The biochemical activity of Pgp4, Pgp5 and Pgp6 was determined in a similar manner to the method used for Pgp1 and Pgp2. Each protein was expressed, purified and incubated with purified PG. The resultant muropeptide profile was examined by HPLC and compared to that of the PG without added enzyme to determine the activity of the enzyme.

Bioinformatic analyses predicted that the three C. jejuni M23 enzymes contain a C-terminal M23 peptidase domain but display various protein domain organizations in the N-terminal region (Figure 1A): Pgp4 has a transmembrane (TM) domain and a coiled-coil region, Pgp5 contains a predicted signal peptide and a Csd3_N domain, and Pgp6 has a TM domain followed by an unknown region. The TM domain and signal peptide play a role in cellular localization in C. jejuni. Protein purification constructs were designed such that the TM domains (Pgp4 and Pgp6) and signal peptide (Pgp5) were not included to ensure cytoplasmic E. coli expression (Figure 1A). The disordered region was removed from the C-terminus of Pgp5 (amino acids 380–386) to increase protein expression levels (Figure 1A). A His₆-tag for protein purification was added to both the N- and C-terminus, with the construct resulting in the highest protein expression levels being selected (C-terminus for Pgp4 and Pgp6 and N-terminus for Pgp5; Figure 1A). The recombinant proteins were expressed in E. coli and purified to >90% purity as determined by SDS-PAGE (Figure 1B). Since these M23 peptidases were predicted to have DD-CPase and/or DD-EPase activity, pentapeptide enriched PG isolated from E. coli strain D456, a triple DD-CPase deletion mutant was used as the substrate (Edwards and Donachie, 1993; Potluri et al., 2010). Purified Pgp4, Pgp5 or Pgp6 was incubated with E. coli D456 PG and the resulting muropeptide profile was examined by HPLC and compared to a no enzyme control (Figure 1C). The addition of Pgp4, Pgp5, and Pgp6 resulted in the cleavage of the terminal D-Ala residue of pentapeptide containing muropeptides (Penta, TetraPenta, and TetraTetraPenta species) demonstrating DD-CPase activity. In addition, no cross-linked muropeptides were detected in the PG of samples to which Pgp4 and Pgp5 were added, indicating that Pgp4 and Pgp5 also had DD-EPase activity cleaving 4–3 cross-links. While formally possible, it is highly unlikely that contaminating E. coli DD-EPases MepA and MepM contributed to the observed activities as we did not detect significant DD-EPase activity with Pgp6 and detected DD-CPase activity for Pgp4 and Pgp6, which cannot be provided by possible contaminating E. coli DD-EPases.

The majority of the M23 peptidase domain proteins contain a characteristic zinc-binding motif H(x)_nD, HxH and are zinc-dependent metallopeptidases (Razew et al., 2022). To examine whether the metal is required for Pgp4, Pgp5, and Pgp6 catalysis, we examined the enzyme activity without metal by adding the metal chelator EDTA. All three enzymes lost activity in the presence of EDTA with the muropeptide peptide profiles being identical to that of the PG only control (Figure 1C). In summary, Pgp4, Pgp5, and Pgp6 are metallopeptidases with Pgp4 and Pgp5 having DD-CPase and DD-EPase activity similar to that of the other known C. jejuni M23 enzyme Pgp3 (Min et al., 2020), while Pgp6 only has DD-CPase activity (Figure 1D).

M23 peptidase domains are zinc metallopeptidases and are common PG hydrolases with various substrate specificities (Razew et al., 2022). For instance, S. aureus LytM is a DD-EPase that cleaves the Gly-Gly bond of pentaglycine cross-bridges in the Gram-positive cell wall (Firczuk et al., 2005); H. pylori Csd1 (An et al., 2016) (note Csd1 functional activity was determined from the muropeptide profile of the mutant and not by biochemical activity assay) and V. cholerae ShyA (Shin et al., 2020) are DD-EPases that cleave between D-Ala and mDap of 4–3 cross-links in Gram-negative PG; and some enzymes such as H. pylori Csd3 (An et al., 2015), C. jejuni Pgp3 (Min et al., 2020), and, as shown in this study, Pgp4 and Pgp5 have both DD-EPase and DD-CPase activity and cleave the D-Ala and mDap bond of 4–3 cross-links and the terminal D-Ala from the pentapeptide. M23 peptidase domain proteins with only DD-CPase activity have not been identified as of yet (see Razew et al. (2022) for a list of enzyme activities of characterized M23 peptidases). We hypothesized that Pgp6 may have a functional motif in its M23 fold to restrict it to monomeric pentapeptides.

