The Clostridium ramosum IgA Proteinase Represents a Novel Type of Metalloendopeptidase*

Clostridium ramosum is part of the normal flora in the human intestine. Some strains produce an IgA proteinase that specifically cleaves human IgA1 and the IgA2m(1) allotype. This prolylendopeptidase was purified from a broth culture supernatant, and N-terminal sequences of the native protein and tryptic fragments thereof were determined. A fragment of the igagene encoding the IgA proteinase was isolated using degenerate primers in PCR, and the complete gene was obtained by inverse PCR. The identity of the iga gene was confirmed by heterologous expression inEscherichia coli. The deduced amino acid sequence indicated a signal peptide of 30 residues and a secreted proteinase of 133,828 Da. A typical Gram-positive cell wall anchor motif was identified in the C terminus. The presence of a putative zinc-binding motif His-Glu-Phe-Gly-His together with inhibition studies indicate that the proteinase belongs to the zinc-dependent metalloproteinases. However, the sequence of the C. ramosumIgA proteinase shows no overall similarity to other proteins except for significant identity around the zinc-binding motif with family M6 of metalloendopeptidases, and the unique sequence of the IgA proteinase in this area presumably establishes a new subfamily. The GC percentage of the iga gene is significantly higher than that for the entire genome of C. ramosum, suggesting that the gene was acquired recently in evolution.

IgA is the major class of immunoglobulin in human mucosal secretions. Two subclasses, IgA1 and IgA2, exist, and IgA2 is found in three allelic forms, A2m(1), IgA2m (2), and A2m (3), among which IgA2m(1) is expressed mainly in Caucasians (1). The IgA1 subclass is predominant in the upper respiratory tract and in serum, whereas more even proportions of IgA1 and IgA2 occur in intestinal and urogenital secretions (2,3). A number of bacterial species that colonize mucosal membranes of man produce IgA1 proteinase. This group of postproline endopeptidases cleaves one of several Pro-Ser or Pro-Thr peptide bonds in the hinge region of human IgA1, including its secretory form, S-IgA1, within an inserted stretch of 13 amino acid residues lacking in IgA2. IgA1 proteinase-producing species include the three leading causes of bacterial meningitis, Neisseria meningitidis, Hemophilus influenzae, and Streptococcus pneumoniae; three commensal streptococci, Streptococcus mitis, Streptococcus oralis, and Streptococcus sanguis; Gemella hemolysans; and several species of Capnocytophaga and Prevotella. In addition, the urogenital pathogens Neisseria gonorrhoeae and Ureaplasma urealyticum produce IgA1 proteinase (reviewed in Refs. 4 and 5).
It is conceivable that IgA1 proteinases enable the bacterial species to escape specific immune defense on mucosal surfaces, although lack of relevant animal models has precluded definitive identification of their exact biological significance (5). Notably, human IgA1 cleaving activity among these bacteria has evolved in at least three independent evolutionary lineages, emphasizing the biological importance of these enzymes. Inhibition studies and molecular characterizations have shown that the Hemophilus and Neisseria IgA1 proteinases are homologous serine-type proteinases, the streptococcal IgA1 proteinases are mutually related metalloendopeptidases, and the enzyme from Prevotella melaninogenica (formerly Bacteroides melaninogenicus) is a cysteine proteinase (reviewed in Ref. 5). An IgA1 proteinase of N. gonorrhoeae has been shown to cleave not only human IgA1 but also LAMP1 (lysosome-associated membrane protein 1) and tumor necrosis factor ␣ (TNF␣) receptor II, features that may contribute to the pathogenesis of infections caused by these bacteria (6,7).
Clostridium ramosum is a strict anaerobic, Gram-positive, spore-forming bacterium. It is part of the commensal flora in the human intestine (8,9), and only rarely has it been associated with disease (10). Some strains of C. ramosum produce an IgA proteinase that cleaves human IgA1 and IgA2 allotype A2m(1) at a Pro-Val peptide bond at positions 221-222 just N-terminal to the hinge region (8,11,12). In IgA2m(2) and IgA2m (3), the Pro at position 221 is substituted by Arg, apparently rendering these allotypes resistant to cleavage by the IgA proteinase. Other Clostridium species may produce proteinase(s) with similar activity (13). The C. ramosum IgA proteinase is inhibited by high concentrations of EDTA, suggesting that it is a metalloproteinase (11). Otherwise, this enzyme has not yet been characterized.
