Peptidoglycan-Modifying Enzyme Pgp1 Is Required for Helical Cell Shape and Pathogenicity Traits in Campylobacter jejuni

The impact of bacterial morphology on virulence and transmission attributes of pathogens is poorly understood. The prevalent enteric pathogen Campylobacter jejuni displays a helical shape postulated as important for colonization and host interactions. However, this had not previously been demonstrated experimentally. C. jejuni is thus a good organism for exploring the role of factors modulating helical morphology on pathogenesis. We identified an uncharacterized gene, designated pgp1 (peptidoglycan peptidase 1), in a calcofluor white-based screen to explore cell envelope properties important for C. jejuni virulence and stress survival. Bioinformatics showed that Pgp1 is conserved primarily in curved and helical bacteria. Deletion of pgp1 resulted in a striking, rod-shaped morphology, making pgp1 the first C. jejuni gene shown to be involved in maintenance of C. jejuni cell shape. Pgp1 contributes to key pathogenic and cell envelope phenotypes. In comparison to wild type, the rod-shaped pgp1 mutant was deficient in chick colonization by over three orders of magnitude and elicited enhanced secretion of the chemokine IL-8 in epithelial cell infections. Both the pgp1 mutant and a pgp1 overexpressing strain – which similarly produced straight or kinked cells – exhibited biofilm and motility defects. Detailed peptidoglycan analyses via HPLC and mass spectrometry, as well as Pgp1 enzyme assays, confirmed Pgp1 as a novel peptidoglycan DL-carboxypeptidase cleaving monomeric tripeptides to dipeptides. Peptidoglycan from the pgp1 mutant activated the host cell receptor Nod1 to a greater extent than did that of wild type. This work provides the first link between a C. jejuni gene and morphology, peptidoglycan biosynthesis, and key host- and transmission-related characteristics.


Introduction
Campylobacter jejuni is a helical, highly motile, Gram-negative e-Proteobacterium and a prevalent zoonotic organism existing asymptomatically in the intestinal tract of birds and other animal species [1][2][3]. However, ingestion of as few as 500 bacteria can result in human disease [4]. C. jejuni is the leading cause of foodborne gastroenteritis in the developed world, causing an acute self-limiting infection of varying severity that can give rise to severe complications such as inflammatory bowel disease, reactive arthritis, and Guillain-Barré syndrome (GBS) [5].
C. jejuni lacks many of the frequently identified virulence factors encoded by other enteric pathogens such as pili, enterotoxins, and specialized secretion mechanisms [6,7]. Genes affecting fundamental aspects of C. jejuni biology in hosts other than humans, such as stress survival, transmission, and asymptomatic colonization, also affect virulence in disease models. We found that C. jejuni strain 81-176 binds calcofluor white (CFW), a compound that reacts with b1-3 and b1-4 carbohydrate linkages and fluoresces under long wave UV light [8,9]. The carbohydrate responsible for CFW reactivity in C. jejuni has not yet been identified, although it was previously shown not to be one of the well-characterized surface polysaccharides expressed by C. jejuni: the capsular polysaccharide, lipooligosaccharide (LOS), N-linked glycoproteins, or O-linked flagellar glycoproteins [10]. C. jejuni mutants with altered CFW reactivity can be readily identified in screens. All C. jejuni CFW hyper-or hyporeactive mutants characterized to date exhibit changes in pathogenesis, virulence, fundamental, and/or stress survival phenotypes ( [10,11]; E. Frirdich and E. C. Gaynor, unpublished). For instance, CFW hyper-reactive mutants overproduce biofilms, while hypo-reactive mutants are biofilmdefective. Other attributes associated with altered CFW reactivity have included defects in colonization, host cell interactions, cell envelope components, and stress survival. A CFW hypofluorescent mutant with a lesion in a novel peptidoglycan peptidase gene (pgp1) serves as the basis of this study.
The helical shape of C. jejuni has long been postulated to be critical for pathogenic attributes such as the ability to burrow through the mucosal layer during infection of zoonotic and human hosts. However, genetic components involved in modulating C. jejuni morphology had not previously been identified. Morphology is maintained in most bacteria by the peptidoglycan (PG) layer [12]. PG is composed of glycan chains consisting of b1-4 linked Nacetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) residues that are cross-linked by short peptides. PG synthesis is a highly regulated process [13,14] taking place at the bacterial inner membrane, where lipid II precursors are polymerized by glycosyltransferase and DD-transpeptidase reactions. In addition to the synthetic enzymes, PG hydrolases are required to cleave bonds in PG or PG fragments for insertion of nascent PG into the mature layer, regulation of cell wall growth, cell separation, PG turnover and recycling, cell lysis, and the release of PG fragments in host-pathogen interactions [15]. Almost every glycosidic and amide bond in PG can be cleaved by one or more specific PG hydrolases, and this redundancy makes it difficult to assign a specific function to a hydrolase [15,16]. It has been speculated that murein synthases and hydrolases are part of a multienzyme complex for PG assembly [16], and several interactions between PG enzymes have been reported [14].
Campylobacter PG contains meso-diaminopimelic acid (meso-Dap) in its peptide side chains and is modified by O-acetylation [17,18], but the detailed muropeptide composition had not previously been elucidated. Bioinformatic analyses identified 3 putative PG synthases in C. jejuni: the penicillin-binding proteins (PBPs) PBP1A (cjj81176_0536), PBP (cjj81176_0550), and PBP2 (cjj81176_0680), but there are no predicted homologs of low-molecular weight PBPs such as DD-endo-or carboxypeptidases. C. jejuni does have homologs of the Csd1 and CcmA endopeptidases and of the Csd3/HdpA endo/carboxypeptidase discovered recently in Heli-cobacter pylori [19,20]. Deletion of each of these genes in H. pylori resulted in curved-rod morphologies with reduced levels of PG cross-linking [19], but preliminary evidence indicates that these enzymes may have a different function in C. jejuni (E. Frirdich and E. C. Gaynor, unpublished).
Here, we report the first C. jejuni gene, pgp1, to be identified with a role in helical morphology. Structural PG analyses of C. jejuni wild-type 81-176 and Dpgp1 mutant strains showed striking alterations in the mutant, with enzyme assays confirming Pgp1 as a novel DL-carboxypeptidase. Furthermore, using the Dpgp1 mutant with a straight morphology, the importance of C. jejuni helical shape and PG in pathogenesis attributes was examined; we identified roles in motility, biofilm formation, chick colonization, and stimulation of host cell proinflammatory mediators Nod1 and IL-8.

