CpaA Is a Glycan-Specific Adamalysin-like Protease Secreted by Acinetobacter baumannii That Inactivates Coagulation Factor XII

Ventilator-associated pneumonia and catheter-related bacteremia are the most common and severe infections caused by Acinetobacter baumannii. Besides the capsule, lipopolysaccharides, and the outer membrane porin OmpA, little is known about the contribution of secreted proteins to A. baumannii survival in vivo. Here we focus on CpaA, a potentially recently acquired virulence factor that inhibits blood coagulation in vitro. We identify coagulation factor XII as a target of CpaA, map the cleavage sites, and show that glycosylation is a prerequisite for CpaA-mediated inactivation of factor XII. We propose adding CpaA to a small, but growing list of bacterial proteases that are specific for highly glycosylated components of the host defense system.

Here, we demonstrate that CpaA increases the clotting time in the aPTT assay through proteolytic cleavage and inactivation of fXII in both human and murine plasma. Specifically, CpaA targets the glycosylated, proline-rich domain of fXII. Deglycosylation of fXII and substitution of one of the glycosylated amino acids in the proline-rich region of fXII reduce the cleavage efficiency of CpaA.

RESULTS
CpaA interferes with the contact activation pathway in human plasma. CpaA, a member of the adamalysin family of secreted metalloproteases, has been found to inhibit clotting in human plasma using the aPTT assay (17). We confirmed that culture supernatant from the WT AB031 strain, but not from the isogenic ΔcpaA mutant, increases the clotting time in an aPTT assay in human plasma (Fig. 1A). Complementation of the ΔcpaA mutant required coexpression of both cpaA and cpaB, the gene coding for a putative chaperone, as no complementation was observed with cpaA alone (Fig. 1B). This is consistent with the results of a recent study that used both cpaA and cpaB to restore secretion of CpaA in an in-frame, unmarked cpaA deletion mutant of A. nosocomialis and may be due to the presence of a cryptic cpaB promoter within cpaA or to differential folding of the transcript, resulting in reduced levels of expression of cpaB in the absence of cpaA (16). Similarly to the ΔcpaA mutant, culture supernatant from the T2SS mutant ΔgspD strain was lacking CpaA (Fig. S2A and B) and had no effect on clotting (Fig. 1A), and while full complementation was not obtained with plasmidencoded gspD due to only partial restoration of CpaA secretion (Fig. S2C), the increase in clotting time in the presence of pgspD compared to the clotting time of the empty FIG 1 Clotting time of human plasma is increased when incubated with culture supernatant from A. baumannii but not from strains deficient in either CpaA or the T2SS. An aPTT assay was performed after incubation of normal human plasma with culture supernatant from A. baumannii strain AB031, AB031 ΔcpaA, or AB031 ΔgspD (A), AB031 or ΔcpaA with either empty vector pMMB960 (p), pcpaA or pcpaAB (B), AB031 or ΔgspD with either empty vector pMMB67 (p) or pgspD (C), and A. baumannii strain 17978 or the T2SS mutant 17978 ΔgspD with either empty vector pMMB67 (p) or pcpaAB (D). The growth medium LB served as a negative control. Values that are significantly different are indicated by ordinary one-way ANOVA (n ϭ 6) are indicated by bars and asterisks as follows: ***, P Յ 0.001; **, P Յ 0.01.