Amino acid positions making up sites of structural or functional importance in a protein are evolutionarily conserved and evolve more slowly than variable positions (Ashkenazy et al., 2010). Therefore, we searched for a highly conserved region in Pgp6 that may indicate a functional site selecting for the monomeric pentapeptide substrate and not the crosslinked substrate by performing a sequence alignment with Pgp6 and its orthologs. Using NCBI blastp,² we found Pgp6 orthologs distributed in the ε- and δ-proteobacteria in the clustered nr database. We manually selected 12 orthologs (query coverage between 92 to 99%; sequence identity between 27 to 46%; and E-value between 5e-151 to 1e-43) (Figure 2A) and used MUSCLE (Edgar, 2004) to align them. The amino acid distribution at each position of Pgp6 orthologs was plotted using WebLogo3 (Crooks et al., 2004). A consensus sequence (Fx₃Nx₃Rx₂Nx₃I) was identified in Pgp6 before the start of the M23 peptidase domain (Figure 2B).

We searched for the consensus sequence (Fx₃Nx₃Rx₂Nx₃I) in Pgp3, Pgp4, and Pgp5, but it was found to be unique to Pgp6. Pgp6 shares higher sequence similarity to Pgp3 (37.40% identity/ 2e-21 E-value) than to Pgp4 (23.6% identity/4e-14 E-value) and Pgp5 (28.2% identity/2e-14 E-value). Therefore, a sequence alignment of Pgp3, Pgp6, and their orthologs from Sulfurimonas autotrophica DSM 16294 and Sulfuricurvum kujiense DSM 16994 was performed (Figure 2C). This suggested that Pgp6 aligns well with Pgp3 except for four insertion sequences: SEQ1(D262-L278), SEQ2 (V293-I298), SEQ3 (Y323-V331), and SEQ4 (K436-N441) (Figure 2C). SEQ1 encompasses part of the consensus sequence (Fx₃Nx₃Rx₂Nx₃I). SEQ3 and SEQ4 contain conserved residues Y324 and W437. The amino acid conservation and its unique occurrence in Pgp6 suggest that these insertion sequences may play a role in limiting the activity of Pgp6 to monomeric substrates, as opposed to Pgp3 that can act on monomeric and dimeric substrates.

To gain more insight into the influence of the insertion sequences on Pgp6 function, we examined the position of the insertion sequences in the Pgp6 protein structure. As there is no available experimental structure for Pgp6, we retrieved a computational model from the Alpha Fold database (Jumper et al., 2021; Varadi et al., 2022). The Pgp6 model (AlphaFold ID A0A0H3PIR6) (Figure 3A, left) had an N-terminal helix (M1-R27), followed by two immunoglobulin-like β-sandwich folds: domain 1 (L28-V138) and domain 2 (K144-Y235), and an M23 peptidase domain (N311-P429). Between domain 1 and domain 2 is a short loop (D139-P143). Domain 2 is connected to the M23 peptidase domain by a linker region (R236-K310) consisting of strand β1 and helices ɑ1ɑ2 (See protein topology of Pgp6 M23 peptidase domain in Supplementary Figure 2). The C-terminal helix (E430-N457) folds back on domain 2. The model consisting of domain 2 to the C-terminal helix (N140 to N457) is predicted to be of high confidence by AlphaFold (Supplementary Figure 3). The per-residue confidence score (pLDDT) is above 96 (this score is between 0 and 100, with 100 meaning the highest confidence) (Jumper et al., 2021). Predicated aligned error (PAE), which represents the distance error (Å) of 2 residues, is low (between 0 and 15 Å), suggesting the relative position and orientation between domains are well defined (Varadi et al., 2022).

