The Cysteine-rich Protein A from Helicobacter pyloriIs a β-Lactamase*

Among the large number of hypothetical proteins within the genomes of Helicobacter pylori, there is a family of unique and highly disulfide-bridged proteins, designated family 12, for which no function could originally be assigned. Sequence analysis revealed that members of this family possess a modular architecture of α/β-units and a stringent pattern of cysteine residues. The H. pylori cysteine-rich protein A (HcpA), which is a member of this family, was expressed and refolded from inclusion bodies. Six pairs of cysteine residues, which are separated by exactly seven residues, form disulfide bridges. HcpA is a β-lactamase. It slowly hydrolyzes 6-aminopenicillinic acid and 7-aminocephalosporanic acid (ACA) derivatives. The turnover for 6-aminopenicillinic acid derivatives is 2–3 times greater than for ACA derivatives. The enzyme is efficiently inhibited by cloxacillin and oxacillin but not by ACA derivatives or metal chelators. We suggest that all family 12 members possess similar activities and might be involved in the synthesis of the cell wall peptidoglycan. They might also be responsible for amoxicillin resistance of certain H. pylori strains.

tified. This group was called family 12 (9), but a functional assignment was impossible. In order to work toward the functional and structural characterization of family 12, we expressed and characterized the gene product of HP0211.
Cao et al. (11) identified the HP0211 gene product in H. pylori culture broth supernatant, verifying that this gene was expressed and secreted into the medium. The protein was termed HcpA (H. pylori cysteine-rich protein A), but a function was not assigned. In a further study, the HP0160 gene product, which is another family 12 member, was identified in H. pylori membrane fractions, and it was shown that this gene product was able to bind penicillin derivatives (12).
Penicillin-binding proteins (PBPs) and ␤-lactamases are enzymes that belong to the same class of proteins. PBPs are involved in the assembly, regulation, and maintenance of the cell wall peptidoglycan. They have in common that they covalently bind penicillins, but their biological functions are more diverse, involving carboxypeptidase, transpeptidase, and transglycosylase activities. In contrast to PBPs, ␤-lactamases hydrolyze ␤-lactame antibiotics. Many ␤-lactamases evolved from PBPs under the selective pressure of ␤-lactame antibiotics (13). Four classes of ␤-lactamases (classes A, B, C, and D) and six PBP classes are known (classes A, B, and C and low and high molecular weight of each) (reviewed in Refs. 14 -17). Most of these proteins contain an active site serine residue, but a minority are zinc-dependent enzymes (␤-lactamase class B).
Here we report the recombinant expression, refolding, and characterization of the HP0211 gene product. This gene codes for a protein that binds and hydrolyzes penicillin derivatives and might be involved in H. pylori antibiotic resistance. Since there are no sequence similarities with known ␤-lactamase or PBP classes, we suggest that HcpA belongs to a new class of ␤-lactamases, designated class E.

MATERIALS AND METHODS
Sequence Analysis-Multiple sequence alignment, dot blot analysis, and data base searches were done with the program package GCG from the University of Wisconsin Genetics Computer Group (18) or with the program ClustalX (19). Leader peptides were identified with the Internet-based neuronal network algorithm SignalP (20), and secondary structure was predicted using the PredictProtein server (21).
Expression Construct-The plasmid GHPDW78 harboring the gene HP0211 was obtained by the American Tissue and Culture Collection. The HP0211 gene was amplified by polymerase chain reaction using the 3Ј-and 5Ј-end primers TACGCTCCCGGGTTAGTGGTGGTGGTGGT-GGT GAAGTTCTATTTTTAATTCCTTGAGAGC and GCACCCCATG-GCAGAGCCAGACGCTAAAG, respectively. The first 23 amino acids corresponding to the leader peptide were deleted, and the C terminus was extended by a His 6 tag. After digestion with the restriction enzymes NcoI and XbaI, the insert was ligated into a pTFT74 expression vector and transformed into Escherichia coli BL21-competent cells. Ampicillin-resistant colonies were screened for expression in 1-ml cultures.