A. baumannii Inactivation of Factor XII ® vector control was statistically significant (Fig. 1C). Lack of full complementation has been reported previously for some T2SS mutants, including gpsD mutants and can be due to expression levels that are either too low or too high that may interfere with the stoichiometry and thus, the function of the T2SS (15,(23)(24)(25)(26). Cloning and overexpression of cpaA and the putative chaperone-encoding cpaB also resulted in extracellular release of CpaA in a T2SS-dependent fashion in the reference strain ATCC 17978, which does not encode its own cpaA and cpaB genes (Fig. S3A). Not surprisingly, culture supernatant from WT ATCC 17978 overexpressing the protease also extended the clotting time in the aPTT assay, while expression of plasmid-encoded cpaA and cpaB in the 17978ΔgspD mutant did not diminish the clotting function (Fig. 1D). The effect on the clotting time in the aPTT assay was dose dependent (Fig. S3B), and the CpaA activity was significantly higher in the culture supernatant of the ATCC 17978 strain than in the original AB031 strain due to overexpression. This material, in which the concentration of CpaA was approximately 0.4 mg/ml, was used in all subsequent analyses to characterize the activity and specificity of CpaA. While we confirmed that CpaA does inhibit clotting in human plasma in the aPTT assay, unlike Tilley et al. (17), we found no significant effect of CpaA in a PT assay (Fig. S4A), which measures the activity of the extrinsic pathway. Despite not having an effect in the PT assay, CpaA does cleave coagulation factor V (fV) (Fig. S4B), but this cleavage neither inactivates nor activates fV. These findings led us to focus on the intrinsic pathway and the aPTT assay.
CpaA targets factor XII. Since the aPTT assay primarily measures the intrinsic coagulation pathway, to identify the target of CpaA, we used a modified aPTT factor assay with a variety of factor-deficient human plasma samples. Normal (healthy) plasma samples were first incubated with CpaA from culture supernatant of ATCC 17978/ pcpaAB or LB medium as a control. The samples were then diluted 1:100 into various human plasma samples lacking specific factors of the intrinsic pathway, and an aPTT assay was performed. In this way, the factor inactivated by CpaA in the normal plasma will not be able to complement the specific factor-deficient plasma. The finding that the clotting time was increased only when CpaA-treated normal plasma was added to fXII-deficient plasma suggested that fXII is cleaved and inactivated by CpaA ( Fig. 2A). These results were confirmed by immunoblotting (Fig. 2B), which showed that CpaA cleaves fXII in human plasma, and by analyzing the effect of CpaA on purified fXII in a fluorogenic assay against a low-molecular-weight substrate (Fig. 2C), showing that CpaA prevents the conversion of fXII into an active serine protease.
Identification of cleavage sites in fXII. Treatment of purified human fXII with increasing amounts of CpaA from culture supernatant of ATCC 17978/pcpaAB followed by SDS-PAGE analysis identified two distinct cleavage events that generated three fXII fragments (Fig. 3A), whereas control culture supernatant from ATCC 17978 with empty FIG 2 CpaA cleaves fXII. (A) Normal human plasma (NHP) was incubated with (ϩ CpaA) or without (Ϫ CpaA) culture supernatant from ATCC 17978/pcpaAB, diluted 1:100 and added to different factor deficient plasma as indicated. Clotting time was determined using an aPTT assay. ***, P Յ 0.001 by ordinary one-way ANOVA; n ϭ 6. (B) NHP was incubated with (ϩ) or without (Ϫ) CpaA present in culture supernatant from ATCC 17978/pcpaAB for 20 min. fXII-deficient plasma incubated without CpaA was used as a control. Samples were subjected to SDS-PAGE and immunoblotting. fXII was detected using anti-fXII antibody. A representative blot is shown. Mw, molecular weight. (C) Purified human fXII was incubated with (ϩ CpaA) or without (Ϫ CpaA) culture supernatant from ATCC 17978/pcpaAB, followed by incubation without or with aPTT reagent to activate fXII and analyzed in a fluorogenic assay against the low-molecular-weight substrate z-GlyGlyArg-AMC to measure fXII proteolytic activity. ****, P Յ 0.0001 by ordinary one-way ANOVA; n ϭ 6.

®
vector had no effect on fXII (Fig. S5). The fragments were blotted, excised, and subjected to N-terminal sequencing by Edman degradation to identify the cleavage sites. The low-molecular-weight fragment 1 sequence indicated that it represents the very N-terminal portion of mature fXII starting with residue Ile20 (residue 1 of mature fXII). Fragment 3, which is a cleavage product of fragment 2, ran as a doublet on SDS-PAGE (Fig. 3A). Both bands of the doublet were found to possess identical N terminus, XXRTPPQSQX. The first two residues and the last residue could not be identified as they are O-linked glycosylated threonines (UniProtKB-P00748 and reference 27). However, as residues 3 through 9 were clearly identified for both fragments, the results suggest that the sequence represents residues 309 to 318 in the glycosylated proline-rich domain of fXII and that fragment 3 is generated by cleavage between Pro308 and Thr309 (Fig. 3B). The reason fragment 3 ran as a doublet with identical N termini is likely due to glycosylation heterogeneity as mass analysis of purified, intact fXII identified three major species with mass differences that were multiples of 657 Da, which is the mass of NeuAc-Hex-HexNAc modification (Fig. S6).