Results
CFW screen for hypofluorescent mutants identifies the C. jejuni 81-176 gene pgp1 As noted, screening for C. jejuni hypofluorescent mutants was previously shown to identify genes involved in pathogenesis-related phenotypes [10]. However, as the Tn7 transposon (Tn) used previously for mutant library construction was reproducibly found to have inserted in intergenic regions (M. K. McLennan and E. C. Gaynor, unpublished), a new random Tn library was generated as part of this study, using the highly efficient mariner system of in vitro Tn mutagenesis developed for C. jejuni [21]. The mariner library was plated onto CFW-containing plates and, from approximately 10, 000 colonies screened, 400 hypofluorescent mutants were isolated. Of the Tn insertions mapped, 8 were in distinct regions of the gene cjj81176_1344 ( Figure 1; the CFW hypofluorescent phenotype is described below). The 1344 gene was named pgp1 (peptidoglycan peptidase 1) to describe its function and identification as the first C. jejuni PG peptidase to be characterized.
The pgp1 gene product is highly conserved in mainly helical and vibrioid bacteria, primarily within the eand d-Proteobacteria but also in a few extremophiles outside the e-Proteobacteria (Table  S1). The H. pylori homolog is described by Sycuro et al. [22]. Pgp1 was annotated as a putative periplasmic protein. However, bioinformatics using conserved domain searches and the threading program PHYRE identified an N-terminal signal peptide, a zinc binding site, and metallocarboxypeptidase catalytic residues and folds. Pgp1 contains a conserved domain at its N-terminus similar to that of the M14 family of metallocarboxypeptidases.
Non-polar insertional deletion of pgp1 and complementation of a pgp1 mutant To explore functional consequences of loss of Pgp1 function in C. jejuni, a non-polar pgp1 targeted deletion strain was constructed (designated Dpgp1 above the gene cluster in Figure 1). To verify lack of polar effects on the downstream gene, an insertional mutation was also created in 1343. The D1343 mutant exhibited wild-type phenotypes for shape, motility, biofilm formation and CFW reactivity (data not shown; these phenotypes for Dpgp1 are described below). The level of 1343 mRNA was also determined to be identical in Dpgp1 and in the wild-type strains by semiquantitative RT-PCR (data not shown) The pgp1 gene is located in the middle of a putative operon, and the location of the pgp1 promoter is unclear from sequence analysis ( Figure 1). Therefore, complementation was first attempted by expressing the pgp1 gene from the cat promoter of the vector pRRC (Cm R ) [23] integrated into an rRNA spacer region of

Author Summary
Bacterial cell shape is dictated by the composition of the cell envelope component peptidoglycan. Some important pathogens have a characteristic helical corkscrew morphology that may help them burrow into mucus overlaying cells to initiate colonization and pathogenicity. One example is Campylobacter jejuni, the leading cause of bacterial-induced diarrheal disease in the developed world. Direct evidence supporting the hypothesis that C. jejuni shape is related to its pathogenicity traits has not previously been provided. We identified a gene encoding a peptidase modifying peptidoglycan that is essential for maintaining the C. jejuni corkscrew shape. We can now connect a C. jejuni gene with morphology and peptidoglycan biosynthesis. Loss of this gene was also found to affect pathogenic attributes such as chicken colonization, biofilms, motility, and activation of host inflammatory mediators. In addition, this is the first study to thoroughly characterize C. jejuni peptidoglycan structure and to identify a gene involved in peptidoglycan maintenance. Our findings highlight an emerging theme in bacterial pathogenesis research: the connection between bacterial cell biology and pathogenesis. Finally, our characterization of C. jejuni cell shape and peptidoglycan provides a starting point for further work in this area in C. jejuni and other bacteria with curved and helical morphologies.
Dpgp1. This construct contained a similar region of pgp1 as shown for pEF20 in Figure 1 (pEF20 will be described further below) and did not complement Dpgp1 (data not shown). However, integration of this same construct into the C. jejuni wild type strain also resulted in morphological and phenotypic alterations (described below), suggesting that pgp1 copy number was likely important for complementation, and that the level of pgp1 expression from the cat promoter was higher than optimal. To generate a complementing strain expressing pgp1 from its native promoter, increasing amounts of the region upstream of pgp1 were cloned with pgp1 into pRRC. These regions were cloned in the opposite direction as cat (to avoid overexpression from the cat promoter) and integrated into an rRNA spacer region of Dpgp1. These constructs yielded varying degrees of complementation, with the pEF35R (Figure 1) construct complementing nearly all Dpgp1 phenotypes to wild-type levels. The complemented strain was verified by PCR and designated Dpgp1c.
The pgp1 gene is required for helical morphology, and Dpgp1 is deficient in CFW reactivity, motility, and biofilm formation Since the pgp1 gene product is highly conserved in helical and vibrioid bacteria, the morphology of the Dpgp1 mutant was examined. The Dpgp1 mutant displayed a striking change compared to wild type and had lost the characteristic helical shape (Figure 2A), adopting a straight morphology ( Figure 2B). Complementation restored the helical shape in approximately 95% of the cells ( Figure 2C).
Consistent with the phenotype of the mariner insertion mutants, the Dpgp1 mutant was hyporeactive to CFW, with wild-type reactivity restored by complementation ( Figure 2D). The Dpgp1 mutant displayed no obvious flagellar structural defects ( Figure 2B), but did exhibit a slight motility defect after incubation for 20 h on 0.4% agar plates, producing halos that were, on average, about 82.5% of wild type ( Figure 2E). Motility was restored to wild-type levels in the complemented strain. Biofilm levels, as assessed by a crystal violet assay, were approximately 1.2-, 1.5-and 1.6-fold lower in the Dpgp1 mutant compared to wild type at days 1, 2, and 3, respectively. Biofilm production was partially restored in Dpgp1c ( Figure 2F). No differences between wild type and Dpgp1 were observed for growth, stress survival, capsule and LOS migration on acrylamide gels, membrane protein composition, and sensitivity to antimicrobial compounds (Table S2).

