This helix includes two aromatic residues, Phe 124 (minimal perturbation, although the flanking residues Thr 123 and Ala 125 have attenuations of 0.40 and 0.46, respectively) and Tyr 128 (attenuation ϭ 0.43; in fact, the stretch between Tyr 128 and Leu 131 shows an average attenuation of 0.51), which are highly conserved in bacterial LMW-PTPs (see Fig. 2A). The perturbations in these regions and the suggestion that the D-loop Tyr 117 contributes to substrate recognition open the possibility that a reorientation of ␣5 increases basestacking interactions with the phosphotyrosine moiety. As stated earlier, the 15 N, 1 H resonance corresponding to Tyr 117 can only be seen at high phosphate concentration. In order to probe the role of Tyr 117 , and possibly of Phe 124 and Tyr 128 , in substrate stabilization, we analyzed 13 C, 1 H HSQC spectra, focusing specifically on the aromatic groups (Fig. 9A). The aromatic resonances corresponding to Phe 124 were weak and could not be analyzed. However, significant attenuations were seen for the ␦and ⑀-resonances corresponding to Tyr 117 in the presence of PHOSpep, although none of the resonances corresponding to Tyr 128 displayed any significant perturbations. No detectable perturbations for any of the aromatic resonances were seen in the presence of WTpep (Fig. 9B). Therefore, it seems likely that Tyr 117 does play a role in substrate recognition, although no definitive conclusion can be reached for the role of Phe 124 . As alluded to earlier by Lescop et al. (34), Leu 40 in the so-called W-loop superimposes well on the specific resi-dues in eukaryotic LMW-PTPs that have been shown to play a central role in substrate recognition. Although the 15 N, 1 H resonances of Leu 40 are not visible in the absence of phosphate, the methyl positions of Leu 40 are significantly attenuated in the presence of PHOSpep (Fig. 9C) but not WTpep (Fig. 9D). Thus, our results provide evidence supporting the conjecture (34) that the tyrosine moiety of the substrate phosphotyrosine is stabilized through ring-stacking interactions with a conserved D-loop aromatic residue, and additional stability is imparted through hydrophobic interactions with the W-loop.
In Vitro Dephosphorylation of Wzc CD by Wzb and Role of the EXXRXXR Motif-In order to ascertain the functional relevance of the NMR-determined binding site for Wzb on Wzc CD , we determined the influence of mutations on the conserved EXXRXXR motif of Wzc CD on its ability to interact with Wzb and to be dephosphorylated by it. For this purpose, we chose a Wzc CD construct bearing a E508A/R511A/R514A triple mutation (Wzc CD,ERR/A ; see Fig. 3). It has been previously shown that E. coli cells complemented with this mutant produced 70% less colanic acid when compared with those carrying wild-type Wzc (19). As shown by the sensorgram in Fig. 10A (the scale is the same as in Fig. 5, C and D), Wzc CD,ERR/A appears to bind very weakly to Wzb. In addition, wild-type Wzc CD is dephosphorylated more rapidly by Wzb compared with Wzc CD,ERR/A (i.e. ϳ94% compared with ϳ69% after a 30-min incubation for a 20:1 ratio) (Fig. 10B). However, not surprisingly, our attempts to quantitatively analyze the dephosphorylation kinetics and obtain K m and k cat values were unsuccessful, given the multiply (and heterogeneously, as mentioned above) phosphorylated YC. A somewhat meaningful set of results could be obtained assuming the presence of two sets of sites on the YC: one set that could be dephosphorylated by Wzb in a highly efficient fashion and a second set that is dephosphorylated with lower efficiency (data not shown). This behavior is somewhat evident by the biexponential behavior displayed by the curves shown in Fig. 10B. To obtain more quantitative K m and k cat values, we created Wzc CD and Wzc CD,ERR/A mutants that contain a single phosphorylatable tyrosine on the C-tail. These mutants (Fig. 3), named Wzc CD,F5 and Wzc CD,ERR/A,F5 , contain a single tyrosine (Tyr 