The results from the N-terminal sequencing of fragment 2 (indicated in Fig. 3A) were not interpretable, possibly due to the presence of several prolines and glycosylated residues. Therefore, we subjected this fragment, as well as fragment 1, to in-gel trypsin digestion and liquid chromatography followed by tandem mass spectrometry (LC/MS/ MS). Identification of glycosylated and nonglycosylated peptides using Bionic and Mascot software and peptide coverage analysis indicated that fragment 1 spans residues 1 to 279 and fragment 2 consists of residues 280 to 610 of fXII (Table 1). These results suggest that CpaA cleaves between Pro279 and Thr280 in the glycosylated proline-rich region of fXII (Fig. 3B). Taken together with the finding that the second site of cleavage also occurs between a proline and threonine, these results suggest that CpaA has a preference for Pro-Thr peptide bonds. A. baumannii Inactivation of Factor XII When normal (healthy) human plasma (NHP) was incubated with increasing amounts of CpaA, followed by SDS-PAGE and immunoblot analysis, plasma fXII was similarly cleaved at two sites by CpaA ( Fig. 3C; note that only full-length fXII and fragments 2 and 3 were detected by the monoclonal anti-fXII, suggesting that the antibody epitope is absent in fragment 1). Furthermore, when these CpaA-treated plasma samples were analyzed in the aPTT assay, the results indicated that inactivation of fXII by CpaA occurs only upon cleavage at the second site, Pro308-Thr309, of fXII (note that the increase in clotting time seen in Fig. 3D correlates with the appearance of fragment 3 in Fig. 3C).
Specificity of CpaA. Both residues Thr280 and Thr309 in human fXII are modified with O-linked glycans (Swiss-Prot entry P00748 and references 27 and 28), suggesting that CpaA may have a preference for O-linked glycosylated residues. To determine whether glycosylation is required for CpaA cleavage, we treated fXII with neuraminidase and O-glycosidase to deglycosylate fXII before incubating it with CpaA. SDS-PAGE analysis indicated that deglycosylation resulted in increased mobility of fXII (Fig. 4A, compare lanes 2 and 4), but more importantly, prevented cleavage by CpaA (Fig. 4A, lane 5). The function of deglycosylated fXII, with and without treatment with CpaA, was also analyzed in a modified aPTT assay in which we determined the ability of purified fXII to restore normal clotting to fXII-deficient plasma. The results in Fig. 4B show that deglycosylated fXII fully restored clotting when added to fXII-deficient plasma, whether  it had been incubated with CpaA or not, providing further support that deglycosylated fXII is resistant to CpaA. Interestingly, mutations in the CpaA-sensitive region of fXII have been identified in patients with hereditary angioedema type III (HAEIII) including two missense mutations at Thr309 and deletion of residues Lys305-Ala321 (29). Inheritance of either Lys or Arg substitutions at Thr309 (thus removing the O-linked glycosylation site) or the deletion are associated with an autosomal dominant form of this disorder in multiple families and leads to increased contact-mediated autoactivation of fXII that results in episodes of swelling of skin, life-threatening upper airway obstruction, and/or abdominal pain (28). When normal plasma (plasma from healthy individuals) and plasma samples from two different patients heterozygous for the fXII-Thr309Lys substitution were incubated with CpaA and then assayed for clotting in the aPTT assay, there was a significant increase in clotting time for normal plasma samples compared to patient plasma samples, suggesting that fXII-Thr309Lys is resistant to inactivation by CpaA (Fig. 5A). In subsequent experiments, we compared cleavage of fXII-Thr309Lys and fXII in patient plasma. While fXII-Thr309Lys was detected with reduced efficiency, the results indicated that both normal fXII and fXII-Thr309Lys in the patient plasma are cleaved by CpaA at the first site to generate fXII fragment 2 and fXII-Thr309Lys fragment 2 ( Fig. 5B; note the mutant fragment demonstrates increased mobility due to the nonglycosylated substitution at position 309). As was shown above, normal fXII is then cleaved a second time to generate fXII fragment 3. However, the fXII-Thr309Lys fragment 2 remains uncleaved following incubation with CpaA ( Fig. 5B and C), which is consistent with our finding that the clotting time of patient plasma, which has approximately 50% wild-type fXII and 50% mutant fXII, is only partially extended by CpaA compared to NHP (Fig. 5A).
