Three-dimensional structure and ligand interactions of the low molecular weight protein tyrosine phosphatase from Campylobacter jejuni

Abstract

A putative low molecular weight protein tyrosine phosphatase (LMW-PTP) was identified in the genome sequence of the bacterial pathogen, Campylobacter jejuni. This novel gene, cj1258, has sequence homology with a distinctive class of phosphatases widely distributed among prokaryotes and eukaryotes. We report here the solution structure of Cj1258 established by high-resolution NMR spectroscopy using NOE-derived distance restraints, hydrogen bond data, and torsion angle restraints. The three-dimensional structure consists of a central four-stranded parallel β-sheet flanked by five α-helices, revealing an overall structural topology similar to those of the eukaryotic LMW-PTPs, such as human HCPTP-A, bovine BPTP, and Saccharomyces cerevisiae LTP1, and to those of the bacterial LMW-PTPs MPtpA from Mycobacterium tuberculosis and YwlE from Bacillus subtilis. The active site of the enzyme is flexible in solution and readily adapts to the binding of ligands, such as the phosphate ion. An NMR-based screen was carried out against a number of potential inhibitors and activators, including phosphonomethylphenylalanine, derivatives of the cinnamic acid, 2-hydroxy-5-nitrobenzaldehyde, cinnamaldehyde, adenine, and hypoxanthine. Despite its bacterial origin, both the three-dimensional structure and ligand-binding properties of Cj1258 suggest that this novel phosphatase may have functional roles close to those of eukaryotic and mammalian tyrosine phosphatases. The three-dimensional structure along with mapping of small-molecule binding will be discussed in the context of developing high-affinity inhibitors of this novel LMW-PTP.

Low molecular weight protein tyrosine phosphatases (LMW-PTPs) form a family of ancient 18-kDa phosphatases that share very little sequence similarity with other PTPs (Zhang 1998, 2003; Jackson and Denu 2001). Representatives of the LMW-PTP group have been found in diverse organisms from bacteria to man, implying their involvement in different physiological functions varying from species to species.

The biological function of LMW-PTPs is particularly ill-defined in prokaryotes. Several lines of evidence have been presented indicating that at least in some Gram-negative bacteria LMW-PTPs are involved in the synthesis and translocation of exopolysaccharides (EPS) and capsular polysaccharides (CPS) (for review, see Cozzone et al. 2004). It was demonstrated that the Wzc/Wzb pair of protein tyrosine kinase (PTK) and LMW-PTP from the Escherichia coli strain K-12 is directly connected with the export of colanic acid (Vincent et al. 1999, 2000), a protective EPS found widely with most E. coli strains. A homologous Wzc/Wzb pair from E. coli K-30 was found to be essential for the assembly of group 1 CPS (Wugeditsch et al. 2001), an important virulence determinant. Other reports associated Etk/Etp, another PTK and LMW-PTP pair from E. coli, with the secretion and assembly of the group 4 CPS in enteropathogenic E. coli. (Ilan et al. 1999; Peleg et al. 2005).

As the EPS and CPS are known pathogenic factors, the components of bacterial tyrosine phosphorylation/dephosphorylation cycles were consequently proposed as hypothetical targets for a new generation of antibiotics (Cozzone et al. 2004). However, the efficacy of such an intervention is difficult to predict, and it was emphasized that this hypothesis needs to be carefully examined. Regardless, developing small-molecule ligands and inhibitors of LMW-PTPs represents an interesting and important problem, particularly in light of the observation that biological functions of LMW-PTPs in prokaryotes are not limited to polysaccharide production. Indeed, in Gram-positive Bacillus subtilis, LMW-PTPs play a role in resistance to ethanol stress (Musumeci et al. 2005) and are not likely to participate in the EPS and CPS synthesis and secretion (Cozzone et al. 2004). In other studies, it was concluded that the Etp LMW-PTP of E. coli regulates bacterial resistance to heat shock by preventing the Etk-mediated phosphorylation of the RNA polymerase σ³² subunit and the anti-σ^E factor RseA (Klein et al. 2003).

Campylobacter jejuni is an important cause of food-borne bacterial enteritis. Though usually self-limiting, C. jejuni infection may lead to the development of an acute peripheral neuropathy, the Guillain-Barré syndrome (Hughes 2004). The annotation of the full-genome sequence of the pathogen (Parkhill et al. 2000) revealed several His kinases and two putative Ser/Thr protein phosphatases; however, no Ser, Thr, or Tyr protein kinases have been found. There is a single putative Tyr phosphatase, Cj1258 (Fig. 1), but a survey of the C. jejuni phosphoproteome found only two proteins phosphorylated at Tyr (S. Voisin, D.C. Watson, L. Tessier, S. Bhatia, J.F. Kelly, and N.M. Young, unpubl.), and C. jejuni does not utilize the polysaccharide export pathway mentioned above. Consequently, the determination of its three-dimensional (3D) structure was undertaken in an attempt to clarify the unknown biological function of Cj1258 in C. jejuni.

Figure 1.

Amino acid sequence comparison between Cj1258 (Campylobacter jejuni) and PtpA (Staphylococcal aureus), LTP1 (Saccharomyces cerevisiae), HCPTP-A (Homo sapiens), BPTP (Bos taurus), TPTP (Tritrichomonas fetus), YwlE (Bacillus subtilus), and MPtpA (Mycobacterium tuberculosis). Catalytic cysteine residues are underlined. The sequences are ranked from top to bottom according to the descending BLAST (http://www.ncbi.nlm.nih.gov/BLAST/) alignment score using the BLOSUM62 matrix (Henikoff and Henikoff 1992) and default on-line parameters (i.e., the sequences with higher similarity to Cj1258 are closer to the top).

It has been demonstrated recently that small molecule inhibitors are of great utility as effective tools for defining the functions of PTPs and, in the case of PTP1B, as lead molecules for the development of therapeutic agents (Xie et al. 2003; Kumar et al. 2005). In particular, these studies showed that small-molecule inhibitors of PTP1B can act as both insulin mimetics and sensitizers, thereby complementing the results obtained from gene knockout experiments (Elchebly et al. 1999; Klaman et al. 2000). In practice, success of a small-molecule-based approach depends on the availability of potent and specific inhibitors. As such, significant efforts in inhibitor design have been expended on targeting PTP1B, and, to a lesser extent, on some other PTPs. In particular, limited attention has been given to the design of specific inhibitors of LMW-PTPs from either eukaryotic or prokaryotic organisms. Interestingly, there were some initial attempts in inhibitor design that have resulted in specific, although weak-binding, inhibitors of the human LMW-PTP HCPTP-A (Zabell et al. 2004). Further optimization of the LMW-PTP inhibitors can take place either by a search for additional binding moieties as part of the fragment-based design strategy (Shuker et al. 1996; Song and Ni 1998; Szczepankiewicz et al. 2003) or by other structure-based approaches (Iversen et al. 2000; Lund et al. 2004). In all, small molecules with weak affinities of specific binding to any putative phosphatase will represent an important step toward the design of high-affinity inhibitors for use either in testing molecular function or as drug leads.

