DegP Initiates Regulated Processing of Filamentous Hemagglutinin in Bordetella bronchiseptica

ABSTRACT Filamentous hemagglutinin (FhaB) is a critical virulence factor for both Bordetella pertussis, the causal agent of whooping cough, and the closely related species Bordetella bronchiseptica. FhaB is an adhesin, suppresses inflammatory cytokine production, and protects against phagocytic cell clearance during infection. Regulated degradation of the FhaB C-terminal prodomain is required to establish a persistent infection in mice. Two proteases, CtpA in the periplasm and SphB1 on the bacterial surface, are known to mediate FhaB processing, and we recently determined that CtpA functions before, and controls the FhaB cleavage site of, SphB1. However, the data indicate that another periplasmic protease must initiate degradation of the prodomain by removing a portion of the FhaB C terminus that inhibits CtpA-mediated degradation. Using a candidate approach, we identified DegP as the initiating protease. Deletion of degP or substitution of its predicted catalytic residue resulted in reduced creation of FHA′ (the main product of FhaB processing) and an accumulation of full-length FhaB in whole-cell lysates. Also, FHA′ was no longer released into culture supernatants in degP mutants. Alterations of the FhaB C terminus that relieve inhibition of CtpA abrogate the need for DegP, consistent with DegP functioning prior to CtpA in the processing pathway. DegP is not required for secretion of FhaB through FhaC or for adherence of the bacteria to host cells, indicating that DegP acts primarily as a protease and not a chaperone for FhaB in B. bronchiseptica. Our results highlight a role for HtrA family proteases in activation of virulence factors in pathogenic bacteria.

model, CtpA degrades FhaB CP to FHA9, which is either released from the bacterial surface or, if the bacteria are grown for extended periods of time, is cleaved by SphB1 to produce FHA, which is then immediately released (Fig. 1A). We hypothesize that cleavage of the PD to form FhaB CP is the critical control point, as it initiates PD degradation. The goal of this study was to identify the initiating protease and to determine its contribution to FhaB processing and function.

RESULTS
BB3749 (degP) appears to encode the initiating protease. To identify candidates for the initiating protease, we searched the genome of wild-type B. bronchiseptica strain RB50 for open reading frames (ORFs) predicted to encode proteases containing N-terminal Sec or TAT signal peptides. The 11 ORFs identified included those encoding both of the previously characterized proteases involved in processing FhaB, sphB1 (21) and ctpA (22), as well as nine predicted to encode endopeptidases and carboxy-terminal-acting peptidases (Table 1). We generated B. bronchiseptica strains containing either an in-frame deletion (DBB0299, DBB3068, DBB2112) or an integrated plasmid disruption (XBB1398, XBB1721, XBB2462, XBB3749, XBB3990, XBB4867) in one of the nine ORFs and examined FhaB/FHA9 in whole-cell lysates (WCL) using a-MCD-specific antisera (Fig. 1C). We included as a control the strain carrying an in-frame deletion of ctpA (DctpA), which has been previously shown to be required for normal FhaB processing (Fig. 1C). Instead of processing FhaB to the 250 kDa FHA9 product, the DctpA strain converts FhaB to the larger FHA 275 polypeptide, likely due to aberrant processing by as-yet-unidentified proteases (22). Two of the mutants displayed altered FhaB processing compared to the wild-type strain. BB0299 (envC) is located directly 59 to, and is cotranscribed with, ctpA (22) and is predicted to encode an M23 family peptidase with a degenerate LytM (dLytM) domain. dLytM homologs in other species, such as EnvC in Escherichia coli, serve as accessory factors that allosterically activate cell wall cleaving hydrolases (24,25). The phenotype of the DenvC Bb mutant is similar, but not identical, to that of the DctpA strain, suggesting that EnvC Bb may function as an accessory factor promoting CtpA activity (Fig. 1C). As envC is in an operon with ctpA, it is also possible that the in-frame deletion mutation in envC could alter expression of ctpA. Characterizing the precise role of EnvC Bb in FhaB processing is beyond the scope of the current work and will be addressed with future studies.