Expression and Isolation of Inclusion Bodies-A 2-liter culture of LB medium containing 100 g/ml ampicillin was inoculated with 40 ml of a stationary overnight culture of HcpA expressing BL21 cells. Cultures * This work was supported by the Baumgartner Foundation (Zü rich, Switzerland). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. were grown at 37°C with constant agitation (270 rpm). After approximately 2 h (A 600 ϭ 0.7), expression was induced with 1 mM isopropyl ␤-D-thiogalactopyranoside. The cultures were grown for an additional 3 h, and cells were harvested by centrifugation (30 min, 2,000 ϫ g, 4°C). Cells were suspended in 10 -20 ml of ice-cold lysis buffer (10 mM Tris/HCl, 2 mM magnesium sulfate, pH 6.8) and ruptured by sonification. 50 g/ml DNase and 65 g/ml RNase were added, and the solution was incubated at 37°C for 30 min. After adding EDTA and CHAPS to final concentrations of 25 mM and 0.25%, respectively, the solution was kept on ice for an additional 30 min. Inclusion bodies were collected by centrifugation (15 min, 20,000 ϫ g, 4°C), and the soluble fraction was discarded. The pellet was washed two times with buffer A (0.1 M Tris/HCl, 20 mM EDTA, pH 6.8) and subsequently buffer B (0.5 M GdnCl in buffer A). Inclusion bodies were solubilized in buffer C (5 M GdnCl, 0.2 M Tris/HCl, 0.1 M DTT, 10 mM EDTA, pH 8.0), and insoluble material was removed by centrifugation. Solubilized inclusion bodies were dialyzed against buffer D (5 M GdnCl, 0.1 M acetic acid).
Refolding and Purification-HcpA was refolded by immobilizing the solubilized inclusion bodies to Ni 2ϩ -NTA-agarose (Qiagen) and removing the guanidinium hydrochloride from the buffer. Protein concentration was determined by the Bradford method (⑀ 595 ϭ 0.084 ml⅐ g Ϫ1 ⅐cm Ϫ1 ) or amino acid analysis. 80 mg of unfolded HcpA was loaded onto 7 ml of Ni 2ϩ -NTA-agarose in buffer D. After adjusting the pH to 8.0, the slurry was filled into a column. A 400-ml gradient from 5 to 0 M GdnCl in 0.1 M Tris/HCl, 3 mM GSH, pH 8.0, refolded the immobilized protein (flow 1 ml/min, 20°C). Protein was eluted with 0.25 M imidazolium, 40 mM MES, pH 6.8, and protein-containing fractions were pooled and dialyzed against buffer E (40 mM sodium acetate, 1 mM EDTA, pH 5.4). To separate misfolded from active protein, the material was further purified on an ampicillin affinity matrix. Ampicillin was bound to activated agarose beads (Affi-Gel ® 10 Gel, Bio-Rad) according to the guidelines of the manufacturer. 25 mg of refolded protein was bound to 2 ml of ampicillin affinity resin, washed with 20 -30 ml of buffer E, and eluted with 10 ml of 0.8 M hydroxylamine in buffer E. Purified HcpA was dialyzed extensively against buffer E.
HPLC Analysis-Protein samples were analyzed on a Nucleosil 300/5 C8 reversed phase column attached to a Hewlett Pakard 1100 Chemstation HPLC system in 0.1% trifluoroacetic acid in water. The sample was eluted with 0.08% trifluoroacetic acid in 84% acetonitrile/water at a flow rate of 1 ml/min and detected at a wavelength of 215 nm. For gel filtration chromatography, a Superdex 200 10/30 column (Amersham Pharmacia Biotech) was equilibrated with phosphate-buffered saline buffer (150 mM sodium chloride, 20 mM sodium phosphate) (pH 5.4) and calibrated with a standard mixture of proteins (Amersham Pharmacia Biotech) at a flow rate of 0.5 ml/min. The following elution profile was used for calibration: blue dextran (2 MDa), 7.8 ml; bovine serum albumin (67 kDa), 13.8 ml; ovalbumin (43 kDa), 14.9 ml; chymotrypsinogen A (25 kDa), 17.1 ml; ribonuclease A (13.7 kDa), 17.8 ml.
