The Hemophilus influenzae Hap Autotransporter Is a Chymotrypsin Clan Serine Protease and Undergoes Autoproteolysis via an Intermolecular Mechanism*

The Hemophilus influenzae Hap adhesin is an autotransporter protein that undergoes an autoproteolytic cleavage event resulting in extracellular release of the adhesin domain (Haps) from the membrane-associated translocator domain (Hapβ). Hap autoproteolysis is mediated by Ser243 and occurs at LN1036-7 and to a lesser extent at more COOH-terminal alternate sites. In the present study, we sought to further define the mechanism of Hap autoproteolysis. Site-directed mutagenesis of residues His98 and Asp140identified a catalytic triad conserved among a subfamily of autotransporters and reminiscent of the SA (chymotrypsin) clan of serine proteases. Amino-terminal amino acid sequencing of histidine-tagged Hapβ species and site-directed mutagenesis established that autoproteolysis occurs at LT1046-7, FA1077-8, and FS1067-8, revealing a consensus target sequence for cleavage that consists of ((Q/R)(A/S)X(L/F)) at the P4 through P1 positions. Examination of a recombinant strain co-expressing a Hap derivative lacking all cleavage sites (HapΔ1036-99) and a Hap derivative lacking proteolytic activity (HapS243A) demonstrated that autoproteolysis occurs by an intermolecular mechanism. Kinetic analysis of Hap autoproteolysis in bacteria expressing Hap under control of an inducible promoter demonstrated that autoproteolysis increases as the density of Hap precursor in the outer membrane increases, confirming intermolecular cleavage and suggesting a novel mechanism for regulation of bacterial adherence and microcolony formation.

Bacterial organisms have evolved a number of pathways for presenting proteins on the cell surface and releasing proteins extracellularly. In pathogenic organisms, these pathways are essential for specific interactions that result in colonization of the host and progression to disease. The bacterial proteins that participate in these interactions are considered virulence fac-tors and influence adherence to host cells, damage to host tissues, and interference with host defense mechanisms.
In Gram-negative bacteria, surface and extracellular proteins must be transported across the inner membrane, the periplasm, and the outer membrane. This goal is achieved by one of five distinct secretion systems. The type V secretion system is used exclusively by the autotransporter family of proteins, typified by the IgA1 proteases of Neisseria and Hemophilus species, the first such proteins to be identified (1). Autotransporters are typically expressed as precursor polypeptides with at least three functional domains, including an NH 2terminal signal sequence, an internal passenger domain, and a COOH-terminal translocator domain. The signal sequence directs export of the polypeptide across the bacterial inner membrane and is then removed by signal peptidase. Subsequently, the translocator domain inserts into the outer membrane and appears to fold into a ␤-barrel structure with a central hydrophilic pore, allowing extrusion of the passenger domain across the membrane. Once the passenger domain is localized on the cell surface, one of several fates is possible. In some cases, the passenger domain remains covalently linked to the membraneassociated ␤-barrel domain. In other cases, it is cleaved but remains cell associated. In still others, it is cleaved and then released extracellularly. Cleavage may involve an autoproteolytic event directed by protease activity in the passenger domain itself or may occur through the action of a separate bacterial protease. Autotransporter passenger domains have been ascribed a wide variety of functions, examples of which include adhesins, toxins, degradative enzymes, and serum resistance factors (2).
Hemophilus influenzae is a Gram-negative bacterium and represents a common cause of human disease, including both localized respiratory tract and systemic (invasive) disease (3). The pathogenesis of H. influenzae disease begins with colonization of the upper respiratory tract. To facilitate colonization, H. influenzae elaborates both pilus and non-pilus proteins called adhesins, which promote adherence to host epithelial cells by interacting with specific host cell surface molecules. The H. influenzae Hap protein is a non-pilus adhesin that promotes adherence and invasion in assays with cultured human epithelial cells and also mediates bacterial aggregation and microcolony formation (4,5). In recent work, we demonstrated that Hap is an autotransporter protein and consists of a 110-kDa passenger domain called Hap s and a 45-kDa translocator domain called Hap ␤ . The passenger domain has serine protease activity and directs autoproteolytic cleavage via Ser 243 , releasing Hap s into the culture supernatant and leaving Hap ␤ embedded in the outer membrane (6). Of note, Ser 243 is part of a GDSGS motif that is present in a subfamily of auto-transporter proteins, including the Neisseria and Hemophilus IgA1 proteases and a diverse group of serine protease autotransporters secreted by members of the Enterobacteriaceae (SPATEs) (7)(8)(9)(10)(11)(12)(13). Hap autoproteolytic cleavage occurs primarily at the peptide bond between Leu 1036 and Asn 1037 , and mutation of these two residues results in nearly complete inhibition of cleavage at this site and increased cleavage at three more COOH-terminal alternate sites (6). Autoproteolysis is blocked by several serine protease-specific inhibitors, including phenylmethylsulfonyl fluoride, Pefabloc, and secretory leukocyte protease inhibitor, a component of human respiratory tract secretions.