Other CpaA targets. FXII is not the only target of CpaA. Even though we did not observe an effect in the PT assay (Fig. S4A), which suggests that fV was neither activated nor inactivated by CpaA, fV is cleaved by CpaA (Fig. S4B). Perhaps there are common  (17). However, we did not observe cleavage of fibrinogen in human plasma (Fig. S7).
CpaA contributes to in vivo fitness in a bacteremia model. We have previously shown that the T2SS supports survival of the A. baumannii reference strain ATCC 17978 in a bacteremia model of leukopenic CBA/J mice (8). Here, we determined whether the T2SS also promotes the in vivo survival of the cpaA-positive A. baumannii strain AB031, which was recently isolated from the bloodstream of a 55-year-old female patient (30). We pooled WT AB031 and the kanamycin-marked ΔgspD mutant cells at a ratio of 1:1 and inoculated CBA/J mice with 10 7 colony-forming units (CFU) intravenously via the tail vein. Twenty-four hours later, the mice were euthanized, and spleens and livers were harvested, homogenized, and plated. In contrast to ATCC 17978 (4), AB031 did not require immunosuppression of mice for infection, and inoculation of 10 7 bacteria resulted in 10 5 CFU/liver/mouse 24 h postinoculation. The CFU of kanamycin-sensitive (WT AB031) and resistant (ΔgspD mutant) bacteria were counted, and competitive indices (CI) were calculated. With competitive indices significantly below 1, the mutant was outcompeted by the WT strain in both liver (CI ϭ 0.033) and spleen (CI ϭ 0.028), indicating that the T2SS contributes to in vivo fitness of AB031 (Fig. 6). Next we evaluated the role of CpaA in in vivo survival by analyzing the colonization of the gentamicin-marked ΔcpaA deletion mutant (17). The ΔcpaA deletion mutant was outcompeted by the WT strain in the liver (CI ϭ 0.045) and to a lesser extent in the spleen (CI ϭ 0.47) (Fig. 6), suggesting that CpaA is expressed and secreted in vivo and that it supports the survival of A. baumannii AB031 in the murine host. These results are consistent with the finding that dissemination of an A. nosocomialis cpaA mutant from the lungs in a pneumonia model of mice is reduced compared to WT A. nosocomialis when inoculated separately (16) and suggests that CpaA is an important Acinetobacter in vivo fitness factor.
CpaA cleaves and inactivates murine fXII with reduced efficiency. CpaA produced and secreted by the WT A. baumannii AB031 strain also interferes with clotting of murine plasma, although the efficiency in murine plasma appears to be reduced compared to human plasma (Fig. 7A). A factor assay similar to the one used in Fig. 2 was performed but with CpaA-treated normal mouse plasma used to reconstitute factor XII-deficient human plasma confirmed that fXII is also the target of CpaA in mouse plasma (Fig. 7B). To further investigate the reduced effect of CpaA on clotting in murine plasma, increasing amounts of CpaA were incubated with murine plasma, followed by aPTT analysis (Fig. 7C). While incubation of 1 l and 4 l of CpaA with human plasma for 20 min extended the clotting time from 40 s to 107 and 269 s (Fig. 3D), respectively,  (Fig. 7C). Even when the amount of CpaA and the incubation time of murine plasma with CpaA was increased to 14 l and 2 h, respectively, murine fXII was not fully cleaved by CpaA (Fig. 7D). These findings indicate that murine fXII is less sensitive to CpaA cleavage than human fXII.