In this study, we present the solution structure of a putative LMW-PTP Cj1258 determined by use of NMR spectroscopy. We then screened a set of small-molecule compounds known to inhibit other PTPs as potential starting points for the design of inhibitors for bacterial LMW-PTPs that closely resemble eukaryotic LMW-PTPs.

Results

Phosphatase activities of Cj1258

Enzymatic activities for Cj1258 were assayed under predetermined optimal pH and temperature conditions at pH 6.5 and 37°C using p-nitrophenyl phosphate as a substrate. The determined K _m value under these conditions was 1.4 mM, which is consistent with those previously reported for other LMW protein phosphatases from Staphylococcal aureus (1.2 and 1.5 mM) and the E. coli Wzb (1 mM) (Vincent et al. 1999; Soulat et al. 2002). In the presence of 5 mM adenine, the K _m value increases by a factor of 4.2 and the maximal velocity by a factor of 1.9. While this level of rate enhancement by adenine bears a closer resemblance to that of the mammalian LMW protein phosphatases rather than the yeast enzyme (Wang et al. 2000a), it should be noted that the K _m value for the p-nitrophenyl phosphate substrate is not a true equilibrium constant that reflects affinity in a simple way. The specificity for phospho-Tyr was determined by incubating Cj1258 with a mixture of phospho-Ser and phospho-Thr proteins extracted from C. jejuni. No changes were seen in the two-dimensional gel map of the Ser- and Thr-phosphorylated proteins (data not shown), indicating that Cj1258 is a phospho-Tyr-specific phosphatase.

Sequence-specific resonance assignments of NMR spectra

The Cj1258 protein was characterized by use of NMR at pH 5.8 and 7.0, and in the presence of 20 mM inorganic phosphate, which is known to occupy and stabilize the active site of LMW-PTPs (Logan et al. 1994; Zhang et al. 1994b). An [¹⁵N,¹H]-HSQC spectrum of Cj1258 at pH 5.8 is shown in Figure 2. A complete sequence-specific assignment was achieved at pH 5.8 for all of the spin systems with the exception of the last three C-terminal His-tag residues, and almost all ¹⁵N, ¹³C, and ¹H chemical shifts. We observed that the initiation methionine residue was not removed, consistent with the specificity of the E. coli methionine aminopeptidase, which is unable to cleave the N-terminal methionine if it is followed by a lysine residue (Hirel et al. 1989). Unassigned atoms and chemical groups include ¹⁵N of the proline residues, primary amines of the N-terminal methionine and lysines, ¹³C′ preceding prolines, and some unobservable or overlapped side chain atoms. At neutral pH, chemical shifts were overall very close to those at pH 5.8. At pH 5.8, the HN/¹⁵N HSQC cross-peaks of Leu9, Gly10, Ile12, Arg14, Gly44, Gly54, Asn68, Gln76, Asn102, Asn105, Ser116, and Asn118 were somewhat broader and weaker than the rest of the resonances, whereas at neutral pH most of them (except Asn102) broadened out and were not found in the spectrum. The chemical shifts assigned at both pH conditions have been deposited to BMRB (deposition no. BMRB-7189).

Figure 2.

[¹⁵N,¹H]-HSQC spectrum of ¹⁵N/¹³C-labeled Cj1258. The protein was in a buffer of 20 mM phosphate, 90% H₂O/10% D₂O, 0.2 mM EDTA, and 0.01% NaN₃ (pH 5.8). Protein concentration was ∼0.7 mM. The spectrum was recorded at 298 K on a Bruker Avance 800 MHz spectrometer. The assignments of the well-resolved backbone HSQC peaks are indicated by the residue numbers.

Interestingly, the ¹⁵N/HN chemical shifts of residues in the highly conserved P-loop were similar to those of bovine LMW-PTP BPTP (Zhou et al. 1994). Since the ¹⁵N/HN chemical shifts are sensitive to the microenvironment dictated by the three-dimensional structure, this observation suggests that the P-loops of Cj1258 and BPTP are conformationally equivalent and may perform the same chemical function.

H/D exchange, backbone–backbone NOE connectivities, and ³J_HNHα coupling constants

Upon reconstitution of lyophilized Cj1258 samples in D₂O, we detected a number of amide protons that exchanged slowly with the solvent, indicating that these protons are well-protected in the three-dimensional structure of Cj1258. Residues with slower exchanging backbone amide protons are displayed in Figure 3. Pairs of residues Lys27/Val122 and Ile86/Asp124 have overlapping [¹⁵N,¹H]-HSQC peaks; therefore, the slower exchanging protons observed for these peaks could not be assigned to specific residues.

Figure 3.

Secondary structure of the recombinant Cj1258 protein. The sequence contains the initiator residue methionine and a His₅ affinity purification tag at the C terminus. Fragments with local backbone RMSD higher than 0.158 Å are indicated with dashed lines. Secondary structure elements (α-helices, β-strands, and the 3₁₀-helix) are shown as bars, arrows, and the basket weave bar, respectively. Asterisks show the residues with slower exchanging backbone amide protons. Pairs of residues Lys²⁷/Val¹²² and Ile⁸⁶/Asp¹²⁴ have overlapping [¹⁵N,¹H]-HSQC peaks; therefore, the slower exchanging protons observed for these peaks were not assigned and indicated with question marks.

Structure calculations

The chemical shifts of the backbone ¹⁵N, ¹³C^α, ¹³C^β, ¹³C′, H^β, and H^α served as input to the TALOS program (Cornilescu et al. 1999) to predict the backbone φ and ψ dihedral angles and derive the initial distribution of secondary-structure elements in Cj1258. The predicted secondary structure is highly similar to that of a typical LMW-PTP with four β-strands and five α-helices (Logan et al. 1994; Su et al. 1994; Zhang et al. 1994b). The assignment of the secondary structure (Fig. 3) was further edited and refined by analyzing the H/D exchange data, ³ J _HNHα coupling constants, backbone–backbone NOE patterns, and finally the calculated structures. The patterns of hydrogen bonding, backbone proton–proton NOE contacts and ³ J _HNHα coupling constants correlated well with secondary structures predicted by TALOS.