Here we describe the purification, cloning, and characterization of the C. ramosum IgA proteinase. Analysis of the deduced translation product of the iga gene encoding the enzyme revealed that it is a metalloendopeptidase with a putative ex-tended zinc-binding motif HEXXHXXXGXXD. The primary structure of the IgA proteinase shows no significant overall similarity to any other known metalloendopeptidase, including any IgA1 proteinase belonging to this class of proteolytic enzymes. Notably, however, the sequence of 30 residues around the zinc-binding motif shows up to 60% identity to the equivalent region of proteinases grouped into the family M6 of metallopeptidases, including PrtV proteinase of Vibrio cholerae and immune inhibitor A of Bacillus thuringiensis. The GC% of the iga gene is significantly higher than that reported for the C. ramosum genome, suggesting that the IgA proteinase gene was acquired recently in evolution through horizontal gene transfer.

EXPERIMENTAL PROCEDURES
Bacterial Strains-C. ramosum strain AK183, which produces an IgA proteinase that cleaves both human IgA1 and the IgA2m(1) allotype, was obtained from Dr. Y. Fujiyama (Kyoto, Japan). Bacteria were grown anaerobically at 37°C in 2ϫ YT medium (14) supplemented with 0.05% sodium thioglycolate. The 2ϫ YT medium was used because it is devoid of high molecular weight proteins that may complicate subsequent purification of the proteinase. Escherichia coli JM 109 (Stratagene, La Jolla, CA) was used as host for propagation of recombinant forms of plasmid pUC19. E. coli One Shot was used for cloning derivatives of the pCR-TOPO vector (Invitrogen, Groningen, The Netherlands), and E. coli BL21(DE3)pLysS (R & D Systems Europe, Abingdon, UK) was used for the expression cloning. The E. coli strains were grown in 2ϫ YT or LB medium (14) supplemented with antibiotics when appropriate.
Preparation of the IgA Proteinase-The cell-free supernatant of a 5-liter culture of C. ramosum was obtained by centrifugation (6,000 ϫ g, 30 min, 4°C). Solid (NH 4 ) 2 SO 4 was added stepwise to 47% saturation, which in a pilot experiment was found to precipitate the IgA proteinase, and NaN 3 was added to a final concentration of 0.05%. The precipitate formed overnight at 4°C was collected by centrifugation (6,000 ϫ g, 45 min, 4°C), dissolved in 0.05 M Tris-HCl, pH 8.2 (buffer T), dialyzed against the same buffer, and stored at Ϫ20°C until used. The ammonium sulfate cut was further fractionated by size exclusion chromatography on a column (1.6 ϫ 50 cm) of Sepharose 12 (prep grade; Amersham Biosciences) equilibrated in buffer T. Eluent fractions containing the IgA proteinase were identified by the capacity to cleave human IgA1 (see below). Fractions with maximal IgA1 cleaving activity were pooled and subjected to anion exchange chromatography on a Mono-Q column (Amersham Biosciences) equilibrated with buffer T. Protein fractions eluted with a gradient of NaCl (0 -1 M) in buffer T were analyzed for IgA1 cleaving activity.
IgA Proteinase Activity Assays-IgA proteinase activity was detected by its capacity to cleave purified human myeloma IgA1. Briefly, IgA proteinase test samples were incubated with myeloma IgA1 at 0.5 mg ml Ϫ1 in buffer T at 37°C overnight, and cleavage was subsequently detected by the presence of characteristic Fc and Fd fragments as revealed by SDS-PAGE (15). This assay was used to determine the susceptibility of the recombinant IgA proteinase to inhibition by human ␣ 2 -macroglobulin and various synthetic compounds including specific inhibitors of serine peptidases (phenylmethylsulfonyl fluoride, 3,4-dichloroisocoumarin, and Pefabloc), cysteine peptidases (E-64 and iodoacetamide), and metallopeptidases (EDTA, 1,10-phenanthroline, phosphoramidon, 2-(N-hydroxycarboxamido)4-methyl pentanoyl-L-Ala-Gly-NH 2 (Zincov), N-benzyloxycarbonyl-Pro-Leu-Glu-hydroxamate, and p-aminobenzoyl-Gly-Pro-D-Leu-D-Ala-hydroxamic). Pefabloc was purchased from Roche Molecular Biochemicals, Zincov was from Merck, 1,7-phenanthroline was from GFS Chemicals (Powell, OH), and the other compounds were from Sigma. The ␣ 2 -macroglobulin was titrated on trypsin prior to use and found to be 60% active, and incubation time before adding IgA1 substrate was up to 5 h.
For quantitative purposes, IgA proteinase activity was titrated using a previously described assay involving enzyme-linked immunosorbent assay technology (16). Briefly, serial dilutions of test sample were incubated with an equal volume of myeloma IgA1 substrate (50 g ml Ϫ1 ), after which reaction mixtures were incubated in enzyme-linked immunosorbent assay wells precoated with antibody specific for Fc␣. Subsequently, wells were incubated with enzyme-conjugated antibody to immunoglobulin light chains and developed with chromogenic substrate. In this assay, cleavage of IgA1 was reflected as a decrease in OD signal relative to the signal measured for wells receiving IgA1 incubated without protease. Based on a regression curve fitted to a plot of OD against dilution, the IgA1 proteinase titer was calculated as the sample dilution corresponding to a 50% decrease in OD.