In addition to harboring serine protease activity, the Hap s passenger domain possesses the adhesive activities responsible for bacterial interaction with epithelial cells and bacterial aggregation. Thus, at first glance release of this domain from the bacterial cell surface seems counterproductive to promoting successful colonization. Indeed, mutation of the Hap active site serine to an alanine results in complete inhibition of autoproteolytic processing, full retention of Hap s on the cell surface, and increased adherence and aggregation (6). On the other hand, release of Hap s may benefit H. influenzae in ways unrelated to its adhesive activities. For example, activity against host substrates has been demonstrated for IgA1 protease as well as several of the SPATE family members (2,7,14,15). Although no such host substrate has been identified yet for Hap, it is possible that Hap proteolytic activity against epithelial tissue components or immune system effectors may facilitate initial colonization. Alternatively, autoproteolytic cleavage of Hap at several sites on the same molecule may release bioactive peptides, as has been documented for IgA1 protease (16). Finally, shedding of an adhesive domain from the cell surface may facilitate disbursement of individual organisms from microcolonies and allow spread to new sites. Such events may be controlled by the presence of host factors that modulate Hap autoproteolytic activity (e.g. secretory leukocyte protease inhibitor) or by features inherent to the mechanism of autoproteolysis itself.
In the present study, we sought to characterize Hap autoproteolysis more fully, including the mechanism of catalysis, the identities of all cleavage sites, and the molecular nature of the enzyme-substrate interaction. Our results demonstrate that Hap belongs to the SA (chymotrypsin) clan and contains a catalytic triad conserved among a subfamily of autotransporter serine proteases. Identification of the three alternate autoproteolytic cleavage sites in Hap revealed a consensus target sequence for Hap enzymatic activity. Interestingly, kinetic analysis and examination of a recombinant strain expressing two different derivatives of Hap established that autoproteolysis occurs at least in part via an intermolecular mechanism. These results provide insights into an expanding family of bacterial serine proteases and suggest a novel mechanism for regulation of bacterial adherence.

EXPERIMENTAL PROCEDURES
Protein Sequence Alignments-Amino acid alignments of protein sequences were performed using ClustalW v1.8 software available on the Baylor College of Medicine Human Genome Sequencing Center website (dot.imgen.bcm.tmc.edu).
Recombinant DNA Methods-DNA ligations, restriction endonuclease digestions, and gel electrophoresis were performed according to standard techniques (22). Plasmids were introduced into E. coli strain DH5␣ by chemical transformation (22). H. influenzae strain DB117 was transformed using the MIV method of Herriott et al. (23).
Construction of Plasmids Encoding Mutated Hap Derivatives-Sitedirected mutagenesis was performed using recombinant PCR 1 techniques. To construct pHapH98A, pHapH117A, pHapD139A, and pHapD140A, pMLD100 (pUC19 containing a 6.7-kb insert with the hap gene from H. influenzae strain N187) was used as a template to amplify a 1.4-kb 5Ј PCR fragment and a 0.9-kb 3Ј PCR fragment that overlapped at an 18-base region encompassing the appropriate mutated site. Each 5Ј fragment was then combined with the corresponding 3Ј fragment in a 1:1 molar ratio to serve as template in generating a 2.3-kb mutated recombinant PCR fragment using external primers from the initial PCR reactions plus a 1:100 dilution of an internal primer corresponding to the coding sequence of the mutated site. These recombinant PCR fragments were digested with AvaI and ClaI and then ligated into AvaI-ClaI-digested pMLD100.
For construction of the primary autoproteolytic cleavage site mutant pHapL1036S, initial PCR products (1.1 kb 5Ј fragment and 1.2 kb 3Ј fragment) were amplified from pMLD100 and used as templates to generate a 2.3-kb mutated recombinant PCR fragment as described above. This recombinant PCR fragment was then digested with NheI and HindIII and ligated into NheI-HindIII-digested pMLD100. The double autoproteolytic cleavage site mutants pHapL1036S::L1046S, pHapL1036S::F1067S, and pHapL1036S::F1077S were constructed in the same manner using pMLD100::HapL1036S as the template in the initial PCR reactions.
To construct plasmid pHapLN1036-7SG::G3H6, a 1.8-kb PCR fragment corresponding to the COOH-terminal 600 residues of Hap was amplified from pDH101 (like pMLD100, but with the hap gene insert in the opposite orientation), incorporating codons for 3 glycine residues, 6 histidine residues, and a BamHI site at the 3Ј end of the hap gene (5Ј end of the PCR product). This PCR fragment was digested with BamHI and NheI and ligated into BamHI-NheI-digested pDH101. The resulting plasmid, pDH101::G3H6, was digested with DsaI and NsiI, liberating a 6.3-kb fragment that was then replaced by ligation with a 6.3-kb insert excised from DsaI-NsiI-digested pDH101::LN1036-7SG (6).