The strain containing a disruption in BB3749 (XBB3749) displayed the expected phenotype for loss of initiating protease activity, i.e., a majority of the FhaB protein remained as the full-length ;371 kDa form (Fig. 1C). We observed partially degraded FhaB polypeptides in XBB3749 WCLs that are not present in lysates of wild-type bacteria, suggesting that in the absence of the BB3749 gene product, unidentified proteases inefficiently degrade FhaB following overnight growth. BB3749 is predicted to encode a high-temperature requirement A (HtrA) family serine protease. Prokaryotic HtrA proteases are extracytoplasmic enzymes initially identified as factors necessary for bacterial survival at elevated temperatures (26)(27)(28). They contain trypsin-like serine protease domains at their N termini, and either one or two protein-binding PDZ domains at their C termini. The protein encoded by BB3749 is predicted to contain a serine protease domain with a conserved catalytic triad (a GNSGG motif and nearby histidine and aspartic acid residues), as well as two PDZ domains ( Fig. 2A). Although annotated as a homolog of mucD, which encodes a protease responsible for regulating mucoidy (due to alginate overproduction) in Pseudomonas aeruginosa (Fig. S1 in the supplemental material), BB3749 is not associated with mucoidy in B. bronchiseptica. Moreover, the amino acid sequence of the protein encoded by BB3749 is 99.8% identical to that of B. pertussis DegP (29). Hence, we will here refer to BB3749 as degP (or degP Bb when distinguishing it from homologs is necessary) and its protein product as DegP (or DegP Bb ). To determine if the FhaB processing defect displayed by the XdegP strain was due to the loss of DegP (and not polar effects of the integrated plasmid), we constructed a strain in which codons 6 to 491 of degP were deleted (DdegP). We also created a derivative of DdegP in which a wild-type copy, driven by the constitutive S12 promoter, was inserted at the attTn7 site (DdegP 1degP). We grew the bacteria overnight in medium containing 20 mM MgSO 4 (which inactivates the BvgAS regulatory system that activates expression of fhaB and other virulence factor-encoding genes) and then outgrew the bacteria in medium without MgSO 4 (BvgAS-activating conditions) for 4 h to induce production of FhaB before collecting WCLs, as previously described (22). As with the XdegP strain, nearly all of the FhaB protein in the DdegP strain was the full-length ;371 kDa form, and normal processing was restored in the complemented strain ( Fig. 2B and C). The intermediate FhaB products that were apparent in WCLs from the XBB3749 overnight cultures were less prominent in the WCLs from DdegP after 4 h of growth under BvgAS-activating conditions, indicating that these aberrant products accumulate only after extended periods of growth (Fig. 2B). For the remainder of this study we examined FhaB processing following 4 h of growth under BvgAS-activating conditions.
To determine if the catalytic activity of DegP is required for FhaB processing, we constructed a strain (iDegP) in which codon 237 was changed to encode alanine instead of the predicted catalytic serine. The FhaB/FHA9 profile of the iDegP mutant was nearly identical to that of the DdegP mutant, indicating that Ser237 and, most likely, the proteolytic activity of DegP are required for efficient FhaB processing ( Fig. 2B and C). DegP Initiates Processing of FhaB in B. bronchiseptica ® DegP is required for B. bronchiseptica growth at elevated temperatures. DegP homologs from other bacterial species, including B. pertussis (29), are required for bacterial survival and growth at elevated temperatures. To determine if B. bronchiseptica also requires DegP to grow under heat stress, we inoculated BG blood agar plates with wild-type, DdegP, iDegP, and DdegP 1degP strains and incubated them at either 37°C or 42°C for 72 h. Both the DdegP and iDegP mutant strains were attenuated for growth at 42°C but not at 37°C (Fig. S2). Failure of the DdegP strain to grow at 42°C was restored by expression of a wild-type copy of degP in trans (DdegP 1degP, Fig. S2). These data indicate that DegP, and its catalytic activity, are required for B. bronchiseptica to grow at 42°C. BvgAS activates (either directly or indirectly) the expression of all genes encoding Bordetella's known protein virulence factors, including fhaB (30). Many of these factors are surface-localized or secreted into the extracellular environment and most, if not all, are produced at a high level when BvgAS is active. The severe growth defect displayed by degP mutants in response to elevated temperature was greatly alleviated in bacteria grown in the presence of 50 mM MgSO 4 , which inactivates BvgAS. Magnesium ions (Mg2 1 ) are crucial for envelope stability (31), so we also grew all four strains on plates containing 50 mM MgCl 2 . Unlike MgSO 4 , which rescued the growth of the mutant strains, addition of MgCl 2 had no effect on bacterial growth at 42°C. Thus, these data suggest the requirement for DegP under heat stress conditions is primarily due to its proteolytic activity on one or more BvgAS-activated virulence factors (29,32).