The Pgp6 model shows that helices ɑ1 and ɑ2 form a helix–loop–helix structure that is situated between the core β-sheet, which consists of seven strands (β2, β5, β13, β9, β8, β7, and β11) and another β-sheet of strands made up of β1, β3 and β4 (Figure 3A, right). The insertion sequences SEQ1 and SEQ3 extend the length of helix ɑ2 as well as strands β3 and β4, respectively. This results in an enclosed surface that wraps the top and one side of the concave surface in the M23 peptidase domain, creating a single-entry point for the substrate (Figure 3B). The β-ɑ-β sandwich is stabilized by numerous hydrophobic contacts, including a hydrophobic cluster formed by residue F271 in SEQ1, Y324 in SEQ3, and a conserved residue F246 in ɑ2 (Figure 3B). In contrast, the equivalent position of ɑ1ɑ2 helix pair in the Pgp3 enzyme with DD-EPase and DD-CPase activity is a flexible linker. Loop L1 of Pgp3 can adopt a huge conformational change to switch from a latent conformation (named as closed form) to a substrate binding conformation (named as open form) (Figure 4) (Min et al., 2020). A comparison of the active site groove between Pgp6 and the open conformation of Pgp3 suggests that Pgp6 has a small substrate entry access, which may limit the conformation and types of substrates. The side chains of Fx₃Nx₃Rx₂Nx₃I in SEQ1 of Pgp6 face toward the zinc binding ligands (Figure 3B). The R279 residue is only 7.5 Å from the zinc binding ligands, suggesting that the positively charged side chain is in a good position to orient the negatively charged PG substrate (e.g., carboxylate group of the terminal D-Ala of the pentapeptide).

The insertion sequences SEQ2 and SEQ4 are remote from the zinc binding center (Figure 3C). SEQ2 belongs to the ɑ2-β2 loop adjacent to a loop (F184-Y194) in domain2. The amide nitrogen and main chain oxygen of residue I298 in SEQ2 are hydrogen-bonded to the main chain of residues P188 and Y189 in domain 2 (Figure 3C). SEQ4 is located at the C-terminal helix, and it participates in the contacts between the M23 peptidase domain and domain 2. The main chain oxygen of residue N441 in SEQ4 makes a hydrogen bond to the side chain hydroxyl group of Ser151 in domain 2; residue W437 has a hydrophobic interaction with the side chain methyl group of Ala159 in domain 2 and a conserved residue W432 (Figure 3C). In summary, from the AlphaFold model of Pgp6, the insertion sequences SEQ1-4 contribute additional contacts within the M23 peptidase domain and between domains with domain 2. We propose that Pgp6 only has DD-CPase activity because of the small active site groove created by the insertion sequences SEQ1 and SEQ3 that selects for the monomeric pentapeptide. The contacts made by SEQ2 and SEQ4 appear to help generate the interdomain orientation of domain 2 and the core of the M23 peptidase domain.

Motility was assayed by measuring halo diameter in soft agar plates (Figure 5A). No flagellar structural defects were observed by electron microscopy (Frirdich et al., 2023). The motility of ∆pgp4 was similar to wild type. However, the complemented strain (∆pgp4c) showed a statistically significant but slight decrease in motility of 11.1% in comparison to wild type and the overexpression strain (81–176 + pgp4) a decrease of 7.6% that was not statistically significant. The ∆pgp5 strain displayed a motility defect which was similar to that reported previously (85.1% of wild type motility) (Stahl et al., 2016). The complement restored motility to wild type levels, while the pgp5 overexpression strain showed an increase of motility of 17.2% in comparison to wild type. The motility of the ∆pgp6 strain was slightly decreased by 7.9% in comparison to wild type. Complementation restored motility to wild type levels with the pgp6 overexpression strain showing a very slight increase in motility of 2.9% in comparison to wild type.

