Identification of Campylobacter jejuni genes involved in commensal colonization of the chick gastrointestinal tract

Chick gastrointestinal colonization capacity of C. jejuni strain 81–176

Campylobacter jejuni strain 81–176 was isolated during an outbreak of C. jejuni disease associated with consumption of raw milk or hand milking of cows (Korlath et al., 1985). This strain elicits gastroenteritis in humans (Black et al., 1988) and reproduces the diarrhoeal syndrome in a ferret model of disease (Doig et al., 1996), but its colonization capacity in chicks has not been studied. We determined whether this strain could naturally colonize the gastrointestinal tracts of chickens, resulting in commensalism using a 1-day-old chick model of infection (Beery et al., 1988; Stern et al., 1988).

Approximately 12–36 h after hatching, chicks were orally infected with a range of doses of C. jejuni strain 81–176 and at 7 days post infection C. jejuni in the caeca were enumerated. With inocula of approximately 900 bacteria and above, all chicks were colonized and the bacterial loads in the caeca ranged from 3 × 10⁸ to 8 × 10⁹ colony-forming units (cfu) per gram of caecal content (Fig. 1). Caecal colonization was observed after inoculation with as low as approximately eight, 20 and 48 organisms (Fig. 1). Four of five chicks inoculated with eight organisms were colonized and three of four colonized chicks contained bacterial loads above 2 × 10⁸ cfu per gram of caecal content after 7 days, similar to the caecal bacterial loads of chicks infected with inocula of higher orders of magnitude (Fig. 1). Colonized chicks appeared healthy and showed no signs of disease, as reported for chicks infected with other C. jejuni strains (Beery et al., 1988; Stern et al., 1988). Chicks gavaged with phosphate-buffered saline (PBS) alone as a negative control did not have C. jejuni in their caeca 7 days post infection.

We next studied the dynamics of colonization by C. jejuni strain 81–176 throughout the chick gastrointestinal tract by giving three different doses of the bacterium orally (∼10⁶, ∼10⁴ and ∼10¹) and determining the number of bacteria in the upper gastrointestinal tract (the proximal and distal small intestines) and the lower gastrointestinal tract (the caeca and large intestines) over time. At day 1 post infection, chicks infected with 3 × 10⁶ bacteria harboured the majority of C. jejuni in the lower gastrointestinal tract, with levels in the caeca and large intestines greater than 3 × 10⁷ and 3 × 10⁶ cfu per gram content respectively (Fig. 2A); the number of C. jejuni in the upper gastrointestinal tract (the proximal and distal small intestines) was at least 100-fold lower. By day 4, the levels of bacteria in all organs increased to an apparent maximal load and somewhat sustained this level of colonization at day 7 (Fig. 2A). At the end of the assay, the caeca of each chick still contained the most C. jejuni, with the large intestines containing a significant amount of bacteria. As with earlier time points, the colonization levels in the upper gastrointestinal tract at day 7 were lower than in the caeca or large intestines (Fig. 2A). Chicks inoculated with lower doses of C. jejuni (1.73 × 10⁴ and 37 bacteria) displayed the same general trend: colonization of primarily the caeca early during the infection period with increasing colonization in all organs over the course of infection (Fig. 2B and C). Compared with infections with a higher level of 81–176 (∼10⁶; Fig. 2A), the time required to reach the maximal bacterial load in each organ was delayed and achieved between days 4 and 7 with an inoculum of 1.73 × 10⁴ bacteria and around day 7 with an inoculum of 37 bacteria (Fig. 2B and C). Overall, the caeca contained the most bacteria at all time-points in each infection; of the 23 colonized chicks in these experiments (Fig. 2A–C), over 90% of the entire bacterial load of C. jejuni 81–176 per chick were in the caeca in 18 chicks (and over 75% in 21 chicks). These results demonstrate that the pathogenic C. jejuni strain 81–176 initiates commensal colonization of the chick intestinal tract, with tropism for the lower gastrointestinal tract – particularly the caeca – similar to what was found with another C. jejuni strain (Beery et al., 1988).