DegP acts first in stepwise processing of the prodomain. Our current model for FhaB processing involves three proteases: two periplasmic enzymes, CtpA and the initiating protease, and one exoprotease, SphB1 (Fig. 1A). We previously determined that CtpA and SphB1 function in series, with CtpA acting prior to, and determining the cleavage site of, SphB1 (22). In fact, our data suggest that SphB1-dependent processing only occurs after extended time in culture and may have no functional relevance under normal (in vivo) circumstances. We have also shown previously that the ECT inhibits CtpA from degrading the PD, as deletion of the ECT (DECT) causes rapid and thorough conversion of FhaB to FHA9 (which we call hyperprocessing) in a CtpA-dependent manner (11,22). We hypothesized that the initiating protease is responsible for removing the ECT to convert full-length FhaB into the CtpA substrate FhaB CP . To test this hypothesis, we first showed (as we have done previously [22]) that addition of an HA epitope N-terminal to the PRR (Fig. 3A, strain FhaB HA-PRR ) has no effect on FhaB processing ( Fig. 3B, both full-length FhaB and FHA9 are present in WCL of the FhaB HA-PRR strain). Deletion of degP in this strain resulted in an accumulation of full-length FhaB HA-PRR and a reduced amount of FHA9, just as it did in wild-type bacteria (Fig. 2), and complementation with wild-type degP restored FhaB processing (Fig. 3B). Deletion of ctpA does not alter the abundance of full-length FhaB in WCL, and results in the formation of FHA 275 ( Fig. 3B and reference 22). FHA 275 is an ;275 kDa polypeptide containing an incompletely degraded PD. We next examined if degP is required for the hyperprocessing of FhaB that occurs in the absence of the ECT. In wild-type bacteria producing FhaB lacking the ECT (FhaB DECT , Fig. 3A), full-length FhaB is barely detectable (Fig. 3C). The same was true for the DdegP strain (Fig. 3C), indicating that DegP does not contribute to the hyperprocessing of FhaB lacking the ECT. In contrast, full-length FhaB DECT was abundant in the DctpA strain (Fig. 3C, yellow band, as this strain also contains an HA epitope immediately N-terminal to the PRR), showing that the hyperprocessing of FhaB DECT is CtpA-dependent (22). These data indicate that DegP acts prior to CtpA and is required for removal of the ECT to allow degradation of the prodomain by CtpA.
Similar to deleting the ECT and, in contrast to inserting an HA epitope N-terminal to the PRR, adding an HA epitope to the C terminus of FhaB (FhaB CT-HA ) results in CtpA-dependent hyperprocessing ( Fig. 3D and reference 22), i.e., this epitope somehow converts the FhaB C terminus into a substrate for CtpA. We reasoned that addition of the HA epitope to the C terminus could impact FhaB processing in two ways: (i) the HA epitope itself could be a substrate for the tail-specific protease CtpA and/or (ii) addition of the HA epitope could disrupt proper folding of the ECT, causing it to be degraded by DegP. Deletion of degP did not prevent hyperprocessing of FhaB CT-HA , indicating that DegP is not required for CtpA-dependent degradation of the full-length FhaB CT-HA and that CtpA proteolytic activity is not dependent on DegP. Consistent with CtpA-dependent degradation of FhaB CT-HA , we observed an accumulation of the full-length polypeptide (371 kDa yellow band) in WCLs from the DctpA strain. We also detected the FhaB CP in DctpA WCLs, which we believe is the product formed following removal of the ECT by the initiating protease (;362 kDa green band). The presence of the FhaB CP in DctpA FhaB CT-HA and not in DctpA FhaB HA-PRR WCLs indicates that the addition of the C-terminal HA epitope promotes removal of the ECT-HA by the initiating protease (DegP). Taken together, these data indicate that DegP acts first in stepwise processing of the PD.
