Phase-Variable Changes in the Position of O-Methyl Phosphoramidate Modifications on the Polysaccharide Capsule of Campylobacter jejuni Modulate Serum Resistance

ABSTRACT

Campylobacter jejuni polysaccharide capsules (CPS) are characterized by the presence of nonstoichiometric O-methyl phosphoramidate (MeOPN) modifications. The lack of stoichiometry is due to phase variation at homopolymeric tracts within the MeOPN transferase genes. C. jejuni strain 81-176 contains two MeOPN transferase genes and has been shown previously to contain MeOPN modifications at the 2 and 6 positions of the galactose (Gal) moiety in the CPS. We demonstrate here that one of the two MeOPN transferases, encoded by CJJ81176_1435, is bifunctional and is responsible for the addition of MeOPN to both the 2 and the 6 positions of Gal. A new MeOPN at the 4 position of Gal was observed in a mutant lacking the CJJ81176_1435 transferase and this was encoded by the CJJ81176_1420 transferase. During routine growth of 81-176, the CJJ81176_1420 transferase was predominantly in an off configuration, while the CJJ81176_1435 transferase was primarily on. However, exposure to normal human serum selected for cells expressing the CJJ81176_1420 transferase. MeOPN modifications appear to block binding of naturally occurring antibodies to the 81-176 CPS. The absence of MeOPN-4-Gal resulted in enhanced sensitivity to serum killing, whereas the loss of MeOPN-2-Gal and MeOPN-6-Gal resulted in enhanced resistance to serum killing, perhaps by allowing more MeOPN to be put onto the 4 position of Gal.

IMPORTANCE Campylobacter jejuni undergoes phase variation in genes encoding surface antigens, leading to the concept that a strain of this organism consists of multiple genotypes that are selected for fitness in various environments. Methyl phosphoramidate modifications on the capsule of C. jejuni block access of preexisting antibodies in normal human sera to the polysaccharide chain, thus preventing activation of the classical arm of the complement cascade. We show that the capsule of strain 81-176 contains more sites of MeOPN modifications than previously recognized and that one site, on the 4 position of galactose, is more critical to complement resistance than the others. Exposure to normal human serum selects for variants in the population expressing this MeOPN modification.

INTRODUCTION

Campylobacter jejuni is among the leading causes of bacterial diarrhea worldwide. The incidence of C. jejuni infection in the developed world ranges from 14.3 to 45.2 cases/100,000 in the United States and Europe, respectively (1). However, in the developing world, the isolation rates range from 5 to 20% (2, 3), and the Global Enteric Multicenter Study (GEMS) study has identified C. jejuni as a significant cause of severe to moderate diarrhea in children in southeast Asia (4). A multisite birth cohort study (MAL-ED) found that Campylobacter spp. were among the leading bacterial causes of diarrhea in the first year of life and were the most frequent cause in the second year of life in developing countries (5). In addition, recent studies have also found an association of C. jejuni infection with malnutrition and growth stunting in pediatric populations in the developing world (6, 7). Moreover, C. jejuni infection is associated with a number of sequelae, including reactive arthritis, irritable bowel syndrome (IBS), and Guillain-Barré syndrome (GBS). GBS follows as many as one in every 1,000 cases of C. jejuni enteritis and is due to molecular mimicry between N-acetyl neuraminic acid-containing lipooligosaccharides (LOS) of most strains of C. jejuni and human peripheral nerves (8). Collectively, these sequelae contribute to the disease burden of C. jejuni beyond that of the acute disease (9).

The C. jejuni polysaccharide capsules (CPS) have been shown to be the major determinant of the serotyping scheme developed by Penner and colleagues that included 47 C. jejuni heat-stable (HS) serotypes (10, 11). The structures of the capsular polysaccharides have been solved for a number of C. jejuni serotypes, with most being characterized by the presence of heptoses in unusual configurations (e.g., ido, gulo, and altro) and nonstoichiometric modifications on the sugars, including ethanolamine, aminoglycerol, and O-methyl phosphoramidate (MeOPN) (12). The MeOPN modification is found in about 75% of CPS types in diverse, apparently serotype-specific linkages and always in nonstoichiometric amounts. The genes for biosynthesis of MeOPN are highly conserved among strains of C. jejuni, while the genes encoding MeOPN transferases are highly conserved at the 5′ end of the gene but divergent at the 3′ end. The conserved 5′ ends are also characterized by the presence of homopolymeric G tracts that undergo phase variation by slip strand mismatch repair during replication, a trait common to genes encoding multiple surface antigens of C. jejuni (13–18). In the case of the MeOPN transferases, a homopolymeric tract of 9 G residues results in a full-length open reading frame (ORF), but slip strand mismatch repair resulting in 8, 10, or 11 G residues results in early truncation of the ORF. Thus, the population consists of mixtures of cells with the MeOPN transferase genes in either on or off configurations, which presumably results in nonstoichiometric levels of MeOPN. These on/off configurations of the MeOPN transferases have been shown to be reversible (19). The high frequency of phase variation in C. jejuni is thought to be due to the lack of mismatch repair enzymes (13, 20).