Functional Characterization-Binding of 6Ј-Flu-Gly-6-APA (22) and Brocillin ® (Molecular Probes, Inc., Eugene, OR) (23) to HcpA was investigated by incubating the protein with the dye-labeled penicillin derivative at molar ratios of 1:60 to 1:3 in phosphate-buffered saline buffer (pH 6.8) for 30 min (37°C). Dye-labeled protein was separated by 15% SDS-PAGE under nonreducing conditions, and the gels were scanned at 530 nm on a Molecular Dynamics Fluorimager 575. ␤-Lactamase activity was assayed spectroscopically by using various chromogenic substrates. Ampicillin, amoxicillin, cefotaxime, cloxacillin, and benzylpenicillin were from Fluka; carbenicillin, cefalotin, cefoxitin, cephaloridine, and oxacillin were from Sigma; and nitrocefine was from Becton Dickinson. Ampicillin sulfone was a kind gift from the group of A. Plü ckthun (Zü rich, Switzerland). All reactions were performed in phosphate-buffered saline buffer (pH 6.0) at 25°C on a Cary 300 UV spectrophotometer. IC 50 values were determined with 200 M nitocefine at a protein concentration of 27 g/ml. Data were processed as described (24).
Folding Characterization-The folding/unfolding behavior was investigated by CD spectroscopy and by binding the fluorescent dye 8-anilino-1-naphtalenesulfonic acid (ANS) to HcpA. CD spectra were recorded at a protein concentration of 5 M in 0 -3 M GdnCl, 30 mM sodium phosphate, pH 6.9, on a Jasco J-751 CD spectrometer. ANS fluorescence was recorded at a protein concentration of 6.5 M in 0 -3 M GdnCl, 0.1 mM ANS, 0.1 M sodium phosphate, pH 6.9, and on a PTI-500 fluorescence spectrometer. The excitation wavelength was 350 nm. Fluorescence at 472 nm was plotted over the GdnCl concentration. The temperature was maintained at 22°C, and the data were fitted to Equation 1 (25). Y obs is the observed signal, a 0/1 and a 2/3 are the intercepts and the slopes at low and high GdnCl concentrations, respectively, a 4 is the free energy of unfolding extrapolated to 0 M GdnCl (⌬G H2O ), and a 5 is the cooperativity of the folding reaction (m). R is the ideal gas constant, and T is the absolute temperature.
Assignment of Disulfide Bridges-11 g of purified HcpA was digested with 0.5 g of trypsin (Promega, in 50 mM acetic acid) in 50 l of 0.1 M Tris/HCl, 10% (v/v) acetonitrile, 2 mM calcium chloride, pH 8.2, for 4 h at 37°C. Disulfide bridges were reduced by adding 0.5 l of 25 mM triscarboxyethylphosphine (Pierce) to 2 l of the digested sample (20 min, 21°C). 2 l of the reduced and oxidized digests were mixed with 2 l of 0.1% trifluoroacetic acid and desalted with a ZipTip C18 (Millipore Corp.). Peptides were eluted from the ZipTip with 3 l of a saturated solution of ␣-cyano-4-hydroxycinnamic acid in 0.1% trifluoroacetic acid (v/v), 60% (v/v) acetonitrile directly onto the MALDI target. MALDI data were recorded on a Bruker Biflex III instrument equipped with a scout ion source. Spectra where acquired with pulsed ion extraction in reflectron mode using a nitrogen laser.

RESULTS
Sequence Analysis-The analysis of the genomes of H. pylori strains 26695 (9) and J99 (10) revealed 95 protein families. According to this analysis, ORFs HP0235, HP0160, HP0211, HP0336, HP00628, HP1098, and HP1117 belong to family 12 (9) (Fig. 1a). The ORFs share between 66 and 22% sequence identity and are rich in cysteine, lysine, and arginine residues. All sequences besides HP0628 and HP0336 possess significant N-terminal signal sequences that guide these proteins into the periplasmic space. Dot plot analysis (18) reveal significant homology within each ORF, indicating that these proteins are composed of several repeats of a common sequence motif. The number of repeats varies between four for HP0336 and nine for HP0235. The sequence motif is approximately 36 residues long and shows a characteristic profile. The most prominent feature of this profile is two cysteine residues that are separated by seven residues (Fig. 1b). The cysteine residues are preceded by alanine, glycine, or serine residues and are present in all family members except HP1117. There is also a conserved alanine residue at position 10 of the profile (Fig. 1b). Residues up to the first cysteine residue are predicted to fold into an ␣-helix, and residues on the C-terminal side of the second cysteine are predicted to fold into a ␤-sheet conformation. The exact boundaries of the secondary structure predictions differ slightly, but the predictions are the same for all ORFs. Data base searches did not reveal sequence homology with any protein of known function.