DegP is required for release of FhaB-derived polypeptides from the bacterial surface. B. bronchiseptica efficiently releases FhaB-derived polypeptides into culture supernatants and the specific polypeptide released depends on the age of the culture and the proteases produced by the bacterial strain (Fig. 4B). Wild-type B. bronchiseptica primarily releases FHA9 during the first 6 h in culture under BvgAS-activating conditions, before SphB1 begins to cleave accumulated FHA9 to cause the formation and immediate release of FHA (22) (Fig. S3). In the absence of CtpA, FHA9 is not produced and, instead, FhaB is converted to the aberrant FHA 275 product, which is retained on the cell surface. FHA 275 cannot be released due to the presence of the complete PNT. However, FhaB or FHA 275 can be cleaved by SphB1 promiscuously at nonpreferred sites to form the smaller polypeptides FHA 1/2 , which are immediately released (20, 22) ( Fig. 4A and B). None of the processed polypeptides (FHA9, FHA, or FHA 1/2 ) were detected in significant amounts in the DdegP mutant, demonstrating that DegP is critically important for release of FhaB-derived polypeptides.
As shown in Fig. 4B and in our previous publication (22), either CtpA or SphB1 is required for release of FhaB-derived polypeptides. In the absence of DegP, however, neither CtpA nor SphB1 is sufficient; FHA9 and/or FHA 1/2 were barely detectable in supernatants of any strain containing the DdegP mutation (Fig. 4B). These results reinforce the conclusion that DegP-mediated cleavage of FhaB occurs first and is required for subsequent degradation by CtpA and then cleavage by SphB1. When considered in DegP Initiates Processing of FhaB in B. bronchiseptica ® conjunction with the WCL data (Fig. 4C), these data are also consistent with our model (12) (Fig. 1 and depicted in greater detail in Fig. S3), which states that the polypeptide formed in the absence of CtpA (FHA 275 ) cannot be released but can serve as a poor substrate for SphB1, cleavage by which generates FHA 1/2 (because the PNT prevents movement through the FhaC channel that would allow the preferred SphB1 cleavage site to be exposed on the surface), which is immediately released (Fig. 4A and B). It is important to note that the supernatant samples have been concentrated substantially compared to the WCLs. Therefore, while levels of FHA 1/2 in supernatants of the DctpA strain appear to be similar to levels of FHA9 in supernatants of wild-type bacteria, no FHA 1/2 was detected in WCLs of the DctpA mutant and a substantial amount of FHA9 was present in WCLs of wild-type bacteria, indicating that in the absence of CtpA, generation of FHA 1/2 by SphB1 is less efficient than generation of FHA9 by wild-type bacteria. In the absence of DegP, FHA 275 is not formed and only a small amount of FHA9 is generated, which cannot be detected in culture supernatants (Fig. 4B).
DegP is not required for translocation of FhaB through FhaC. DegP could be indirectly required for FhaB processing by promoting FhaB translocation through FhaC. To determine if loss of DegP prevents efficient translocation of FhaB to the cell surface, we used a dot blot assay to compare relative amounts of FhaB on the surface of intact bacteria following 4 h of growth under BvgAS-activating conditions (Fig. 5). In this assay, only surface-exposed epitopes are detectable by antibodies when cells are intact. When the bacteria are boiled to disrupt the membranes, internal epitopes can additionally be detected by antibodies. The a-MCD antibody (green) detected the MCD in unboiled bacteria, indicating that the MCD was located external to the outer membrane. However, the HA tag was only detected after boiling the bacteria, indicating that the PD remains localized in the periplasm. As expected, neither the MCD nor the HA tag was detected in the DfhaB strain. The protease mutants, DctpA and DdegP, were similar to wild-type bacteria in that the MCD was detected regardless of boiling, while the HA tag was only detected postboiling. These results indicate that neither CtpA nor DegP is required for FhaB to be secreted through the outer membrane or for the PD to remain within the periplasm. Therefore, the reduction in FhaB processing that occurs in degP mutant strains is not due to mislocalization or instability of FhaB.