Pgp1 overexpression also affects helical morphology
To assess the effects of pgp1 overexpression, pgp1 was expressed from either the cat or aphA-3 promoter in pRRC (data not shown) or pRRK (pEF20 in Figure 1), respectively, and integrated at an rRNA spacer region of wild-type C. jejuni. Both pRRC and pRRK derivatives had the same effects on wild type, so only the results for pEF20 are shown. The pgp1 overexpressing strain displayed an altered cell shape in approximately 50% of the population, producing kinked and straight cells among the helical cells ( Figure 3A). This strain also exhibited reduced motility and a defect in biofilm formation ( Figure 3B & C). Reverse transcriptasequantitative PCR (RT-qPCR) confirmed that the levels of pgp1 mRNA were 5.1-fold higher in the overexpressing strain than in wild type. In addition, there was a 1.2-fold increase in pgp1 mRNA levels in the complementing strain Dpgp1c in comparison to wild type, which may explain the partial complementation of some phenotypes. Expression of pgp1 in E. coli had no effect on cell shape (data not shown).

Peptidoglycan analysis
Muropeptide profile of wild-type C. jejuni 81-176 reveals numerous stem peptide species. Given the shape phenotype of the pgp1 mutant strains and the putative peptidase domain of Pgp1, we hypothesized that Pgp1 cleaves the PG stem peptide. Since C. jejuni PG had not been fully characterized, we first analyzed the muropeptide composition of wild type. The HPLC analysis is shown in Figure 4A, and the structure of individual muropeptides ( Figure 4E) was confirmed by co-elution with known muropeptides from E. coli [24] and by LTQ-FT-MS/MS (Table 1; raw data in Tables S3 and S4) [25].
The most abundant muropeptides were the monomers with di-, tri-and tetrapeptides, as well as dimeric 4-3 cross-linked TetraTetra and TetraTri species. Only one very minor peak with a pentapeptide side chain was detected in C. jejuni, a TetraPenta dimer with a Gly residue at position 5, as opposed to the more common D-Ala. Muropeptides with 3-3 crosslinks or Lys-Arg residues indicating bound lipoprotein were not detected and thus are either not present or present in very small amounts. The degree of cross-linking at 29.5% is consistent with the value of 30% published previously [17]. The average glycan chain length determined by the fraction of 1,6-anhydromuramic acid containing muropeptides from chain ends was 14.7 dissacharide units. The PG glycan backbone of two different strains of C. jejuni was reported to be modified with O-acetyl groups at levels of 55.7% and 62% relative to MurNAc content [18]. The fact that we observed lower levels of O-acetylation in our analyses (5.7%) is likely due to the loss of most of the labile O-acetyl groups in the standard alkaline muropeptide reduction procedure applied in this study. the approximate location of this insertion is shown above the gene cluster and is denoted by the Dpgp1 strain designation. The regions cloned into the integrative vectors pRRC (pEF35R; Cm R ) or pRRK (pEF20; Km R ) used for complementation and overexpression, respectively, are shown below the gene cluster. An R after the plasmid name indicates that the region is cloned in the opposite direction as the antibiotic resistance cassette promoter. doi:10.1371/journal.ppat.1002602.g001 Loss of pgp1 and Pgp1 overexpression result in altered PG muropeptide profiles. PG from the Dpgp1 mutant, complement (Dpgp1c), and pgp1 overexpression (81-176+pgp1) strains was isolated, and the muropeptide profiles were analyzed by HPLC as above ( Figure 4A-D, Tables 1 and S4). The Dpgp1 mutant exhibited a number of changes from wild type, most significantly a striking increase in tripeptides and reduced dipeptides and tetrapeptides. The overexpression strain (81-176+pgp1) exhibited the opposite profile from Dpgp1, with a decrease in total tripeptides, an increase in total dipeptides, and to a lesser extent, an increase in tetrapeptides and Gly-containing pentapeptides. Complementation restored many of the changes in Dpgp1 to near-wild-type levels; however, pgp1 may be modestly overexpressed in Dpgp1c based on some similarities to the 81-176+pgp1 profile. The overall degree of crosslinking did not vary significantly between the different strains. Together with sequence analysis (see above), the PG data suggests Pgp1 functions as a DL-carboxypeptidase, hydrolyzing the D-Glumeso-Dap bond in tripeptides to form dipeptides. Loss of most Oacetyl groups in the PG preparation techniques used in this study precluded exact quantification of this PG glycan backbone modification in the Dpgp1 mutant. As PG O-acetylation confers resistance to lysozyme [26,27], it would be expected that if Pgp1 affects O-acetylation, the Dpgp1 mutant would display a change in lysozyme sensitivity. In the presence of the metal chelator EDTA to permeabilize the outer membrane, no difference in MIC for lysozyme was observed between wild type and Dpgp1 strains (Table  S2). Biochemical analyses (described below) confirming Pgp1 as a DL-carboxypeptidase cleaving the PG stem peptide further suggested no functional involvement of Pgp1 in O-acetylation, thus additional experiments to assess this were not pursued.
Pgp1 is a metal-dependent DL-carboxypeptidase cleaving monomeric disaccharide tripeptides to dipeptides. Pgp1 protein was expressed and purified from E. coli ( Figure 5A) and incubated with purified Dpgp1 PG to assess the biochemical function of the protein. The resultant PG was analyzed as above, and the muropeptide profiles were compared to that of Dpgp1 ( Figure 5B-D, Tables 1 and S4). In the presence of ZnCl 2 , the addition of Pgp1 resulted in a striking increase in monomeric dipeptides with a concomitant decrease in monomeric tripeptides, characteristic of a metal-dependent DL-carboxypeptidase ( Figure 5C; Table 1). This activity was eliminated in the absence of ZnCl 2 with added EDTA, establishing the dependence of Pgp1 on divalent ions ( Figure 5D). This confirmed Pgp1 as a metal-dependent DL-carboxypeptidase cleaving monomeric disaccharide tripeptides to dipeptides ( Figure 5E).