Construction of pHap⌬1036-99 was carried out by first amplifying initial PCR fragments corresponding to residues 650 -1035 (1.1-kb 5Ј fragment) and 1100 -1395 (1.1-kb 3Ј fragment) of Hap. The primers used to create these PCR products were engineered such that the two fragments overlapped at an 18-base region, allowing amplification of a recombinant PCR product in which the codons for residues 1035 and 1100 were immediately juxtaposed, deleting the intervening sequence. This recombinant PCR fragment was digested with NheI and HindIII and ligated into NheI-HindIII-digested pMLD100.
Plasmid pJS106::Trc was constructed by first amplifying a 1.6-kb PCR fragment that contained the lacI q gene, trc promoter, and polylinker cloning region from pTrc99A. The primers used in this amplification were engineered such that the resulting PCR product contained HindIII and PstI sites at the 5Ј end and a HindIII site at the 3Ј end, allowing digestion of this fragment with HindIII and ligation into HindIII-digested pDH101. The resulting plasmid, pDH101::Trc, was then digested with NcoI and BglII, excising a fragment containing the pTrc99A polylinker cloning region, the hap upstream region, and 1.4 kb of the hap coding sequence. Next, a 1.4-kb PCR fragment corresponding to the excised hap coding sequence was amplified from pDH101, incorporating an NcoI site at the 5Ј end. This fragment was digested with NcoI and BglII and ligated into NcoI-BglII-digested pDH101::Trc, thereby juxtaposing the hap start codon downstream of the trc promoter and ribosome-binding site.
Following the introduction of mutated fragments into the complete hap gene, mutations were verified by nucleotide sequencing. Subse-quently, the mutated hap genes were excised as 6.7-kb PstI fragments and then ligated into PstI-digested pGJB103. Exceptions included the mutated hap gene encoding Hap⌬1036-99, which was ligated into PstI-digested pLS88, and the mutated hap gene encoding HapLN1036-7SG::G3H6, which was excised as a BamHI-PstI fragment and ligated into BglII-PstI-digested pGJB103. All constructs were first transformed into E. coli DH5␣ and then into H. influenzae strain DB117.
Analysis of Bacterial Culture Supernatant and Outer Membrane Fractions-DB117 expressing wild type Hap or mutant derivatives was grown to an optical density at 600 nm (A 600 ) of 0.8. Sarkosyl-insoluble outer membrane proteins were isolated by the method of Carlone et al. (24), and extracellular proteins were precipitated from culture supernatants with 10% trichloroacetic acid as described previously (4). Outer membrane fractions were resuspended in 25 l of 10 mM HEPES, pH 7.4, plus 25 l of 2 ϫ Laemmli buffer, while precipitated extracellular proteins were resuspended in 10 l of 1 M Tris, pH 9.0, plus 10 l of 2 ϫ Laemmli buffer. Protein samples were resolved by SDS-PAGE using 7.5% SDS-polyacrylamide gels (25). To ensure that comparable amounts of protein were analyzed, similar volumes from cultures of similar density were loaded into each lane. Resolved proteins were electrotransferred to a nitrocellulose membrane and then detected by immunoblot analysis using antiserum Rab730 diluted 1:500 (6). An anti-rabbit IgG antiserum conjugated to horseradish peroxidase (Sigma) was used as the secondary antibody, and detection of antibody binding was accomplished by incubation of the membrane in a chemiluminescent substrate solution (Pierce) and exposure to film.
Kinetic Analysis of Autoproteolysis during Induced Expression of Hap-DB117 expressing Hap under the control of a IPTG-inducible promoter (DB117/pJS106::Trc) was grown to an A 600 of 1.0, at which time IPTG was added to a final concentration of 0.1 mM. A 10-ml aliquot was removed from the culture every 10 min, beginning with the addition of IPTG, and proteins were isolated from the culture supernatants and outer membrane fractions as described above. In order to compensate for any differences in protein amounts caused by increases in culture density over the course of the experiment or by variations in purification efficiency, volumes loaded for resolution by SDS-PAGE were adjusted according to culture density for secreted proteins and according to final protein concentration (Bio-Rad protein assay) for outer membrane proteins. After resolution by SDS-PAGE, transfer to nitrocellulose, and detection by immunoblotting, protein band intensities were quantitated by scanning densitometry using an LKB Ultroscan XL laser densitometer (Bromo, Sweden). To further ensure that protein from an equivalent number of bacterial cells had been loaded in each lane, levels of major outer membrane protein P4 (OMP P4) in each of the outer membrane fractions were quantified by scanning densitometry after detection by an anti-P4 monoclonal antibody, EPR5-2.1 (generously provided by B. Green, Wyeth-Lederle Vaccines), and anti-mouse IgG secondary antiserum conjugated to horseradish peroxidase (Sigma).