Biofilm formation was assessed by crystal violet assays and measured over 3 days (Figure 5B). Biofilm levels of ∆pgp4 were similar to wild type at day 1 and then 1.8-, and 2.7-fold lower at days 2 and 3, respectively. Biofilm production was restored in the complemented strain (∆pgp4c) and slightly above wild type by day 3. The pgp4 overexpression strain produced higher levels of biofilm than wild type at day 2 but not day 3. Biofilm levels of ∆pgp5 were 1.5-fold higher than wild type at days 2 and 3. Wild type biofilm levels were seen in the complemented and pgp5 overexpression strain. Biofilm levels of ∆pgp6 were 1.4-fold lower than wild type at day 2 and then 1.2-fold higher at day 3 (although this increase was not significant). The complement showed wild type biofilm levels with the pgp6 overexpression displaying 1.4-fold lower than wild type levels at day 2 but not day 3.

C. jejuni autoagglutination is a measure of virulence. The autoagglutination kinetics were determined by monitoring the levels of autoagglutination at 3, 6, and 24 h in PBS at room temperature (Figure 5C). The ∆pgp4 and ∆pgp6 mutants showed higher levels of autoagglutination than wild type at 3, 6 and 24 h. This was not complemented to wild type levels in ∆pgp4c and ∆pgp6c (except for ∆pgp4c at 3 h). The pgp4 overexpression strain displayed decreased autoagglutination at 6, while the pgp6 overexpression strain displayed slightly increased autoagglutination at 3 h. The ∆pgp5 mutant autoagglutination was less than wild type at 3 h but was slightly increased in comparison to wild type by 24 h. The ∆pgp5 complemented strain showed increased autoagglutination in comparison to wild type at 6 h and 24 h with the levels being higher than the ∆pgp5 mutant at 24 h. The pgp5 overexpression strain had an even more pronounced increase in autoagglutination than ∆pgp5c at 3 h, 6 h and 24 h.

No difference in hydrophobicity was detected between the strains by either the SAT or BATCH method (data not shown) of examining hydrophobicity. Using the SAT method, all strains precipitated/aggregated in an initial ammonium sulfate concentration of 0.0313 M.

CFW fluorescence was determined by visualizing bacterial growth on agar plates containing CFW under long wave UV (Figure 5D). After 48 h of growth, the ∆pgp1 strain is hypofluorescent and the ∆spoT strain hyperfluorescent in comparison to wild type and are used as controls. The ∆pgp4 strain was hypo-fluorescent, while the complement and pgp4 overexpression strain were similar to wild type. Despite displaying an increase in biofilm formation, ∆pgp5 showed wild type CFW fluorescence, as did ∆pgp5c and the pgp5 overexpression strain. The ∆pgp6, ∆pgp6c and pgp6 overexpression strains all showed wild type fluorescence.

Some PG hydrolases are preferentially used by the cell under certain growth conditions such as acidic pH (Egan et al., 2020). Therefore, we wanted to determine whether one of Pgp4, Pgp5 or Pgp6 was more important for growth at low pH by assessing acid survival. The ability of the pgp4, pgp5 and pgp6 mutant, complemented and overexpression strains to survive on MH-TV agar plates at pH 4.5, 5.0, 5.5, 6.0, 6.5 and 7.0 was compared to that of wild type. At pH 4.5, there was very little to no growth. A difference in survival of the strains was only seen at pH 5.0 so only the results from that pH are shown (Figure 6). At pH 5.0, the survival of ∆pgp5c, 81–176 + pgp5, and 81–176 + pgp6 strains was reduced at least 10-fold in comparison to wild type. pH survival experiments carried out in broth produced inconsistent results and were not included.

In order to assess the integrity of the outer membrane of ∆pgp4, ∆pgp5, ∆pgp6, the minimum inhibitory concentration that reduced growth by 50% of the mutants to detergents, antimicrobial compounds, chelating agents and salts was determined and compared to that of wild type. There were no large differences in sensitivity of the mutants in comparison to wild type (Supplementary Table 4). Interestingly, the MIC to MES (2-(N-morpholino) ethanesulfonic acid) decreased at pH 5.0 in comparison to pH 7.0. MES is used as a buffering agent with a buffering capacity of 5.5–7.0 due to its good solubility in water, high stability and minimal interference with biological compounds. Our studies examining pH survival in MH media made up with 50 mM and 100 mM MES as a buffering agent indicated that MES may be having an effect on C. jejuni survival which was confirmed by determining the C. jejuni MIC to MES. Because of the decrease in MIC for MES at pH 5.0, the MIC of all compounds tested was determined at pH 5.0 (Supplementary Table 4). Interestingly, the MIC at pH 5.0 in comparison to pH 7.0 for all the detergents decreased as did that of EDTA, while the MIC of polymyxin increased 2-fold.