715 ) in the YC, with all other C-tail tyrosine residues mutated to phenylalanine (Y705F, Y708F, Y710F, Y711F, and Y713F). Tyr 715 from each Wzc CD subunit was found to engage the active site of a neighboring subunit in the crystal structure (19) of the Wzc CD oligomer. Hence, retaining Tyr 715 was the obvious choice for the enzymatic assays. The dephosphorylation data corresponding to these mutants could be fitted to sin- gle-exponential functions (e.g. for a 35:1 kinase/phosphatase ratio, the dephosphorylation rate is 1.81 Ϯ 0.06 min Ϫ1 for Wzc CD,F5 and 0.28 Ϯ 0.20 min Ϫ1 for Wzc CD,ERR/A,F5 ) (Fig. 10C). Analysis of the enzyme kinetics in a Hanes-Woolf framework (Fig. 10D) confirmed that the decreased dephosphorylation rate was indeed the result of the reduced ability of Wzc CD,ERR/A,F5 (K m ϭ 0.84 Ϯ 0.28 M) compared with Wzc CD,F5 (K m ϭ 0.04 Ϯ 0.31 M) to form an enzyme-substrate complex with Wzb. On the other hand, the k cat for Wzc CD,F5 is only ϳ3.5-fold higher than that for Wzc CD,ERR/A,F5 . One may therefore conclude that mutations in the conserved EXXRXXR motif on the ␣2 helix of Wzc CD lead to a reduction in its affinity for Wzb, resulting in a decrease in the efficiency of dephosphorylation. 40 and Tyr 117 -In order to test the importance of the W-loop Leu 40 and the D-loop Tyr 117 in the Wzb catalyzed dephosphorylation of WzcCD in vitro, we created Wzb mutants Wzb L40A , Wzb Y117A , and the double mutant Wzb L40A/Y117A . Although the activity of Wzb L40A appeared to be marginally lower than that of wild-type Wzb, Wzb Y117A (and the Wzb L40A/Y117A double mutant) was significantly less efficient in dephosphorylating Wzc CD (see Fig. 11). This suggests that although Leu 40 may play a secondary role in interacting with the phosphotyrosine substrate, as suggested by the NMR studies described above, its involvement is less critical than that of Tyr 117 .

DISCUSSION
Maintenance of the natural equilibrium between the YC high and YC low states of Wzc, a direct result of the activity of the cytosolic LMW-PTP Wzb, appears to be critical for the production of colanic acid and survival under stress (43). Here, we have used a variety of methods to ascertain the structural elements that mediate Wzc/Wzb interactions that are central in regulating the BY-kinase signaling pathway. Using SPR measurements, we have found that a majority of the binding energy for the interactions of Wzc with Wzb is provided by elements external to the phosphorylated YC of the former. Perturbations in solu-tion NMR spectra reveal that the Wzb binding surface on Wzc involves a region that does not include any of the Wzc catalytic elements but involves the ␣2 helix bearing a motif (EXXRXXR) that is highly conserved in BY-kinases, in addition to elements that are spatially proximate to it. In fact, most of the perturbations, including and around the EXXRXXR motif, involve extensive line broadening rather than chemical shift perturbations. We attribute these effects to the direct binding event that occurs on the intermediate exchange regime on the chemical shift time scale. The chemical shift perturbations, indicative of fast exchange, occur peripheral to the binding site and likely  . Dephosphorylation of Wzc CD by wild-type and mutated Wzb. Mutation of Tyr 117 on the Wzb D-loop to alanine significantly reduces its ability to dephosphorylate Wzc CD , whereas mutation of the W-loop Leu 40 to alanine has only a minor effect. Also shown are data for the L40A/Y117A double mutant. Data for the catalytically dead Wzb mutant (Wzb C9S ) and wild-type Wzb are also shown for comparison. A, autoradiogram obtained after incubation of phosphorylated Wzc CD with Wzb and mutants thereof for a 10-min period. B, residual phosphorylation after 10-min (black) and 25-min (red) incubation periods. The intensities are normalized against the 10-min data for Wzb C9S . The dotted lines serve as guides for the residual phosphorylation levels when using wild-type Wzb.