To examine the effect of reducing agents on enzyme activity, purified IgA proteinase was incubated with dithiothreitol (1 mM) or ␤-mercaptoethanol (1 mM) for 1 h at room temperature, and the activities of such treated proteinases were titrated along with an untreated control sample of the enzyme. To prevent interference of reducing agents with the enzyme-linked immunosorbent assay, reaction mixtures were diluted 1:15 with phosphate-buffered saline prior to analysis.
To evaluate potential proteolytic activity of the IgA proteinase against other substrates, several proteins at 1 mg ml Ϫ1 were incubated in a volume of 50 l at 37°C for 24 h with an amount of the partially purified recombinant proteinase capable of cleaving completely 0.5 mg of human IgA1 within 2 h. The proteins used were human IgG, IgD, IgE, IgM, ␣ 2 -macroglobulin, ␣ 1 -proteinase, ␣ 1 -antichymotrypsin (all from Athens Research and Technology, Athens, GA) and fibrinogen, bovine albumin, carboxymethylated lysozyme, collagen type I and IV, oxidized insulin ␣-chain, and gelatin from Sigma. After incubation, the reaction was stopped by boiling in reducing sample buffer, and the integrity of the proteins was analyzed by SDS-PAGE.
Amino Acid Sequencing-The most active fractions eluted from the Mono-Q column were subjected to reducing SDS-PAGE, blotted onto ProBlott membranes (PerkinElmer Life Sciences), and stained with Coomassie Blue. The band corresponding to the presumed IgA proteinase was excised, and the N-terminal sequence was determined using an ABI 477A/120A protein sequencer (PerkinElmer Life Sciences). For generation of tryptic peptides, the band in the polyacrylamide gel stained with Coomassie Blue was excised from several lanes and digested in situ as described (17). In brief, after washing with a mixture of ammonium bicarbonate and acetonitrile, pieces of the gel were shrunk with acetonitrile and dried completely. A solution containing modified trypsin (Promega, Madison, WI) was allowed to soak into the gel pieces. After incubation overnight at 37°C, generated peptides were recovered by extraction and separated by narrow bore RP-HPLC using the SMART system (Amersham Biosciences). Amino acid sequencing of selected peptides was performed as described above.
Southern Blot Analysis-Unless otherwise stated, the DNA manipulations were performed according to Sambrook et al. (14). Whole-cell DNA from C. ramosum was isolated, digested with EcoRI, and subjected to agarose gel electrophoresis and Southern blot analysis including hybridization at high and low stringency as described previously (18). As probes we used a 5.1-kb fragment of the S. sanguis strain ATCC 10556 iga gene (18) and a PCR product of genomic DNA from C. ramosum strain AK183 amplified with the primers 5Ј-AACGTGTTT-TCGGGCAGATGA-3Ј and 5Ј-TGATAGTCTTGCATCGCTTTC-3Ј, identified in the present study, using the Expand Long Template PCR System as recommended by the supplier (Roche Molecular Biochemicals). The DNA probes were purified after agarose gel electrophoresis (QIAEX II Gel Extraction Kit, Qiagen, Valencia, CA) and labeled with [ 32 P]dCTP (Random Labeling Kit, Roche Molecular Biochemicals).
Sequencing the IgA Proteinase Gene-Several degenerate primers with sequences corresponding to reverse translation of the obtained amino acid sequences of the peptides analyzed were purchased from DNA Technology (Aarhus, Denmark). The primers were combined in pairs of a forward and a reverse one and used in PCRs containing 100 ng of genomic DNA and 30 pmol of each primer using Ready To Go PCR beads (Amersham Pharmacia Biotech) and subjected to the following cycling parameters for 30 cycles: 94°C for 1 min, 55°C for 1 min, 72°C for 2 min, with an initial denaturation step at 94°C for 5 min and a final extension at 72°C for 8 min. The resulting PCR products were cloned into the E. coli plasmid vector pCR-TOPO using the TOPO-TA Cloning KIT (Invitrogen). For inverse PCR, the genomic DNA was digested with either MspI or EcoRI, and 50 ng of the resulting fragments was circularized in a 20-l reaction volume (Rapid DNA Ligation Kit; Roche Molecular Biochemicals). The self-ligated mixture was purified with Wizard Minicolumns (Promega) and used as template in the inverse PCR with primers pointing outwards using either Ready To Go PCR beads or the Expand High Fidelity PCR System (Roche Molecular Biochemicals) and the same cycling parameters as described above except that the annealing temperature was 60°C. The inverse PCR products were cloned into E. coli plasmids pCR-TOPO or pUC19. The ExoIII/S1 Deletion Kit (MBI, Fermentas, Lithuania) was used for construction of plasmid clones with nested, unidirectional deletions for sequencing the insert of recombinant plasmids using the universal M13 primers.