Inactivation of ompP2, ompP5, and rec1 in Strain Rd-To inactivate the ompP2 gene in H. influenzae strain Rd, the cat cartridge from plasmid pUC⌬ECAT was inserted into a PvuII site contained within the ompP2 gene from type b strain DL42 on plasmid pEJH91-1-35 L3 (26). The inactivated ompP2 gene was excised by digestion with PstI, purified by agarose gel electrophoresis, and used to transform strain Rd. To inactivate the ompP5 gene in Rd/ompP2, a 1.5-kb PCR fragment encoding the 3Ј half of the Rd ompP5 gene plus 1 kb of downstream sequence was amplified from Rd chromosomal DNA, incorporating DraIII sites at each end to facilitate ligation of the fragment into DraIII-digested pUC4K (Amersham Pharmacia Biotech). The resulting 5.6-kb plasmid was then digested with PstI, excising a 2.8-kb fragment containing the kanamycin resistance cassette from pUC4K followed by the cloned portion of ompP2 and downstream sequence. This 2.8-kb fragment was then ligated into PstI-digested pUC19 (New England Biolabs), creating plasmid pUC19::kan R -P5-3Ј. A 1.5-kb PCR fragment encoding the 5Јhalf of the Rd ompP5 gene plus 1 kb of upstream sequence was amplified from Rd chromosomal DNA, incorporating an EcoRI site at the 5Ј end and an XbaI site at the 3Ј end to facilitate ligation of the fragment into EcoRI-XbaI-digested pUC19:: kan R -P5-3Ј. The resulting 5.9-kb plasmid, pUC19::P5-5Ј-kan R -P5-3Ј, was linearized by digestion with XmnI and used to transform Rd/ompP2. Finally, the rec1 gene of Rd/ ompP2/ompP5 was inactivated by transformation with DNA from strain BC200/rec1, which possesses a mutant rec1 gene linked to streptomycin resistance (27). Streptomycin-resistant transformants were screened for the loss of recombination activity by assaying their ability to survive exposure to 44 ergs/cm 2 of ultraviolet radiation at 254 nm.

Identification of Hap Autoproteolytic Cleavage Sites-Plasmid
pHapLN1036-7SG::G3H6 was transformed into strain Rd/ompP2/ompP5/ rec1. A 1-liter culture of Rd/ompP2/ompP5/rec1/pHapLN1036-7SG::G3H6 was grown to an A 600 of 0.8, and Sarkosyl-insoluble outer membrane proteins were purified using a scaled-up version of the protocol described by Carlone et al. (24). The pellet consisting of the outer membrane fraction was resuspended in 5 ml of 20 mM Tris, 6 M guanidinium chloride, 100 mM sodium chloride, pH 8.0. 1 ml of a 50% slurry of Talon beads (CLONTECH) was added to the sample, and the mixture was incubated at 4°C for 3 h. The Talon beads were then washed 4 times with 5 ml of 20 mM Tris, 8 M urea, 100 mM sodium chloride, 10 mM imidazole, pH 8.0, and histidine-tagged outer membrane proteins were eluted from the beads in 0.5 ml of 20 mM Tris, 8 M urea, 100 mM sodium chloride, 10 mM imidazole, 100 mM EDTA, pH 8.0. The eluted proteins were resolved by SDS-PAGE using 7.5% SDS-polyacrylamide gels and electrotransferred to a polyvinylidene difluoride membrane (28). After staining with Coomassie Brilliant Blue R-250, three protein bands migrating at 43, 41, and 39 kDa, respectively, were excised from the membrane and submitted to Midwest Analytical (St. Louis, MO) for aminoterminal sequence determination performed by automated Edman degradation using a Perkin-Elmer Applied Biosystems model 477A sequencing system.