Coccoid formation during aging was monitored to compare the morphological transition between strains, as done previously (Frirdich et al., 2019; Frirdich et al., 2023). Cells were grown at 37°C on solid media under microaerophilic conditions and coccoid formation was examined at day 1, day 2, day 4 and day 8 by DIC microscopy (Figure 7A). Samples were taken from the center of the plate (sampling from different areas of the plate can show some variability). The percentage of helical, coccoid, and cells transitioning to the coccoid form was quantified from the DIC images (Figure 7B). The mutant strains were all impaired in the coccoid transition with a higher percentage of cells in the transitioning state and less in the coccoid form and a very slightly higher amount of cells still in the helical form. At day 8, 4.3% of wild type were in the transitioning state in contrast to 30.5% in ∆pgp4, 20.2% in ∆pgp5 and 15.2% in ∆pgp6. The numbers of coccoid cells were 95.0% for wild type, 66.5% for ∆pgp4, 78.6% for ∆pgp5 and 80.0% for ∆pgp6 and the remaining helical cells were 0.6% for wild type, 3.0% for ∆pgp4, 1.2% for ∆pgp5 and 4.8% for ∆pgp6. The complemented and overexpression strains were similar to wild type.

A gentamicin protection assay was used to assess the adherence, invasion, and intracellular survival properties of the C. jejuni pgp4, pgp5 and pgp6 mutant, complemented and overexpression strains in INT407 epithelial cells (Figure 8). Only the pgp5 mutant showed a slight reduction in adherence, invasion and intracellular survival, as demonstrated previously (Stahl et al., 2016). The decrease in ∆pgp5 adherence and invasion (t = 3 h) of 12.4% (in comparison to wild type) and intracellular survival (t = 7 h) of 8.5% was restored by complementation, while the reduction in invasion (t = 5 h) of 14.8% was partially restored by 8.0%. The adherence and invasion of ∆pgp4c and pgp4 overexpression strains measured 3 h post-infection was 21.5 and 17.8% less than wild type, with only the difference between the complement and wild type being statistically significant. No differences at the subsequent invasion and intracellular survival time points were seen. The pgp6 overexpression strain showed a statistically decrease of 17.6% in intracellular survival at 7 h post-infection in comparison to wild type.

C. jejuni motility is a key virulence determinant in host colonization facilitating penetration of the highly viscous mucosal layer of the gastrointestinal tract (Tikhomirova et al., 2024). Motility is also required for chemotaxis. Chemotaxis is important for C. jejuni host invasion and environmental survival (Korolik, 2019). In soft agar, the motility of rod shaped ∆pgp1 and ∆pgp2 mutants was 82.5% and 73.7% of wild type, respectively. In the curved rod mutants, the motility of ∆pgp4 was similar to wild type and that of ∆pgp6 92.1% of wild type, while that of ∆pgp5 was more pronounced being 85.1% of wild type and more similar to that of ∆pgp1. The morphology of the ∆pgp5 mutant had an increase in angularity (or mean curvature), while the ∆pgp4 and ∆pgp6 had decreased angularity, potentially contributing to the variation in motility defects (see Supplementary Figure 1 for morphology) (Frirdich et al., 2023). C. jejuni has amphitrichous flagella or a flagellum at each cell pole. A recent study on how C. jejuni coordinates the rotation of each opposing flagellum during movement has identified an unappreciated role for the C. jejuni helical shape in C. jejuni flagellar motility (Cohen et al., 2020). In high viscosity media, C. jejuni wraps the leading left handed flagellar filament around the right handed helix of the cell body to increase swimming speeds (Cohen et al., 2020). Directional changes involve unwrapping of the filament from the cell body with the helical shape of opposite handedness to the filament being required for efficient unwrapping, as the rod shaped ∆pgp1 mutant was defective for flagellar unwrapping (Cohen et al., 2020). Flagella unwrapping ability of the curved rod mutants with different helical shapes may be altered and could affect motility.