Plasmid DNAs for sequencing were prepared as recommended by the supplier of the sequencing kit, and PCR products were purified with Wizard Minicolumns (Promega). Individual sequence reactions on plas-mid DNA were performed with the Taq DyeDeoxy-Terminator cycle sequencing Kit (PerkinElmer Life Sciences), whereas Thermo Sequenase Dye Terminator Cycle Sequencing Kit (Amersham Biosciences) was used for sequencing PCR products. Sequence reactions were analyzed with an ABI PRISM 377 DNA sequencer (PerkinElmer Life Sciences). As sequencing primers, we used the universal M13 sequencing primers as well as oligonucleotides designed on the basis of preceding sequences. The DNA sequence was determined for both strands of the iga gene. Computer analysis of the sequences was performed with programs included in the GCG package (Genetics Computer Group, Madison, WI). BLAST and PSI-BLAST at NCBI (available on the World Wide Web at www.ncbi.nlm.nih.gov/BLAST/) were used for data base searching.
Expression of the ORF 1 -For expression in E. coli of the ORF, we used the vector pGEX-5T, which is designed to express a recombinant fusion protein consisting of a histidine hexapeptide and glutathione S-transferase followed by the amino acid sequence of interest (19). The primers 5Ј-CCGATGACCATTGGATCCGCATCAAAGC-3Ј and 5Ј-GCT-TAAAGGTCTATTCTCGAGTTATTCAGCG-3Ј were used in a PCR on genomic DNA from strain AK183 to amplify a fragment encoding the presumed secreted form of the IgA proteinase, and, in addition, the primers add a BamHI and an XhoI restriction site, respectively. For the PCR, we used the Pwo polymerase as recommended by the supplier (Roche Molecular Biochemicals). The vector and the PCR product were digested with BamHI and XhoI, ligated, and transformed into E. coli JM109. Colonies harboring the correct recombinant plasmid, termed pGEX-5T-iga, were identified by restriction analysis of plasmid DNA. The plasmid DNA was subsequently used to transform E. coli BL21(DE3)pLysS, and expression of the recombinant protein was induced in a culture with an A 600 of 0.6 by adding IPTG to a final concentration of 1 mM. After an additional 4 h of growth, the cells were pelleted, resuspended in phosphate-buffered saline, pH 7.3, and disrupted by mild sonication. The supernatant of the lysate was tested for IgA1 cleaving activity as described above.
The recombinant IgA proteinase was partially purified using affinity chromatography on glutathione-Sepharose (Amersham Pharmacia Biotech). Briefly, the E. coli cells from 3 liters of culture grown for 4 h at 25°C after induction of expression by IPTG were suspended in 50 ml of phosphate-buffered saline and disrupted using a French press. Cell debris was removed by ultracentrifugation (105,000 ϫ g for 60 min), and recombinant IgA proteinase was separated using an on-column cleavage and purification procedure as recommended by the manufacturer. Thrombin was removed using benzamidine-Sepharose (Amersham Biosciences).
Site-directed Mutagenesis-Three mutants of the C. ramosum IgA proteinase, H539A, D550A, and E551A, were generated. Mutator oligonucleotide primers were designed to introduce restriction enzyme sites to facilitate subsequent screening for mutated DNA products.