C. jejuni CPS have been shown to be important for pathogenesis. Nonencapsulated mutants were attenuated in a ferret model of diarrheal disease and were reduced in their ability to colonize chickens, mice, and piglets (21–25), although there are conflicting reports on the role of MeOPN in pathogenesis of a Galleria mellonella model of disease (26, 27) and in the invasion of intestinal epithelial cells in vitro (24, 28). Capsules are a major factor in resistance to complement-mediated killing and, in the case of C. jejuni 81-176 (HS23/36), decoration of the CPS with MeOPN is essential for complement resistance. Thus, mutants of 81-176 expressing a CPS lacking MeOPN were as sensitive to complement-mediated killing as a mutant lacking CPS (23, 24, 28). CPS of two serotypes, HS2 and HS23/36, have also been shown to have immunomodulatory effects in vitro (23, 27), and the immunomodulatory effects for the 81-176 CPS have been confirmed in vivo in a mouse model (29). MeOPN has also been shown to serve as a phage receptor (19, 30).

The 81-176 CPS is a repeating trisaccharide of 3-substituted galactose (Gal), 2-substituted 3-O-methyl-6-deoxy-altro-heptose (Hep), and 3-substituted N-acetylglucosamine (GlcNAc): [→3)-α-Gal-(1→2)-6d-3Me-α-altro-Hep-(1→3)-β-GlcNAc-(1→]. Strain 81-176 contains genes predicted to encode two putative MeOPN transferases, CJJ81176_1420 and CJJ81176_1435 (23), here abbreviated to CJJ1420 and CJJ1435, and MeOPN has been reported on both the 2 position of galactose (MeOPN-2-Gal) and the 6 position of galactose (MeOPN-6-Gal) (Fig. 1) (31; P. Guerry, J. Yuening, and M. A. Monteiro, U.S. provisional patent application 62/075,399). Jiao et al. (32) have shown that antibodies generated against an 81-176 capsule conjugate vaccine (33) recognized synthetic MeOPN-6-Gal. Here, we identify a third site of MeOPN modification on the 81-176 CPS, at the 4 position of galactose (MeOPN-4-Gal), and show that CJJ1420 encodes the transferase responsible for this activity. We also show that MeOPN appears to mediate resistance to complement by blocking binding of preexisting, antiglycan antibodies present in normal human sera (NHS). MeOPN-4-Gal appears to be the major modification responsible for resistance to complement-mediated killing, although the CJJ1420 gene appears to be primarily in an off configuration during in vitro culture. Growth in the presence of NHS selected for a subpopulation of cells in which CJJ1420 has phase varied to the on configuration.

FIG 1.

Structure of the 81-176 CPS. The R group can be either H or MeOPN.

RESULTS

MeOPN modifications on the 81-176 CPS.

Previously, using mass spectrometry we detected a nonstoichiometric MeOPN unit at the 2 position of galactose (MeOPN-2-Gal) in 81-176 CPS (31), with a ³¹P resonance similar to that in Fig. 2A (peak Y). Here, we confirmed this MeOPN-2-Gal linkage by NMR (Fig. 3A) through the detection of a cross-peak between the ³¹P resonance Y (δ_P 14.45) of MeOPN and H-2 (δ_H 4.52) of the galactose unit in a ¹H-³¹P correlation experiment.

FIG 2.

³¹P-NMR of CPS purified from wild-type 81-176 (A and B) and CJJ1435⁻ (3636) (C) strains.

FIG 3.

1D slices obtained from 2D ¹H-³¹P HMBC NMR experiments showing the attachment of MeOPN to positions 2 (A) and 6 (B) of Gal in wild-type 81-176 CPS (293 K) and position 4 of Gal in CJJ1435⁻ mutant CPS (315 K) (C). Chemical shifts may change slightly if MeOPN is present at the nonreducing repeating block.

In some 81-176 CPS preparations, albeit of lower intensity, the ³¹P-NMR spectrum displayed an additional resonance (Fig. 2B) at δ_P 14.15 (designated peak Z). A similar peak (data not shown) was also observed in the ³¹P-NMR of the CJJ1420 mutant, called 3477 (see Table 1), which exhibited a cross-peak (Fig. 3B) between the phosphorous of MeOPN and H-6 resonances of some of the CPS galactose units, which resonated very near the methyl resonances of MeOPN (δ_H 3.75 to 3.81). The NMR data suggested that peak Z in the 81-176 wild type and in the CJJ1420⁻ strain corresponded to a nonstoichiometric placement of MeOPN at position 6 of galactose (MeOPN-6-Gal), which is consistent with the data using synthetic MeOPN-6-Gal (32).

TABLE 1.