CpaA cleaves human fXII expressed in mice. To determine whether differences in the primary sequences of human and mouse fXII or whether other differences between human and mouse plasma are responsible for the difference in susceptibility to CpaA, we transiently expressed human fXII in mice and tested it for cleavage by CpaA. The cDNA for human fXII was cloned into pLIVE, an in vivo expression vector that allowed for expression of human fXII from the mouse albumin promoter in the liver following hydrodynamic tail vein injection of the plasmid DNA. Plasma samples from two different mice were obtained before (day 0) and after DNA injection (either day 1 or day 3), and the human fXII was tested for sensitivity to CpaA cleavage. Samples were subjected to SDS-PAGE and immunoblotted with sheep anti-human fXII antibodies that do not cross-react with mouse fXII (see day 0 samples in Fig. 8). Using the same amount of CpaA previously used for cleavage of fXII in human plasma and incubating for 20 min, we found that human fXII expressed in mice was fully sensitive to CpaA, A. baumannii Inactivation of Factor XII ® suggesting that the difference in CpaA sensitivity between human and mouse fXII are likely due to differences in their primary amino acid sequences (Fig. 9).

DISCUSSION
Our studies focused on the putative function of the metalloprotease CpaA, which is secreted by the T2SS in A. baumannii. Coagulation factor fXII was recognized as a substrate for CpaA in both human and murine plasma. N-terminal sequencing and mass spectrometry of human fXII fragments generated following treatment with CpaA localized the cleavage sites to Pro279Thr280 and Pro308Thr309 located in the prolinerich domain of fXII. These findings suggest that CpaA has specificity for Pro and Thr at the P1 and P1' positions, respectively. Interestingly, both Thr280 and Thr309 are glycosylated residues, and deglycosylation of fXII prevents cleavage at both sites. Similarly, substitution of the glycosylated Thr309 with lysine protects fXII-Thr309Lys in plasma from patients with HAEIII from CpaA-mediated inactivation. If CpaA contributes to human disease, individuals with this mutation as well as the Thr309Arg substitution and the K305-A321 deletion may be partially protected from the effect of CpaA during A. baumannii infection (28,29,31,32).
Glycosylation generally protects proteins from proteolysis; however, in the case of fXII, glycosylation makes it a target for proteolysis, as deglycosylation of fXII prevents cleavage. This finding adds CpaA to a small, but growing list of bacterial proteases with specificity toward glycosylated residues. For example, IgA proteases produced by pathogens such as Streptococcus pneumoniae, Haemophilus influenzae, and Neisseria species specifically target proline-threonine and proline-serine peptide bonds in the glycosylated hinge region of human IgA1 (33). Another example, StcE from enterohemorrhagic Escherichia coli cleaves the glycosylated N-terminal domain of C1-esterase inhibitor (34).
The primary amino acid sequences of human and mouse fXII are 71% identical; however, the glycosylated proline-rich domain greatly differs between the two fXII FIG 8 Human fXII expressed in mice is sensitive to CpaA. Plasma from two different mice injected hydrodynamically with pLIVE expressing human fXII cDNA were collected on day 3 (mouse 1) and day 1 (mouse 2) and incubated without (Ϫ) or with (ϩ) CpaA-containing culture supernatants from ATCC 17978/pcpaAB. Control plasma samples were collected from each mouse prior to plasmid injection (day 0) and used as a negative control. Purified human fXII (hfXII; 10 ng) is included as a positive control on the left. A representative blot is shown.

FIG 9
Low sequence conservation between the proline-rich domain of human and mouse fXII. The primary sequence, including the N-terminal signal peptide of human fXII (hfXII) was aligned with that of mouse fXII (mfXII). The sequence between the conserved cysteines at positions 276 and 340 (highlighted in magenta) containing the proline-rich domain of hfXII is shown on the top. Note that this region is 18 residues shorter in mfXII (bottom). The sites cleaved by CpaA in hfXII are marked in green. species. Between Cys276 and Cys340 in human fXII, there is no significant homology observed with mouse fXII and the mouse sequence is 18 residues shorter (Fig. 9). While mouse fXII does have sites equivalent to Pro279Thr280 and Pro308Thr309, other differences in the proline-rich domain may be responsible for the reduced cleavage efficiency of mouse fXII by CpaA.