Identification of C. jejuni genes involved in colonization of chick caeca by signature-tagged transposon mutagenesis

In order to identify genes of C. jejuni 81–176 involved in colonization of the chick caeca, we employed signature-tagged transposon mutagenesis (STM; Hensel et al., 1995; Shea et al., 2000). We adapted the Himar1-based solo transposon (Hendrixson et al., 2001) with 82 different DNA signature tags, created 82 different transposon mutant libraries, and constructed pools containing 74–82 different transposon mutants of C. jejuni 81–176, with one mutant coming from a different transposon mutant library. We orally infected 1-day-old chicks with different pools and recovered the bacteria from the caeca 7 days post infection, as the caeca contained a high majority of bacteria in the chicks at this time during the infection period (Fig. 2A–C). We tested 1550 transposon mutants and obtained 187 putative mutants attenuated for chick caecal colonization.

To test the colonization capacity of the putative mutants, we attempted to perform a competition assay by mixing each transposon mutant with streptomycin-resistant 81–176 Sm^R (DRH212; Hendrixson et al., 2001) in a 1:1 ratio and determining the levels of each strain in the caeca 7 days post infection. Control experiments in which 81–176 was competed with 81–176 Sm^R demonstrated that the number of bacteria of each strain obtained from the chick caeca was highly variable. This finding suggested that results from competition experiments were not reliable, and thus we switched to using infections with only one C. jejuni strain per chick. As an alternative method for determining the caecal colonization capacity of the putative mutants, we orally gavaged three to five chicks with approximately 10⁴ cfu of each mutant and determined the number of bacteria 7 days post infection. This inoculum was chosen because it is the lowest that we have found routinely gives about 1–2 × 10⁹ cfu per gram caecal content at day 7 post infection for strain 81–176. A transposon mutant was considered attenuated for caecal colonization if it colonized the caeca of all infected chicks at least 10-fold lower than wild-type 81–176 (∼ < 2 × 10⁸ cfu of the transposon mutant per gram caecal content). Of the 187 putative mutants from our primary screen, 29 mutants were attenuated for caecal colonization by this measure (Table 1 and Fig. 3A–C); many of the remaining mutants had significantly less colonization defects ranging from two- to fivefold and were not characterized further. These 29 mutants represent 22 different genes required for wild-type colonization of the chick caeca.

Seventeen of the 29 mutants exhibited a non-motile or altered flagellar motility phenotype in Mueller–Hinton (MH) motility medium (Table 1; Fig. 3A and B; data not shown). Previous studies found that motility of C. jejuni is required for wild-type levels of caecal colonization (Nachamkin et al., 1993; Wassenaar et al., 1993), validating our screening procedure for identifying mutants of C. jejuni attenuated for colonization. Included among these genes involved in motility and chick colonization are genes encoding transcription factors, σ²⁸ and σ⁵⁴, required for transcription of specific flagellar genes (Hendrixson et al., 2001; Hendrixson and DiRita, 2003); flagellar secretory or structural components (fliR and flgK); flagellar motor proteins (motA and motB); a chemotaxis regulatory protein (cheY ); a protein required for proper flagellar function (pflA; Yao et al., 1994); and four proteins with no significant homologies to proteins with known functions (Cj0248, Cj0454c, Cj0618 and Cj0883c; Parkhill et al., 2000). Cj0248 and Cj0883c have previously been identified as being required for motility (Hendrixson et al., 2001). Mutation of many of these genes result in 10- to 1000-fold reductions in the bacterial loads in the chick caeca (Fig. 3A and B). The cheY and flgK mutants displayed a more dramatic defect in colonization; recovery of these mutants after infection ranged from 100-fold less than wild type to below the limit of our detection ability (<100 cfu per gram caecal content; Fig. 3A).

Twelve colonization mutants displayed wild-type motility in MH motility medium (Table 1 and data not shown); these mutants represent 10 different genes required for wild-type levels of caecal colonization. Four mutants contained transposon insertions in three different genes, pglE, pglF and pglH, which are part of a multigene locus encoding a general protein glycosylation system responsible for adding α-linked N-acetylgalactosamine residues to many different proteins in C. jejuni including flagellin (Szymanksi et al., 1999; Linton et al., 2002). These mutants typically displayed a 100- to 1000-fold reduction in colonization, but in a few chicks displayed over a 10⁶-fold reduction in colonization (Fig. 3C). We also identified transposon insertions in attenuated mutants in Cj0456c, encoding a protein of unknown function, and two genes encoding proteins that probably function in amino acid transport, livJ and Cj0903c (Parkhill et al., 2000). Mutation of Cj0456c reduced caecal colonization capacity of 81–176 approximately 100-fold, whereas mutation of livJ or Cj0903 caused equivalent to more dramatic defects in colonization (Fig. 3C).