RESULTS
The Hap Catalytic Triad Consists of His 98 , Asp 140 , and Ser 243 -In considering the mechanism of Hap autoproteolysis, we first sought to identify the residues that participate with Ser 243 in the Hap catalytic site. Examination of the Hap predicted amino acid sequence revealed two aspartic acid residues ϳ100 residues amino-terminal to Ser 243 (Asp 139 and Asp 140 ) and two histidine residues 120 -140 residues amino-terminal to Ser 243 (His 98 and His 117 ), reminiscent of the SA (chymotrypsin) clan of serine proteases. Alignment of Hap sequences from eight different clinical isolates of nontypable H. influenzae revealed absolute conservation of all four of these residues (data not shown). With this information in mind, we changed each of these residues individually to an alanine using sitedirected mutagenesis and then expressed the resulting mutant proteins in H. influenzae strain DB117. As shown in Fig. 1, examination of outer membrane proteins from DB117/ pHapD140A and DB117/pHapH98A demonstrated accumulation of the 155-kDa full-length Hap protein and absence of the 39 -45-kDa Hap ␤ cleavage products, identical to earlier observations with DB117/pHapS243A. Similarly, examination of proteins in culture supernatants from these two strains demonstrated the absence of the 110-kDa secreted Hap s protein (Fig. 1). In contrast, mutation of either Asp 139 or His 117 had no effect on autoproteolysis. Taken together, these data suggest that the Hap catalytic triad is composed of Ser 243 , Asp 140 , and His 98 . Alternatively, mutation of these residues may have affected autoproteolysis by disrupting the tertiary structure around the active site. An amino acid alignment of 10 autotransporter serine proteases that share the GDSGS sequence surrounding the putative active site serine revealed absolute conservation of all three of these residues (Fig. 2), providing further evidence that these amino acids represent the catalytic triad.
The P1 Residue Is Critical for Recognition of the Hap Primary Autoproteolytic Cleavage Site-Previous studies demonstrated that site-directed mutagenesis of Hap residues Leu 1036 and Asn 1037 virtually eliminated autoproteolytic cleavage of the LN1036-7 peptide bond and resulted in increased abundance of three bands, 39 -43 kDa in size, in immunoblots of outer membrane proteins from bacteria expressing this Hap derivative, representing cleavage at more COOH-terminal alternate sites. Given that serine protease enzymatic activity typically relies on recognition of residues immediately NH 2terminal to the cleaved peptide bond, we sought to determine whether mutation of the P1 residue alone would inhibit Hap autoproteolysis at the LN1036-7 site. As shown in Fig. 3A, examination of outer membrane proteins from DB117/ HapL1036S revealed a marked decrease in abundance of the 45-kDa Hap ␤ species, indicating nearly complete elimination of cleavage at the LN1036-7 site. At the same time, the 110 -116-kDa secreted Hap s proteins were present in slightly decreased quantity in culture supernatants (Fig. 3B). In this and other immunoblots, the Hap s proteins resulting from autoproteolytic processing at multiple sites appeared to migrate as a single band due to their close proximity in size. A very small amount of the 45-kDa cleavage product was produced, similar to the situation with HapLN1036-7SG, suggesting that additional amino acids beyond the P1 residue may also be critical for target site recognition.
Identification of Alternative Hap Autoproteolytic Cleavage Sites Reveals a Conserved Motif-Next we set out to identify the alternate cleavage sites responsible for production of the 39 -43-kDa Hap ␤ minor cleavage products. To facilitate recovery of these proteins for NH 2 -terminal amino acid sequencing, we first constructed a Hap derivative that contains both the mutated primary cleavage site (LN1036-7SG) and a 3xGly-6xHis tag at the C terminus. Ultimately this derivative was expressed in an H. influenzae strain Rd mutant that does not express major outer membrane proteins P2 and P5, which consistently contaminated early preparations of His-tagged Hap ␤ minor cleavage products from strain DB117. Talon bead affinity purification of 6xHis-tagged proteins from the outer membrane of Rd/ompP2/ompP5/rec1/pHapLN1036-7SG::G3H6 lead to the recovery of three protein species migrating at 43, 39, and 41 kDa, in order of abundance (not shown). NH 2 -terminal amino acid sequencing of the 43-kDa band revealed the presence of equivalent amounts of two proteins, one beginning with Thr 1047 (TAETQK) and the other beginning with Ala 1048 (AETQKS), indicating that autoproteolytic cleavage at the secondary site occurs either between Leu 1046 and Thr 1047 , between Thr 1047 and Ala 1048 , or both. NH 2 -terminal amino acid sequencing of the 39-and 41-kDa bands resulted in unambiguous assignment of the tertiary site between Phe 1077 and Ala 1078 (ALEAAL) and the quaternary site between Phe 1067 and Ser 1068 (SDPLLP). Alignment of the five residues flanking each side of the primary, secondary, tertiary, and quaternary autoproteolytic cleavage sites allowed construction of a consensus target sequence for Hap enzymatic activity, namely ((Q/R)(S/ A)X(L/F)) at the P4 through P1 positions (Fig. 4). There is no apparent similarity of sequence at the PЈ positions. To confirm the conclusions resulting from NH 2 -terminal amino acid sequencing of the His-tagged Hap ␤ species and to assess whether the P1 residue is critical for recognition of the alternate cleavage sites, we performed site-directed mutagenesis. As shown in Fig. 3, mutation of the P1 residues at both the primary and secondary sites (HapL1036S, L1046S) resulted in nearly complete elimination of both the 45-and 43-kDa cleavage products and a corresponding decrease in abundance of the 110 -116 kDa secreted Hap s proteins. The finding that mutation of Leu 1046 disrupted cleavage at the secondary site suggests that this site probably occurs between Leu 1046 and Thr 1047 , although it is also possible that Leu 1046 is important as the P2 residue for cleavage between Thr 1047 and Ala 1048 . Unlike the situation with mutation of the primary site alone, mutation of both the primary and secondary sites did not result in increased abundance of tertiary and quaternary cleavage products. Mutation of the P1 residues at both the primary and tertiary sites (HapL1036S, F1077S) resulted in elimination of the 45-and 39-kDa cleavage products, while mutation of the P1 residues at both the primary and quaternary sites (HapL1036S, F1067S) resulted in elimination of the 45-and 41-kDa cleavage products (Fig. 3A). The amounts of 110 -116-kDa secreted Hap s proteins in culture supernatants from DB117 expressing these two mutants were not appreciably different from what was observed for DB117/pHapL1036S (Fig. 3B), further demonstrating that cleavage at the tertiary and quaternary sites contributes little to Hap autoproteolysis, even in the context of the primary site mutant.