Our previous work demonstrated the advantage of possessing the T2SS for colonization of ATCC 17978 (8). We now show that this is also true for the AB031 strain. However, in contrast to ATCC 17978, neutropenia is not required for dissemination and colonization of AB031 in the liver and spleen, suggesting that AB031 is better able to resist host defense mechanisms. The importance of the T2SS in multiple strains of A. baumannii as well as A. nosocomialis for increased survival in vivo is indicative of a widespread role of the system and its substrates in the species as a whole. While the defect in colonization was equivalent for the spleen and liver for the T2SS mutant, there was a difference between the organs for the ΔcpaA mutant with the liver showing a greater difference than the spleen. This requires further investigation; however, it is interesting to note that the liver is the primary site of fXII synthesis (35), and it is possible that A. baumannii is exposed to higher levels of fXII in this organ, thus decreasing the competitiveness of the ΔcpaA mutant. An alternative explanation is that the WT and mutant strains occupy a much smaller space in the spleen and the mutant is complemented in trans with secreted CpaA from neighboring WT bacteria. However, the latter scenario is less likely, as a similar competition assay for WT and a ΔlipaA mutant to determine the contribution of a secreted lipase to in vivo fitness did not indicate a reduced requirement for lipase in the spleen (8).
The role of fXII is not yet fully understood, since while fXII deficiency demonstrates a significant defect in the aPTT assay of coagulation, it is not associated with bleeding (36). However, recent studies have shown that fXII promotes intravascular thrombus formation (21,37,38). CpaA may inactivate fXII to prevent the formation of intravascular clots that would otherwise trap A. baumannii during bacteremia. Without the hindrance of a clot, A. baumannii may disseminate and gain access to other organs. Another example of a pathogen that is hypothesized to evade clot-mediated immobilization is group A streptococcus (GAS). GAS produces streptokinase, which can degrade the major constituent of the clot, fibrin, through conversion of plasminogen to plasmin, thus enabling GAS to escape and disseminate (39). Cleavage of fXII by CpaA may also affect the kallikrein/kininogen pathway. fXIIa is the primary activator of prekallikrein to kallikrein which, in turn, liberates bradykinin from high-molecularweight kininogen, and in mice deficient in fXII, bradykinin levels are significantly reduced (40). Bradykinin is a peptide hormone with a wide array of roles. It promotes efflux of immune cells by increasing vascular permeability and vasodilation by binding to B-2 receptors (41). Bradykinin, in its metabolized form, binds B-1 receptors (41). Engagement of both the B-2 and B-1 receptors leads to release of immune and inflammation mediators that are immune cell dependent (42). By inactivating fXII, CpaA may reduce the generation of bradykinin, potentially reducing the recruitment of immune mediators, including neutrophils, and thus support survival of A. baumannii. In a very recently reported study of a sterile inflammation model, a direct role for the zymogen form of fXII in neutrophil trafficking was identified that is independent of fXIIa protease function or the kallikrein/kininogen pathway (22). Cleavage of fXII by CpaA could therefore affect the upregulation of neutrophil functions and prevent important processes such as chemotaxis and formation of neutrophil extracellular traps, which are critical for innate antibacterial processes. Finally, fXII is also associated with the classical pathway of the complement system as active fXII is capable of activating C1 esterase, and therefore, CpaA-mediated inactivation of fXII could also interfere with a branch of the acquired immune system.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Acinetobacter baumannii ATCC 17978 and AB031 and their isogenic mutants were cultured in Luria-Bertani (LB) broth at 37°C. Carbenicillin (100 g/ml) or kanamy-A. baumannii Inactivation of Factor XII ® cin (50 g/ml) was used for plasmid maintenance. The bacterial strains and plasmids used in this study are given in Table 2.
Construction of ⌬gspD strains. Chromosomal DNA isolated from the AB031 strain was used as the template for PCR. PCRs were carried out with Phusion DNA polymerase (Thermo Fisher). Primers were synthesized by IDT Technologies.