Capsular mutants of 81-176 used in this study^a

Strain	Genotype	Designation	Strain background	Source or reference
3390	mpnC::cat	mpnC	Wild type	22
3477	CJJ1420::aph3	CJJ1420−	Wild type	This study
3498	CJJ1420::aph3 hipO::CJJ1420R*+cat	CJJ1420−/C	3477	This study
3636	CJJ1435::cat	CJJ1435−	Wild type	This study
3637	CJJ1435::cat astA::CJJ1435R*+aph3	CJJ1435−/C	3636	This study
3479	CJJ1420::aph3 CJJ1435::cat	CJJ1420− CJJ1435−	3477	This study
3501	hipO::CJJ1420R*+cat	CJJ1420R	Wild type	This study
3718	hipO::CJJ1420R*+cat CJJ1435::apr	CJJ1420R CJJ1435−	3501	This study

^aR*, homopolymeric tract of G's that was subjected to phase variation and was repaired as described in Materials and Methods.

The ³¹P-NMR spectrum of (Fig. 2C) strain 3636 (CJJ1435⁻) did not show either peak Y or peak Z but yielded a previously unseen phosphorous resonance at δ_P 14.73 (designated peak X in Fig. 2C). A two-dimensional (2D) ¹H-³¹P-NMR experiment showed a connection between peak X and a proton resonance at δ_H 4.88 (Fig. 4). Since the ³¹P-NMR spectrum from a double transferase mutant, strain 3479 (CJJ1420⁻ CJJ1435⁻), showed no MeOPN resonances (data not shown), this activity must be encoded by CJJ1420. A new strain was constructed for use in additional structural analyses in which a repaired, overexpressed allele of CJJ1420 (termed CJJ1420R for repaired) was introduced into the hipO gene of a CJJ1435 mutant. This CJJ1420R CJJ1435⁻ strain was called 3718 (see Materials and Methods and Table 1).

FIG 4.

Spectrum of a ¹H-¹³C HSQC experiment showing the assignment of ¹H and ¹³C resonances of the CPS from mutant 3718.

Characterization of the capsule and MeOPN linkage in the C. jejuni CJJ1420R CJJ1435⁻ strain.

Sugar composition and linkage analysis of CPS from the CJJ1420R CJJ1435⁻ strain revealed that, as in 81-176 wild-type CPS, 3-substituted Gal and 3-substituted GlcNAc were part of the trisaccharide repeating block. However, the majority of the heptose in the CPS from the CJJ1420R CJJ1435⁻ mutant was present as the nonmethylated 2-substituted 6-deoxy-altro-heptose (6d-altro-Hep) in place of the 2-substituted 6-deoxy-3-O-methyl-altro-heptose derivative typically found in 81-176 wild-type CPS.

A more noteworthy structural deviation of the CPS from the CJJ1420R CJJ1435⁻ mutant was revealed by ³¹P-NMR spectroscopy, which displayed the previously the unseen resonance X at δ_P 14.72 (Fig. 2C). This ³¹P resonance did not belong to the previously characterized MeOPN substitutions at positions 2 and 6 of Gal and thus pointed toward the fact that the CJJ1420R CJJ1435⁻ strain produced a CPS with another MeOPN substitution. An accompanying 2D ¹H-³¹P heteronuclear correlation (HMBC) experiment indeed revealed a new interconnectivity between the new MeOPN ³¹P resonance at 14.72 ppm and a CPS ¹H resonance at 4.92 ppm (Fig. 3C). Using a 1D ¹H-¹H selective total correlation spectroscopy (TOCSY), the peak at δ_H 4.92 was irradiated revealing its connection to two ring proton resonances at δ_H 3.932 and δ_H 4.203 and to an anomeric resonance at δ_H 5.057. The anomeric resonance at δ_H 5.057 was in turn also irradiated, and its relationship to the ring resonances at δ_H 4.920, δ_H 4.203, and δ_H 3.932 was confirmed. These data, combined with a 2D ¹H-¹H COSY experiment, resulted in the assignment of ring protons: δ_H-1 5.057, δ_H-2 3.932, δ_H-3 4.203, δ_H-4 4.920, and δ_H-5 4.250. The new MeOPN linkage thus involves position 4 of this ring system.

The monosaccharide ring carbons associated with the CPS trisaccharide repeat were assigned through 2D ¹H-¹³C HSQC. Figure 4 shows the three anomeric cross-peaks in which “A” represents the H-1/C-1 of 6d-α-altro-Hep, “B” that of H-1/C/1 of α-Gal, and “C” that of H-1/C-1 of β-GlcNAc. Ring system A (6d-α-altro-Hep) carbons were located at δ 101.6 (A-1), 85.2 (A-2), 72.6 (A-3), 74.0 (A-4), 70.1 (A-5), 36.4 (A-6), 36.5 (A-6′), and 61.0 (A-7). The downfield carbon shift of A-2 at δ 85.2 agreed with the assignment of H-2 of the 2-substituted 6d-α-altro-Hep. Ring system B (α-Gal) carbons were assigned at δ 99.6 (B-1), 70.2 (B-2), 79.2 (B-3), 79.0 (B-4), and 71.6 (B-5). It could also be observed here that the C-3 of the 3-substituted α-Gal, at the downfield position of δ_C 79.2, matched H-3 (δ_H-3 4.203) of the previously described ring system containing MeOPN. Moreover, the associated C-4 at δ_C 79.0 (δ_H-4 4.920) of ring B (α-Gal) was characterized as that carrying the MeOPN in CPS from the CJJ1420R CJJ1435⁻ strain. The sole unit of the trisaccharide repeat in the β-configuration, that of 3-substituted β-GlcNAc (ring system C) contained C-1 at δ 105.0, C-2 at δ 59.7, C-3 at δ 78.0, and the CH₃ group of the N-acetyl moiety δ 25.1.