The effect of pgp1 on host-related phenotypes
Dpgp1 is defective for in vivo chick colonization. A C. jejuni chick colonization model was used to determine the role of Figure 2. A pgp1 mutant has a straight morphology and defects in CFW reactivity, motility, and biofilms. Negatively stained TEM images of A, the helical C. jejuni 81-176 strain; B, the straight Dpgp1 strain with intact flagella; and C, the complemented strain Dpgp1c with restored helical morphology in 95% of the population. D, Dpgp1 is hypofluorescent after 48 h of growth on plates containing 0.002% CFW which is restored by complementation. E, Dpgp1exhibits a slight motility defect, as assayed by measuring halo diameters in soft agar plates. Standard error of the mean was calculated from 12 measurements. F, Dpgp1 is defective for biofilm formation, partially complemented in Dpgp1c. Biofilm formation was assessed by crystal violet staining of standing cultures in borosilicate tubes and quantification of dissolved crystal violet at 570 nm. Standard errors of the mean were calculated from triplicate cultures and are representative of three independent experiments. The asterisk (*) indicates a statistically significant difference using the unpaired Student's t-test, with * or ** indicating p,0.05 or p,0.01, respectively. Numerous other phenotypes showed no difference from wild type (Table S2). doi:10.1371/journal.ppat.1002602.g002 pgp1 in vivo. One-day old chicks were infected orally with a dose of 10 4 C. jejuni wild-type 81-76, Dpgp1, or Dpgp1c, and cecal contents were assessed for C. jejuni after 6 days. The Dpgp1 mutant exhibited a statistically significant (p 0.0009) 2.9-log decrease in average levels of colonization compared to wild type ( Figure 6A). Complementation restored colonization levels to within 0.5-log of wild type ( Figure 6A). As a control to explore whether the slight motility defect of Dpgp1 may be responsible for the reduction in chick colonization, we also infected chicks with a DcarB mutant [10]. carB encodes carbamoylphosphate synthase and is directly downstream of an intergenic Tn7 insertion in a CFW hypofluorescent mutant (dim13) isolated in a previous screen [10]. A DcarB mutant had wild-type helical morphology, but displayed a slight motility defect of approximately 86.0% that of wild-type ( Figure S1) and a hypofluorescent phenotype on CFW similar to Dpgp1 [10]. However, unlike Dpgp1, the DcarB mutant showed no statistically significant change in chick colonization ( Figure S1).  Figure  S2). This was also seen at 5 h and 9 h (data not shown), but not at 21 h post gentamicin treatment. No differences in IC survival were observed in either macrophage cell line tested (data not shown).
Because of the modest motility defect of Dpgp1 in soft agar, we also wished to assess whether attachment and invasion might be altered in a tissue culture medium of higher viscosity. MEM supplemented with carboxymethylcellulose (CMC) was previously used to demonstrate increased host cell attachment and invasion for some strains of C. jejuni in a higher viscosity medium [28]. INT407 cells were infected with C. jejuni 81-176 and Dpgp1 in MEM containing CMC. No significant difference in invasion was seen after 1 h or 3 h with the addition of 0.6%, 1% or 2% CMC ( Figure S2). Furthermore, CMC addition decreased the total levels of invaded bacteria for both the wild type and mutant strains.
PG changes in C. jejuni Dpgp1 result in an altered epithelial cell Nod1 response. PG is recognized by Nod-like receptors of the innate immune system. To test whether the PG changes resulting from a pgp1 mutation altered Nod receptor stimulation, we measured expression of a NF-kB lgk luciferase reporter transfected in human embryonic kidney HEK293T cells along with either the human (h)Nod1, mouse (m)Nod1, or human (h)Nod2 receptor ( Figure 6B-D). We found that PG from the Dpgp1 mutant produced a statistically significant increase in mNod1 activation compared to wild type. In contrast, Dpgp1 mutant PG did not produce a statistically significant change in stimulation of hNod1, despite a modest increase from wild type. C.   (Table S3) to determine the structures shown in E. G, N-acetylglucosamine; M, reduced N-acetylmuramic acid; L-Ala, L-alanine; D-iGlu, D-isoglutamic acid; D-Glu, D-glutamic acid; meso-DAP, meso-jejuni wild type PG stimulated Nod2; however, there was no statistically significant difference in activation levels between the wild type and Dpgp1 mutant PG.
Secretion of interleukin-8 (IL-8) is increased by exposure of epithelial cells to Dpgp1. It was previously shown that epithelial cells produce IL-8 in response to C. jejuni infection [29].
To examine the effect of the Dpgp1 mutation on IL-8 secretion, the INT407 human epithelial cell line was infected with C. jejuni wild type and Dpgp1 strains, and the levels of IL-8 secreted into the supernatant were measured by ELISA ( Figure 6E). The data shown in Figure 6E are representative of three separate experiments. IL-8 secretion from cells exposed to Dpgp1 after 8 h was reproducibly higher than wild type, and in some experiments was statistically significant (not shown). After 24 h, release of IL-8 in response to Dpgp1 was 1.5-fold higher than wild type and statistically significant in every experimental trial.