The initial global fold of Cj1258 was generated using only NOE distance and angle restrains. Hydrogen bond donors and acceptors were assigned using the amide exchange data (Fig. 3) combined with secondary structure information and the initial calculated structures. The identified hydrogen bonds were included in further calculations as pairs of hydrogen bond distance restraints. Amino acid residues can be divided into two distinct groups with good or poor backbone definition in the 10 best structures (Table 1) (PDB ID 2GI4) of 50 generated by the calculations. Figure 4 shows the profiles of local backbone atom root-mean-square deviation (RMSD), local heavy atom RMSD, and angle order parameters (Koradi et al. 1996). Abrupt increases in local backbone RMSD are localized between elements of canonic secondary structures and are a consequence of the lack of NOE and angular restraints within these regions. An averaged local backbone RMSD over the entire sequence is 0.158 Å and residues with local RMSD higher than the average are indicated in Figure 3. Overall, the three-dimensional structure of Cj1258 is well defined with an average pairwise global backbone RMSD of 0.95 ± 0.14 Å (Table 2). Therefore, the structural definition of Cj1258 is slightly better than those reported for the solution structures of LMW-PTPs from eukaryotic organisms (Logan et al. 1994; Gustafson et al. 2005). Upon excluding poorer defined regions (with local RMSD values higher than 0.158 Å), the pairwise global RMSD decreased to 0.83 ± 0.09 Å.

Table 1.

Overview of the structure parameters of the 10 best structures of Cj1258 calculated by CNS/ARIA

Figure 4.

Local structural parameters of the 10 best Cj1258 structures as reported by MOLMOL. (Top) Angular order parameters for φ (solid line) and ψ (dotted line) dihedral angles. (Bottom) Local average pairwise RMSD for backbone (solid line) and heavy atoms (dotted line) based on tripeptide fragments with the residue in question as the central amino acid.

Table 2.

Detailed comparisons among the 10 best structures calculated for Cj1258

Figure 5 displays a cluster of the 10 best structures superimposed over the backbone atoms of structurally well-defined parts of Cj1258 and a ribbon representation of the lowest energy conformation in the cluster. The three-dimensional structure represents an expected α/β fold typical of a LMW-PTP with a twisted central four-stranded parallel β-sheet (residues Lys3–Ile7, Phe36–Ala40, Phe84–Thr87, and Val107–Leu108), surrounded with five α-helices, i.e., residues Arg14–Ala29, Tyr53–Gln61, Gln76–Glu81, Asn90–Asn99, and Asp132–Ser150. The fourth β-strand of Cj1258 was unusually short, spanning only two residues, and it is flanked by two helical turns of four and three residues, i.e., Asn102–Asn105 and Ile110–Asp112. The first turn (residues Asn102–Asn105) was assigned to a 3₁₀ conformation based on the absence of strong H^α-H^β (i, i + 3) NOEs in the ¹³C-edited NOESY-HSQC data, whereas the second turn was difficult to assign as a result of resonance overlap. Loops forming the active site, i.e., the phosphate binding P-loop (residues Cys8–Ser15), the E-loop corresponding to the W-loop in BPTP (residues Thr42–His52), and the D-loop (residues Phe113–Asn130), are more poorly defined by NMR data than the elements of secondary structure as judged by the local RMSD values (Fig. 4). Due to the lack of unambiguous NOE contacts, residues Asn30–Phe35, connecting α1 and β2, display higher than average local RMSD values in contradiction with the observation of lowered amide hydrogen exchange rates for Leu31, Glu32, Glu34, and Phe35 (Fig. 3).

Figure 5.

Stereoview of the Cj1258 three-dimensional structure. (A) Cluster of the 10 lowest energy Cj1258 structures superimposed using the backbone atoms of well-defined residues. Well-defined residues were chosen based on the values of local backbone RMSD (fragments with local backbone RMSD higher than 0.158 Å were excluded). The flexible poly-His C-terminal tag is not displayed. (B) Ribbon representation of the lowest energy conformer.

Structure comparisons with other PTPs

The three-dimensional solution structure of Cj1258 was compared (Table 3) with the crystal and solution structures of yeast (LTP1 from Saccharomyces cerevisiae), bacterial (MPtpA from Mycobacterium tuberculosis, and YwlE from B. subtilis), bovine (BPTP), human (HCPTP-A) LMW-PTPs, and a LMW-PTP from a primitive eukaryotic parasite (TPTP from Tritrichomonas fetus). The sequence of every chosen LMW-PTP, when aligned with the sequence of Cj1258, had a number of amino acid insertions and deletions. Therefore, to avoid ambiguities of structural comparison, we calculated the pairwise RMSD values by superimposing only clearly homologous regions at the level of three-dimensional structure, namely, core structures formed by the regular α/β elements or loops surrounding the active sites. Surprisingly, the backbone atom coordinates of the secondary structure elements of Cj1258 deviate less from those of bovine and human LMW-PTPs than from those of MPtpA, YwlE, TPTP, and LTP1.

Table 3.

Comparisons among the 10 best structures calculated for Cj1258, and with those of other PTPs

To assess the level of structural similarity between the active sites of the various enzymes, we calculated the RMSD values by superposition of the backbone atoms of the P-loops (residues Cys8–Arg14 of Cj1258) or of all the loops surrounding and forming the active site (residues Cys8–Arg14, Thr42–Gly54, Thr87–Asn92, and Val122–Tyr127 of Cj1258). The structural differences between the P-loops of crystallized phosphatases (BPTP, HCPTP-A, MptpA, and LTP1) and Cj1258 (∼0.7 Å) were close to the precision of the P-loop structural definition in Cj1258 (∼0.35 Å). A slightly higher difference (∼1.0 Å) was observed between the P-loops of Cj1258 and the distant tyrosine phosphatase PTP1B. When the relative orientation of the loops forming the active site in the crystallized phosphatases was compared with that of Cj1258, Cj1258 displayed significant and approximately equal RMSD from all other LMW-PTPs. The loops deviated even more if the structure of Cj1258 was compared with solution structures of LMW-PTPs (TPTP and YwlE). These apparent structural differences can be explained by a mobile character of the surface loops in solution, resulting in too few NOEs and lack of atomic structure definition.