MeOPN is the immunodominant epitope recognized by an anti-81-176 conjugate vaccine.

Jiao et al. (32) showed that anticonjugate antibodies reacted with synthetic MeOPN-6-Gal. We examined reactivity of rabbit hyperimmune serum to an 81-176-CRM197 conjugate vaccine, CJCV1, by enzyme-linked immunosorbent assay (ELISA) to CPS isolated from wild-type 81-176 or mutants. The results (Fig. 5A) indicated that the reaction of anti-CJCV1 antibody was strongest to the wild-type CPS (titer: 5.9 × 10⁶). There was a marked reduction in titer to CPS purified from the CJJ1420⁻ mutant that expressed MeOPN-2-Gal and MeOPN-6-Gal (titer, 6.6 × 10⁵) and an even greater reduction to CPS purified from the CJJ1435⁻ mutant that expressed only MeOPN-4-Gal (endpoint titer 600). The difference in these latter two titers suggests either that there was very little MeOPN-4-Gal present in the immunizing conjugate vaccine or that the epitope was poorly immunogenic. Interestingly, the endpoint titer (8100) to CPS from the mpnC mutant that lacked all MeOPN was higher than that of the CJJ1435⁻ mutant, suggesting that the presence of MeOPN-4-Gal prevented the binding of antibodies specific for the polysaccharide chain.

FIG 5.

Endpoint ELISA titers to CPS from wild-type 81-176 and mutants. (A) Endpoint titers of a rabbit polyclonal serum to wild-type and mutant capsules. (B) Endpoint titers of five pools of human sera purchased commercially to wild-type and mutant capsules. The pool shown in red was the pool of sera used in Fig. 6.

Role of MeOPN in resistance to complement-mediated killing.

Although van Alphen et al. (24) reported that the population of their strain of 81-176 had the CJJ1420 gene in an off configuration, they constructed a double mutant mutated in both putative transferase genes and showed that this mutant was sensitive to complement killing, an observation consistent with earlier work with the mpnC mutant (23). When the variable regions of both MeOPN transferases were sequenced from the population of our version of strain 81-176, CJJ1420 was also in an off configuration, while CJJ1435 was on. However, when we determined the sequences of the variable regions of both transferases from 50 individual colonies of 81-176, 24% of the colonies expressed CJJ1420 in an on configuration (12/50), while 82% of the colonies expressed CJJ1435 in an on configuration (41/50). Only 6% of the colonies (3/50) expressed both genes in on configurations.

We compared serum resistance levels of the CJJ1420⁻ mutant, the CJJ1435⁻ mutant, and a double mutant lacking both transferases (strain 3479; see Table 1) using increasing amounts of NHS. The results (Fig. 6) indicated that at all concentrations of sera, the CJJ1435⁻ mutant, expressing MeOPN-4-Gal, was significantly more resistant than the wild type and that the CJJ1420 mutant expressing MeOPN-2-Gal and MeOPN-6-Gal was significantly more sensitive than wild-type 81-176 at concentrations of NHS ranging from 5 to 15%. When both mutants were complemented with their respective, repaired alleles (called CJJ1420⁻/C or CJJ1435⁻/C), the serum resistance returned to levels comparable to that of the wild type. However, mutation of both MeOPN transferases (strain 3479) resulted in enhanced sensitivity over the CJJ1420⁻ mutant and showed levels of sensitivity similar to those reported previously for another double transferase mutant (24) and for the mpnC mutant (23).

FIG 6.

Resistance of C. jejuni strains to increasing amounts of NHS. Bacteria were exposed to increasing amounts of NHS for 1 h at 37°C, and survivors were enumerated by plate counts. The CJJ1435⁻ mutant was significantly different from the wild type at all four concentrations of NHS (P < 0.05). The CJJ1420⁻ mutant was significantly less serum resistant than the wild type at 5% NHS (P < 0.05), 10% NHS (P < 0.005), and 15% NHS (P < 0.05). There was no significant difference between the complements of the two mutants and the wild type at any concentration of NHS. The resistance of the double transferase mutant (CJJ1420⁻ CJJ1435⁻) was significantly lower than that of the wild type at 5% NHS (P < 0.0005), 10% NHS (P < 0.005), and 15% NHS (P < 0.05). CJJ1420⁻/C and CJJ1435⁻/C strains are the complements of the CJJ1420⁻ and CJJ1435⁻ mutants, respectively.

Phase variation of MeOPN transferases.