Purification and Refolding-As anticipated due to the presence of 12 cysteine residues, recombinant HcpA accumulated in an insoluble form as inclusion bodies in E. coli. Several refolding protocols were tested. When the protein was refolded by dilution, the protein was soluble but precipitated upon concentration. We therefore refolded the protein when it was immobilized on a Ni 2ϩ -NTA-agarose column. This method had two advantages. It combined the purification and the refolding steps, and the protein was eluted at a relatively high concentration of 3-4 mg/ml. At the beginning of this investigation, the function of HcpA was unknown, and the refolding was done under standard conditions. When a functional assay became available, we optimized the refolding conditions such that the specific activity was maximized. A basic pH (pH Ͼ 7) and the presence of reduced glutathione were beneficial, but the addition of arginine and oxidized glutathione reduced the specific activity and were therefore omitted. The protein was pure, according to SDS-PAGE and HPLC analysis as indicated in Figs. 2, a and b, and 3a and was concentrated to 5-10 mg/ml. The yield of soluble protein was between 25 and 30%. To separate traces of misfolded molecules from the native protein, we added an affinity chromatography step. HcpA was bound to ampicillin that was immobilized on agarose beads by its primary amino group. After extensive washing, the bond between the protein and ampicillin was cleaved by hydroxylamine treatment.
Biochemical Characterization-The oligomerization state of purified HcpA was analyzed by gel filtration chromatography. The protein eluted as a single peak at 16.7 Ϯ 0.1 ml, indicating that the protein is monomeric in solution (data not shown). The HcpA sequence contains 12 cysteine residues. In order to determine if these cysteine residues form disulfide bridges, we analyzed the reaction with 5,5Ј-dithiobis-2-nitrobenzoic acid (DTNB). As shown in Fig. 3b, DTNB is not reduced by refolded HcpA under native conditions or in the presence of 3 M GdnCl. However, DTNB reacted with HcpA prior to refolding. This indicates that there are no free SH groups present on the surface of refolded HcpA or buried in the hydrophobic core. All cysteine residues must be involved in disulfide bridges. This is also supported by the different retention times of native and reduced protein in reversed phase HPLC (Fig. 3a) and by the molecular mass of refolded HcpA. In contrast to the native protein that elutes after 20.6 min as a single peak with a small shoulder, the reduced protein ( Assignment of Disulfide Bridges-Disulfide bridges in HcpA were assigned by tryptic digestion and subsequent reduction of the peptide mixture with triscarboxyethylphosphine. The MALDI spectra of oxidized peptides contained signals for five peptides that included cysteine pairs (Table I). The mass differences between oxidized and reduced peptides confirmed that cysteine pairs Cys 56 /Cys 64 , Cys 92 /Cys 100 , Cys 128 /Cys 136 , Cys 164 / Cys 172 , and Cys 196 /Cys 204 formed disulfide bridges a, b, c, f, and g in refolded HcpA. However, no unique peptide containing a disulfide bridge between Cys 232 and Cys 240 (disulfide h) was observed, although DTNB titration strongly suggested the presence of six disulfide bridges. Cys 232 and Cys 240 should also form a disulfide bridge, because no free SH groups were detected in refolded HcpA, and all cysteine residues besides these two residues were assigned to disulfide bridges. In addition, the observed intramolecular sequence similarity suggested that the disulfide connectivity is the same for all ␣/␤-units. HcpA contained 34 theoretical trypsin cleavage sites, and five of them cluster around the proposed disulfide bridge h. Perhaps the resulting peptides are simply too small for MALDI mass spectroscopy, or the digestion products are too heterogeneous.
