DNA Rearrangements in the Flagellin Locus of an flaA Mutant of Campylobacter jejuni during Colonization of Chicken Ceca

Abstract

Campylobacter jejuni is an enteropathogen for humans but commensal for chickens. In both hosts, the flagella and motility are important colonization factors. The flagellin gene is duplicated in Campylobacter, but only one flagellin gene, flaA, is sufficient for motility. We found that, during colonization of the chicken intestine, a nonmotile flaA mutant of C. jejuni underwent rearrangements within its flagellin locus, thereby regaining its motility and colonization capacity. In contrast, in vitro motile revertants isolated from liquid culture showed different flagellin DNA rearrangements than after reversion in the chicken.

Campylobacter spp. have two almost identical flagellin genes which encode the structural subunit of the flagellum. The flagella of Campylobacter are highly immunogenic surface structures (3, 10), essential for motility and virulence (11, 13, 14). The two flagellin genes, flaA and flaB, 1.7 kb each, are located adjacent to one another, but they have different promoters (5, 12). The flaB gene is not needed for motility, and therefore it is speculated that flaB serves as a depot for antigenic variation (1, 15) or has a function in motility under different circumstances (15, 16). The significance of this gene duplication has been the subject of several studies, and both differential expression (15) and recombinations (1, 16) were detected in vitro. However, so far, no evidence was found that these processes also may occur in vivo.

Mutant R1 of Campylobacter jejuni 81116 has an inactivated flaA gene (flaA::Km^r), makes very short, nonfunctional flagella (13), and exhibits reduced colonization levels in chicken ceca (14). We performed an experiment in order to determine the genetic stability and the colonization behavior of R1, also in the presence of wild-type campylobacters, during colonization of chicken ceca over a period of 6 weeks. In addition, genetic stability and reversion to motility were tested in vitro.

Sixty specific-pathogen-free chickens (layers) received a dose of 2.5 × 10⁵ or 5 × 10⁵ CFU of R1 bacteria (grown in brucella broth with 1% yeast extract and kanamycin [30 μg/ml]) on days 1 and 15 of age, respectively. A control group of 60 animals received no R1 bacteria. Both groups were orally inoculated with wild-type C. jejuni 81116 (10⁵ CFU) at day 29 of age. Every week after the beginning of the experiment, 10 animals of each group were sacrificed, and colonization was determined as the number of CFU per gram of cecal content. Serial dilutions were plated on selective Campylobacter agar medium (Difco), with or without kanamycin (30 μg/ml), and incubated under microaerobic conditions. The results of reisolation are summarized in Table 1. One week after the first dose, all Campylobacter colonies recovered from the ceca were still kanamycin resistant and colonized at levels of about 10⁷ CFU/g of cecal content, which is comparable to the results obtained by Wassenaar et al. (14) but higher than that reported by Nachamkin et al. (11). This might be due to differences between the C. jejuni strains or the breed of animals. At later time points, campylobacters were reisolated that had become kanamycin sensitive and were able to colonize at relatively high levels (Km^s Col⁺). In addition, several animals contained kanamycin-resistant, well-colonizing (Km^r Col⁺) campylobacters. These results are in constrast with previous findings (11, 14). However, bacteria of both phenotypes were clearly motile compared to R1 bacteria using dark-field light microscopy, but the level of motility was not quantified. The bacteria with these phenotypes were called pseudorevertants, and apparently these bacteria had adapted to the conditions in the chicken ceca and were able to colonize at wild-type levels.

TABLE 1.

Reisolation of Campylobacter colonies from chicken ceca after colonization with the flaA mutant R1 and challenge with wild-type bacteria^a

Chicken no.	Kanamycin present in agar medium	No. of bacteria in cecum (CFU/g)
		Day 7	Day 15	Day 21	Day 29	Day 35	Day 42
1	Yes	9 × 107	<1 × 104	<1 × 103	1 × 108*	6 × 106*	1 × 106
	No	1 × 108	4 × 107	5 × 106	2 × 108*	4 × 108*	2 × 107
2	Yes	1 × 107	<1 × 104	<1 × 103	<1 × 103*	2 × 108*	<1 × 103
	No	2 × 107	1 × 106	4 × 108	5 × 108*	3 × 108*	2 × 106
3	Yes	2 × 107	<1 × 104	<1 × 103	<1 × 103	1 × 103	<1 × 103
	No	6 × 107	1 × 107	2 × 107	1 × 107	3 × 108	1 × 108
4	Yes	3 × 105	<1 × 104	<1 × 103	<1 × 103	1 × 104	2 × 108
	No	1 × 106	2 × 107	3 × 106	2 × 106	4 × 108	1 × 108
5	Yes	3 × 107	<1 × 104	<1 × 103	<1 × 103	<1 × 103*	<1 × 103
	No	2 × 107	1 × 107	3 × 107	2 × 108	1 × 108*	1 × 106
6	Yes	2 × 107	<1 × 104	<1 × 103	<1 × 103	1 × 107*	2 × 106
	No	4 × 107	<1 × 104	4 × 107	2 × 106	3 × 108*	2 × 106
7	Yes	2 × 107	1 × 105	<1 × 103	<1 × 103	1 × 105	<1 × 103
	No	1 × 107	1 × 107	3 × 109	4 × 103	1 × 109	2 × 108
8	Yes	1 × 107	<1 × 104	<1 × 103	<1 × 103	<1 × 103*	5 × 108
	No	2 × 107	2 × 107	5 × 107	1 × 109	2 × 108*	4 × 108
9	Yes	1 × 107	<1 × 104	6 × 108	<1 × 103	1 × 103	2 × 107
	No	1 × 107	2 × 108	2 × 108	2 × 108	1 × 108	1 × 108
10	Yes	4 × 107	<1 × 104	<1 × 103	<1 × 103	5 × 104*	<1 × 103
	No	4 × 106	1 × 107	1 × 107	4 × 107	1 × 108*	<1 × 103