Transposon insertions in two adjacent genes Cj0019c and Cj0020c resulted in generally large decreases in caecal colonization (Fig. 3C). Cj0019c encodes a putative MCP; homologues of these proteins in other bacteria detect specific environmental components and transduce signals to the flagellar motor to alter the direction of motility (Blair, 1995; Falke and Hazelbauer, 2001). Transposon insertions in Cj0019c did not affect in vitro motility however (data not shown), and resulted in caecal colonization levels in six of seven chicks approximately 10⁶-fold lower than wild-type 81–176 and below the limit of detection in three other chicks (Fig. 3C). Cj0020c, which is upstream of Cj0019c and may be co-transcribed with this gene, encodes a putative periplasmic cytochrome c peroxidase. A transposon insertion in this gene resulted in 20-fold to more than a 10⁶-fold reduction in caecal colonization (Fig. 3C).

To ensure that the caecal colonization defects of the above mutants were specific for in vivo growth conditions, we analysed the growth curve of each mutant in MH biphasic medium. The growth curves of all the mutants over a 48 h period were similar to that of wild-type C. jejuni 81–176 (data not shown), suggesting that the colonization defects were due specifically to in vivo growth and not the result of a general growth defect.

Analysis of colonization defect by defined motility mutants

Some of the chick colonization mutants defective for motility had transposon insertions in genes encoding σ-factors required for transcription of flagellar genes including rpoN (encoding σ⁵⁴), which is required for transcription of flgDE2 (encoding putative flagellar hook-associated proteins) and flaB (encoding a minor flagellin), and fliA (encoding σ²⁸), which is involved in transcription of flaA (encoding the major flagellin; Hendrixson et al., 2001; Hendrixson and DiRita, 2003). In addition, we found transposon insertions in fliR and Cj0883c, which is directly upstream of flhA (Parkhill et al., 2000); a transposon insertion in Cj0883c may have polar effects on the transcription of flhA. Both fliR and flhA encode putative components of the flagellar secretory apparatus, which is presumably required to secrete flagellar proteins, and, in conjunction with the FlgRS two-component regulatory system, is required for σ⁵⁴-dependent transcription of flagellar genes (Hendrixson and DiRita, 2003).

To determine the colonization capacity of C. jejuni mutants with defined deletions of flagellar genes, we infected chicks with derivatives of C. jejuni 81–176 Sm^R (DRH212) containing in-frame deletions of fliA, rpoN, flgR, flgS, flhA or fliR (Hendrixson et al., 2001; Hendrixson and DiRita, 2003). All flagellar regulatory mutants demonstrated colonization capacities approximately 100- to 10,000-fold lower than wild-type strains (81–176 or 81–176 Sm^R) in four separate chicks (the ΔrpoN mutant colonized to wild-type levels in one of four chicks; Fig. 4). The phenotypes of the fliA, rpoN and fliR deletion mutants verify those of the corresponding transposon mutants identified in our screen. Additionaly, because flhA is downstream of Cj0883c and the ΔflhA mutant was defective for colonization, the transposon insertion originally isolated in the Cj0883c mutant may have polar effects on the expression of flhA. These data also indicate that the FlgRS two-component regulatory system governing σ⁵⁴-dependent transcription of flagellar genes is involved in commensal colonization, and along with the additional characterization of the above flagellar gene deletion mutants, emphasize the importance of flagellar biosynthesis and motility in efficiently promoting commensal colonization of the chick caeca.