Functional Characterization-Krishnamurthy et al. (12) reported the ability of the HP0160 gene product to bind the dye-labeled penicillin derivative 6Ј-Flu-6-APA. We repeated similar experiments with the refolded and purified HcpA protein and the chromogenic penicillin derivatives 6Ј-Flu-Gly-6-APA (22) and Brocillin (23). As shown in Fig. 4, a and b, both substances bound to HcpA in a concentration-dependent man- ner. Since both substances had only the penicillin moiety in common, it was unlikely that binding was mediated by the chromogenic groups. However, we also tested the ability of HcpA to hydrolyze other ␤-lactame substances. The results for a panel of frequently used ␤-lactame antibiotics are summarized in Table II. HcpA hydrolyzed several APA and ACA derivatives, indicating that besides its penicillin binding activity HcpA possesses also a ␤-lactamase activity. Using the chromogenic substrate nitrocefine, we analyzed the pH and temperature dependences of the hydrolysis (Fig. 5). The maximum activity was observed for pH 5.5. At acidic or basic conditions, HcpA is approximately 3-5 times less active. The temperature optimum was between 35 and 40°C. HcpA was efficiently inhibited by APA derivatives such as carbenicillin, cloxacillin, and oxacillin. ACA derivatives like cefotaxime, cefoxitin, and cefalotin were much less effective. Metal chelators such as EDTA and dipicolinic acid had no significant inhibitory activity.
Folding Characterization-To decide if the protein was in a native-like conformation, we analyzed the folding/unfolding behavior by CD and fluorescence spectroscopy. Since HcpA does not contain any tryptophane residues, we investigated the fluorescence quenching of ANS as a function of the guanidinium hydrochloride concentration. In the absence of GdnCl, HcpA possesses a CD spectrum that indicates a high ␣-helix content (Fig. 6a). From the CD spectrum, the contents of ␣-helix, ␤-sheet, turn, and random coil were determined to be 46, 30, 10, and 14%, respectively. From the sequence, the ␣-helix, ␤-sheet, and random coil contents were predicted to be 51, 13, and 36%, respectively. In the presence of 3 M GdnCl, the ␣-helix signal vanished completely from the CD spectrum. By plotting the CD signal at 222 nm over the GdnCl concentration, the free energy of unfolding (⌬G H2O ) and the cooperativity parameter (m) were determined from the intercepts and the slopes of the curves at the transition phase. Similar values were obtained when folding/unfolding was monitored by ANS fluorescence. From the titration curves shown in Fig. 6, a and

DISCUSSION
The protein HcpA was expressed, refolded from inclusion bodies, and purified by Ni 2ϩ -NTA and affinity chromatography. According to SDS-PAGE and reversed phase HPLC analysis, the protein is pure, which is an indispensable prerequisite for structural and functional studies. However, the refolding step could have yielded a mixture of soluble proteins with different disulfide connections and exactly identical electrophoretic mobilities. This possibility is supported by the observation that the electrophoretic mobilities of reduced and oxidized HcpA are identical within the limits of error (Fig. 2b). Nonreducing SDS-PAGE analysis is therefore unsuitable to distinguish HcpA species with different disulfide connections. Reversed phase HPLC seems to be more suitable, because the reduced protein elutes after a significantly different retention time from the reversed phase HPLC column. Because of the different retention times of oxidized and reduced HcpA, it is unlikely that disulfide-scrambled or partially reduced molecules show exactly identical retention times in reversed phase HPLC. Based on the reversed phase HPLC chromatogram, we conclude that there is one predominant molecular species present in the protein solution.
Because the function of HcpA was unknown at the beginning of this investigation, we tried to assess the proper folding by CD and ANS fluorescence spectroscopy. Particularly, the content of ␣-helices derived from the CD spectrum agrees very well with the value that was predicted from the amino acid se-  quence. The values for ␤-sheets and random coils agree much less, because ␤-sheets are more difficult to predict (26) and show a much weaker signal in the CD spectrum. Titration of HcpA with GdnCl yielded the free energy of unfolding (⌬G H2O ) and the cooperativity parameter of the folding reaction (m). It was shown that m is a function of the number of disulfide bridges and the surface that is buried upon folding, which is proportional to the chain length (27). Under the assumption that HcpA contains six disulfide bridges, the cooperativity parameter was calculated to be 18.7 kJ/(mol⅐M), which is in agreement with the experimentally obtained m values of 15.1 and 16.6 kJ/(mol⅐M). The m value for the GdnCl induced unfolding of ␣-chymotrypsin (241 residues, five disulfide bridges) was determined to be 17.2 kJ/(mol⅐M) (28) and agrees even better than the theoretical value. The free energy of unfolding ⌬G H2O depends on the exact experimental conditions and the technique that was used for its determination (29); therefore, it is an unsuitable folding indicator. Because of the good agreement between biophysical properties, such as ␣-helix content and cooperativity of unfolding, with the theoretical values, we assume that the refolded HcpA possesses a native structure.