^aBacteria giving values marked with an asterisk (two individual colonies and mixed cultures of more than 100 colonies) were analyzed at the DNA level by Southern blot analysis and PCR (see Table 2). 

In order to determine which changes had occurred in the flagellin locus, we analyzed those pseudorevertants reisolated at days 29 and 35 (Table 1) by Southern blot hybridization and PCR (Table 2). For Southern blot analysis and PCR, two individual colonies as well as mixed cultures (>100 colonies) were tested, and all showed the same results. In the Southern blot, the Km^s Col⁺ phenotype showed only one band of 2.5 kb after hybridization and had a PCR fragment of 2.5 kb, indicating a deletion of 2 kb, about the size of one flagellin gene, suggesting that intrachromosomal recombination between the 5′ end of flaA and the 5′ end of flaB led to the loss of the kanamycin resistance cassette as well as the 3′ end of flaA, the intergenic region, and the 5′ end of flaB, thereby creating a flaA-flaB chimeric gene similar to that reported by Alm et al. (1). We have designated this genotype RM1 (Fig. 1).

TABLE 2.

Characteristics of campylobacters: flaA mutant R1 and wild type (WT), reisolated pseudorevertants from chicken ceca on days 29 (before infection) and 35 (after infection), and reisolated pseudorevertants from in vitro cultures

Campylobacter origin	Kanamycin, colonization, and motility phenotypes	Positive bands (kb) in Southern blot analysisa	Length (kb) of flagellin PCR fragmentb	Genotype
WT and R1
WT	Kms Col+	2.5 + 1.9	4.5	WT
R1	Kmr Col±	2.0 + 1.9 + 1.15 + 1.03	6	R1
Reisolated pseudorevertants from chicken ceca
Day 29	Kmr Col+	2.5 + 1.9 + 1.15/1.18c + 1.03	8	RM2
	Kms Col+	2.5	2.5	RM1
Day 35	Kmr Col+	2.5 + 1.9 + 1.15/1.18 + 1.03	NT	RM2
	Kmr Col+	2.5 + 1.9 + 1.15/1.18 + 1.03	8	RM2
	Kms Col+	2.5	NT	RM1
	Kmr Col+	2.5 + 1.9 + 1.15/1.18 + 1.03	8	RM2
	Kms Col+	2.5 + 1.9	NT	WT
	Kmr Col+	2.5 + 1.9 + 1.15/1.18 + 1.03	NT	RM2
Reisolated pseudorevertants after in vitro growth
With kanamycin	Kmr Mot+	NT	6d	RM3
Without kanamycin	Kms Mot+	NT	2.5	RM1

^aThe probe used contained both kanamycin resistance gene and flaA sequences (see Fig. 1) 

^bPCR with primers that amplify the complete flagellin locus: TAACAACCAATCGTGGTGC (approximately 500 bp upstream of flaA) and TCTTGCATGAAATTTTAGGAC (approximately 100 bp downstream of flaB). NT, not tested. 

^c1.15/1.18 represents a double band (see Fig. 1). 

^dAdditional PCRs were performed with primers upstream of flaA and downstream of flaB in combination with primers that anneal to the ends of the kanamycin resistance gene and point outwards (ATACCTTAGCAGGAGACATT and GGGCGGACAAGTGGTATGAC), which showed that the kanamycin gene is located in flaB. 

FIG. 1.

Schematic representation of flagellin genes in C. jejuni 81116 genotypes. flaA is black, flaB is gray, and the Km^r gene is white. The flagellin gene organization of strain R1 (top) and the wild type (WT, bottom) have been determined before (12, 13). The gene organization of RM1 is based on PCR and Southern blot analysis (Table 2) and partial sequence data (not shown). The map of RM2 is based on sequence analysis of the 8-kb PCR fragment. The map of RM3 is based on PCR analysis with multiple primer sets (Table 2). EcoRV restriction sites are shown as short vertical lines, and the sizes of the EcoRV restriction fragments correspond to the sizes of hybridizing fragments indicated in Table 2. The flaA-Km^r PCR fragment of R1, obtained with a flaA (primers A and C)-specific primer set (16), was used as a probe in Southern blot analysis (see Table 2).