For the H539A mutation, the codon CAC for histidine was replaced by GCC, a codon for alanine. In addition, a silent mutation in codon Ala 538 was introduced by changing GCA into GCG, creating an Eco52I restriction site, CGGCCG. Outward facing primers for the H539A mutagenesis were 5Ј-AACCGTGTCCAAACTCGGCCGCAAAAGTTT-3Ј and 5Ј-TGCTCGGTCTCGGTGATGAATA-3Ј. For the D550A mutation, the codon GAT for aspartic acid was replaced by GCT, a codon for alanine. In addition, a silent mutation in codon Gly 549 was introduced by changing GGT into GGG, creating an Eco88I restriction site, CTC-GGG. Outward facing primers for the D550A mutagenesis were 5Ј-AT-CCGTTACTGTATTCAGCCCCGAGACCGA-3Ј and 5Ј-ATTTGCTTGA-CGATAAGGAACTTAA-3Ј. For the E551A mutation, the codon GAA for glutamic acid was replaced by GCA, a codon for alanine. This change created an Mph1103I restriction site, ATGCAT. Outward facing primers for the E551A mutagenesis were 5Ј-CAAATATCCGTTACTGTATG-CATCACCGAG-3Ј and 5Ј-CTTGACGATAAGGAACTTAAATCAC-3Ј. A 1.2-kb SalI-HindIII fragment of plasmid pGEX-5T-iga was cloned into HindIII-and SalI-digested pTZ19R, generating pSH1200. Desired mutations and restriction sites were introduced into pSH1200 by inverse PCR using Pfu polymerase as recommended by the supplier (Fermentas) and the primer pairs described above. Following temperature cycling, the PCR products were treated with DpnI restriction endonuclease to digest parental DNA template methylated by the Dam methylase. After heat inactivation of DpnI, the DNAs were used to transform E. coli JM109 supercompetent cells. The mutants were selected by restriction analysis of plasmid DNA and confirmed by se-quence analysis. The 1.2-kb SalI-HindIII fragments with desired mutations were cloned into SalI-and HindIII-restricted pGEX-5T-iga. The resulting plasmids, termed pGEX-iga-H539A, pGEX-iga-D550A, and pGEX-iga-E551A, had single amino acid substitutions at amino acids His 539 to Ala, Asp 550 to Ala, and Glu 551 to Ala, respectively. The plasmids were transformed into E. coli BL21(DE3)pLysS, and expression of recombinant proteins was induced as described above.

Purification, Characterization, and Amino Acid Sequence
Analysis of the C. ramosum IgA Proteinase-Differences in the proportion of secreted compared with cell-associated forms of the IgA1 proteinase produced have been observed for different species and strains of bacteria (20). Here we found that in the early stationary phase the majority of the IgA1 cleaving activity in C. ramosum strain AK183 was secreted into the medium (data not shown). The C. ramosum IgA proteinase was purified from culture supernatant by a combination of ammonium sulfate precipitation, size exclusion, and anion exchange chromatography. Eluent fractions containing the proteinase were identified by their ability to cleave human IgA1, releasing intact Fc and Fd fragments (analyzed by SDS-PAGE), and the activity in fractions was determined by titration of the ability to cleave human IgA1 (analyzed by the enzyme-linked immunosorbent assay-based assay). The titer of IgA proteinase activity in the initial 5 liters of culture supernatant was 8, and in the peak activity fraction (0.5 ml) upon anion exchange it was 128. This modest increase in activity suggested a loss of enzyme activity during the process of purification. Because C. ramosum is a strictly anaerobic bacterium, we speculated that the enzyme might regain its activity if subjected to reducing conditions. However, we found that preincubation with neither 1 mM ␤-mercaptoethanol nor 1 mM dithiothreitol had any influence on its capacity to cleave human IgA1. The loss of activity during purification remains unexplained.
The IgA proteinase was active at neutral pH, and it retained activity upon storage at Ϫ20°C for several weeks. It has been previously shown that 100 mM EDTA inhibits the activity of the C. ramosum IgA proteinase (11). We found that the IgA1 cleaving activity was completely inhibited by 0.5 mM EDTA, suggesting that the enzyme is a metalloproteinase. More detailed enzymatic characterization of the IgA proteinase activity was performed using the partially purified recombinant form of the enzyme (see below).
Reducing SDS-PAGE analysis of the Mono-Q fractions revealed that the intensity of a band corresponding to a protein of 130 kDa correlated with the IgA1 proteinase activity, suggesting that this band represented the IgA proteinase (Fig. 1). The 130-kDa protein as well as tryptic peptides derived from it and purified by HPLC were subjected to N-terminal amino acid sequence analysis. The N-terminal sequence of the 130-kDa protein was determined to be AXKPDIKVXDYVKMGVYNN, while the N-terminal sequences EYGFHYFISPSD, FEDGX- 1 The abbreviations used are: ORF, open reading frame; IPTG, isopropyl-1-thio-␤-D-galactopyranoside. EIPNTAGG, and EYTGAY were obtained for three of the tryptic peptides. None of the sequences shared significant similarity to other bacterial IgA1 proteinases or to any known proteins, as revealed by searching the GenBank TM data base.