Hap Autoproteolysis Occurs via an Intermolecular Mechanism-In considering the mechanism of Hap autoproteolysis, it is possible that a given molecule cleaves itself. Alternatively, one Hap molecule may cleave an adjacent neighboring molecule. To address the possibility that intermolecular cleavage occurs, we simultaneously expressed two Hap derivatives in the same bacterial cell. The first derivative, HapS243A, lacks proteolytic activity and therefore is unable to mediate autoproteolytic release of the Hap s passenger domain when expressed by itself (Fig. 5, A and B). The second derivative, Hap⌬1036-99, is proteolytically active but lacks a 64-residue region containing all four autoproteolytic cleavage sites, thus eliminating autoproteolysis when expressed on its own (Fig. 5, A and B). When these two derivatives were expressed together on the same bacterial cell surface, we observed four 39 -45-kDa Hap ␤ species in the outer membrane and 110 -116-kDa Hap s species in culture supernatants, representing cleavage of Hap s from Hap ␤ at all four sites (Fig. 5, A and 5). These results demonstrate that proteolytically active Hap⌬1036-99 molecules were recognizing and acting on cleavage sites in nearby HapS243A molecules. Examination of fractions from a co-culture prepared with both DB117/HapS243A and DB117/Hap⌬1036-99 revealed no Hap ␤ in outer membranes and no Hap s in the culture supernatant, suggesting that intermolecular autoproteolysis likely occurs only between Hap molecules on the same bacterial cell and not between Hap molecules on different cells (not shown).
To extend the observation that intermolecular cleavage occurs, we examined the possibility that released Hap s is capable of cleaving the full-length Hap precursor on the cell surface. Consistent with previous results, we found that incubation of purified Hap s (final concentration 2 g/ml) with DB117/ HapS243A resulted in cleavage of HapS243 at the primary site ( Fig. 6) (6). However, comparison with outer membrane fractions from DB117 expressing wild type Hap revealed relatively reduced cleavage at the primary site (reduced amount of the 45-kDa Hap ␤ protein) and no cleavage at the 3 alternate sites (no smaller Hap ␤ species) (Fig. 6). Next, we incubated purified Hap s with DB117 expressing a Hap derivative containing only the ␤-barrel domain (Hap ␤ , beginning after the primary cleavage site). Again, there was no evidence of cleavage at the 3 alternate sites (Fig. 6). Given that the amount of purified Hap s used in these digestion experiments was much greater than the quantity of Hap s released into culture supernatants of DB117 expressing wild-type Hap, it seems unlikely that intermolecu- Hap Autoproteolysis Is Dependent on the Density of Hap on the Bacterial Surface-Although our experiments assessing cleavage of HapS243A by Hap⌬1036-99 demonstrated intermolecular autoproteolysis, they did not exclude the possibility of an intramolecular mechanism as well. To address the extent to which intermolecular cleavage occurs during processing of wild type Hap, we examined the kinetics of autoproteolysis as a function of the amount of Hap protein present on the cell surface. Autoproteolysis due to a purely intermolecular mechanism would be predicted to increase as a function of protein density on the bacterial surface, while autoproteolytic cleavage due to a purely intramolecular mechanism would be expected to occur at a constant rate independent of protein density. In performing these studies, we expressed wild type Hap in strain DB117 under control of an IPTG-inducible promoter. In order to measure the effect of increasing levels of Hap expression on the rate of autoproteolytic cleavage, we quantitated Hap s in culture supernatants and Hap precursor and Hap ␤ in outer membrane fractions. Briefly, bacteria were grown to early stationary phase, and then IPTG was added to the culture medium to induce maximal expression of Hap. Samples were collected every 10 min over the next 70 min for isolation of culture supernatant and outer membrane fractions. Equivalent amounts of these fractions were analyzed by immunoblotting, and quantities of Hap precursor, Hap s , and Hap ␤ were determined by scanning densitometry (Fig. 7A). Values were normalized based on levels of a constitutively expressed major outer membrane protein, OMP P4 (data not shown). As shown in Fig. 7B, the amount of Hap precursor plus Hap ␤ present in the outer membranes increased linearly over time beginning at the point of induction. Because these two protein species represent the total amount of Hap on the cell surface (cleaved and uncleaved forms), we concluded that IPTG induction resulted in a linear increase in Hap expression over time (first-order kinetics). In contrast, Hap ␤ and Hap s increased in quantity at exponential rates. Together these results demonstrate that autoproteolysis of wild type Hap increases exponentially in response to linear increases in the quantity of Hap on the cell surface, suggesting that a density-dependent intermolecular mechanism is the predominant means of autoproteolysis.