Biofilm formation ensures C. jejuni transmission and persistence in the environment and in the human gastrointestinal tract (Tikhomirova et al., 2024). Motility does have a role in biofilm formation in C. jejuni (Puning et al., 2021; Ma et al., 2022). The rod-shaped ∆pgp1 and ∆pgp2 mutants with reduced motility also displayed reduced biofilm formation (Frirdich et al., 2012; Frirdich et al., 2014). This correlation was restricted to the rod-shaped mutants. A mutant in the O-acetylpeptidoglycan esterase ape1 with altered cell curvature had a slight motility defect but was a hyper-biofilm former (Ha et al., 2016). Similarly, ∆pgp5, with a more similar morphology to ∆ape1 than ∆pgp4 and ∆pgp6, also had a motility defect and produced higher biofilm levels. Despite the lack of a motility defect, biofilm formation was reduced in ∆pgp4 to greater levels than ∆pgp1 and ∆pgp2. There was no change in ∆pgp6 biofilm formation even though this mutant had slightly reduced motility. Surface hydrophobicity is also a critical factor influencing biofilm formation in C. jejuni (Ma et al., 2022). This was unchanged in ∆pgp4, ∆pgp5 and ∆pgp6 (as well as ∆pgp1 and ∆pgp2). These results indicate that biofilm formation with the curved rod mutants was affected by a factor other than motility and cell surface hydrophobicity.

CFW is a fluorescent dye binding β (1,3) and β (1,4) glycosidic bonds on the cell surface. We had previously identified CFW hypo- and hyper-reactive (dim and brt) mutants with changes in cell surface attributes and defective for various aspects of pathogenesis such as biofilm formation (McLennan et al., 2008; Naito et al., 2010; Frirdich et al., 2012; Frirdich et al., 2014). A correlation was established between CFW reactivity and biofilm formation with CFW hyperfluorescent strains being hyperbiofilm forming and conversely, CFW hypofluorescent strains (such as ∆pgp1 and ∆pgp2) being hypobiofilm forming (McLennan et al., 2008; Naito et al., 2010; Frirdich et al., 2012; Frirdich et al., 2014). This held true for the ∆pgp4 mutant that had reduced biofilm formation and was hypofluorescent on CFW, but not ∆pgp5 that produced increased biofilms but no change in CFW reactivity. Biofilms produced by the ∆pgp6 mutant were not significantly different to wild type at day 3 and showed wild type CFW reactivity. The exact C. jejuni polymer recognized by CFW in C. jejuni CFW fluorescence assays is unknown, but CFW does recognize the β(1–4) linkages of the PG backbone (Frirdich et al., 2017). Our current hypothesis is that CFW accessibility to the PG backbone is what dictates CFW fluorescence (Frirdich et al., 2017). Passage on CFW was found to select for mutations in pgp1, pgp2 and pgp4 that are all hypofluorescent on CFW, thereby being more resistant to the stress caused by CFW (Frirdich et al., 2017).