Characterization of the dynamics of colonization of Cj0019c and Cj0020c deletion mutants in the chick gastrointestinal tract

Two genes identified in our screen, Cj0019c and Cj0020c, are adjacent to each other in the C. jejuni genome (Parkhill et al., 2000). As mentioned above, Cj0019c displays homology to numerous bacterial MCP proteins that are responsible for detecting environmental cues and transducing signals to the flagellar motor to change the direction of motility towards chemoattractants and away from chemorepellants. Cj0020c shows significant homology to bacterial periplasmic cytochrome c peroxidases, which bind two c-type haeme compounds to convert hydrogen peroxide to water (Rönnberg and Ellfolk, 1979; Ellfolk et al., 1983; Goodhew et al., 1990). Transcription of these two genes may be linked as the last four codons of Cj0020c overlap the beginning of the Cj0019c coding sequence.

To verify the colonization defect originally observed with the Cj0019c and Cj0020c transposon mutants, we deleted each gene from C. jejuni strain 81–176 Sm^R (DRH212; Hendrixson et al. 2001). For Cj0019c, we fused in frame codon four to the stop codon, deleting the intervening 588 codons. For Cj0020c, we fused in frame the start codon to the last 12 codons, deleting the intervening 292 codons.

The mutants were given orally to chicks at two different doses and caecal bacterial loads were determined 7 days post infection. At an inoculum of approximately 10⁴, colonization of the ΔCj0019c mutant was undetectable in four of five chicks and the colonization level in the remaining chick was slightly lower than 81–176 Sm^R(Fig. 5). Although increasing the inoculum to approximately 5 × 10⁵ bacteria resulted in increased number of chicks colonized, the levels of colonization were 100-fold lower than what was seen with 81–176 Sm^R given at a similar inoculum (Fig. 5). The caecal colonization capacity for the ΔCj0020c mutant was more dramatically altered relative to wild type. With an inoculum of 10⁴ bacteria, two chicks were colonized with the ΔCj0020c mutant at levels 10- to 10,000-fold lower than wild-type and three chicks appeared not to be colonized at all; raising the inoculum by approximately 50-fold resulted in four of five chicks being colonized, but the bacterial loads in the caeca did not significantly increase. These deletion mutants confirmed results obtained with the respective transposon mutants isolated in the STM random screening procedure.

To determinine whether Cj0019c and Cj0020c are required for specific tropism for the caeca or for general gastrointestinal colonization of chicks, we further characterized their colonization capacity by analysing colonization dynamics of each in the chick gastrointestinal tract over time. When inoculated at a dose of 5.6 × 10⁵, the ΔCj0019c mutant was detected in all intestinal organs of each chick at day 1 (Fig. 6A). However, colonization decreased over time to undetectable levels in the proximal small intestines and 10- to 100-fold in the large intestines (compare Fig. 6A with the colonization dynamics of wild-type 81–176 in Fig. 2A). The colonization levels in the caeca and distal small intestines remained relatively constant. Again, the lower gastrointestinal tract – the caeca and the large intestines – contained the highest amounts of the ΔCj0019c mutant. Compared with the chicks infected with similar numbers of wild type 81–176 (Fig. 2A), the colonization level of the ΔCj0019c mutant was generally 100-fold lower in all organs. When the inoculum of the ΔCj0019c mutant was decreased to 10⁴, significant colonization was not detected in the organs until day 4 and colonization increased in all organs through day 7, but remained at least 100-fold lower in all organs compared with wild-type 81–176 (data not shown).

Dynamics of colonization by the ΔCj0020c mutant were different from those of the ΔCj0019c mutant. When inoculated at a dose of 8.2 × 10⁵, this mutant colonized all intestinal organs of chicks at day 1 and the amount recovered increased until day 4 (Fig. 6B). However, after day 4 the level of colonization in all organs stalled or began to decrease; the colonization levels in all organs at day 7 were 10- to 100-fold lower than wild-type 81–176 used at the same inoculum (compare Fig. 6B to Fig. 2A). When the inoculum of the ΔCj0020c mutant was decreased to 10⁴, colonization was not consistently detected in all organs at days 1 and 4, and by day 7 only one chick (out of three) showed detectable caecal colonization (data not shown); no other organs harboured detectable C. jejuni, suggesting that the infection was almost cleared. These results suggest that Cj0019c and Cj0020c are not required for specific tissue tropism and are instead required for wild-type colonization of the entire gastrointestinal tract. Because of these results, we propose to annotate Cj0020c and Cj0019c as docA and docB, respectively, for determinant of chick colonization.