The iga Gene Sequence from C. ramosum Strain AK183-Although the molecular mass and catalytic mechanism of the C. ramosum IgA proteinase was similar to that of the IgA1 proteinase from streptococcal species (18,(21)(22)(23), the C. ramosum iga gene encoding the IgA proteinase showed no homology to the streptococcal iga genes. Even when using hybridization at very low stringency conditions, genomic DNA from C. ramosum strain AK183 did not hybridize with the iga gene from S. sanguis in a Southern blot analysis (data not shown). To isolate the C. ramosum iga gene, the N-terminal amino acid sequences obtained for the putative IgA proteinase and the tryptic fragments of it were used to design degenerate primers for PCR amplification of a part of the C. ramosum iga gene using genomic DNA from strain AK183 as template. Forward primer 5Ј-ATGGGIGTITAYAAYAAY-3Ј was deduced from reverse translation of the amino acid sequence MGVYNN from the N-terminal sequence of the mature protein, and reverse primer 5Ј-RAARTARTGRAAICCRTAYTC-3Ј was deduced from the sequence EYGFHYF obtained for one of the tryptic peptides. A single amplicon of ϳ1.2 kb was produced. The nucleotide sequence of this fragment was determined and used for design of primers for inverse PCR to obtain the complete iga gene sequence. Combined, a sequence of 4242 nucleotides was determined. To correct for errors that may occur due to imperfect fidelity of the DNA polymerases in the PCRs and which would be carried over in the cloning procedures applied in the sequencing strategy (see "Experimental Procedures"), the sequence obtained was used to design primers for PCR amplification of overlapping fragments of the AK183 iga gene. Direct sequencing of the amplicons revealed a total of five errors in the initial sequence.
The sequence contained a large ORF with the potential of encoding a protein of 1,234 amino acids. The N-terminal sequence of the 130-kDa protein as well as the N-terminal sequences of the tryptic fragments of it were all identified within the primary structure deduced from the ORF (Fig. 2). The ORF was preceded by typical promoter elements (Fig. 2). The sequence GGAAGT, six nucleotides upstream of the proposed ATG start codon, is similar to the Shine and Dalgarno sequence GGAGGT (24) and is in a suitable position for a ribosome binding site (25). Thirty-five nucleotides upstream of the proposed ATG start codon, the sequences TATAATA and TTGAC separated by 17 nucleotides match the Ϫ10 and Ϫ35 promoter elements, respectively (26,27). Another possible ATG start codon was located 15 nucleotides downstream of the first one. A possible transcriptional terminator in the form of an inverted repeat structure was identified downstream of the ORF (Fig. 2).
A Southern blot analysis of genomic DNA of C. ramosum AK183 restricted with PstI, which has no recognition sites in the iga gene sequence determined, and hybridized with a 4-kb fragment containing the ORF showed a single band of 14 kb (results not shown), suggesting that the iga gene is a single copy gene in C. ramosum strain AK183.
Interestingly, the GC percentage of the iga gene was 43 compared with an overall GC percentage of 26 in the C. ramosum genome (28). This difference strongly suggests that the IgA proteinase gene in C. ramosum was acquired recently in evolution through horizontal gene transfer from another bacterium with a higher GC percentage.
Expression of the IgA Proteinase in E. coli and Characterization of the Recombinant Protein-To verify that the ORF in fact represented the C. ramosum iga gene, we performed heterologous expression in E. coli. The sequence encoding the presumed mature IgA proteinase (positions 537-4151 in Fig. 2) was amplified by PCR and cloned into the E. coli expression vector pGEX-5T. This vector is designed to express a recombinant fusion protein consisting of a histidine hexapeptide and glutathione S-transferase followed by the amino acid sequence of interest. The plasmid construct, termed pGEX-5T-iga, was transformed into E. coli BL21(DE3)pLysS. Intracellular expression of the fusion protein was induced by IPTG, and after incubation the cells were disrupted by sonication. The resulting lysate showed IgA proteinase activity (Fig. 3), demonstrating that the ORF sequenced was the iga gene. In addition, Nterminal sequencing of the Fc fragment generated by the recombinant fusion protein revealed the sequence VPSTP. This sequence is identical to that previously reported for Fc induced by the C. ramosum IgA proteinase (11), indicating that the specificity of the recombinant proteinase was identical to the native one.
Relatively high expression of the active recombinant form of the IgA proteinase enabled us to perform a more detailed characterization of the enzyme activity. First, we confirmed that the IgA proteinase is highly specific for human IgA, since none of the other human immunoglobulins, including IgG, IgD, IgE, and IgM, were susceptible to cleavage. In addition, none of the other proteins tested (fibrinogen, albumin, collagen type I and IV, and two serpins, ␣ 1 -proteinase and ␣ 1 -antichymotrypsin) were cleaved by the IgA proteinase even after incubation for 24 h at an enzyme concentration sufficient to cleave 0.5 mg of human IgA1 in 2 h (data not shown). Especially significant was the lack of an effect on serpins, since this group of proteins possess a surface-exposed loop (the reactive site loop), which is readily cleaved even by nontarget proteinases with a restricted specificity like periodontain (29) and collagenase (30). These data, together with the lack of activity against unstructured polypeptides such as gelatin, carboxymethylated lysozyme (Fig. 4), and oxidized insulin ␣-chain, indicate an exquisite specificity of the C. ramosum IgA proteinase. Such a narrow specificity limited to the hinge region of the IgA molecule is also a common feature of other IgA1 proteinases (4,5).