C. jejuni autoagglutination is associated with full-length, glycosylated flagella, but not motility directly (Misawa and Blaser, 2000; Golden and Acheson, 2002; Guerry et al., 2006; Ewing et al., 2009). The structure of the C. jejuni flagellar motor is also important for mediating flagellar disentanglement which affects agglutination levels (Cohen et al., 2024). By electron microscopy, the ∆pgp4, ∆pgp5, and ∆pgp6 mutants did not display any obvious changes in flagellar structure or number (Frirdich et al., 2023). However, more subtle flagellar changes such as in the flagellar motor or changes in flagellar glycosylation could explain the differences in agglutination. Also, the C. jejuni helical shape minimizes interactions between the cell body and flagellar filament that contribute to autoagglutination (Cohen et al., 2020). The changes to the helical shape in the ∆pgp4, ∆pgp5, and ∆pgp6 mutants could thereby affect the autoagglutination ability of these strains. The ∆pgp4 and ∆pgp6 mutants which have a curved morphology with decreased angularity in comparison to wild type (Frirdich et al., 2023) agglutinated faster than wild type. The ∆pgp5 mutant with morphology with increased angularity in comparison to wild type (Frirdich et al., 2023) displayed altered autoagglutination kinetics agglutinating more slowly than wild type but then agglutinating to slightly higher levels than wild type by 24 h. It is possible that the particular morphology of ∆pgp5 with an increased angularity affects its autoagglutination kinetics. The pgp4 overexpression strain also showed decreased autoagglutination, but only at the 6 h time point. Like ∆pgp5, this strain had increased angularity, but an overall different morphology to that of ∆pgp5 (Frirdich et al., 2023). All the complemented strains and overexpression strains had altered autoagglutination kinetics at certain timepoints. While they did not have changes in angularity to explain the autoagglutination results, they may have slightly altered morphologies due to PG hydrolase expression differences that could affect agglutination. C. jejuni autoagglutination has been correlated with bacterial hydrophobicity and adherence to INT407 cells (Misawa and Blaser, 2000; Golden and Acheson, 2002; Guerry et al., 2006). Cell surface hydrophobicity was unaffected in all the mutants. Decreased adherence to INT407 cells was seen with the ∆pgp5 mutant, which did have decreased agglutination at 3 h.

C. jejuni transitions to a coccoid form as it ages and under environmental stresses such as starvation, temperature, pH, and osmolarity (Svensson et al., 2008; Jackson et al., 2009; Ikeda and Karlyshev, 2012; Frirdich and Gaynor, 2013; Frirdich et al., 2019). The exact role of C. jejuni coccoid formation in environmental survival and pathogenesis is still unclear. However, C. jejuni do become coccoid within epithelial cells and coccoid C. jejuni are non-motile and non-infectious (Frirdich et al., 2019). Previously, the DL-CPase Pgp1 and the amidase AmiA were shown to be key for coccoid formation, but not the LD-CPase Pgp2 (Frirdich et al., 2019). The ∆pgp1, ∆amiA and ∆pgp1∆amiA mutants had higher levels of helical cells after 8 days of culture in comparison to wild type and were defective in initiation of the coccoid form (Frirdich et al., 2019). In contrast, the ∆pgp4, ∆pgp5 and ∆pgp6 were inhibited in completion of the transitioning state to the coccoid form with an accumulation of transitioning cells in comparison to wild type after 8 days of growth. Coccoid PG showed increased levels of dipeptides and decreased tri- and tetrapeptides with no change in the degree of cross-linking and the level of monomers, dimers and trimers (Frirdich et al., 2019). These changes are not a direct result of the DD-CPase and/or DD-EPase activity of Pgp4, Pgp5 and Pgp6.

Acid survival is an important factor affecting the ability of an enteric pathogen to survive passage through the gastrointestinal tract. As stated previously, some PG hydrolases are important for growth in acidic pH (Egan et al., 2020). Loss of pgp4, pgp5 and pgp6, as well as pgp1 and pgp2 (data not shown), did not affect survival at low pH. This is supported by previous studies. In C. jejuni strain 11168, expression of pgp4, pgp5 or pgp6 were unchanged in response to acid shock (Reid et al., 2008a) or during growth under mildly acidic conditions (pH 5.5, 6.0 and 6.5) (Reid et al., 2008b). Also, no transposon mutants in pgp4, pgp5 or pgp6 were identified as having impaired growth at pH 5.5 or pH 6.0 (Reid et al., 2008b). However, a proteomics study using C. jejuni 11168 showed the reduced abundance of the Pgp6 protein when C. jejuni was grown under mildly acidic conditions (pH 5.8) and subjected to acid shock (pH 4.0, 2 h) (Ramires et al., 2023). This correlates with our results that higher levels of Pgp6 in the pgp6 overexpression strain were slightly detrimental to acid survival at pH 5.0. The same was seen with Pgp5 with the complemented and overexpression strain both having slightly reduced growth at pH 5.0. The levels of Pgp5 in the complemented strain could be higher than those in the wild type possibly explaining the defect in this strain. How Pgp5 and Pgp6 can adversely affect survival at acidic pH remains to be established.