Ligand-binding characteristics of Cj1258

Several commercially available small compounds known to inhibit or activate tyrosine phosphatases (Marseigne and Roques 1988; Dissing et al. 1993; Logan et al. 1994; Zhang et al. 1994a, b; Moran et al. 1995; Wang et al. 2000a; Fu et al. 2002) were tested for their specific binding to Cj1258. In certain titration experiments, the presence of 2 mM phosphate was required to identify ligand-binding effects on the resonance peaks that are otherwise not seen in the spectra. Resonance-specific changes were observed in the [¹⁵N,¹H]-HSQC spectra of Cj1258 in the presence of putative inhibitors/activators and these changes were mapped to affected amino acid residues in the three-dimensional structure of Cj1258 (a typical mapping of Cj1258 interactions with 4-acetamidocinnamic acid is shown in Figure 6; a complete mapping and tabulation of spectral changes for every ligand can be found in the Supplemental Material). Only well-resolved [¹⁵N,¹H]-HSQC peaks were followed, and residues with significant overlaps or not visible throughout the titrations appeared as unaltered in the presented data and the analysis.

Figure 6.

Mapping of 4-acetamidocinnamic acid binding onto the three-dimensional structure of Cj1258. (A) Active site loops (P-loop in red, D-loop in blue, and E-loop in yellow) and α5-helix (in green). Other panels display interaction of Cj1258 with 4-acetamidocinnamic acid (B) and 4-acetamidocinnamic acid in the presence of 2 mM phosphate (C). The largest observed chemical shift with respect to the ligand-free spectrum was chosen as the reference value. Residues experiencing chemical shifts of 66% or higher of the reference shift are mapped in red, residues experiencing the chemical shifts between 33% and 66% of the reference shift are mapped in yellow, residues experiencing extreme broadening or sharpening are mapped in blue, and residues experiencing partial broadening or sharpening are mapped in cyan. Two views of the molecule separated by a 90° rotation are shown in each row.

Chemical-shift changes (in ppm) caused by saturating concentrations of cinnamaldehyde and 2-hydroxy-5-nitrobenzaldehyde were either small [δ(ppm) < 0.015 for cinnamaldehyde] or the affected residues were dispersed over the entire surface of Cj1258 (for 2-hydroxy-5-nitrobenzaldehyde), suggesting weak and nonspecific molecular interactions. Other compounds, which can be divided into two classes, showed specific interactions with Cj1258 around the active-site region. One of the two classes consisted of phosphate, 4-(phosphonomethyl)-phenylalanine, trans-cinnamic acid, 2-carboxycinnamic acid, and 4-acetamidocinnamic acid. These inhibitors represent a product (the phosphate anion) or nonhydrolyzable mimetics of the substrate phosphotyrosine. Two other ligands, adenine and hypoxanthine, belong to the family of purine-type effectors of PTPs (Dissing et al. 1993; Ostanin et al. 1995; Wang et al. 2000a), which bind close to the active site and can modulate the activity of these enzymes.

Inorganic phosphate is the smallest competitive inhibitor of LMW-PTPs among those tested, with an expected K _d of ∼1 mM (Zhang and Van Etten 1990, 1991; Taddei et al. 1994). Consistent with the observations reported for the BPTP (Logan et al. 1994) and TPTP from T. fetus (Gustafson et al. 2005) interaction of Cj1258 with inorganic phosphate in acetate buffer either sharpened or made visible the [¹⁵N,¹H]-HSQC cross-peaks of the P-loop residues Leu9–Gly10, Ile12–Cys13, and Ser15 (cross-peaks corresponding to Cys8 and Asn11 were sufficiently narrow and observed in the absence of phosphate). A sharpening of the resonances of residues Gly54, Lys72, Trp126, and Asn130 was also detected. Inorganic phosphate caused significant chemical-shift perturbations (>0.1 ppm) for residues Ile7, Cys8, Ala18, His52, Thr55, Asn57, Asn92, Tyr127, Gly129, Phe131, Asp132, Ile137, Leu140, and Lys143. Notably, two loops forming the active site, i.e., the P- and D-loops, were the most affected by binding of the phosphate ion. Phosphate ions also affect the chemical shifts of residues in helices α2 and α5.

The effects of phosphotyrosine mimetic compounds on the P-loop were established in the presence of 2 mM phosphate. In the presence of 2 mM phosphate, residues Cys8–Ile12 and Ser15 of the P-loop became visible in the [¹⁵N,¹H]-HSQC spectra. Spectral responses of Cj1258 to inhibitor titrations were less sensitive when phosphate was present (presumably caused by the competition for the active site); however, more residues could be used as reporters of binding to the P-loop. Upon addition of the inhibitors, differential shifting and/or broadening were observed for Cys8–Ile12, Ser15 for trans-cinnamic acid (CA) and 2-carboxycinnamic acid (CCA2) or Cys8–Asn11, Ser15 for 4-acetamidocinnamic acid (ACA4), or Cys8–Gly10, Ile12 for 4-(phosphonomethyl)-phenylalanine (P-Phe). In the absence of phosphate, all four mimetic compounds caused changes of peaks from residues Cys8 and Asn11, confirming the above results obtained in the presence of 2 mM phosphate. Additionally, the Cys13 peak was sharpened by CA, implying some conformational stabilization of the P-loop upon ligand binding.

Putative inhibitors, CA, CCA2, and ACA4 had a consistently profound effect on residues Ser116, Tyr127 [δ(ppm) > 0.1], and Ser128 [δ(ppm) > 0.04] of the D-loop. As well, significant Trp126 broadening was detected when phosphate was present and the residue was therefore observable. Similarly, the addition of P-Phe noticeably influenced residues Trp126–Ser128, but Ser116 was not affected by the inhibitor. Interaction of Trp126–Tyr127 with the benzene ring of the inhibitors could be expected on the basis of the reported structural studies of BPTP, HCPTP-A, and LTP1 in complexes with HEPES, MES, and pNPP (Zhang et al. 1997, 1998; Wang et al. 2000b). In BPTP and HCPTP-A, residues Tyr131 and Tyr132, corresponding to Trp126 and Tyr127 of Cj1258, form a hydrophobic wall that interacts with the piperazine (HEPES) and morpholine (MES) rings analogous to the benzene ring of the CA derivatives (Zhang et al. 1997, 1998). In LTP1, the side chain of Trp134 (Trp126 in Cj1258) packs against the HEPES piperazine ring (for the wild-type protein) or closely interacts with the pNPP nitro group (for the inactive C13A mutant) (Wang et al. 2000b).

On the other side of the active site, the E-loop is affected markedly by all four compounds, and the majority of residues constituting the E-loop are shifted in the NMR spectra. It is interesting to note that Glu45 was affected much less by the smallest compound CA, as opposed to the larger CCA2, ACA4, and P-Phe. In fact, in complex with P-Phe, Glu45 exhibited the strongest chemical shift as compared with other residues of the E-loop. On the other hand, Met51 was shifted significantly [δ(ppm) > 0.08] for all three CA-based ligands, whereas no change was caused by P-Phe binding. These observations are consistent with the proposal that the loop connecting β2 and α2 of LMW-PTPs (i.e., the E-loop in Cj1258) is responsible for substrate specificity (Cirri et al. 1996; Zhang et al. 1998), and suggest that Glu45 and Met51 may play an indirect role in the recognition of the small molecular weight inhibitors.