Analysis of other MCP and MCP-domain containing proteins for chick caecal colonization

Because docB encodes a putative MCP involved in establishing colonization of the chick gastrointestinal tract, we thought it relevant to determine whether other MCP-like proteins in C. jejuni were required for this process. Twelve genes (including docB) are proposed to encode MCPs or proteins containing MCP domains in the C. jejuni NCTC11168 genome (Parkhill et al., 2000; Marchant et al., 2002). Three of these proteins in C. jejuni include CetA and CetB, which are required for energy taxis (specific migration towards the energy-generating substrates fumarate and sodium pyruvate; Hendrixson et al., 2001), and Cj1191c, which shows high homology to CetB but is not required for energy taxis (Hendrixson et al., 2001). The other eight MCP or MCP-domain-containing proteins have so far remained uncharacterized.

We inactivated each gene encoding an MCP or MCP-domain-containing protein (except for Cj0246c, which we were unable to inactivate) either by inserting a cat-rpsL or a kan-rpsL cassette or by deletion mutagenesis in C. jejuni 81–176 Sm^R (DRH212). All mutants displayed wild-type motility in MH motility medium (data not shown). The mutants were used to orally infect 1-day-old chicks at inocula of approximately 10⁴ and the bacteria in the caeca were enumerated 7 days post infection. Only one other MCP, encoded by Cj0262c, appeared to be required for wild-type colonization (Fig. 7). Of the eight chicks infected with one of two isolated Cj0262c mutants, only one chick became colonized close to the level at which chicks infected with wild-type were. Four chicks had bacterial loads in the caeca 10- to 100-fold lower than wild type and three chicks contained bacterial loads at or below the limit of detection in the assay (<100 cfu per gram caecal content). Mutation of the remaining 10 genes encoding an MCP or MCP-domain-containing protein did not significantly alter the colonization capacity of C. jejuni. Of note, the genome sequence of C. jejuni NCTC11168 contains a stop codon in Cj0951c that causes a premature stop codon. In C. jejuni strain 81–176, this mutation is not present and the coding sequence therefore extends downstream, resulting in an open reading frame of ∼ 1600 nucleotides encoding an MCP with domains similar to those of other bacteria. Nevertheless, this protein in C. jejuni 81–176 does not function in chick colonization (Fig. 7). These experiments also demonstrate that Cj1191c and the energy taxis system mediated by CetA and CetB are not required for colonization, as mutation of the these genes did not alter the colonization capacity of C. jejuni. These results indicate that only two MCPs, DocB and Cj0262c (which we term DocC), are required for wild-type level of colonization of the chick gastrointestinal tract.

Bacterial strains and plasmids

All bacterial strains and plasmids used in this study are located in Table S1 (see Supplementary materials). For details regarding the construction of plasmids and strains for producing specific C. jejuni mutants, see Supplementary materials. Campylobacter jejuni was grown in microaerophilic conditions at 37°C on Mueller–Hinton (MH) agar as previously described (Hendrixson et al., 2001). For C. jejuni, antibiotics were used at the following concentrations: trimethoprim, 10 µg ml⁻¹; cefoperazone 30 µl ml⁻¹; kanamycin, 50 µg ml⁻¹; chloramphenicol, 15 µg ml⁻¹; and streptomycin, 0.5, 1 or 2 µg ml⁻¹. E. coli DH5α and DH5αλpir were grown in Luria–Bertani (LB) agar or broth. For E. coli, antibiotics were used in the following concentrations: ampicillin, 100 µg ml⁻¹; kanamycin, 50 µg ml⁻¹; and chloramphenicol, 15 µg ml⁻¹.

Construction of signature-tagged solo transposon mutants and mutant pools

The solo transposon was amplified from pFalcon (Hendrixson et al., 2001) by PCR using primers that added 5′PmeI restriction sites. The amplified fragment contained 98 bp of sequence 5′ and 217 bp of sequence 3′ to solo. The PmeI-digested PCR fragment was ligated into pUC19 that had been digested with EcoRI and HindIII and blunt-ended with T4 DNA polymerase (Invitrogen) to generate pFalcon2.