represent conformational changes occurring upon Wzb binding. A more detailed parsing of these effects can only be achieved upon the availability of an atomic resolution model of the Wzc CD -Wzb complex and a detailed analysis of the changes in dynamics upon complex formation. The critical role of the EXXRXXR motif in binding Wzb is confirmed by binding studies and biochemical assays, which reveal that mutations in this motif greatly reduce the ability of Wzc CD to interact with and be dephosphorylated by Wzb. As discussed above, many lines of data indicate that the EXXRXXR motif plays a key role in the formation of asymmetric oligomers of Wzc CD , allowing the YC of one molecule to access the catalytic site of another, a necessary condition to allow the intermolecular autophosphorylation. Our present results add a new wrinkle to this model, suggesting that not only does Wzb generate YC low states from the YC high states through its enzymatic activity, but it also probably prevents the reassociation (and consequent phosphorylation) of the Wzc CD monomers. One may speculate that the purpose of this protection is to allow the active site of Wzc CD , which is occluded in the oligomeric state, to be accessible to downstream targets, such as Ugd (32). The studies presented here involve the isolated CD of Wzc. It is quite possible that conformation of Wzc CD and its interactions with Wzb may be modulated by periplasmic events involving the ectodomain of Wzc and indeed its trans-membrane segments. Transmission of extracytoplasmic events through the membrane to modulate the activity of cytoplasmic kinase domains has been extensively investigated in eukaryotic receptor tyrosine kinases, most notably the epidermal growth factor receptor (58,59). In addition, cytoplasmic events, including post-translation modifications, such as phosphorylation, could also modify the interactions between Wzc CD and Wzb. This mode of regulation has been seen in eukaryotic tyrosine phosphatases. For example, in T-cells, the hematopoietic PTP docks onto the mitogen-activated protein kinase (MAPK), ERK2, using a conserved D-site sequence (60). The hematopoietic PTP stays bound to ERK2 even after dephosphorylating the phosphorylated tyrosine on the activation loop of ERK2 and sequesters the inactive kinase to the cytosol (61). Detachment occurs upon phosphorylation of a D-site serine on hematopoietic PTP by protein kinase A (PKA) (62). This is an intriguing scenario, especially because it has recently been shown that Etp, the cognate LMW-PTP for Etk, is phosphorylated on a D-loop tyrosine by a yet unknown cellular kinase that is neither Etk nor Wzc (63). However, we have not yet found compelling evidence that the corresponding tyrosine (Tyr 117 ) in Wzb is phosphorylated in the cellular milieu.
As mentioned previously, the interaction of Wzb with Wzc CD does not seem to require the phosphorylated YC. The use of motifs for docking interactions external to the substrate moiety to enhance affinity and specificity has been well documented in eukaryotic phosphatases (64,65). Our results further demonstrate that the catalytic site of Wzb, especially the P-loop bearing the conserved CXXXXXRS motif, undergoes conformational changes upon docking interactions with Wzc CD⌬C that lacks the YC. This contrasts with interactions on the Wzc CD side that do not seem to display any discernable allosteric effects at its active site. Thus, docking interactions outside the Wzb catalytic site appear to trigger conformational changes in the catalytic P-loop, perhaps priming it for catalysis. Wzb undergoes significant conformational changes in the presence of phosphate, with most of these changes localized on the catalytic P-loop, on the D-loop, and, to a lesser extent, on the W-loop. Stehle et al. (66) have recently solved the structure of MptpA from Mycobacterium tuberculosis in the phosphatefree form. This structure, which represents the first LMW-PTP structure in the absence of phosphate, provides a means to compare the conformational changes expected upon phosphate binding given that the crystal structure of MptpA bound to a phosphomimetic chloride ion exists (67). In this study, Stehle et al. (66) show that MptpA adopts an open structure in the absence of phosphate with the D-and P-loops more distant from each other, whereas the presence of chloride induces a more closed conformation, greatly reducing the distance between the two loops. The perturbations seen in the Wzb Dand P-loops in our case are consistent with these observations. We further note the presence of extensive conformational exchange in the P-loop in the absence of phosphate, as reflected by the missing resonances. This conformational sampling is greatly reduced in the presence of phosphate, a fact that has also been previously demonstrated (34). Thus, the phosphate group of the phosphotyrosine moiety appears critical in substrate recognition by Wzb. The tyrosine moiety is stabilized largely by ring-stacking interactions with the D-loop tyrosine (Tyr 117 ) and hydrophobic interactions with a W-loop leucine (Leu 40 ). Notably, the Tyr 117 position is strictly conserved in bacterial LMW-PTPs (Phe 120 in Bacillus subtilis; see Fig. 2A), whereas the Leu 40 position, although less strictly conserved, always contains a large hydrophobic residue. The central role of Tyr 117 in stabilizing the substrate, in what is likely an optimal orientation for catalysis, is confirmed by the fact that the phosphotyrosine phosphatase activity of Wzb is significantly reduced in a Y117A mutant. Considering all of the results presented here, it seems that docking interactions between Wzc and Wzb bring the phosphorylated YC close to the catalytic site of Wzb. Dephosphorylation by Wzb proceeds due to the high local concentration of the substrate, phosphorylated YC, despite the low inherent affinity of Wzb for the isolated phospho-YC. Thus, YC dephosphorylation by Wzb may therefore be appropriately described as proximity-mediated (68).
In conclusion, we have, for the very first time, presented a comprehensive analysis of the interactions between the catalytic domain of the BY-kinase Wzc with its cognate phosphatase, Wzb. We have identified elements within the catalytic domain of Wzc that are crucial for the recognition of Wzb. In addition, our studies provide insight into the recognition by Wzb of the phosphorylated YC of Wzc, its natural substrate. These results represent a major step forward in obtaining a comprehensive understanding of the regulation of the BY-kinase signaling pathway by bacterial PTPs.