It has been assumed that the C. ramosum IgA proteinase is a metalloproteinase solely on the basis of its inhibition by EDTA (11). To extend enzyme characterization, we investigated the effect of a broad range of class-selective inhibitors of serine proteinases, cysteine proteinases, and metalloproteinases on the activity of the partially purified recombinant IgA proteinase. At 1 mM concentration, none of the compounds specific for the first two groups of peptidases, including phenylmethylsulfonyl fluoride, 3,4-dichloroisocoumarin, Pefabloc, E-64, and iodoacetamide, had any effect on IgA1 cleavage (results not shown). Also, among several metalloproteinase inhibitors tested, only metal-chelating compounds, 1,10-phenanthroline and EDTA, inhibited the enzyme activity (Fig. 5, lanes  3 and 4). Significantly, a nonchelating isomer of phenanthroline (1,7-phenanthroline) had no effect (Fig. 5, lane 8). Neither phosphoramidon nor Zincov, a compound specifically designed to inhibit metalloproteinases, had any effect on IgA1 cleavage (Fig. 5, lanes 5 and 6). The IgA proteinase activity was also insensitive to inhibition by other hydroxamate-based compounds such as N-benzyloxycarbonyl-Pro-Leu-Glu-hydroxamate (Fig. 5, lane 7) and p-aminobenzoyl-Gly-Pro-D-Leu-D-Alahydroxamic acid as well as ␣ 2 -macroglobulin (results not shown). Taken together, the inhibition profile of the C. ramosum IgA metalloproteinase reiterates the unique character of the enzyme.
The Amino Acid Sequence of the C. ramosum IgA Protein- ase-The deduced amino acid sequence of this novel proteinase, when compared with the N-terminal sequence determined for the secreted protein, indicates that the first 30 amino acids of the primary translation product comprise the signal peptide. This is in perfect agreement with the predictions made on the basis of the primary structure inferred from the iga gene sequence by the computer program SignalP (31). Taking the signal peptide into account, the deduced mature IgA proteinase contains 1,204 amino acids, has a calculated M r of 133,828, and has an isoelectric point of 5.79. This is in agreement with the size of the purified proteinase observed in SDS-PAGE (130 kDa).
In the C terminus, we identified a putative cell wall sorting signal that in other Gram-positive bacteria has been found to target surface proteins to the cell wall (32). The sequence SPQTG at positions 1196 -1200 presumably constitutes the sortase recognition site. It was previously reported that in Clostridium difficile the sortase appears to recognize SPXTG or PPXTG instead of the conventional LPXTG motif (33). In the C. ramosum IgA proteinase, a small spacer, DNSN, separated this motif from a transmembrane domain, IFLWFALLFVSAAGVT-GITAY, followed by a positively charged tail, NKKKKEHAE, at the C terminus. These features are in agreement with other presumed substrates for sortase-like proteins (33). Provided that the anchor motif is functional, the sortase cleaves at the Thr-Glu peptide bond in the recognition site and covalently links the threonine, and thereby the N-terminal part of the protein, to peptidoglycan in the cell wall (32,34). However, we found that the majority of IgA proteinase activity in C. ramosum AK183 was released into the medium. Release of surface proteins with a typical Gram-positive cell wall anchor motif has been reported for the ␣ and ␤ antigens present in the c protein complex of Streptococcus agalactiae (35)(36)(37), and Streptococcus mutans sheds surface antigen P1 and secretes exo-␤-D-fructosidase (38,39). The release of anchored surface proteins may be brought about by turnover of the peptidoglycan layer or by proteolytic cleavage of the proteins next to the anchoring. Provided that the cell wall anchor sorting signal in the C. ramosum IgA proteinase is functional, the mechanism by which the proteinase is released from the cell wall remains to be elucidated.
A putative zinc-binding motif was identified at positions 539 -543 followed by an aspartic acid residue seven positions downstream in the sequence HEXXHXXXGXXD and resembling the extended zinc-binding site typical for the metzincin group of metallopeptidases, but as a significant difference in the IgA proteinase there are four instead of three residues between the second His and Gly (40,41). In all members of this clan with the exception of leishmanolysins, the third zinc ligand is His or Asp, located invariably six residues downstream of the HEXXH motif (42). In case of C. ramosum IgA proteinase, there are seven residues separating the second (His) and the third (Asp) zinc ligand. Nevertheless, the sequence encompassing the zinc-binding motif is remarkably similar to that of the PrtV proteinase of V. cholerae and the immune inhibitor A of B. thuringiensis (Fig. 6), each of which is a proteolytic member of clan MA. This significant similarity includes the presence of the conserved Gly, which allows the formation of the ␤-turn necessary to bring the zinc ligands together in this group of metalloproteinases (43,44). Therefore, it can be predicted that His 539 , His 543 , and Asp 550 of the C. ramosum IgA proteinase polypeptide chain form the metal binding site, while Glu 540 is the active site residue.