The serum killing data suggested that expression of MeOPN-4-Gal enhanced serum resistance. We reasoned that exposure to high levels of NHS would select for cells in the population in which the CJJ1420 gene was in the on configuration. An aliquot of an overnight culture of 81-176 was plated for single colonies on Mueller-Hinton agar, and another aliquot was exposed to 20% NHS for 1 h prior to plating for single colonies. The variable regions of the CJJ1420 and CJJ1435 genes were sequenced from these individual colonies. The results indicated that without exposure to NHS the CJJ1420 gene was in the on configuration in 9.5% of the 42 colonies sequenced and CJJ1435 was on in 90.5% of the colonies sequenced, which is consistent with the data presented above. In contrast, after exposure to NHS, CJJ1420 was on in 100% of the 43 colonies sequenced and CJJ1435 was on in 53.5% of the 43 colonies sequenced. Without exposure to NHS, 4.8% of the colonies were on for both genes, while after exposure to NHS, 53.5% of the colonies were on for both genes. No colonies were off for both genes.

NHS contains antibodies to the 81-176 polysaccharide chain.

ELISAs were performed on five commercially available NHS samples, including the one used in the experiments described above, against CPS purified from 81-176 wild type and the mutants. The results, shown in Fig. 5B, indicated that there were preexisting antibodies in NHS to the 81-176 CPS (mean titer, 800), but that the titer against CPS from all of the mutants defective in MeOPN was significantly higher than the titer to wild-type CPS. CPS from the mpnC mutant showed the highest titer (26,400), suggesting that the presence of MeOPN on the wild-type CPS blocks the attachment of preexisting anti-glycan antibodies to the CPS. Reactivity against CPS from the CJJ1420⁻ mutant was also significantly higher than that from wild-type CPS, a finding consistent with the loss of the MeOPN-4-Gal that is expressed in a minority of the cells in the population. The reactivity to CPS from the CJJ1435⁻ mutant was higher than that of CPS from the CJJ1420⁻ mutant and slightly lower than that of the CPS from the mpnC mutant, observations consistent with loss of the MeOPN-2-Gal and MeOPN-6-Gal modifications that are expressed in the majority of the cells in the population. Collectively, these data suggest that NHS contains antibodies that cross-react with the polysaccharide chain but not MeOPN.

DISCUSSION

We have demonstrated that, in addition to the two previously reported sites of MeOPN modification, the 81-176 CPS can be modified at a third site, MeOPN-4-Gal. It appears that the transferase encoded by CJJ1435 is bifunctional and is responsible for addition of MeOPN to both the 2 and the 6 positions of Gal, although modification at Gal-2 appears to be the preferred site based on the relative ³¹P-NMR signals. To our knowledge, this is the first report of a bifunctional MeOPN transferase. Mutation of CJJ1435 not only resulted in the loss of MeOPN-2-Gal and MeOPN-6-Gal but also resulted in the appearance of a new ³¹P-NMR signal that was shown to correspond to MeOPN-4-Gal, which is encoded by CJJ1420. When grown in vitro, most 81-176 cells expressed CJJ1435, and only a subset of the population (9.5 to 24%) expressed CJJ1420. The MeOPN-4-Gal ³¹P-NMR signal was initially observed in a CJJ1435 mutant and was characterized in a strain in which the CJJ1420 transferase was overexpressed in a CJJ1435⁻ mutant background. Thus, the inability to transfer MeOPN to 2-Gal and 6-Gal appeared to enhance modification at the 4 position of Gal, perhaps due to an increased pool of MeOPN in the cell. Interestingly, CPS from the CJJ1420R CJJ1435⁻ mutant also contained a majority of 6d-altro-Hep in place of the typical 3-O-methyl-6d-altro-Hep normally found in 81-176. The reason for this change remains uncertain, but a similar shift in Hep composition in the 81-176 CPS was observed previously in a deep rough LOS mutant (31).

Complement-mediated killing of C. jejuni has been reported to occur primarily by the classical pathway (24, 34), and it is thought that the CPS likely functions to shield the cell from naturally occurring antibodies in NHS that cross-react with surface proteins. Data presented here, however, suggest that MeOPN moieties also serve to protect the polysaccharide chain from preexisting antiglycan antibodies in NHS. The presence of MeOPN on the wild-type CPS inhibited the binding of these naturally occurring antibodies, as measured by ELISA, compared to the CPS from the mpnC mutant lacking all MeOPN. However, the CJJ1435⁻ mutant lacking the major modifications at the 2 and 6 positions of Gal bound more antibody than did the CJJ1420⁻ mutant lacking the minor MeOPN-4-Gal modification. This appears to be inconsistent with the observation that the CJJ1420⁻ mutant, lacking MeOPN-4-Gal, was more sensitive to complement-mediated killing than the wild type, and the CJJ1435⁻ mutant lacking MeOPN-2-Gal and MeOPN-6-Gal was more serum resistant than the wild type. The preexisting antibodies to the 81-176 polysaccharide chain in NHS are likely directed toward the rather common β-d-GlcpNAc-(1-3)-α-d-Galp linkage (altro-Hep is a rare sugar). The importance of modification at the 4 position of Gal to serum resistance may relate to the fact that it is the closest site of modification to the GlcNAc-(1-3)-Gal linkage and may be more effective at impeding binding of cross-reacting antiglycan antibodies (Fig. 1). However, alterations in the level and site of MeOPN modification may also alter the secondary or tertiary structure of the capsule and affect the accessibility of antibodies to the glycan chain differently on the intact cell in the serum killing assay than that measured against purified CPS in an ELISA.