DegP is not required for FhaB-mediated adherence to host cells in vitro. While the dot blot analysis indicates that DegP is not required for translocation of FhaB to the bacterial surface, it cannot distinguish between active or inactive conformations of FhaB. We performed in vitro adherence assays to determine if the FhaB on the surface of DdegP bacteria is capable of promoting bacterial adherence to human epithelial cells in vitro. Adherence of the DdegP strain was similar to that of wild-type B. bronchiseptica, indicating that DegP is not required for translocation of FhaB to the bacterial surface or for the MCD to fold into an adherence-competent conformation (Fig. 6). These data, along with the finding that strains that produce catalytically inactive DegP are defective for FhaB processing, are consistent with the primary function of DegP being that of a protease, rather than a chaperone, at least with regard to FhaB.
DegP does not signal through the RseA-RpoE pathway to control FhaB processing indirectly. Some HtrA proteins, such as E. coli DegS (DegS Ec ), are not degradative proteases. Instead, they transduce signals from the periplasm to cytoplasmic effectors, controlling the expression of genes encoding stress response factors indirectly. In the E. coli RseA-RpoE signal transduction system, the anti-sigma factor RseA, which is embedded in the inner membrane, binds and sequesters the RpoE sigma factor (s E) in the cytoplasm, preventing s E from binding to and directing RNA polymerase to activate expression of target genes (33). In E. coli undergoing exponential growth under standard laboratory conditions, DegS Ec homotrimers associated with the inner membrane are in an inactive conformation. When E. coli is exposed to conditions that disrupt the folding and localization of proteins in the periplasm, misfolded proteins bind DegS Ec 's single PDZ domain, resulting in allosteric activation of DegS Ec 's proteolytic activity (34). When active, DegS Ec cleaves a specific site on a periplasmic region of RseA, initiating a proteolytic cascade that causes the release of s E and the subsequent activation of stress response genes (33). Collectively, this process is referred to as regulated intramembrane proteolysis (RIP) (35). Among the genes activated in the E. coli s E regulon is degP (36-39). Although the genomes of the Bordetella reference strains Tohama I (B. pertussis) and RB50 (B. bronchiseptica) encode functional Rse-RpoE systems (40-42), DegP Initiates Processing of FhaB in B. bronchiseptica ® they do not encode DegS homologs, as neither of the two HtrA homologs produced by these strains (DegP and the BB4867 gene product) are predicted to include TM domains capable of inserting in the inner membrane. Therefore, the protease responsible for initiating regulated proteolysis of RseA in Bordetella is not known. Unlike E. coli, in which degP is located distally on the chromosome from rpoE, degP is in an operon with rpoE, rseA, and rseB in B. pertussis and B. bronchiseptica (Fig. 7A). These observations led us to ask if DegP could regulate FhaB processing indirectly, by cleaving RseA Bb and subsequently increasing s E-dependent expression of a gene encoding yet another protease. RpoE is thought to be essential in B. pertussis (41) but is dispensable for growth of B. bronchiseptica under standard laboratory conditions (42). To determine if DegP signals through the RseA-RpoE pathway, we deleted rpoE in wild-type, DdegP, and iDegP B. bronchiseptica (Fig. 7B). If DegP signals through RseA-RpoE to activate expression of a gene encoding another protease, then loss of RpoE would cause the DrpoE strain to exhibit a similar defect in FhaB/FHA processing seen in WCLs from DdegP and iDegP mutants. Instead, we observed that deletion of DrpoE had no effect  on FhaB processing by any of the strains under the conditions examined, indicating that DegP is not indirectly regulating degradation of the FhaB PD via RpoE (Fig. 7).