To experimentally verify the prediction that His 539 and Asp 550 constitute part of the zinc binding motif and are therefore indispensable for the enzyme activity, we constructed and expressed mutant forms of the IgA proteinase in which these residues were individually replaced by alanine. As expected, neither of these two mutants possessed IgA1 cleaving activity (Fig. 7, lanes 4 and 5). Notably, however, the E551A mutant was fully active (Fig. 7, lane 6). These data corroborate the alignment-based predictions of the zinc-binding and catalytic residues (Fig. 6) and indicate that the IgA proteinase of C. ramosum can be included into clan MA. The endopeptidases from clan MA are also known as metzincins, because there is a conserved Met in a turn that underlines the active site (41,45). However, this Met is absent in the IgA proteinase, indicating the uniqueness of this enzyme, which most likely establishes a new subfamily of metallopeptidases in family M6 of clan MA.
Currently, the family M6 consists of only three members listed in the MEROPS Data Base (available on the World Wide Web at www.merops.co.uk) exemplified by PrtV proteinase of V. cholerae and immune inhibitor A of B. thuringiensis. A PSI-BLAST search in the unfinished genomes, using these proteinase sequences for comparison, revealed, however, that similar putative proteinases are encoded in the genomes of Bacillus stearothermophilus, Streptomyces coelicolor, Clostridium acetobutylicum, Shewanella putrefaciens, and Bacillus anthracis (Fig. 6). The latter species can potentially express at least four different metalloproteinases that are homologous, with one being almost identical to immune inhibitor A (94% identity in the primary structure), the only member of the M6 family with defined biological function. This metalloendopeptidase cleaves the antibacterial proteins attacin and cecropin found in insects, disabling the immune system of lepidoptera infected by B. thuringiensis (46,47). In this respect, it is interesting to note that the IgA proteinase has an analogous function, since its activity is also specifically aimed at the host antimicrobial defense mechanisms.
The importance of IgA1 cleavage by mucosal pathogens or commensals in their ability to escape immune defense seems apparent but is difficult to establish due to the lack of a relevant animal model (5). The biological significance of IgA1 proteinase activity can, however, be inferred indirectly from the fact that nature developed these specific proteinases based on three different catalytic mechanisms. Moreover, it is now apparent that within the metalloproteinase class, the specificity to cleave human IgA is present in two evolutionary lineages, with the C. ramosum enzyme capable of cleaving both IgA1 and IgA2m(1) molecules. This seems to be a major advantage for this bacterium, because in the gut environment, a natural habitat of C. ramosum, both isotypes of IgA occur in comparable amounts. It remains to be examined whether strains of C. ramosum producing this proteinase preferentially colonize subjects homozygous for the IgA2m(1) allotype. Besides, the expression of the recombinant IgA proteinase facilitates production of a large amount of the active, recombinant protein in a pure form for further studies of this intriguing molecule.
FIG. 6. Conserved region in C. ramosum IgA proteinase. A comparison is shown of the region around the predicted active site and zinc-binding domains (indicated by A and Z, respectively) of C. ramosum IgA proteinase (IgAPrt), B. thuringiensis immune inhibitor A (InA; accession number X55436), and V. cholerae PrtV (PrtV; accession number Y00557), with a hypothetical secreted proteinase of S. coelicolor (ScPprt; accession number CAB51001) and those of PrtV-related proteinases obtained from the conceptual translation of sequences retrieved from genome data bases (available on the World Wide Web at www.ncbi.nlm.nih.gov/Microb_blast/unfinishedgenome.html): BaPrt1 from B. anthracis, BaPrt2 from B. anthracis, BstePrt from B. stearothermophilus, ClaPrt from C. acetobutylicum, and SputPrt from S. putrefaciens. The sequences were aligned using the ClustalW multiple sequence alignment tool. The arrows above the sequences indicate Gly and Met residues conserved in the metzincin family of metallopeptidases. The asterisks indicate identical residues, and dots indicate conserved residues with similar properties in members of the M6 family of metzincins. Gaps (dashes) have been introduced to optimize alignment. The numbers of the first and last amino acid in the alignment are indicated for each protein. In this assay compared with Figs. 3 and 5 we used a different preparation of human IgA1 with a distinct glycosylation, and therefore the fragments migrate slightly differently in the gel. Below the SDS-PAGE gel, a Western blot analysis of appropriate lysates probed with anti-His 6 antibodies demonstrates that the lack of IgA1 cleaving activity was not due to deficiency in expression of the mutated proteinase. The position of the recombinant proteinase is indicated by an arrow.