The biological activity is of course the best indicator for the successful refolding of a protein. Therefore, the identification of penicillin binding activity (12) in one of the family 12 members was very important for the characterization of HcpA. Chromogenic penicillin derivatives, such as 6Ј-Flu-Gly-6-APA and Brocillin migrate with the protein even under denaturating conditions on an SDS-PAGE gel. The stability of the complex is an indication of a covalent interaction between the penicillin moiety and the protein. We utilized this stability for the application of an additional affinity purification step. The covalent bond between ampicillin and the protein was cleaved by hydroxylamine (30). The addition of an affinity chromatography step ensured that only active HcpA was used for further characterization.
The biological functions of PBP are quite diverse. They include transpeptidase, transglycosylase, and carboxypeptidase activities. These activities consist of an acylation and a deacylation step. HcpA possesses both activities, which is shown by its ability to hydrolyze APA and ACA derivatives (Table II). The K m and k cat values for the hydrolysis of the chromogenic ACA derivative nitrocefine were 47 M and 0.28 min Ϫ1 , respectively. The corresponding values for the E. coli RTEM ␤-lactamase are 45 M and 48,000 min Ϫ1 (31). Similar values have been determined for many other ␤-lactamases (reviewed in Ref. 32). For the nitrocefine hydrolysis, the K m values of HcpA and E. coli RTEM ␤-lactamase are similar, but the E. coli enzyme is several orders of magnitude more active. Different explanations are possible for the low catalytic activity. One possibility might be that only a minor part of the protein is correctly folded. In contrast to K m , which is independent of the protein concentration, k cat depends very much on the exact concentration of active protein. If this were the reason for the low activity, it would have been impossible to separate the HcpA-penicillin complexes by gel electrophoresis. It is more likely that the deacylation is the rate-limiting step of the reaction, which would explain the binding of HcpA to the ampicillin affinity resin and the stability of the HcpA-Brocillin and HcpA-6Ј-Flu-Gly-6-APA complexes. The activity might also be affected by the His 6 tag and the deletion of the N-terminal leader peptide. For the E. coli RTEM ␤-lactamases, it was shown that the deletion of the leader peptide increased the catalytic activity (31). The HP0160 gene product that was isolated from H. pylori membranes also lacks the leader peptide (residues 1-25). Therefore, we assume that the activity of refolded HcpA is similar to the activity of the native protein.  The nitrocefine hydrolysis is pH-and temperature-dependent. The highest activities were found for pH 5.5 and 37°C. These values reflect the adaptation of H. pylori to its biological niche. The optimal temperature for hydrolysis corresponds to the body temperature of its host. Although the pH of the gastric mucosa is strongly acidic (pH 1-2), H. pylori survives these harsh conditions by the formation of a local microenvironment. The urease activity is responsible for a much higher pH on the surface of the bacteria (33), and the HcpA activity was obviously optimized for the pH of the local microenvironment rather than for the pH of the gastric juice.
Penicillin binding proteins and ␤-lactamases belong to the same protein superfamily and have been classified according to sequence similarity (14) or substrate specificity (32). The overall sequence similarity is very low, although the topologies of secondary structural elements and the active sites of ␤-lactamases are similar. The three-dimensional structures of PBPs and ␤-lactamases share a common ␣/␤-sandwich domain that hosts the penicillin binding site. So far, four classes of ␤-lactamase sequences (denoted A-D) and six classes of PBP sequences (denoted A-C and the low and high molecular weight of each) are known (14). There is no detectable sequence similarity between any family 12 sequence and the sequences in the PBP/␤-lactamase superfamily. Therefore, we propose to place HcpA and the HP0160 gene product (12) into a new class of ␤-lactamases designated class E. Although there is no sequence similarity, the suggested modular architecture of the family 12 proteins is not substantially different from known ␤-lactamase structures. The gene products of HP0235, HP0160, HP1098, HP0211, HP0628, and HP0336 consist of 9, 7, 7, 6, 6, and 4 ␣/␤-units, respectively, with a disulfide bridge between the ␣-helix and the ␤-sheet. These ␣/␤-units could fold into a ␣/␤sandwich domain that is similar to the known ␤-lactamase structures. However, a regular architecture of secondary structural elements as is predicted for the family 12 proteins has not been observed for any ␤-lactamase or PBP so far. ␤-Lactamases are recognized by a significant sequence motif that consists of three loci. Locus 1 contains the motif SXXK, including the active site serine. Loci 2 and 3 possess the more diverse patterns ((S/Y)X(N/S/D/C) and (K/L)(T/S)G) and are separated from locus 1 by up to a few hundred residues. Several sequence patterns that fit the motifs for loci 1 and 2 are present in the HcpA sequence (Fig. 1a), but locus 3 is missing completely. If these residues were important for the ␤-lactamase activity, they should have been conserved throughout the family 12 sequences or at least in the sequence of HP0160, which is the only family member where a similar function was confirmed. Since this is not the case, it is uncertain if the predicted residues are involved in the active site and if the classical ␤-lactamase sequence motif is valid for this particular class of enzymes.