DISCUSSION
Identification of His 98 , Asp 140 , and Ser 243 as a set of three residues involved in the Hap catalytic site provides strong evidence that Hap is a classical serine protease with a catalytic triad. The order in which these residues appear in the Hap primary sequence and the distances between them suggest that Hap is a member of the SA (chymotrypsin) clan of serine proteases (29). This clan comprises a diverse spectrum of proteases including the eukaryotic chymotrypsins, trypsins, and leukocyte elastase of the S1 family, as well as several other families of microbial proteases. Members of the SA clan share a conserved catalytic site and a similar three-dimensional structure.
To extend our conclusion that Hap is a member of the SA clan, we used the 3D-PSSM structural modeling program (Imperial Cancer Research Fund Fold Recognition Server, www.bmm.icnet.uk/servers/3dpssm) to predict the secondary structural elements within the 300-residue Hap NH 2 -terminal domain and to compare predicted folding patterns with a library of known protein structures. The results of this analysis strongly suggest that the Hap NH 2 -terminal domain could be modeled to fit the three-dimensional structure of bovine chymotrypsinogen C, an S1 family protease with similarities in sequence and structure to both chymotrypsin and elastase (35). Chymotrypsinogen C and the Hap NH 2 -terminal domain exhibit roughly 19% identity at the amino acid level. Importantly, with few exceptions, hydrophobic residues predicted to form the structural core of bovine chymotrypsinogen C are conserved in the Hap sequence. Our analysis also predicts that Hap could be modeled using the structure of trypsin from the fungal organism Streptomyces griseus, although conservation of predicted core residues is not as extensive between Hap and S. griseus trypsin as it is between Hap and bovine chymotrypsinogen C (36). Ongoing efforts to crystallize the secreted Hap s protein may eventually confirm these structural predictions.
Further analysis of the Hap amino acid and nucleotide sequences highlights several features that may help to evaluate the evolutionary relationship between Hap and other SA clan members. For example, the AGT codon present at the Hap active site serine is reminiscent of AGY codons utilized by physiologically complex proteases of higher metazoans, such as blood clotting factors, and contrasts to the TCN codons conserved among more "primordial" clan members, such as typsin, chymotypsin, and elastase (34). Furthermore, whereas most SA clan proteases have either a proline or tyrosine at position 225 as a determinant of Na ϩ -activated allostery, Hap has an aspartic acid at the corresponding position. Other notable differences between functionally important residues in S1 family proteases and their counterparts in the Hap sequence include substitutions for conserved serines at positions 189 and 214 by alanine and arginine, respectively, and the absence of four cysteine residues at positions 1, 122, 191, and 220 that form disulfide bonds critical to the stability of the catalytic pocket. These dissimilarities may reflect how selective pressures contributing to the evolution of Hap differed from those affecting eukaryotic S1 proteases and further suggest that the Hap catalytic pocket may represent a distinct and possibly novel structure.
Alignment of the amino acid sequences of Hap and nine other autotransporter proteases revealed sequence similarities throughout the NH 2 -terminal domains and absolute conservation of all three catalytic triad residues, suggesting that these proteins comprise a subfamily of autotransporter proteases that share a common structure at the catalytic site. Included in this subgroup are Hap, the Neisseria and Hemophilus IgA1 proteases, and the SPATEs, a closely related set of putative bacterial virulence factors with distinct enzymatic functions. The existing classification system has tentatively assigned these autotransporters to the S6 family within the SA clan, based solely on conservation of a GDSGS motif and identification of the active site serine within this motif in three members of the subfamily (30,31). Our present study provides the first conclusive biochemical evidence for assignment of an autotransporter to the SA clan and further suggests that Hap belongs to a subset of autotransporters related to but distinct from the S1 family proteases. Of note, several groups have described members of a second family of autotransporter serine proteases that possess a presumed catalytic motif seemingly distinct from that of Hap and its family members (32,33). Based on the conservation of aspartic acid, histidine, and serine residues among its members in the same order and relative positions as the active site residues in subtilisin, this second family is referred to as the subtilases, although biochemical data to support this claim have not been reported.