Discussion
Bacterial cell shape programs and their impact on biology are becoming increasingly well-characterized for model organisms like E. coli and Caulobacter crescentus [30,31]. However, studies linking bacterial shape and pathogenesis are in their infancy. This is the first report to address the morphogenesis program of the prevalent foodborne pathogen C. jejuni. Furthermore, apart from three annotated uncharacterized PBPs, no other C. jejuni PG-modifying enzymes have been identified. This is in spite of the fact that, as our study demonstrates, C. jejuni produces numerous muropeptide diaminopimelic acid; Gly. Glycine; Ac, O-acetyl groups at the C-6 hydroxyl group of MurNAc; Anh, 1,6-anhydro group at MurNAc. The asterisk (*) indicates that it is not known on which MurNAc residue the modification occurs. doi: 10 species indicative of additional unidentified and potentially novel PG hydrolases. Our identification and characterization of pgp1 provides the first link between cell shape and PG in C. jejuni. Pgp1 is conserved primarily in bacteria with curvature (Table S1). Although functions have not yet been ascribed for other putative shaperelated proteins in C. jejuni, it is of note that helical eand dproteobacteria also typically contain the elongase and cytoskeletal components MreB, MreC, and RodA, but not MreD or RodZ. As such, unique PG and morphogenesis programs are likely to be identified both within and among these organisms, which in turn will lend new insight into their biology and pathogenesis.
Enzyme assays demonstrated that Pgp1 is a metal-dependent DLcarboxypeptidase cleaving monomeric disaccharide tripeptides to disaccharide dipeptides. This is consistent with Pgp1 harboring an N-terminal M14 metallocarboxypeptidase domain present in two other PG dipeptidyl peptidases, E. coli MpaA and Bacillus sphaericus ENP1, which cleave tri-and tetrapeptides to dipeptides [32][33][34][35]. This also correlates with the reduction in dipeptides and increase in tripeptides in Dpgp1 PG. Decreased tetrapeptides in Dpgp1 may be an indirect consequence of the increased level of tripeptides affecting activity of DD-carboxypeptidases, or an indirect effect on other PG hydrolases via disruption of a putative enzyme complex. Efforts to identify Pgp1 binding partners that would support the latter hypothesis are underway. Our muropeptide analysis of C. jejuni 81-176 wild-type PG also revealed a muropeptide composition expected for a Gram-negative species, with several interesting differences from the related helical organism H. pylori [19,36] and the rod-shaped E. coli [14,24] (Table S5). For instance, unlike H. pylori, C. jejuni contains very low levels of pentapeptides, suggesting high DD-carboxypeptidase activity. C. jejuni PG also has a higher level of cross-linking than either other species, and a relatively short average glycan chain length.
With pgp1 being the first C. jejuni gene directly linked with helical morphology, it allowed for the first direct comparison between a helical wild type strain and an isogenic targeted rod-shaped mutant in critical aspects of C. jejuni pathogenesis and survival. The modest motility defect of the rod-shaped Dpgp1 mutant may in part account for the modest biofilm defect [37][38][39]. Further corroborating the pgp1-biofilm connection are our observations that pgp1 is up-regulated in the hyperbiofilm-forming DcprS strain (S. Svensson and E. C. Gaynor, unpublished) and that Dpgp1 is hyporeactive to CFW [10]; the latter observation in turn continues to support a link between C. jejuni CFW reactivity and key biological and pathogenic processes. Isolated PG, which has a b1-4 disaccharide backbone, binds to CFW (E. Frirdich and E. C. Gaynor, unpublished), suggesting that changes in PG architecture in Dpgp1 may account for decreased CFW reactivity. However, as PG isolated from both wild type and mutant strains bound CFW equally well, a related explanation is that PG alterations affect the accessibility of CFW to binding sites on periplasmic PG molecules.
The Dpgp1 mutant also allowed us to explore roles for shape and PG composition on host interactions. Surprisingly, the Dpgp1 mutant was not defective for host cell attachment or invasion in vitro and even displayed a slight increase in short-term intracellular survival. Increasing the viscosity of the media did not affect the ability of the mutant to invade epithelial cells ( Figure S2). It is of note that although attachment and invasion to host cells increased for some C. jejuni strains in higher viscosity medium [28], this was not the case for our highly invasive wild-type strain 81-176, suggesting varying invasion programs among different C. jejuni strains.
Intracellular C. jejuni was previously shown to induce host epithelial cell inflammatory responses via activation of the cytoplasmic Nod receptors [40,41], which recognize distinct PG muropeptide molecules. The minimal molecule recognized by hNod2 is muramyl dipeptide (MDP), a structure common to PG from both Gram-negative and Gram-positive organisms [42]. hNod1 and mNod1 recognize DAP-containing muropeptides restricted to Gram-negative organisms. hNod1 exhibits a preference for tripeptides and mNod1 for tetrapeptides; however mNod1 also, to a lesser degree, recognizes DAP-containing tripeptides ( [43]; J. Lee and S. Girardin, unpublished). Luciferase assays are a sensitive method to probe the capacity of a PG preparation to trigger Nod1 or Nod2 and provide complementary information to PG compositional data by HPLC and MS analyses. Using these assays, only mNod1 exhibited a statistically significant difference in activation by Dpgp1 PG compared to wild type. It is possible that C. jejuni modifies its PG in a manner that affects Nod signaling, as modifications of the PG backbone have been shown to affect how PG fragments are sensed by host cell receptors [44]. This poses an interesting question for future work. Our data also indicate that wild-type PG can activate hNod2, consistent with findings by Al-Sayeqh et al. (2010) who also showed NF-kB activation in HEK293T cells transfected with Nod2 and infected with live C. jejuni. While Zilbauer et al. (2007) did not observe C. jejuni activation of Nod2, this may reflect differences between cell lines used in these assays, and/or differences in reporter sensitivity.
Nod activation is an important aspect of innate immunity against C. jejuni [40,41]; however, the mechanism by which C. jejuni Figure 6. The role of cell shape and PG composition on host related phenotypes. A, Dpgp1 is defective for chick colonization. Each point represents the log CFU/g cecal contents of an individual chick 6 days following infection with 10 4 CFUs of the indicated C. jejuni strains. The geometric mean is denoted by a black bar. B-D, to assay the ability of C. jejuni wild type and Dpgp1 PG to activate Nod proteins, human embryonic kidney cells (HEK293T) were co-transfected with either C. jejuni 81-176 or Dpgp1 PG at 0.1 mg/mL, vectors for a NF-kB luciferase reporter, and either human Nod1 (B, hNod1), mouse Nod1 (C, mNod1) or human Nod2 (D, hNod2). Nod activation was determined by measuring the activity of the NF-kB luciferase reporter in comparison to the non-stimulated (NS) negative control. Positive controls used were TriDAP, FK565, and MDP. Data represent the mean 6 SEM of three independent experiments. E, Deletion of pgp1 increases IL-8 secretion in the INT407 epithelial cell line. Quantification of IL-8 levels was performed by