Other specific spectral changes induced by the CA-based compounds include significant chemical shifts in the α5 helix. Spectral changes in α5 were almost undetectable upon interaction of Cj1258 with P-Phe, whereas CCA2 produced chemical shifts of ∼0.1 ppm in Tyr135–Ile137, Leu140, and Lys143.

Adenine (AD) and hypoxanthine (HX) are known activity modulators of LMW-PTPs (Dissing et al. 1993; Ostanin et al. 1995). In a three-dimensional crystal structure of the complex formed by LTP1 with adenine and inorganic phosphate, the adenine molecule was found to be sandwiched between Trp134 and Tyr51 residues of the D-loop and a loop corresponding to the E-loop of Cj1258 (Wang et al. 2000a). Hydrogen bonds connect the adenine molecule to residues His52, Asp132, and an ordered water molecule next to the phosphate ion. Potential interactions of adenine/hypoxanthine with Cj1258 were therefore tested both in the absence and presence of a comparatively high concentration of phosphate (20 mM). Overall, phosphate diminishes the effects of adenine/hypoxanthine binding, implying a negative impact of the bound phosphate ion on the affinity of Cj1258 for the modulators. Both molecules altered the P-loop resonances under both phosphate-loaded and phosphate-free conditions (Cys8 and Asn11, no phosphate, AD/HX; Cys8–Ser15, 20 mM phosphate, AD; Leu9–Ser15 20 mM phosphate, HX). Chemical shifts caused by hypoxanthine in the P-loop were larger than those caused by adenine. Apparently, both hypoxanthine and adenine produce an indirect effect on the P-loop conformation, which may explain the changes in k_cat and K _m observed in the presence of the activity modulators.

In Cj1258, Glu45 sequentially corresponds to Tyr51 of LTP1, the residue stacking adenine on the E-loop side (Wang et al. 2000a). Surprisingly, neither Glu45 nor its neighbor His46 of Cj1258 displayed large chemical shifts upon adenine binding, even though Glu45 was one of the most affected residues in the E-loop upon binding of CA derivatives and P-Phe. The largest chemical shifts [δ(ppm) > 0.06 for the phosphate-free sample] were detected for residues Thr42, Gly44, and Met51. As opposed to adenine, binding of hypoxanthine affected the E-loop of Cj1258 to a much lesser extent. This is in agreement with the prediction made on the basis of the established molecular contacts of the adenine molecule in LTP1 (Wang et al. 2000a). Residue His46 of the E-loop is expected to form a hydrogen bond with Oɛ1 of Glu49, leading to an elevated pK _a due to double protonation, and consequently, it is not able to form a hydrogen bond with the protonated N1 atom of hypoxanthine. As in the complex with adenine, in the presence of hypoxanthine, the most shifted residues were Thr42, Gly44, and Met51 [δ(ppm) 0.022, 0.024, and 0.045, respectively].

As expected, both molecules altered the chemical shifts of Trp126–Ser128. The adenine/hypoxanthine-induced shifts of Trp126 were only detected in the presence of phosphate, when the residue is seen in the HSQC spectrum. One residue of the D-loop and several of the α5 helix were only affected in the phosphate-free solution: Ser116 [δ(ppm) = 0.109 for AD binding; 0.048 for HX binding], Asp132 [δ(ppm) = 0.053 for AD binding; 0.060 for HX binding], Tyr135 (broadened for AD binding), Leu140 [δ(ppm) = 0.019 for HX binding], Lys143 [δ(ppm) = 0.094 for AD binding; 0.130 for HX binding], and Leu146 [δ(ppm) = 0.051 for HX binding].

Discussion

Bacterial LMW-PTPs are ubiquitous regulators of tyrosine phosphorylation. Despite their abundance, the spectrum of functions of LMW-PTP in prokaryotes has yet to be elucidated. Several studies demonstrate a role of LMW-PTPs in CPS/EPS synthesis and export, and in stress resistance (Ilan et al. 1999; Vincent et al. 1999, 2000; Wugeditsch et al. 2001; Klein et al. 2003; Musumeci et al. 2005; Peleg et al. 2005). Cj1258 is a single protein assigned to the LMW-PTP family in C. jejuni (Parkhill et al. 2000), a common cause of infectious enterocolitis. The physiological function of the protein is not yet established; however, it shares >60% sequence similarity with LMW-PTPs from human pathogens, such as the HIV-associated Mycoplasma penetrans (Sasaki et al. 2002); Pasteurella multocida (May et al. 2001); Neisseria meningitides, a causative agent of meningitis and septicemia (Tettelin et al. 2000); Bacillus anthracis, the etiological agent of anthrax (Read et al. 2003); and its evolutionary neighbor Bacillus cereus (Ivanova et al. 2003). No bacterial PTKs have been associated with Cj1258, and its functional annotation is still ambiguous, but there are known examples of bacterial proteins homologous to LMW-PTPs being recruited for alternative functions (Bennett et al. 2001; Zegers et al. 2001).

We have determined the three-dimensional solution structure of Cj1258 and found that it has the canonic structure topology of LMW-PTPs containing a central four-strand parallel β-sheet and five α-helices. We also found that the fourth strand of the β-sheet is unusually short and is flanked by two helical turns. This somewhat unusual structural feature is not likely to play a functional role in the enzymatic activity, as this β-strand is remote from the active site. Line broadening and poorer structural definition of some residues in the P-, E-, and D-loops suggest the presence of conformational exchange on a microsecond–millisecond time scale. A superposition of the secondary structures of Cj1258 with those of other LMW-PTPs demonstrated that Cj1258 deviates less from human (HCPTP-A) and bovine (BPTP) LMW-PTPs than from those of the lower organisms S. cerevisiae (LTP1), T. fetus (TPTP), B. subtilus (YwlE), and M. tuberculosis (MPtpA), as judged by RMSD values. The solution three-dimensional structure of the LMW-PTP YwlE from the Gram-positive bacterium, B. subtilus has been reported recently (Xu et al. 2006). The YwlE phosphatase was proposed to play a role in stress resistance (Musumeci et al. 2005), providing an example of the diverse LMW-PTP functions in bacteria. The YwlE phosphatase was shown to have an α/β core packing typical of all known LMW-PTPs, including Cj1258. Similarly to Cj1258, the fourth β-strand was flanked by two short helices, a 3₁₀-helix preceding the β-strand and an α-helix following the β-strand. On the other hand, contrary to Cj1258, the backbone atom coordinates of the YwlE core fold are closer to LTP1 and MPtpA than to BPTP and HCPTP-A (Xu et al. 2006). It is worth noting that sequence comparison (Fig. 1) shows closer similarity of Cj1258 to eukaryotic PTPs than to MPtpA and YwlE. The functional significance of this finding is not clear, particularly in light of the observation that surface loops forming the active site region are mobile in solution and are capable of accommodating a range of small-molecule compounds.