To generate the AB031ΔgspD strain, we used a previous plasmid construct, pCVDΔgspD (8). This construct was conjugated from the Escherichia coli strain SY327pir into the A. baumannii AB031 strain. Transconjugates in which pCVDΔgspD had recombined into the A. baumannii genome were selected on LB agar containing carbenicillin. To select for the second recombination event, individual colonies were cultured overnight in LB broth, diluted, cultured to late log phase, and spread on LB agar containing 3% sucrose. Sucrose-and kanamycin-resistant and carbenicillin-sensitive isolates were considered positive for the recombination event.
Construction of pcpaA and pcpaAB plasmids. The cpaA and cpaB genes were amplified from chromosomal AB031 DNA using primers 5=-GAGGAGCTCTGGTTTGCTAACCTGC-3= and 5=-GAGGCATGC TCTCTACCAGAACCGTT-3=. The 2,532-bp product was digested with SacI and SphI and ligated into a low-copy-number, broad-host vector pMMB67EH to make pcpaAB. The construct was verified by sequencing and conjugated from the E. coli strain MC1061 into WT and ΔgspD A. baumannii 17978 strains to make WTpcpaAB and ΔgspDpcpaAB.
In vivo competition assay. Competition assays were carried out as described previously (8). Briefly, overnight cultures of strains AB031 and AB031ΔgspD or AB031ΔcpaA were suspended in phosphatebuffered saline (PBS). Inocula of 10 7 cells of equal amounts of AB031 and AB031ΔgspD were coadministered via tail vein injection into 8-week-old female CBA/J mice (Jackson Laboratory). Spleens and livers were removed after mice were euthanized 24 h postinoculation. Organs were homogenized in PBS and plated onto LB agar with and without kanamycin. CFU counts were determined and used to calculate the competitive index (CI) as follows: CI ϭ ([mutant CFU/WT CFU]/[mutant input CFU/WT input CFU]).
Concentration of CpaA in supernatant. Cultures (10 ml) of each A. baumannii strain were grown for 16 h in LB. The supernatant was separated from the cells by centrifugation at 1,250 ϫ g for 10 min and filter sterilized. The supernatants were concentrated 100ϫ using Amicon Ultra Centrifugal filters with a 30-kDa cutoff. To determine the approximate concentration of CpaA present in concentrated supernatant of 17978/pcpaAB cultures, the supernatant was subjected to SDS-PAGE and Coomassie blue staining and the amount of CpaA was compared to known amounts of BSA, a protein of similar size. Activated partial thromboplastin time (aPTT) assay. Concentrated cell-free supernatants were mixed with normal (healthy) human pooled plasma samples (George King) for 20 min at 37°C. An equal volume of aPTT reagent (Pacific Hemostasis) was added. After incubation for 5 min, CaCl 2 (final 10 mM) was added. Clot times were determined using an Amelung KC4A Micro Coagulation analyzer.
To identify the target of CpaA, a modified aPTT assay was performed by first incubating concentrated AB031 supernatant with normal plasma. After incubation, the plasma was diluted 1:100 and added to human plasma samples derived from donors deficient in congenital coagulation factor (George King), and an aPTT assay was performed as described above. Replacement of cpaA with aacC1 (Gm r ) 17 AB031ΔgspD Replacement of gspD with aph-3 (Km r ) This study ATCC 17978 Wild type for T2SS ATCC ATCC 17978ΔgspD Replacement of gspD with aph-3 (Km r ) 8 described (43). Blood was collected from the retro-orbital plexus or the inferior vena cava and anticoagulated with sodium citrate at a 1:9 (vol/vol) ratio. Statistical analysis. Ordinary one-way ANOVAs were performed on the aPPT and modified aPTT tests. A Wilcoxon signed rank test was used to analyze the in vivo murine data.
Ethics statement. All mouse care and procedures were performed according to the protocols (PRO00007111, PRO00005191, and PRO00007879) approved by the Institutional Animal Care and Use Committee at the University of Michigan. These protocols are in complete compliance with the guidelines for humane use and care of laboratory animals mandated by the National Institutes of Health.
Patients. The plasma samples from patients heterozygous for fXII-Thr309Lys were obtained with informed consent at Grenoble University Hospital Centre, Grenoble, France.