ELISA. Data represent the mean 6 SEM of three independent experiments. The asterisk (*) indicates a statistically significant difference using the unpaired Student's t-test, with * or *** indicating p,0.05 or p,0.001, respectively. doi:10.1371/journal.ppat.1002602.g006 PG fragments may reach the cytosol to activate Nod1 and/or Nod2 has not been elucidated. C. jejuni can be internalized into intestinal epithelial cells, surviving within an intracellular membrane-bound compartment known as the C. jejuni-containing vacuole (CCV) [45]. An oligopeptide transporter expressed in the early endosome has been implicated in the transport of Nod ligands from the endosome to the cytoplasm in HEK293T cells [46]; a similar system could allow cytoplasmic delivery of PG fragments from C. jejuni in the CCV. Another possibility is that during growth in the extracellular environment C. jejuni may shed part of its PG, as shown for other bacteria [44,47,48]; released Nod1 and Nod2-stimulatory molecules could then be transported to the cytoplasm by oligopeptide transporters or taken up by phagocytosis or clathrin-dependent endocytosis and transported to the cytosol to activate Nods [44,[49][50][51]. Bacterial outer membrane vesicles may also play a role in delivering C. jejuni Nod ligands to the cytoplasm [52].
Production of IL-8 and other proinflammatory mediators by intestinal epithelial cells infected with C. jejuni is thought to be key to the development of diarrhea and clearance of infection. Zilbauer et al. (2007) suggested that C. jejuni Nod1 activation is the primary signaling event required for IL-8 expression in Caco-2 intestinal epithelial cells. However, other work indicates that C. jejuni-induced IL-8 secretion can be triggered by other pathways in addition to the activation of Nod1, such as through Toll-like receptors (TLRs) and a pathway independent of Nods and TLRs that has yet to be identified [40,[53][54][55][56][57][58]. Thus while the increase in IL-8 secretion in response to Dpgp1 may be due to the modest increase in hNod1 activation observed in our luciferase assays, it cannot be ruled out that deletion of pgp1 leads to a change in another, as-yet unidentified factor stimulating IL-8 expression. Future work is planned to address these hypotheses.
Our colonization data are in agreement with the longstanding hypothesis that the corkscrew morphology of C. jejuni is critical for burrowing into the mucus layer; however, the impact of shape and PG on colonization may be multifactorial. For instance, motility is a key factor in colonization [59][60][61][62][63][64], thus the Dpgp1 colonization defect may in part be due to its decreased motility. However, a DcarB mutant with a similar motility defect was not deficient for chick colonization ( Figure S1). When the strongly colonizing C. jejuni strain 305/94 was re-isolated from chickens, it displayed a more pronounced motility defect than Dpgp1 despite having normal flagella, and had lost its helical shape, exhibiting a similar straight morphology as Dpgp1 [65]. It is intriguing to hypothesize that not only shape, but also the underlying structure of the PG may play an important role in colonization. We have found that a straight phenotype can arise from various changes in muropeptides based on preliminary analysis of other straight mutants identified in our laboratory (E. Frirdich, J. Vermeulen, and E. C. Gaynor, unpublished). PG changes or changes resulting from the loss of pgp1 could cause as-yet uncharacterized alterations to the cell surface affecting the ability of C. jejuni to survive in the chicken cecum, or affect host cell interactions and stimulation of innate immune receptors. Our understanding of the chicken innate immune system and how it responds to C. jejuni is in its infancy. The chicken genome encodes Nod1 but not Nod2, although chickens possess an ortholog of the NLRP3/NALP3 Nod-like receptor that is similar to Nod2 and also binds MDP [66,67]. NLRP3 may substitute for Nod2 in chickens. Further insight into mechanisms by which C. jejuni survives commensally in the chicken cecum requires additional studies of these and other innate immune system factors.
Identification and characterization of pgp1 provides a critical first step in understanding how shape and PG modifications impact C. jejuni pathogenesis. However, the shape program itself in C. jejuni is likely to be somewhat complicated. For instance, to the best of our knowledge, Pgp1 is the only PG modifying enzyme to date where overexpression causes cell straightening, indicating that the precise dose of Pgp1 is important for maintenance of morphology. One interpretation of this finding is that the proper ratio of monomeric tripeptides to dipeptides, which is disrupted in both the Dpgp1 mutant and pgp1-overexpressing strains, may be required for proper shape determination. It is also possible that excess Pgp1 could, as suggested for loss of Pgp1 above, indirectly affect other PG hydrolases by disrupting the stoichiometry of a putative PG biosynthetic/modification complex. Additional complexities in the shape program are evidenced by reports of straight Campylobacter strains in the literature for which the morphological change has not yet been attributed to mutations in specific gene product(s). For instance, some C. jejuni flagellar mutants appeared to have a straight morphology [68][69][70]; however, other mutants with lesions in the same flagellar genes remained helical ( [71,72]; E. Frirdich and E. C. Gaynor, unpublished). Two reports also cited that passage through the chicken gut or chick embryos resulted in straight C. jejuni isolates [65,73]. Conversely, laboratory passage has led to the isolation of poorly colonizing rod-shaped C. jejuni and C. coli strains [74,75]. In the 11168-GS (straight) and 11168-O (helical) strains of the 2004 study, sequence and expression levels of pgp1 were found to be identical (E. Frirdich and E. C. Gaynor, unpublished). Collectively, this suggests not only that genes other than pgp1 affecting morphology remain to be discovered, but also that mechanisms such as phase variation and/ or epigenetics may be involved.
The pgp1 homolog in H. pylori (named csd4) was identified in a screen for cell shape mutants and is described by Sycuro et al. [22]. Morphology, PG profile, and enzymatic activity indicate conserved functions for the gene products in both organisms. While some enzymes and their effects on morphology are conserved between these related helical bacteria, there are also likely to be differences in the overall PG remodeling and shape determining programs, particularly given their very different wildtype muropeptide profiles.
The Dpgp1 mutant will be a valuable tool to continue to study the effects of the loss of C. jejuni helical shape on its biology and pathogenesis. So far, pgp1 has been found to be important for all aspects of the C. jejuni life cycle. As this study represents the first identification of a gene involved in C. jejuni helical shape and of a role for PG in shape determination, it will also provide the basis for work characterizing additional enzymes involved in C. jejuni PG biosynthesis and shape determination. These studies have now been made easier by the availability of PG structural data for the wild-type C. jejuni strain 81-176 published as part of this work. Future detailed biochemical and structural studies on Pgp1 will also provide interesting insight into the function of this key protein in C. jejuni physiology.