Because sequence similarity is not applicable to assign HcpA to one of the known ␤-lactamase families, we investigated the substrate specificity for various ␤-lactame substrates. According to the substrate and inhibitor profiles, four main groups have been distinguished (32). The distinction was made based on the preference for either APA or ACA derivatives and the inhibition by either EDTA or clavulanic acid. HcpA slowly hydrolyzes APA as well as ACA derivatives. The turnover for APA derivatives is typically in the range between 1.0 and 0.5 min Ϫ1 , which is 2-3 times faster than for ACA derivatives. HcpA is therefore regarded as a penicillinase rather than a cephalosporinase. Cephotaxime is neither a good substrate nor a potent inhibitor for this enzyme. The enzyme is not inhibited by EDTA or dipicolinic acid, indicating that HcpA is a metalindependent ␤-lactamase. Because clavulanic acid was not available to us, we used the closely related inhibitor ampicillin sulfone instead. In contrast to clavulanic acid, ampicillin sulfone is less efficient but shares the same binding mechanism (34). Ampicillin sulfone is a rather weak inhibitor with an IC 50 value in the upper micromolar range. In contrast to ampicillin sulfone, cloxacillin and oxacillin are much more potent. This inhibitor profile is distinct from all profiles of the ␤-lactamase class 2. Bush and co-workers (32) suggest that penicillinases that are metal-independent and are not efficiently inhibited by clavulanic acid, such as HcpA, should belong to the activity class 4.
At the moment, we can only speculate about the functions of HcpA in vivo. The most probable biological functions of HcpA and the family 12 proteins in general are the maintenance of the cell wall proteoglycan through the bacterial life cycle. H. pylori appears in two distinct morphologies. Spiral-shaped bacteria are the predominant form in the stomach, but cocoidal morphologies were also observed in damaged tissue (35,36). Associated with the transition from spiral to cocoidal shaped morphologies is a switch in the muropeptide composition, which is substantially different from other bacteria such as E. coli (37). The remodeling of the muropeptide and its special composition requires a set of specific enzymes. Since family 12 members are unique to H. pylori, they are potential candidates for this task.
PDBs are classical targets for antibacterial drugs, and the closely related ␤-lactamases are responsible for drug resistance. Antibiotic resistance also occurs in the treatment of H. pylori with amoxicillin. Dore et al. (38) investigated the binding of [ 3 H]benzylpenicillin to PBPs from amoxicillin-susceptible and -resistant H. pylori strains by SDS-PAGE analysis. Besides three high molecular weight PBPs, they identified a novel 30-kDa protein only in amoxicillin-susceptible strains. In amoxicillin-resistant strains, this protein was not detected. These results indicate that the 30-kDa protein, which was not further characterized, was either no longer expressed or was mutated such that the protein-[ 3 H]benzylpenicillin complex became less stable. Because HcpA binds benzylpenicillin and possesses the right molecular weight, it could be involved in ampicillin resistance. However, the presence of ␤-lactamases in significant amounts could not be detected in vivo, which can be explained by the moderate catalytic activity of HcpA in vitro. We conclude that HcpA is the prototype of a new family of ␤-lactamases that are so far unique to H. pylori and might fulfill important functions in cell wall maintenance and development. Therefore, HcpA might be a suitable target for the development of drugs against H. pylori.