Identification of the four Hap autoproteolytic cleavage sites provides evidence that Hap may be most like chymotrypsins in terms of substrate specificity. Site-directed mutagenesis confirmed that the residue at the P1 position of Hap autoproteolytic cleavage sites is critical for recognition of substrate by the Hap catalytic pocket, a feature typical of serine proteases in general. Members of the SA clan have differing substrate specificities at the P1 position, with examples including bulky hydrophobic residues such as phenylalanine and leucine in chymotrypsin-like proteases, basic residues such as arginine and lysine in trypsin-like proteases, and small hydrophobic residues such as alanine in elastase-like proteases. The presence of either leucine or phenylalanine at the P1 positions of the Hap autoproteolytic cleavage sites suggests that the Hap catalytic site substrate pocket can accommodate bulky hydrophobic residues, analogous to chymotrypsin. The presence of leucine at the P1 position of the Hap primary and secondary autoproteolytic cleavage sites suggests a preference for this residue over phenylalanine, contrasting with most chymotrypsin-like proteases but resembling bovine chymotrypsinogen C, which cleaves preferentially after leucine residues, presumably due to the influence of a threonine residue (Thr 926 ) in the specificity pocket (35). Cleavage after large hydrophobic residues has also been demonstrated for several other IgA1 protease-like autotransporters. The Neisseria and Hemophilus IgA1 proteases cleave after proline residues within the hinge region of IgA1, and in vitro digestion of chromagenic peptide substrates by SepA from Shigella flexneri suggested that this enzyme acts on target sites with phenylalanine at the P1 position (37,38). Further examination of Hap substrate specificity using chromogenic peptide substrates should help to address whether Hap is truly a chymotrypsin-like protease, keeping in mind that recognition of the cleavage site may require the presence of additional residues beyond those present in commonly available substrate reagents. For many SA clan proteases, substrate recognition and catalytic efficiency are dependent on residues both NH 2 -terminal to and COOH-terminal to the P1 position. At this point it remains unclear whether substrate positions P2 through P4 affect Hap enzymatic activity, although homologies between residues at the P2 and P4 positions among the autoproteolytic cleavage sites suggest this possibility.
Intermolecular mechanisms of autoproteolytic processing have been demonstrated for a number of proteases. Typically, the processing event serves to remove an inhibitory peptide chain from a zymogen, thus converting it into an active enzyme (19, 39 -42). Autoproteolysis of Hap differs from these examples in that Hap enzymatic activity requires no modification of the protease domain subsequent to expression of the precursor on the cell surface. Rather, processing serves to alter localization of the passenger domain from the cell surface to the extracellular environment. Release of Hap s via a predominantly intermolecular mechanism might therefore allow H. influenzae organisms to regulate the percentage of Hap s passenger domain associated with the cell surface by modifying the expression level of precursor molecules. In this model, one could imagine that during the initial stages of colonization, bacteria might express a relatively low level of Hap on the cell surface such that enough passenger domain is present to promote adherence to epithelial structures and formation of microcolonies but not enough to facilitate appreciable levels of autoproteolysis. Once colonization reaches a stage at which available local binding sites are saturated with dense colonies of organisms or a host immune response is elicited, bacteria might then increase Hap expression to levels that promote processing, resulting in extracellular release of Hap s . Such a transition might benefit the bacteria in a variety of ways, for example, by removing an antigenic structure from the cell surface, by promoting disbursement of organisms from large colonies and facilitating spread to new sites, and by allowing the secreted protease domain to encounter host substrates involved in subsequent stages of pathogenesis. Regulation of this transition might be further influenced by inhibitors of Hap enzymatic activity, namely secretory leukocyte protease inhibitor, which is present in normal respiratory secretions and is increased in quantity in the setting of inflammation (5).
In summary, the H. influenzae Hap autotransporter is a member of the SA (chymotrypsin) clan of serine proteases with a catalytic triad that consists of His 98 , Asp 140 , and Ser 243 . These residues are conserved among a subset of autotransporters that appear to constitute a distinct family within the SA clan. Furthermore, Hap cleaves after large hydrophobic residues, suggesting chymotrypsin-like substrate specificity. Autoproteolytic release of the Hap s passenger domain from the bacterial cell surface occurs at least in part by intermolecular cleavage of one precursor molecule by a neighboring molecule on the same cell. This mechanism of autoproteolysis is consistent with the density-dependent processing observed during controlled expression of wild type Hap and suggests a novel mechanism for regulation of bacterial adherence. Further investigation of Hap autoproteolysis and Hap substrate specificity may provide clues as to the ultimate purpose for release of the Hap passenger domain and general insights into the diverse functions of serine proteases.