Although NHS contain preexisting antibodies to the polysaccharide chain, the MeOPN modifications appear to be immunodominant epitopes on 81-176-CRM197 conjugate vaccines. ³¹P-NMR analyses of the CPS used to generate the conjugate used to generate the hyperimmune serum used in this study indicated the presence of MeOPN-2-Gal only, and at the time of its synthesis, we were unaware of the additional modifications (data not shown). Moreover, subsequent examination of the seed strain used to produce the vaccine revealed that the levels of expression of CJJ1420 and CJJ1435 were similar to those reported here for the laboratory stock of 81-176 (data not shown). As shown in Fig. 5A, the endpoint titer of rabbit hyperimmune serum to this conjugate was >2 logs higher against wild-type CPS than CPS from the mpnC mutant. Interestingly, the CPS of the CJJ1435⁻ mutant, expressing only MeOPN-4-Gal, had a lower ELISA titer to the anticonjugate antibody than CPS from the mpnC mutant, again suggesting the ability of the MeOPN modifications to block access to the glycan. The immunodominance of MeOPN in conjugate vaccines appears to be comparable to the immunodominance of nonstoichiometric O-acetyl groups on the polysaccharide conjugates based on other bacterial pathogens (33–38). Nonstoichiometric modifications to sugars confer considerable heterogeneity to polysaccharide chains and can affect immunogenicity (39). This heterogeneity is more complex for C. jejuni, since phase variation modulates both the level and the position of MeOPN modifications. It has been reported that early in infection with C. jejuni patient sera could induce low levels of complement-mediated killing of multiple C. jejuni strains, but that, after 48 h of infection, patients developed higher-level serum bactericidal titers that were strain specific (34), an observation that may relate to MeOPN-sugar specific antibody responses. We are exploring the possibility that antibodies directed to MeOPN-sugar moieties in conjugate vaccines can induce serum bactericidal killing.

C. jejuni is characterized by variability of surface antigens (13). Phase variation of genes affecting LOS, CPS, and flagella are well documented (14–16, 24). Recent studies have also shown that, in addition to phase variation, high-frequency mutations can occur in genes that affect motility (40, 41). More recently, extensive variations, including insertions, deletions, and missense mutations of two genes, apt and purF, involved in stress responses of C. jejuni have been reported (42). Different alleles of these two genes were associated with various survival abilities under different stress conditions. Collectively, these observations support the suggestion that C. jejuni is a quasispecies containing multiple genotypes that can be selected based on their relative fitness in a particular environment. Phase variation of the MeOPN transferases in C. jejuni 81-176 provides another example of this bet-hedging phenomenon. The organism is generally considered to be relatively serum sensitive (43) and, when grown in vitro, the MeOPN transferases of strain 81-176 are in a configuration that does not allow for maximal complement resistance, meaning that the MeOPN-4-Gal transferase is predominantly in an off configuration. Exposure to NHS selected for the minor population of cells that were expressing MeOPN-4-Gal. Thus, the levels of serum resistance measured in vitro for a population may not reflect the levels of resistance that can be achieved in vivo. C. jejuni is an invasive pathogen and would be exposed to increasing levels of NHS as it invaded through the intestinal epithelium. Thus, it may be that only a subpopulation of cells is capable of survival following invasion.

MATERIALS AND METHODS

Strains and growth conditions.

All work was done in the 81-176 strain of C. jejuni. Mutants of this strain are listed in Table 1. C. jejuni was routinely cultivated on Mueller-Hinton agar at 37°C under microaerobic conditions. Multiple working freezer stocks were made of each mutant to minimize in vitro passage. Media were supplemented with antibiotics as needed. For capsule extraction, cells were grown in porcine brain heart infusion broth (Difco).

Oligonucleotide primers.

All oligonucleotide primers used are listed in Table 2 and were synthesized by Life Technologies.

TABLE 2.