DISCUSSION
Our results indicate that the HtrA family protease DegP initiates degradation of the FhaB prodomain in B. bronchiseptica. Three key findings support this conclusion. (i) FhaB processing was dramatically reduced, but not completely abrogated, in strains lacking DegP (DdegP) or producing DegP in which the predicted catalytic serine was replaced with alanine (S237A, iDegP). The small amount of FHA9 detected in WCLs of degP mutants was likely due to the prodomain being converted to a CtpA substrate by housekeeping proteases, which is revealed only in the absence of DegP. (ii) DegP was dispensable for degradation of FhaB polypeptides with altered C termini (DECT and CT-HA) that are efficiently degraded by CtpA (Fig. 3). These data are consistent with our previous studies, which determined that the native FhaB C terminus is a poor substrate for CtpA (22), and indicate that, in wild-type bacteria, DegP cleaves the prodomain first, generating a C terminus that is then efficiently degraded by CtpA. CtpA-dependent degradation of the DECT and CT-HA polypeptides in the absence of DegP also indicate that accumulation of full-length in WCLs from degP mutants is not due to the loss of CtpA activity. (iii) DegP was not required for translocation of FhaB through FhaC (Fig. 5) or for FhaB to mediate adherence to epithelial cells (Fig. 6), consistent with DegP functioning as a protease and not a chaperone for FhaB. Previous studies determined that if FhaB is stalled in the periplasm (i.e., in fhaC mutants), it is degraded (43). The stability of full-length FhaB and the surface-localization of the MCD in degP mutants, therefore, indicate that DegP is not required to preserve FhaB stability or to deliver the TPS domain to FhaC in the periplasm. Although it is formally possible that DegP regulates FhaB processing indirectly by activating another protease, the most parsimonious explanation for our findings is that DegP acts on FhaB directly.
FhaB is required to mediate adherence to mammalian cells, prevent bacterial clearance by innate immune effectors, and control inflammatory signaling during infection. While processed FHA9 is sufficient for adherence, full-length FhaB and regulated degradation of the prodomain is required specifically for B. bronchiseptica to resist phagocytic cell clearance; mutants with altered prodomain processing are cleared rapidly from the lower respiratory tract but are not altered for adherence in vitro or in vivo or for suppression of inflammatory cytokines (11). Consistent with these data, the DdegP strain adhered to human A549 epithelial cells as efficiently as do wild-type bacteria (Fig. 6), confirming that regulated prodomain processing is not required for FhaB to function as an adhesin. The fact that FhaB was exported to the cell surface (Fig. 5) and able to mediate adherence in degP mutants (Fig. 6) indicates that DegP does not function as a chaperone required to stabilize FhaB in the periplasm or to bring the N terminus of the TPS domain to FhaC. Furthermore, if DegP was an important chaperone for the FhaB prodomain, then lack of DegP would result in hyperprocessing, similar to the DECT strain. However, full-length FhaB was stabilized in degP mutants, indicating that DegP does not promote folding or stability of the prodomain.
Although B. bronchiseptica consistently produces and releases FHA9 (and sometimes FHA) from the bacterial surface during in vitro culture, neither the mechanism by which release occurs, nor its biological relevance, is understood. Our previous data suggested that either CtpA or SphB1 are sufficient for release (22). However, our new data indicate that their sufficiency is dependent on DegP, as no FhaB-derived polypeptides are released by degP mutants (Fig. 4B). Future studies to determine how DegP and CtpA (and potentially EnvC) cooperate to degrade the prodomain should also reveal the mechanism underlying retention versus release of FhaB-derived polypeptides, another important outstanding question for TPS systems, in general.
DegP's proteolytic activity is required for initiating FhaB processing in B. bronchiseptica, but it is not known whether the DegP homolog from B. pertussis is involved in degradation of the FhaB prodomain in that organism. A previous study that examined DegP in B. pertussis determined that this protease may have both catalytic and noncatalytic functions which are required for maintaining envelope integrity and are involved with FhaB secretion (29). However, that study did not examine prodomain degradation (29). Comparison between our study and Baud et al. highlights some key differences between B. pertussis and B. bronchiseptica, namely, that B. pertussis degP mutants are more susceptible to elevated temperature than are B. bronchiseptica degP mutants. B. pertussis degP mutants are attenuated for growth at 37°C (29), while B. bronchiseptica strains lacking DegP present a growth defect at 42°C but not at 37°C (Fig. S2). Baud et al. determined that growth at 37°C of the degP mutants could be partially restored by expression of catalytically inactive DegP (S237A) in trans, which, combined with examination of binding interactions between DegP and nonnative N-terminal FhaB polypeptides in vitro, led them to conclude that DegP functions as a chaperone in B. pertussis (29). In contrast, we found that DegP's proteolytic activity is required for both FhaB processing and resistance to heat stress. Our iDegP strains, which produce DegP S237A from the native locus, were as attenuated for growth at 42°C and for FhaB processing as was the DdegP strain ( Fig. S2 and Fig. 2), consistent with DegP in B. bronchiseptica functioning primarily as a protease for both FhaB processing and resistance to envelope stress.