Interestingly, Murphy and Belas (9) have isolated similar spontaneous flagellin recombinants of Proteus mirabilis after in vitro growth and also after urinary tract infection of mice. They suggest that flagellin sequences are not lost after recombination but are retained elsewhere on the chromosome, presumably at a silent locus. Recently, a third Campylobacter flagellin-like gene has been described by Chan et al. (4). But if that were the case in the RM1 genotype, Southern hybridization would have shown the presence of these (additional) flagellin sequences.

Molecular analysis of the Km^r Col⁺ phenotypes exhibited a unique hybridization pattern of four bands, different from those of R1, the wild type, and RM1. This genotype was named RM2. PCR of the flagellin locus gave a fragment of 8 kb which was sequenced completely (GenBank accession number AF202168). The flagellin locus of RM2 consists of one chimeric flaA-flaB gene, one chimeric flaB-flaA gene which is interrupted by the kanamycin resistance gene, and one wild-type flaB gene (Fig. 1). Clearly, RM2 is not the result of one simple recombination because duplication of flagellin sequences took place. Although no selective pressure from antibiotics was present during colonization of chicken ceca for maintaining the kanamycin resistance cassette, RM2 still possessed the Km^r gene. Possible mechanisms for the appearance of RM2 are (i) two recombinational events between sister chromosomes during one cell division (8); (ii) one recombination between sister chromosomes (8) followed by natural transformation (1, 15) with RM1 or wild-type DNA and subsequent exchange (two crossovers) of the kanamycin resistance cassette with the chimeric gene from RM1 or flaB of the wild type; or (iii) natural transformation followed by recombinations (1, 15). RM2-like mutants were, however, not identified in the P. mirabilis study (9) mentioned above. The RM2 genotype resembles that of a mutant that was previously described by Alm et al. (1), and they indicated that natural transformation could be an important mechanism. In our animal experiment, the number of animals with the RM2 genotype increased from one to five after inoculation with wild-type C. jejuni bacteria. This might give the impression that wild-type bacteria are involved in flagellin gene transfer—but not a prerequisite—in the origination of RM2. It is also possible that it is a matter of the time during which recombinations can take place and in the animal selection for the motile phenotypes.

The recombinations that were detected led to the synthesis, under control of the flaA promoter, of a chimeric FlaA-FlaB flagellin protein almost identical to FlaA (only four amino acid substitutions). Since Campylobacter is commensal in chickens, there is no antibody response that abolishes motility and thereby eliminates Campylobacter bacteria from the chicken intestine. Thus, the appearance of RM1 and RM2 is not caused by negative immune selection (2) but rather by positive selection for the colonization capacity of the new phenotypes.

In vitro (pseudo)reversion was tested by growth of R1 bacteria in liquid medium with and without kanamycin. After several time points, the motility of the bacteria was observed by dark-field light microscopy, and after 4 weeks (one passage/week), motile cells were observed. By means of plating in motility 0.4% agar medium, single colonies showing a motile phenotype were isolated, and the flagellin gene locus was analyzed by PCR. Surprisingly, we found the RM1 genotype (only in cultures without kanamycin) but not the RM2 genotype. Instead, another genotype was identifed in cultures with kanamycin, which we called RM3, not found after R1 colonization of chicken ceca. PCR analysis was done on two colonies with primers upstream of flaA and downstream of flaB (Table 2) in combination with primers that anneal to the ends of the kanamycin resistance gene and point outwards. This showed that the flagellin locus in RM3 was changed and the Km^r cassette had moved from flaA to the flaB gene (Fig. 1), as described before by Wassenaar et al. (16).

The most striking finding in this study is the difference between the rearrangements after in vitro versus in vivo pseudoreversion: only in animals was a duplication of flagellin sequences detected. It seems that the circumstances in the chicken intestine apply a different selection pressure on the flagellin locus than in vitro conditions. Wild-type bacteria also maintain both flagellin genes, although any mutant with an active flaA promoter and an intact flagellin coding region is still motile and able to colonize the ceca of chickens. The question remains what the possible function is for the flaB gene or the structural features of a FlaB flagellum. These data affirm the differential role of the two flagellin genes in the complex epidemiological life cycle of Campylobacter.

This is the first time that recombinations, deletions, and duplications between flagellin gene sequences of Campylobacter have been demonstrated in a natural host. In a study by Wassenaar et al. (17), no evidence for recombination between flagellin gene sequences could be found. This may be due to the short colonization period (5 days) after which bacteria were reisolated from the chickens. Recently, by sequencing many flaA and flaB genes, Harrington et al. (7) presented indirect evidence that recombination takes place. Finally, Hanninen et al. (6) described recombinations between Campylobacter strains after cocolonization of these strains in the chicken intestine. There is increasing evidence that genetic recombination can occur in Campylobacter in the chicken intestine, which is a frequently encountered niche. This observation will have an important impact on our understanding of the molecular epidemiology of campylobacters and the use of live campylobacters as vaccines.