The mobility of the active site loops is an important structural feature related to the catalytic mechanism, specificity of LMW-PTPs, and to inhibitor design. This issue has been discussed in detail in a recent report on the solution structure of TPTP, a LMW-PTP from the primitive eukaryotic bovine parasite T. fetus (Gustafson et al. 2005). The structure of TPTP was determined in the presence of 20 mM phosphate, which is an inhibitor of PTPs (K _i ^TPTP = 7.6 mM (Thomas et al. 2002)). A comparatively low-inhibitor concentration used in that study allowed additional observations regarding the mobility and hydrogen bonding network around the active site. The discussion and hypotheses made by Gustafson et al. (2005) can be extended also to Cj1258 as follows. In the crystal structure of BPTP (Su et al. 1994; Zhang et al. 1994b), phosphate binding is facilitated by a stabilization of the conserved active site Asn15 residue (Asn11 in Cj1258 and Asn14 in TPTP) in a left-handed conformation. The unfavorable left-handed conformation of the Asn residue is maintained by hydrogen bonds with a conserved Ser43 (Ser39 in Cj1258 and Ser37 in TPTP) and His72 from the long loop connecting the α2 and α3 helices (His66 in Cj1258 and Gln67 in TPTP). In solution, residue His72 of BPTP, Gln67 of TPTP, and His66 of Cj1258 are found to be too far from the corresponding Asn residue to form a hydrogen bond (∼8–9 Å). Nevertheless, it was hypothesized that the differences in the crystal and solution structures of BPTP may be due to a transient nature of the Asn–His hydrogen bond in solution as opposed to a stable one in crystal. There may also exist a transient hydrogen bond in the solution structures of TPTP (Gustafson et al. 2005) and Cj1258 (this study). A number of calculated structures of both TPTP and Cj1258 containing left-handed Asn residues in the active site witness in favor of this hypothesis, because a strained, left-handed conformation of the Asn residue is a hallmark of the proper hydrogen bonding within the active site. Such a transient nature of the Asn–His/Gln hydrogen bond may be attributed to only a partial occupancy of the phosphate-binding site at low phosphate concentrations (Gustafson et al. 2005). Similarly, the P-loop broadening and poor structural definition of the loop connecting the helices α2 and α3 in Cj1258 is a consequence of a comparatively low occupancy of the phosphate-binding site.

Interaction of Cj1258 with phosphotyrosine mimetics and purine-type modulators of the LMW-PTP catalytic activity is entirely in line with the studies reported for eukaryotic LMW-PTPs. Cj1258 binds to competitive PTP inhibitors via the phosphate-binding P-loop, E-loop, and the hydrophobic tip (Trp126–Tyr127) of the D-loop regions. The molecular interactions are somewhat indiscriminate, since the Cj1258 protein can adapt to a number of chemical variants of the small-molecule ligands. In this regard, it is interesting that CA-based inhibitors interact with Cj1258, since they were originally developed for the mammalian PTP1B (Moran et al. 1995) and to our knowledge were never reported to have been tested for interactions with LMW-PTPs. The molecular adaptations appear to involve residues of the E-loop, which is believed to be essential for substrate specificity (Cirri et al. 1996; Zhang et al. 1998), as well as Ser116 of the D-loop and the α5 helix. The apparent effect of inhibitors on the spectral characteristics of Ser116 and residues of the α5 helix highlights the importance of long-range protein structural rearrangements in response to binding and their potential contribution to substrate specificity. Ser116 is located on a tip of a D-loop protrusion (residues Ser114–Leu117) occluding the α5 helix in Cj1258. Homologous protrusions are observed in other D-loops of LMW PTPs (Logan et al. 1994; Zhang et al. 1994b, 1998; Wang et al. 2000a; Madhurantakam et al. 2005) and their positions with respect to the α5 helix vary significantly. Moreover, a comparison of the solution (Logan et al. 1994) and crystal (Zhang et al. 1994b) three-dimensional structures of bovine BPTP shows a great degree of conformational adaptability of the D-loop protrusion to the environment. It is possible that interactions of the D-loop protrusion with the α5 helix also influence substrate specificity.

Although the E-loop region of Cj1258 contains Glu45 instead of a hydrophobic aromatic residue, which could stack with adenine in the active site, adenine still binds to the protein and enhances its phosphatase activity. Moreover, neither Glu45 nor His46 are strongly affected by adenine binding, suggesting an alternative mode of interaction between the E-loop and adenine. On the other hand, a weaker influence of hypoxanthine on the chemical shifts of the E-loop residues may indicate the existence of a hydrogen bond between the N1 atom of adenine and a hydrogen donor of the E-loop, possibly Nɛ2 of His46. Activation of phosphatases by adenine has been attributed to holding an oriented water molecule in the proximity of the phosphorus atom of the phosphoenzyme catalytic intermediate (Wang et al. 2000a). Additionally, the observed indirect influence of adenine on the P-loop conformation may contribute to the activation. Further studies are required for the precise positioning of the adenine molecule into the active site of Cj1258.

The observed interactions of small molecules with the active site of Cj1258 suggest strategies for the design of high-affinity inhibitors. Previous design efforts of PTP inhibitors utilized iterative structure-based affinity and selectivity optimization guided by X-ray protein crystallography, molecular modeling, and enzyme kinetic analyses (Iversen et al. 2000; Lund et al. 2004). The Cj1258 protein and its LMW-PTP homologs display highly adaptable loop structures in the vicinity of the active site, potentially responding to small molecule binding with a reorganization of the structure and hydrogen bonding network. This conformational flexibility may represent a significant challenge for molecular modeling and limit the success of affinity predictions. On the other hand, the fragment-based methodology (Shuker et al. 1996; Song and Ni 1998) may prove to be more effective for the Cj1258 system as shown for the design of other PTP inhibitors (Puius et al. 1997; Szczepankiewicz et al. 2003; Zabell et al. 2004). More specifically, for LMW-PTPs, linking molecular fragments corresponding to phosphotyrosine mimetics and purine derivatives may lead to higher affinity molecules. A similar strategy can also utilize the small set of synthetic compounds with putative binding to both the phosphate and purine-binding sites of HCPTP-A (Zabell et al. 2004).