In a follow-up study, Baud et al. determined that DegP can be both soluble and associated with bacterial membranes as trimers, that membrane associated DegP has a higher affinity for nonnative FHA polypeptides (lacking most of the MCD and all of the prodomain), and that membrane-associated DegP has proteolytic activity (44). These findings provide insight into how DegP could be regulating degradation of the FhaB prodomain. We hypothesize that membrane-associated DegP trimers could form a stable complex with the FhaB prodomain until a signal causes a conformational change in the prodomain such that DegP recognizes the prodomain as a substrate and initiates degradation. The signal that initiates degradation of the prodomain is unknown; however, previous work from our lab and others suggests that FhaB may be part of a novel toxin delivery system. The adenylyl cyclase toxin (ACT)-known to inactivate phagocytic cells (45)(46)(47)-binds FhaB on the bacterial surface (48,49), and we hypothesize that regulated degradation of the FhaB prodomain controls delivery of surface-associated ACT to those cells. Our current model states that ACT forms a stable complex with full-length FhaB on the bacterial surface until these FhaB-ACT complexes interact with the ACT receptor (CR3, CD11b/CD18, Mac1) on phagocytic cells (46,50). We propose that this interaction between FhaB-ACT and CR3 could generate a signal that is propagated through the FhaB MCD, resulting in a conformational change in the prodomain that allows DegP to initiate degradation of the prodomain. Previous data indicate the FhaB ECT-which is 100% identical at the protein level between FhaB homologs in B. pertussis, B. bronchiseptica and Bordetella parapertussis-likely adopts a conformation that stabilizes the prodomain by preventing degradation by CtpA (22). It is interesting to speculate that FhaB-ACT binding the CR3 receptor could trigger the ECT to unfold and therefore become a substrate for DegP, consistent with the preference for DegP homologs, such as those from E. coli and B. pertussis, to degrade unfolded proteins (29,51). It is also possible that the signal that initiates FhaB processing is envelope stress that B. bronchiseptica faces during infection, which could induce conformational changes in the FhaB prodomain that cause it to become a substrate for DegP. Future studies will focus on identifying the signal that initiates FhaB processing and determining if regulated processing of FhaB is involved in ACT delivery to specific host cells.

MATERIALS AND METHODS
Bacterial strains. Bacterial strains and plasmids are listed in Table S1 in the supplemental material. In-frame deletions were created by allelic exchange using derivatives of the pSS4245 plasmid (52). For complementation, degP was inserted into a derivative of the pUC18 plasmid that contained the S12 promoter and integrated into the chromosome at the att:Tn7 site (53). E. coli strains were used to amplify vectors (DH5a) and to transform B. bronchiseptica (RHO3). Mutations were confirmed by PCR and/or sequencing. As we have published in the past, our "wild-type" strain for FhaB studies is RBX11, a derivative of RB50 in which fhaB is more genetically tractable because the strain lacks fhaS, a gene highly homologous to fhaB but that plays no role in virulence (43).
Culture media and conditions. Bordetella bronchiseptica strains were streaked from 280°C stocks onto Bordet-Gengou agar (BD Biosciences) supplemented with 6% defibrinated sheep blood (HemoStat Laboratories) and grown at 37°C for 2 to 3 days. Because some combinations of protease deletions, such as DctpA DdegP, resulted in growth defects that were minimized if BvgAS was not active, protease dual mutants were streaked onto blood plates supplemented with 50 mM MgSO 4 . Colonies were picked from these plates and cultured overnight in Stainer-Scholte broth (SS) containing MgSO 4 (50 mM) to prevent production of Bvg-induced virulence factors, including FhaB. Bacteria were pelleted and washed with cold, sterile Dulbecco's phosphate-buffered saline (DPBS, Gibco) and then cultured further for 4 h in fresh Stainer-Scholte broth lacking MgSO 4 to initiate production and processing of FhaB. This culture method is useful for examining early FhaB production and processing and to compare differences found across bacterial mutants that may be hidden after 16 h of overnight growth (22). For experiments described in Fig. 1, bacteria were simply grown overnight in Bvg-inducing conditions (SS without MgSO 4 ). Escherichia coli strains were grown at 37°C in lysogeny broth or on lysogeny broth agar. Where appropriate, medium was supplemented with streptomycin (20 mg/ml), kanamycin (50 mg/ml), and MgSO 4 (50 mM). For heat tolerance experiments (Fig. S2), the indicated strains were grown in Bvg1 phase liquid cultures (SS broth) overnight, normalized to an optical density at 600 nm (OD 600 ) of 1, and then 10-fold serial dilutions of each culture were spotted (5 ml per spot) onto BG blood agar plates alone or onto plates supplemented with 50 mM MgSO 4 or 50 mM MgCl 2 .