Primers used in this study

Primer	Sequence (5′–3′)
pg12.13	GGAATTCGATGATTATTTTATAGATATTGGTGTGCCTGAGG
pg12.14	CCCTCGAGGGGATATTACTATCGACTATATCGTAACTATTACAACC
pg12.25	CCAGCTGAACTTGCTTGGGAGATG
pg12.26	GGGATATTACTATCGACTATATCGTAACTATTACAACC
pg10.07	GTGTGATGTGGTGGTTACGTTGAATTCGGG
pg10.08	CTCAAATCTATAGTAAGTGGCATGATTAACATGCCAAGC
pg14.67	CATCCTTATCCTTCATTACTTGATCC
pg14.68	CGTGGAACATGTTTATTTATCATATGC
pg12.31	CATGAAAATCCTGAGCTTGGTTTTGATG
pg12.32	GTATTTTAAAACTAGCTTCGCATAATAAC
pg12.33	GCGCCCATGGGTTAACGGAGCACTTCCATGACCACCTCTTCC
pg12.34	GCGCCCATGGTCTAGAAGATCTCCTATTTATGCTGCTTCTTTGCTTCTGG
pg12.29	CGGGATCCAAAGGAGAAACCCTATGTATAACCCAAACTCAGC
pg12.30	GGAATTCGTAAAATCCCCTTGTTTCATATTGATTCCTTTCTCTAATTTTAAACAC
pg12.37	GCTATGATTGAGTTTACAAACAATGGAGGAGGATATATAGCATTATTTAAAAAACTC
pg12.38	GAGTTTTTTAAATAATGCTATATATCCTCCTCCATTGTTTGTAAACTCAATCATAGC
pg14.35	GGAATTCCTATATTATAAGATAATAACACAATTCGCCTCCTATG
pg14.03	CGGGATCCAGGAGAAACCCTATGTATAACCCAAACTCAGC
pg14.09	GCTATGATTGAGTTTACAAACAATGGAGGAGGATATATAGCATTATTTAAAAAACTC
pg14.10	AGTTTTTTAAATAATGCTATATACCTCCTCCTTTGTTTGTAAACTCAATCATAGC
pg12.17	ATGTATAACCCAAACTCAGCTATAGAAAGAG
pg15.13	GAGAATTGAGGATACTATGTCCAGTTAATCC
pg15.14	GCTTTCTCTCCTGTTCCATGGCCTCC

NMR and gas chromatography-mass spectrometry (GC-MS) analyses.

¹H-, ¹³C-, and ³¹P-NMR experiments were recorded using a Bruker AMX 400 spectrometer equipped with a Cryoprobe. Experiments were run at 293 or 315 K. Heteronuclear single quantum correlation spectroscopy (HSQC) and heteronuclear multiple bond correlation spectroscopy (HMBC) experiments were performed using Bruker TopSpin 3.0 software. Prior to analysis samples were lyophilized with D₂O (99.9%) three times. The HOD resonance at δ_H 4.821 was used as the internal standard for ¹H experiments. A standard of trimethylpropanoic acid in D₂O was used to establish a reference for the HOD signal. Orthophosphoric acid (δ_P 0.0) was used as the external reference for all ³¹P experiments.

Monosaccharides were characterized as alditol acetate derivatives. The CPS was first digested with 4 M trifluoroacetic acid at 105°C, and then the monomers were reduced with NaBD₄ in water overnight at room temperature. The alditols were acetylated with acetic anhydride at 105°C. The resulting alditol acetates were extracted using dichloromethane and analyzed by GC-MS in a ThermoFinigan PolarisQ ion trap equipped with a DB-17 capillary column.

Rabbit polyclonal antisera.

Rabbit hyperimmune polyclonal antibodies were generated against the 81-176/CRM197 conjugate vaccine, CJCV1 (32).

PCR.

All PCR products generated for cloning or sequence analysis were amplified using Phusion high-fidelity polymerase (New England BioLabs). All other PCRs used Taq polymerase (Applied Biosystems/Life Technologies).

Mutation of CJJ1420.

CJJ1420 was cloned into pCRScript using primers pg12.13 and pg12.14 that introduced EcoR and XhoI sites, respectively. This plasmid was subjected to transposon mutagenesis using Tnp Km (Epicentre) and individual kanamycin-resistant (Km^r) transposon insertions were sequenced with primers internal to the transposon to determine the site of insertion. A nonpolar transposon insertion at bp 367 of the 1,779-bp gene was used to electroporate 81-176 to obtain Km^r using methods previously described. The putative mutation was confirmed by PCR using the primers pg12.25 and pg12.26 that bracket the insertion point of the kanamycin gene, and this mutant was called strain 3477.

Mutation of CJJ1435.

CJJ1435 was cloned into pCRScript using primers pg10.07 and pg10.08. The cat cassette from pRY109 (34) was cloned into a unique NcoI site located at bp 747 of the 1,813-bp gene. Clones were partially sequenced to determine orientation of the cat cassette, and one in which the gene was inserted in the same orientation as CJJ1435 was used to electroporate 81-176 to obtain chloramphenicol resistance (Cm^r). Putative clones were confirmed by PCR using pg14.67 and pg14.68 that bracket the NcoI site of insertion, and the resulting mutant was called strain 3636.

Construction of a double mutant in both putative MeOPN transferases.

Strain 3477 (CJJ1420::aph3) was electroporated to obtain Cm^r with the same plasmid used to generate strain 3636, thus generating a double mutant, strain 3479 (Table 1).

Construction of a hipO insertion vector.

The hipO gene of 81-176 (CJJ1003), encoding the nonessential enzyme benzoylglycine amidohydrolase, was cloned into pCRScript using primer set pg12.31 and pg12.32. A unique XbaI site was introduced in the center of the hipO gene by inverse PCR with primer sets pg12.33 and pg12.34. This plasmid was called pCPE3490.