One of the potential difficulties for the fragment-based approach is the negative impact of fragment binding to one of the phosphate and purine-binding sites on the other. For example, we observed a weaker NMR signal response to adenine and hypoxanthine when 20 mM phosphate was present. Also, the affinities of the modulators for free human LMW-PTPs are different from the affinities for the corresponding enzyme–substrate complexes (Dissing et al. 1993). Therefore, it will be essential to carry out NMR-based screening (Shuker et al. 1996) of molecule pairs chosen among the derivitized phosphotyrosine- and purine-like compounds.

Overall, the three-dimensional structure of Cj1258 and its interaction with small molecules confirm the assignment of its function as phosphotyrosine dephosphorylation. Whether it has an internal substrate or is involved in pathogenesis in other manners (DeVinney et al. 2000) remains to be determined. The gene was present in >95% of the C. jejuni strains that have been genotyped by DNA microarrays, which included both animal commensal and human pathogenic strains (Taboada et al. 2004). This level of conservation is seen for less than one-third of all the genes in the genome strain, NCTC11168. Cj1258 homologs are also present in all four of the other sequenced Campylobacter genomes, i.e., RM1221, C. upsaliensis, C. lari, and C. coli, but not in Helicobacter pylori. In other microarray experiments, the expression of its gene was decreased in vivo in an ileum model versus in vitro (Stintzi et al. 2005) and immediately after addition of iron to iron-limited cells (Palyada et al. 2004). When the Cj1258 gene was knocked out by insertional inactivation, no obvious phenotypic changes were seen and general gene expression was not significantly different from the control (C.D. Carillo and J.H.E. Nash, pers. comm.). The same expression data do not entirely rule out a role in pathogenesis, since the expression levels of genes involved in infection processes may vary with the stage of infection, and the infectivity of the Cj1258 knockout mutant has yet to be examined. Notwithstanding these results, the high level of genome conservation points toward Cj1258 as having an intrinsic and potentially important role in the organism.

Materials and methods

Cloning, overexpression, and purification of Cj1258

The DNA encoding the putative protein tyrosine phosphatase Cj1258 was obtained from the genome strain C. jejuni 11,168 by PCR amplification with the PWO polymerase according to the manufacturer's recommendations (Roche). The primers used incorporated either the Nde1 or Sal1 cloning sites and a 5X His tag, followed by two termination codons in the 3P PCR primer. The primer pair used was: 5′-GCTAGCTAGCTACATATGAAAAAAATACTCTTTATATGCTTAGGC-3′ (forward sequence) and 5′-GCTAGCTAGCTAGTCGACTTATTAGTGATGGTGATGGTGTTTTGATAAAAAACAAGTAAATTTTTACAAGC-3′ (reverse sequence). The PCR product was gel purified, and cloning sites generated by double digestion with Nde1 and Sal1 restriction enzymes according to the manufacturer's suggested protocol (New England Biolabs, Inc.). The gene was cloned into the expression plasmid pCWori⁺ and the DNA construct maintained in E. coli AD202. Cells were grown at 37°C in 1 L of 2YT medium supplemented with 150 μg/mL ampicillin. Overexpression of Cj1258 was induced through the addition of 1 mM isopropyl β-D-thiogalactopyranoside (IPTG) at an A₆₀₀ of 0.5, and the growth continued for 6 h. Cells were harvested by centrifugation at 10,000g for 15 min, resuspended in 10 mM HEPES buffer (pH 7.0) containing protease inhibitors (Roche), and lysed by mechanical disruption. The cell lysate was clarified by centrifugation at 27,000g for 30 min, and cell debris was discarded. Total membrane and soluble protein fractions were obtained from clarified cell extracts by centrifugation at 100,000g for 60 min. Following adjustment to 500 mM NaCl and 50 mM imidazole, the soluble protein fraction was loaded directly onto a 5-mL HisTrap column (Amersham Biosciences) previously equilibrated with 10 mM HEPES, 500 mM NaCl, and 50 mM Imidazole (pH 7.0). Cj1258 was eluted using a linear gradient of 50–500 mM imidazole in 10 mM HEPES and 500 mM NaCl (pH 7.0), and fractions containing the protein of interest, as determined by SDS-PAGE, were pooled and desalted three times by concentration and reconstitution with 10 mM HEPES buffer (pH 7.0) using an Amicon filtration membrane with a 10,000 MW cutoff (Millipore).

Enzyme kinetics determined by a continuous assay

The phosphatase activity of Cj1258 was determined at 37°C at the predetermined optimum pH of 6.5 in 100 mM sodium citrate using p-nitrophenyl phosphate as the substrate. After a preliminary set of experiments, enzyme kinetics were followed using 12 substrate concentrations spanning the concentration range of approximately 0.3× to 30× the K_m value. Kinetic reaction mixtures were thermally equilibrated for 10 min at 37°C prior to reaction initiation by the addition of 0.1 μg of purified Cj1258 in a total assay volume of 200 μL. The enzymatic rate was calculated from the observed increase of adsorption at A₄₀₅ and the kinetic parameters were determined from fitting of initial velocities to Michaelis-Menten kinetics using a GraphPad Prism Version 3 program. Triplicate sets of data were obtained at each substrate concentration.

NMR sample preparation

All putative phosphatase inhibitors were purchased from Sigma and used without further purifications. The Cj1258 phosphatase was uniformly labeled with either ¹⁵N or both ¹⁵N and ¹³C by growing cells in the ¹⁵N- or ¹⁵N/¹³C-labeled rich medium Bio-Express-1000 (Cambridge Isotope Labs), respectively. A volume of 2 mL of LB medium containing 100 μg/mL ampicillin was inoculated with a single colony and shaken at 37°C overnight. The cells were collected by centrifugation and transferred into 200 mL of the 0.5X Bio-Express medium containing 100 μg/mL ampicillin. The cell culture was shaken at 37°C and induced with IPTG when OD₆₀₀ reached 0.7–0.8. The cells were harvested by centrifugation 5 h after the induction. The Cj1258 protein was recovered from the cytoplasmic fraction and purified with Ni-NTA agarose affinity resin (QIAGEN, Inc.) by the manufacturer's standard protocol. Briefly, the cell pellet was suspended in lysis buffer (50 mM phosphate buffer, 300 mM NaCl, and 10 mM imidazole at pH 7.4) in the presence of a cocktail of protease inhibitors (Complete Lysis Kit, Roche Applied Science), and the cells were lysed by sonication. The cytoplasmic fraction was separated from insoluble cell debris by centrifugation at 100,000g. The supernatant was incubated for 1 h at room temperature with the Ni-NTA resin. The resin was washed with a solution of 20 mM buffered imidazole, and the protein was eluted with 250 mM imidazole in the lysis buffer.