A549 human epithelial cells were cultured as described by the supplier (ATCC) in Ham's F12 medium (Gibco) with 10% fetal bovine serum (FBS) (VWR). Cells were grown until they reached 80% confluence in T75 tissue culture treated flasks (Corning) before passaging, and the cells were passaged no more than 15 times out of a freeze.
Immunoblots. To examine cell-associated proteins by Western blotting, whole-cell lysates were prepared by boiling pelleted cells in Laemmli buffer and shearing by passage through a 26G needle. For released proteins, culture supernatants were filtered through 0.2-mm filters and proteins were precipitated using 10% trichloroacetic acid, rinsed with cold acetone, resuspended in 1 M Tris-HCl (pH 8.8) and Laemmli buffer mixture, and boiled. Proteins were separated by SDS-PAGE using 4% or 5% polyacrylamide gels. Proteins were transferred to nitrocellulose membranes (GE Healthcare) and the membranes were probed with a rabbit polyclonal antibody generated against the FHA mature C-terminal domain (MCD) (10) and a mouse monoclonal antibody generated against an HA epitope (BioLegend). Corresponding a-rabbit and a-mouse IRDye secondary antibodies (LI-COR Biosciences) were used to detect proteins using a LI-COR Odyssey Classic Blot Imager (LI-COR Biosciences). Protein quantification was performed using LI-COR Image Studio version 5 software. B. bronchiseptica sample volumes were normalized based on optical density of the cultures.
To compare cell surface exoproteins versus internal proteins by dot blot, 100 ml of 0.5 OD 600 /ml of bacteria were spotted onto nitrocellulose membranes using a 96-well vacuum manifold. The membranes were probed using anti-MCD and anti-HA antibodies. Signal detected on intact bacteria corresponded to surface-exposed epitopes, and boiling disrupted the membrane to allow additional antibody interaction with otherwise inaccessible internal epitopes.
Bacterial adherence assay. These assays were performed as previously described with minor alterations (5). A549 cells were seeded on sterile coverslips in 12-well plates. Once cells reached a confluence of 50% to 80%, the seeding medium was removed and the cells were inoculated with the indicated B. bronchiseptica strains (wild type [wt], DfhaB, DdegP, and DdegP 1 degP) in SS broth at a multiplicity of infection (MOI) of 100 CFU per epithelial cell (2.5 Â 10 7 CFU/ml of bacteria). One set of wells was treated with SS broth alone (mock infection control). The bacterial density of each inoculum was confirmed by plating on BG blood agar. The 12-well plates were centrifuged at 500 Â g for 5 min at room temperature and then incubated for 15 min at 37°C with 5% CO 2 . The inoculum was removed from each well and the cells were washed four times with Hanks' balanced salt solution (Difco). Cells were then fixed with 0.5 ml of ice-cold methanol for 5 min, after which the methanol was removed and the coverslips were air dried for 10 min. The cells were then stained for 15 min with 0.5 ml of 1:20 diluted Giemsa stain (Sigma-Aldrich). The stain was removed and the coverslips were washed twice with water before being air dried and mounted onto slides with permount (Sigma-Aldrich). The slides were imaged at 60Â magnification (Keyence).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank the Cotter lab for helpful discussions, especially Jessica Beauchamp for critically reading the manuscript and providing editorial suggestions. We thank the NIH for financially supporting this work (grants R01-AI129541 and R01-AI153160) and we