Eighty-two unique DNA signature tags, each contained in a separate pUTmini-Tn5Km2 derivative (Martindale et al., 2000) were individually amplified by PCR with primers P3 and P5 (Hensel et al., 1995), digested with KpnI, and ligated into KpnI-digested pFalcon2. After cloning each signature tag, a collection of 82 pFalcon2 derivatives each containing a different signature tag in the solo transposon was obtained. Each plasmid was purified and used in in vitro transposition reactions. Transposition reactions were performed as previously described (Hendrixson et al., 2001; Hendrixson and DiRita, 2003) using 500 ng Himar1 C9 transposase purified from DH5α/pMALC9 (Akerley and Lampe, 2002), 2 µg of purified C. jejuni strain 81–176 chromosomal DNA, and 1 µg of each pFalcon2 derivative containing a signature-tagged solo transposon. After transposition, the transposed DNA was repaired and transformed into C. jejuni 81–176 as previously described (Hendrixson et al., 2001). Transformants were recovered on MH agar containing 50 µg ml⁻¹ kanamycin. As a result, 82 C. jejuni 81–176 signature-tagged solo mutant libraries were created in which each library contained mutants with a unique signature-tagged transposon. Combined, the mutant libraries totalled over 59 000 transposon mutants.

For construction of signature-tagged pools of C. jejuni 81–176 mutants, each of the 82 signature-tagged solo mutant libraries was streaked on MH agar containing 50 µg ml⁻¹ kanamycin and grown for 48 h at 37°C under microaerophilic conditions. To construct a mutant pool, a mutant from each library was patched onto two identically sectored MH agar plates and grown for 48 h at 37°C under micraerophilic conditions. All 82 signature-tagged 81–176 solo mutants of one pool from one plate were recovered, mixed together, and stored at −80°C in MH broth containing 15% glycerol. The mutants on the other plate were each arrayed in a 96-well plate in MH broth containing 15% glycerol and stored at −80°C. This procedure was repeated to generate 20 different pools of signature-tagged 81–176 solo mutants.

Chick colonization assays

White leghorn strain Δ chicken eggs were acquired from a local farm and incubated in an egg incubator (Sportsman Incubator Model 1202; Georgia Quail Farms) for 21 days at 37.8°C with the appropriate humidity and rotation of eggs according to manufacture's instructions until the chicks hatched from the eggs.

For testing the caecal colonization capacity of C. jejuni 81–176 and derivatives, each strain was streaked on MH agar and grown at 37°C under microaerophilic conditions for 48 h. Strains were streaked heavily onto three MH agar plates and grown at 37°C under microaerophilic conditions for 16 h. Each strain was resuspended, diluted in MH broth to an OD₆₀₀ = 0.4, and then diluted in PBS appropriately to obtain the proper inoculum. Twelve to 36 hours after hatching, chicks were divided into groups of three to 10 and infected orally with 100 µl of each inoculum for each strain. Dilutions of each inoculum were plated on MH agar to determine the number of bacteria in each inoculum. Each group of chicks was housed separately in brooders and given water and food ad libitum. Chicks were sacrificed at day 1, 4 or 7 post infection and the appropriate organs were removed. The contents of each particular organ were collected, weighed, and resuspended in PBS to a final concentration of 0.1 g of organ content per ml. Ten-fold serial dilutions of each sample were made and plated on MH agar containing 10 µg ml⁻¹ trimetho-prim and 30 µg ml⁻¹ cefoperazone to select for growth of C. jejuni. Plates were incubated for 48 h at 37°C under microaerophilic conditions and the colonies recovered were counted. The colonization capacity of C. jejuni in an organ from each chick was reported as the number of cfu per gram of organ contents.

Screening of signature-tagged transposon mutants for mutants attenuated in caecal colonization

Hybridization blots for screening of mutants present in the input and output pools from infections with signature-tagged mutant  pools  were  prepared  by  amplifying  each  of  the  82 tags individually from the respective pFalcon2-containing signature-tagged solo derivative by PCR with P3 and P5 primers (Hensel et al., 1995). Each 40-bp unique signature tag was purified after HindIII-digestion and eluted as previously described (Merrell et al., 2002) and specifically arrayed onto Duralon nitrocellulose membranes (Stratagene). Membrane-bound DNA was denatured with 0.4 M NaOH for 8 min, rinsed with 0.5 M Tris-HCl (pH 7.0) for 5 min, rinsed twice in 2× SSC for 5 min, and cross-linked to the membranes in a UV Stratalinker (Stratagene). Membranes were stored in 2× SSC at 4°C.