Ethics statement
Animal experiments were carried out in strict accordance with the University of Michigan Committee on Use and Care of Animals (UCUCA). Animal infection and euthanasia protocols were approved by the University of Michigan UCUCA and assigned approval number 10462. Oral gavage was carried out under humane guidelines using an approved protocol judged not to cause distress or harm to the animals. Euthanasia was carried out under humane guidelines using a lethal dose of isofluorane. All animal use procedures are in compliance with University guidelines, State and Federal regulations, and the standards of the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The University of Michigan Animal Welfare Assurance Number on file with the NIH Office of Laboratory Animal Welfare (OLAW) is A3114-01, and the University is fully accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International (AAALAC, Intl.).

Bacterial strains and growth conditions
Bacterial strains and plasmids used in this study and their construction are described in Text S1. C. jejuni strains were grown at 38uC in Mueller-Hinton (MH; Oxoid) broth or 8.5% (w/v) agar supplemented with vancomycin (10 mg/mL) and trimethoprim (5 mg/mL) (unless otherwise indicated) under microaerobic/ capnophilic conditions (6% O 2 , 12% CO 2 ) in a Sanyo tri-gas incubator for plates or using the Oxoid CampyGen system for broth cultures. Growth media were supplemented with chloramphenicol (Cm; 20 mg/mL) or kanamycin (Km; 50 mg/mL), where appropriate. E. coli strains used for plasmid construction were grown at 37uC in Luria-Bertani (LB; Sigma) broth or 7.5% agar (w/v) and supplemented with ampicillin (100 mg/mL), Cm (15 mg/mL), or Km (25 mg/mL), as necessary.
RNA extraction, cDNA synthesis and RT-qPCR RNA was extracted from log phase broth bacteria (OD 0.3), and cDNA was generated from the RNA, as described previously [74,76]. The expression ratio of pgp1 was determined relative to the C. jejuni gyrA gene encoding DNA gyrase subunit A. The primer and TaqMan probe sequences used for qPCR are described in Text S1. Each qPCR reaction mixture consisted of 25 ng of cDNA, 950 nM of each primer, 250 nM TaqMan probe, and 10 ml 26 TaqMan Gene Expression Master Mix (Applied Biosystems) in a total volume of 20 ml. Duplicate reactions for the gene of interest and the housekeeping gene were run in a Stratagene Mx3000P real-time PCR system for 2 min at 60uC, 10 min at 95uC, and then 40 cycles of 15 s at 95uC and 1 min at 60uC. The quantitative PCR cycle threshold (C T ) results were analyzed by the comparative C T method (DDC T method).

Microscopy
Transmission electron microscopy (TEM) was carried out on 18 h broth cultures. Samples were fixed in a final concentration of 2.5% (v/v) of gluteraldehyde for 2-3 h on ice. Cells were then harvested, resuspended in an equal volume of H 2 O, and stored at 4uC. For imaging, 2 mL of bacteria was spotted onto parafilm to which 4 mL of 0.5% uranyl acetate was added for 1 min. A formavar-carbon film on 300 mesh copper grid (Canemco, Lakefield, Quebec, Canada) was added to the bacteria-uranyl acetate spot for 2 min. The grid was then removed, dried, washed ten times in sterile water, dried again and visualized on a Hitachi H7600 TEM equipped with a side mount AMT Advantage (1 mega-pixel) CCD camera (Hamamatsu ORCA) at the UBC Bioimaging facility (The University of British Columbia, Vancouver, BC, Canada).