For the last step of NMR sample preparation, the protein was transferred into a 20-mM phosphate buffer (pH 5.8 or 7.0) containing 90% H₂O/10% D₂O, 0.2 mM EDTA, and 0.01% NaN₃ using an Amicon Centriprep YM-10 concentration cell. To replace water with D₂O, buffered protein samples were lyophilized and reconstituted in the equivalent volume of 99.96% D₂O (Cambridge Isotope Labs). For titrations with small molecules, 11 or 22 μL of a concentrated stock solution of ¹⁵N-labeled Cj1258 in a buffer of 20 mM sodium acetate-d₃, 0.4 mM sodium phosphate, 0.2 mM EDTA, and 0.01% NaN₃ (pH 5.8) was added to inhibitor dissolved in 0.45 mL of a solution containing 50 mM sodium acetate-d₃, 90% H₂O/10% D₂O, 0.2 mM EDTA, and 0.01% NaN₃ (pH 5.8). When necessary, the final phosphate concentration was adjusted to 2 mM.

NMR spectroscopy and resonance assignments

NMR spectra were recorded at 298 K on a Bruker Avance-500 MHz spectrometer equipped with a cryoprobe and on a Bruker Avance-800 MHz spectrometer. The complete data set included HBHAcoNH, HNCACB, HNCO, CBCAcoNH with gradient enhancement, H(C)CH-COSY, 3D ¹⁵N-edited NOESY, 3D ¹³C-edited NOESY, and aromatic side chain ¹³C-edited 3D NOESY spectra (Gehring and Ekiel 1998; Sattler et al. 1999). The 3D ¹⁵N-edited NOESY was recorded with spectral widths of 11,494 Hz (F1, ¹H), 1784 Hz (F2, ¹⁵N), and 9602 Hz (F3, ¹H) and with 2048, 40, and 160 sampling points along the t₁, t₂, and t₃ time dimensions, respectively. The NOE mixing time was 120 msec. The 3D ¹³C-edited NOESY with a mixing time of 120 msec was carried out as a 1024 × 64 × 136 complex data matrix with the ¹³C carrier frequency set to 40 ppm and a spectral width of 65 ppm in the ¹³C dimension. For aromatic side chain atoms, the 3D ¹³C-edited NOESY was carried out with a mixing time of 120 msec and the ¹³C carrier frequency set to 130 ppm. The 3D HNHA experiment was used to measure ³ J _HNHα coupling constants (Kuboniwa et al. 1994). All NMR spectra were processed with NMRPipe (Delaglio et al. 1995) and converted for viewing and analysis by use of the NMRView software program (Johnson and Blevins 1994).

Assignments of the backbone ¹H, ¹⁵N, and ¹³C chemical shifts were derived from two-dimensional [¹⁵N,¹H]- and [¹³C,¹H]-HSQC spectra, combined with the set of three-dimensional HNCACB, CBCAcoNH, HNCO, and HBHAcoNH spectra. Sequential assignments were confirmed by the analysis of the HBHAcoNH and ¹⁵N-edited NOESY spectra. Aliphatic nuclei were assigned primarily through the analysis of an H(C)CH-COSY spectrum of Cj1258. The 3D ¹⁵N- and ¹³C-edited NOESY spectra were also utilized to confirm and obtain additional assignments, correlating the previously assigned H^N and H^α protons with their respective intraresidue side chain nuclei (H^α, H^β, H^γ, etc.). Proton assignments of the aromatic side chains were derived from the aromatic side chain ¹³C-edited 3D NOESY spectrum, combined with the 3D ¹⁵N- and ¹³C-edited NOESY spectra and the [¹³C,¹H]-HSQC spectrum of the aromatic region.

Chemical shift differences δ(ppm) induced by small-molecule binding were measured at 500 MHz and calculated as δ² = [Δδ(¹H)]² + [Δδ(¹⁵N) × α_N]², where the scale factor α_N ∼ 0.17 was estimated as a ratio of ¹H and ¹⁵N chemical shift dispersion for the backbone amides (Farmer et al. 1996).

Structure calculations

Automated assignment and structure calculations were performed using an ARIA-CNS protocol (Brunger et al. 1998; Linge et al. 2001). Distance restraints for the structure calculations were derived from the intensities of NOE cross-peaks in the 3D ¹⁵N- edited NOESY spectrum, ¹³C-edited NOESY spectra, and aromatic side chain ¹³C-edited 3D NOESY spectrum recorded both in H₂O/D₂O and D₂O. An error of ±1 Hz was assumed in converting 64 ³ J _HNHα coupling constants to dihedral φ angle restraints, using the parameterization method of Vuister and Bax (1993). Chemical shift-based dihedral angle restraints were obtained using the backbone torsion angle predictor TALOS (Cornilescu et al. 1999). The dihedral φ and ψ angles (a total of 198) were allowed to vary within the margins calculated by doubling the standard deviation ranges predicted by TALOS. Slowly exchanging amide protons were determined from the amide H/D exchange data obtained upon reconstitution of the lyophilized samples in D₂O. To detect these protons, we followed [¹⁵N,¹H]-HSQC spectra of the reconstituted protein for 2 d starting immediately after the D₂O addition.

The first rounds of structure calculations were used to remove assignment errors, noise, and other artifacts and to produce an initial set of global folds for Cj1258. The set of the initial structures was refined in the following rounds of calculations by lowering the ambiguity cutoff parameter and including 50 hydrogen bond and other manually assigned distance restraints in the restraint list. A set of 50 structures was generated at the beginning of every iteration. The final list of restraints contained 994 unambiguous and 158 ambiguous distant restraints divided into 452 sequential, 336 medium-range (1 < |i-j| < 4), and 364 long-range contacts. The 10 best structures with the lowest total energy were analyzed and visualized with the Sybyl, InsightII (Biosym), and MOLMOL software programs (Koradi et al. 1996).

Electronic supplemental material

Supplemental material for this article includes backbone ¹⁵N NMR relaxation parameters for the ¹⁵N-labeled Cj1258 protein (Supplemental Fig. A), mapping of small-molecule binding onto the three-dimensional structure of Cj1258 (Supplemental Fig. B), and tabulated values of chemical shifts δ(ppm) upon titration of Cj1258 with active site ligands (Supplemental Table A).