Construction of strains expressing repaired alleles of CJJ1420 and CJJ1435.

The CJJ1420::aph3 mutant was complemented with a repaired allele as follows. The wild-type CJJ1420 gene was PCR amplified using primers pg12.29 and pg12.30, which introduced BamHI and EcoR1 sites, respectively, and the resulting amplicon was cloned into BamHI and EcoRI digested pCPE108, which contains the σ²⁸ promoter from flaA cloned between the XbaI and BamHI sites of pBluescript (44). The phase-variable G9 tract within CJJ1420 was repaired by QuikChange (Life Technologies) mutagenesis such that the G9 was changed to GGAGGAGGA using the primers pg12.37 and pg12.38. The entire insert was moved as an EcoRI-NotI fragment into pBluescript, and a SmaI-ended cat cassette from pRY109 (34) was inserted into the EcoRV site 3′ to the repaired CJJ1420 gene. The entire construction (σ²⁸-CJJ81176_1420+cat) was PCR amplified with forward and reverse primers and cloned into the unique XbaI site within the hipO gene in pCPE3490 (described above) that had been blunted. This construction, called pCPE3494, was used to electroporate 3477, the CJJ1420::cat mutant, to obtain Km^r, generating strain 3498 (designated the CJJ1420⁻/C strain).

The CJJ1435::cat mutant was complemented using a similar approach. Plasmid pCPE108 was modified to contain an aph3 gene at the XhoI site in the polylinker, generating pCPE3583. CJJ1435 was PCR amplified using primers pg14.35 and pg14.03, which introduced BamHI and EcoRI sites, respectively, and cloned into BamHI and EcoRI-digested pCPE3583. The phase-variable G9 tract located within the coding region of CJJ1435 was subjected to site-directed mutagenesis as described above using the primers pg14.09 and pg14.10. The repaired CJJ1435 gene and the adjacent aph3 gene were PCR amplified using forward and reverse primers and cloned into an EcoRV site on a plasmid containing the astA of strain 81-176, as previously described (44, 45). This plasmid was used to electroporate the CJJ1435 mutant, 3636, to obtain Km^r, generating strain 3637 (designated the CJJ1435⁻/C strain).

A strain was also constructed that overexpressed CJJ1420 in a CJJ1435 mutant background for NMR studies. Plasmid pCPE3494 that was used to construct the complement of the CJJ1420 mutant (described above) was electroporated into wild-type 81-176 to generate strain 3501. An apramycin cassette from plasmid pAC1 (46) was inserted into the unique NcoI site in the clone of CJJ1435 described above. This clone was electroporated into strain 3501 to generate strain 3718 (Table 1).

Anti-CPS ELISAs.

To determine the anti-CPS response in hyperimmune rabbits or NHS, Carbo-BIND plates (Corning, Corning, NY) were coated with 100 μl of oxidized CPS from wild-type or mutant strains (2 μl/ml in sodium acetate buffer [pH 5.5]) for 1 h at room temperature according to the manufacturer's instructions. Plates were washed with 1× phosphate-buffered saline–0.05% Tween 20 (PBST), blocked with 5% fetal calf sera in PBST (5% FCS-PBST) for rabbit or 1% casein for NHS for 1 h at 37°C, and washed again with PBST. All sera were serially diluted in blocking buffer in duplicate and incubated for 1.5 h at 37°C. After washing, horseradish peroxidase-conjugated goat anti-rabbit IgG (Sigma-Aldrich, St. Louis, MO) or goat anti-human IgG (KPL, Gaithersburg, MD) was diluted in blocking buffer and added at 100 μl per well for 1 h at 37°C before washing. ABTS-peroxidase substrate (KPL) for rabbit or TMB (eBiosciences, San Diego, CA) for NHS was used as a detection reagent and the optical density at 405 nm (OD₄₀₅) or OD₄₅₀, respectively, was measured. The mean OD of negative-control wells (coating buffer alone) plus three standard deviations was used to determine the endpoint titer.

Phase variation of MeOPN transferases.

To determine the percentage of a population expressing each of the phase-variable genes encoding the CJJ1420 and CJJ1435 transferases, frozen working stocks of bacteria were thawed, serially diluted and plated for single colonies on Mueller-Hinton agar at 37°C for ∼36 to 40 h. Individual colonies were picked and resuspended in 20 μl of water. The suspensions were boiled for 10 min and centrifuged for 5 min in a microfuge. The supernatants were removed and 1 μl of each was used as a template for PCR. The variable regions of the two MeOPN transferases were PCR amplified with pg12.17, which maps to the conserved region, and pg15.13, which is specific for CJJ81176_1420 or pg15.14, which is specific for CJJ81176_1435. The resulting PCR products were purified and sequenced with pg.12.17.

Complement killing.

Pooled NHS were purchased from Sigma and a single lot was used for all experiments. Assays were done as described by Maue et al. (23), except that a range of NHS was used. Assays were repeated two to nine times for each strain. Statistics were prepared using GraphPad Prism.