Phenotypic characterization: CFW, motility, and biofilm formation
Phenotypic assays were carried out with strains grown in shaking MH-TV broth or biphasic cultures grown for 18 h. CFW fluorescence was assayed as described previously [10]. For motility, cultures were diluted to an OD 600 of 0.2 in MH-TV and 2 ml was point inoculated into MH-TV plates containing 0.4% agar. Plates were incubated for 20 h and the halo diameter was measured. Biofilm formation was assayed using crystal violet as described previously [11] with the exception that the absorbance was measured at 570 nm.
Peptidoglycan isolation and muropeptide analysis C. jejuni strains were passaged once from frozen stocks and then passaged to 20-25 MH plates and grown for 20 h to obtain logphase bacteria at a final OD of 200-600. Cells were collected into cold MH broth by scraping, harvested by centrifugation at 8 0006 g for 15 min and then resuspended in 6 mL ice cold H 2 O. Cells were lysed by dropwise addition to 6 mL 8% SDS boiling under reflux. PG was purified from the cell lysate, digested with the muramidase cellosyl (kindly provided by Hoechst, Frankfurt, Germany), and the resulting muropeptides were reduced with sodium borohydride and separated by HPLC as described [77]. Muropeptide fractions were collected, concentrated in a Speed-Vac, acidified by 1% trifluoroacetic acid, and analysed by offline electrospray mass spectrometry on a Finnigan LTQ-FT mass spectrometer (ThermoElectron, Bremen, Germany) at the Newcastle University Pinnacle facility as described [25]. Muropeptide structures were assigned based on (i) comparison with retention times of known muropeptides from H. pylori, Caulobacter crescentus and E. coli and (ii) the obtained MS data (Table S3) and MS/MS fragmentation patterns (not shown).

Expression, purification and enzymatic activity of Pgp1
The C. jejuni 81-176 pgp1 gene was cloned for expression in E. coli without its signal peptide (amino acids 16-464 of the protein) in frame with the thioredoxin-and His-tag of the pET32a vector to give plasmid pEF46. A detailed description of the cloning of the expression construct, and the expression and purification protocol is included in Text S1. For enzyme assays, the purified protein was dialysed against 0.05 M Tris-Cl, pH 7.5 containing 0.01 M ZnCl 2 , 0.3 M NaCl and 20% glycerol. Purified Dpgp1 PG (1.0 mg/ml) was incubated with Trx-His 6 -Pgp1 (5 mM) in 0.02 M NaH 2 PO 4 , pH 4.8, 0.005 M ZnCl 2 , and 0.1 M NaCl for 4 h at 37uC on a Thermomixer at 750 rpm. A control sample received no enzyme and another enzyme sample contained 0.01 M EDTA and no ZnCl 2 . The samples were incubated with 10 mg of cellosyl (Hoechst, Frankfurt am Main, Germany) for 1 h, boiled for 10 min and centrifuged at room temperature for 15 min at 16 000 g. The muropeptides present in the supernatant were reduced with sodium borohydride and analyzed by HPLC, as described [24].

Chick colonization
Chick colonization was performed as described previously [61,76], with an infective dose of 10 4 CFU. Chicken experiments were carried out under protocol #10462 approved by the University of Michigan Committee on Care and Use of Animals (UCUCA).

In vitro invasion and intracellular survival in epithelial and macrophage cell lines
The human epithelial cell lines T84, Caco-2 and INT407 and the murine RAW 264.7 and human Thp-1 macrophage cell lines were used for C. jejuni infections. Carboxymethylcellulose (CMC) was added to the tissue culture media to increase viscosity. A detailed description of the tissue culture infections is included in Text S1.

Interleukin-8 quantification
The INT407 human epithelial cell line was seeded at approximately 1610 5 cells/ml in MEM supplemented with 10% FBS into 24-well tissue culture plates and allowed to grow for 20-24 h prior to infection. The cells were washed three times with MEM and either left uninfected or infected with either C. jejuni wild-type strain 81-176 or Dpgp1 at an O.D. 600 of 0.002/mL taken from an 18 h shaking broth culture. Supernatants were collected after 8 h and 24 h, centrifuged for 10 min to pellet residual cells and bacteria, and frozen at 280uC until assayed. The concentration of IL-8 present in the supernatants was measured by the human IL-8 ELISA kit (Invitrogen, Camarillo, CA). Figure S1 C. jejuni DcarB has a motility defect, but is unaffected for chick colonization. A, negatively stained TEM images of DcarB showing the full-length flagella and helical morphology. B, DcarB exhibits a slight motility defect, as assayed by measuring halo diameters in soft agar plates. Standard error of the mean was calculated from 8 measurements. The asterisks (***) indicate a statistically significant difference (p,0.001) using the unpaired Student's t-test. C, DcarB shows no defect for chick colonization. Each point represents the log CFU/g cecal contents of an individual chick 6 days following infection with 10 4 CFUs of the indicated C. jejuni strains. The geometric mean is denoted by a black bar. (TIF) Figure S2 C. jejuni Dpgp1 shows slightly enhanced intracellular survival, but no defect in invasion in the presence of carboxymethylcellulose (CMC). A, a gentami-cin (Gm) protection assay was used to assess the invasion and intracellular survival ability of Dpgp1 in Caco-2 intestinal epithelial cells. Gm was added 3 h post-infection with bacterial strains. After 2 h the Gm was washed off and the cells were incubated with fresh MEM containing 3% FBS and a low dose of Gm. B, infections were carried out in the presence of MEM containing 1% and 2% CMC to examine the ability of Dpgp1 to invade INT407 epithelial cells in higher viscosity media. CFUs were determined for each well by lysing the cells with water and plating the dilutions onto MH-TV plates. Standard errors of the mean were calculated from triplicate readings and are representative of three independent experiments. The asterisk (*) indicates a statistically significant difference (p,0.05) using the unpaired Student's t-test. (TIF)