To screen each pool for mutants attenuated for colonization of the chick caeca, each pool was streaked onto MH agar and grown for 48 h at 37°C under microaerophilic conditions. Each pool was then streaked heavily onto five MH agar plates and grown for 16 h at 37°C under microaerophilic conditions and then resuspended in PBS, pelleted, and diluted in PBS to OD₆₀₀ = 1. Bacteria were concentrated fivefold and 100 µl was used to orally infect each of three chicks (approximately 10⁹ bacteria). The remaining inoculum suspension was used to purify chromosomal DNA from each pool (input pool). Seven days post infection, chicks were sacrificed and the contents of both caeca from each chick were collected, combined, weighed and resuspended in PBS to a final concentration of 0.1 g caecal content per ml. Dilutions of bacteria were plated on MH agar containing 10 µg ml⁻¹ trimethoprim and 30 µg ml⁻¹ cefoperazone to recover approximately 10 000 bacteria per chick. The output pool bacteria obtained from chicks infected with identical pools were combined and chromosomal DNA was purified.

For identification of putative mutants attenuated in caecal colonization, the signature tags from the input and output pool chromosomal DNAs from each infection were amplified with P2 and P4 primers (Hensel et al., 1995), Pfu polymerase (Stratagene), and DIG DNA labelling mix (Roche) to generate dioxygenin (DIG)-dUTP labelled probes. The probes were then used in dot blot hybridizations with the signature tag-arrayed membranes. Hybridization and development of dot blots were performed using the DIG DNA Labeling and Detection Kit (Roche) according to manufacture's instructions. Tags present in the input pool but absent in the output pool represent putative mutants deficient for efficient caecal colonization. Occasionally, the primary screening of an individual pool revealed between 20 and 50 putative mutants absent from the output pools, suggesting a great loss of many mutants, presumably the result of a bottleneck in the colonization process. These putative mutants from an individual pool were combined into a second pool consisting of 20–50 mutants and screened again for caecal colonization in chicks as described above. The second screening procedure often reduced the number of putative mutants in a pool to below 20 individual mutants.

Each putative mutant was recovered from the frozen arrayed pool and individually tested for caecal colonization in chicks at an inoculum of approximately 10⁴. The number of bacteria per gram of caecal colonization 7 days post infection was determined as described above. A putative mutant was considered attenuated for colonization if the caecal bacterial loads in all chicks infected with the mutant were at least 10-fold lower than those of chicks infected with wild-type C. jejuni 81–176 (caecal colonization capacity < 2 × 10⁸ cfu of mutant per gram caecal content compared with ∼ 2 × 10⁹ cfu of wild type per gram caecal content). Identification of the site of the transposon insertion in each attenuated mutant was determined by DNA sequencing. The generated DNA sequences were compared with the genomic sequence of C. jejuni NCTC 11168 (Parkhill et al., 2000) to determine the site of the transposon insertion in C. jejuni 81–176.

Motility assays

Motility phenotypes of strains were tested in MH motility media containing 0.4% agar as previously described (Hendrixson et al., 2001).

In vitro growth curve determination

To determine the in vitro growth rates of C. jejuni mutants, strains were streaked on MH agar and grown for 48 h at 37°C under microaerophilic conditions. Each strain was streaked heavily onto three MH agar plates and the plates were incubated for 16 h at 37°C under microaerophilic conditions. Growth from each strain was resuspended in MH broth and diluted to a final concentration of approximately 2 × 10⁵ cfu per ml. For each strain, 20 mls of MH agar were placed in sterile T75 tissue culture flasks. After the agar had solidified, the inoculated MH broth was placed in the tissue culture flasks to create a biphasic medium. Growth was monitored both spectrophotometrically by OD₆₀₀ readings and by determining the number bacteria per ml by plating dilutions of the biphasic media on MH agar at 8, 24, 32 and 48 h.