Similar to ∆pgp1 and ∆pgp2, the ∆pgp4, ∆pgp5 and ∆pgp6 mutants were not more sensitive to compounds that would indicate differences in outer membrane stability in comparison to wild type. Changes in C. jejuni outer membrane permeability occurred when the pH was dropped from 7.0 to 5.0, signified by increased susceptibility to the chelating agent EDTA and detergents. At pH 5.0, polymyxin B resistance increased two-fold. Polymyxin is a cationic peptide that binds to the negatively charged lipopolysaccharide. Increased resistance to polymyxin B at pH 5.0 could indicate a change in the charge or fluidity of the outer membrane (Trimble et al., 2016). Interestingly, all strains including wild type were more sensitive to MES at pH 5.0 in comparison to pH 7.0 with an MIC₅₀ of 50 mM at pH 5.0. MES at a concentration of 100 mM was often used to study C. jejuni responses to acidic pH and acid survival (Reid et al., 2008a; Reid et al., 2008b) due to its buffering capacity at pH 5.5–7.0. These results may be confounded by the negative affects of MES at low pH.

Only the ∆pgp5 mutant had a decrease albeit minor in adherence, invasion and intracellular survival in cultured epithelial cells replicating our previous results (although with complementation to wild type levels in this work) (Stahl et al., 2016). This mutant also had the most significant motility defect and impaired autoagglutination, factors important in targeting and adherence to host cells, respectively. In a previous study, ∆pgp5 PG showed decreased secretion of the proinflammatory chemokine IL-8 from epithelial cells (Stahl et al., 2016). In that study, the ∆pgp5 mutant along with the rod-shaped ∆pgp1 and ∆pgp2 mutants were used to examine the role of helical shape in colonization ability and inflammation in a Sigirr−/− mouse model of C. jejuni intestinal infection (described in the Introduction) (Stahl et al., 2016). The ∆pgp5 mutant was selected as the representative curved rod mutant as it had the most drastic and consistent changes to the curvature within the bacterial population (Stahl et al., 2016). In the Sigirr−/− mouse model, despite the impaired host in vitro phenotypes, the ∆pgp5 mutant could penetrate the mucus overlaying the intestinal epithelium and crypts, colonize and trigger an inflammatory response similar to wild type and unlike ∆pgp1 and ∆pgp2 that were non-pathogenic (Stahl et al., 2016). Interestingly, ∆pgp1 and ∆pgp2 were not defective for adherence, invasion and intracellular survival in epithelial cells (Frirdich et al., 2012; Frirdich et al., 2014). As shown in this study, neither were the pgp4 nor the pgp6 mutants. The ∆pgp4c complemented strain had reduced adherence and invasion (at the 3 h time point), but then showed wild type invasion at 5 h and intracellular survival at 7 h. This is a confounding observation as decreased adherence would be expected to accompany a decrease in invasion and intracellular survival but these were similar to wild type. One possible explanation is that the decreased adherence is a result of the decreased motility of this strain (Figure 5A). The ∆pgp4c complemented strain may have an increased ability to invade and survive intracellularly in comparison to wild type that makes up for the adherence defect and then results in wild type invasion and intracellular survival at later timepoints. The pgp6 overexpression strain was slightly reduced in intracellular survival at 7 h post-infection. Interestingly, this strain was also sensitive to low pH. However, C. jejuni survives intracellularly by avoiding delivery to the acidic lysosomes (Watson and Galan, 2008), so the importance of pH resistance in C. jejuni intracellular survival is unknown.

Another group identified the pgp4 (1105) gene as having a role in C. jejuni cell curvature in a transposon mutant library visual screen for cell shape mutants (Esson et al., 2017). They found that the mutant had a curved rod phenotype visually similar to that of our ∆pgp4 mutant, although morphological analyses were not carried out. Their ∆pgp4 mutant in the same 81–176 wild type strain as the one used in our experiments had a motility defect and reduced adherence and invasion in vitro in cultured Caco2 cells (Esson et al., 2017), while our ∆pgp4 mutant had neither a motility nor an adherence or invasion defect in INT407 cells. The cause of phenotypic differences between our two ∆pgp4 mutant strains is unclear. Esson et al. (2017) showed that their ∆pgp4 had wild type chick colonization.