Genetic relatedness among Campylobacter jejuni serotyped isolates of diverse origin as determined by numerical analysis of amplified fragment length polymorphism (AFLP) profiles


  • Present address: E.M. Nielsen, Statens Serum Institut, Copenhagen, Denmark.

S.L.W. On, Danish Veterinary Institute, Bülowsvej 27, DK-1790 Copenhagen V, Denmark (e-mail:


Aims:  To use amplified fragment length polymorphism (AFLP) analysis to evaluate the genetic relatedness among 254 Campylobacter jejuni reference and field strains of diverse origin representing all defined ‘Penner’ serotypes for this species.

Methods and Results:  Field strains (n = 207) from human diarrhoea and diverse animal and environmental sources were collected mainly through a National surveillance programme in Denmark and serotyped by use of the established ‘Penner’ scheme. Genetic relationships among these isolates, and the archetypal serotype reference strains, were assessed by numerical analysis of AFLP profiles derived from genomic DNA. Extensive genetic diversity was seen among the strains examined; however, 43 groups of isolates were identified at the 92% similarity (S-) level. Thirteen groups contained isolates from a single host, possibly representing genotypes of ‘low risk’ to human health. The remaining 30 groups contained isolates from humans, chickens and associated food products, cattle, sheep, turkeys, ostriches and/or dogs. Strains assigned to serotypes 2, 6/7, 11 and 12 formed major clusters at the 77·6% S-level. Most other serotypes did not form homogeneous clusters.

Conclusions:  High-resolution genotyping applied to strains from a comprehensive range of sources provides evidence for multiple sources of sporadic C. jejuni infection. The results suggest that public health protection measures should be directed at all foods of animal origin.

Significance and Impact of the Study:  The genetic relatedness among all ‘Penner’ serotypes of C. jejuni is assessed by AFLP analysis. In addition, further evidence of epidemic and host-specific clones of C. jejuni is provided.


Campylobacter spp. are considered to be the most prolific bacterial cause of human gastroenteritis worldwide, with C. jejuni subsp. jejuni (hereafter C. jejuni) the most frequently reported species from cases of human diarrhoea (Skirrow 1994). Neurological disorders such as Guillain-Barré syndrome (GBS) and Miller–Fisher syndrome are important postinfection complications, with GBS currently the most common cause of flaccid paralysis as the eradication of polio in developed countries (Nachamkin et al. 1998).

Between 1100 and 2300 cases of campylobacteriosis per 100 000 persons has been estimated to occur in the USA, UK and the Netherlands (Friedman et al. 2000). The incidence of Campylobacter infection has increased significantly in many developed countries since 1990: in New Zealand, a ninefold increase in reported cases from 1990 to 1998 has been noted (Friedman et al. 2000). The reasons for this burden of disease are unknown, but the development of effective intervention strategies to reduce the incidence of infection is certainly hindered by the complex epidemiology of the organism. The majority of human Campylobacter infections occur sporadically, with the major vehicle of infection assumed to be contaminated food, and in particular chicken (Skirrow 1994; Friedman et al. 2000) However, a substantial body of evidence has accumulated to suggest that other reservoirs of infection including nonpoultry meats and domestic pets are also important sources of human disease (On et al. 1998; Dingle et al. 2001; Fitzgerald et al. 2001; Wolfs et al. 2001; Dingle et al. 2002; Gillespie et al. 2002). Nonetheless, there are relatively few studies of the possible importance of nonpoultry reservoirs of human C. jejuni infection.

The use of typing methods, and in particular genotyping methods, have proven valuable tools for epidemiological investigations of Campylobacter spp. (Newell et al. 2000). We have previously shown that a high-resolution method based on the detection of whole genomic amplified fragment length polymorphisms (AFLP) is an effective tool for identifying epidemiological and clonal relationships of most campylobacters (On and Harrington 2000). AFLP genotyping involves PCR amplification of fragments among neighbouring restriction sites in the whole genetic content of the organism under study (Savelkoul et al. 1999). This combination of restriction enzyme analysis and PCR generates thousands of DNA fragments of which 40–400 are selectively detected. Consequently, the method is a sensitive measure of genetic diversity, yet is relatively insensitive to the various genomic phenomena that are known to affect the stability of the Campylobacter genome [reviewed by (Wassenaar et al. 2000)].

In this study we combine the use of AFLP with an established heat-stable serotyping scheme (Penner et al. 1983) to investigate genetic and clonal relationships between C. jejuni strains from an extensive (and, to our knowledge, unequalled in a single study) range of sources in Denmark.

Materials and methods

Bacterial strains

A total of 254 C. jejuni isolates were studied. A total of 207 were Danish isolates obtained during 1996–2001 from sporadic cases of human enteritis (n = 33), human and water strains from a water-borne outbreak (Engberg et al. 1998) (n = 5), dogs (n = 12) (Hald and Madsen 1997), chickens (n = 41), turkeys (n = 25), ostriches (n = 23), cattle (n = 28), sheep (n = 14), a pig (n = 1) and foods (retail chicken products) (n = 25). Nonhuman isolates were collected through National (Danish) surveillance programmes and were, as far as possible, chosen to be geographically, chronologically and/or serotypically distinct. However, the 41 chicken isolates were obtained from 36 independent samples, while the 12 canine strains were obtained from 10 independent sampling events. In addition, all 47 C. jejuni reference isolates used in the ‘Penner’ heat-stable serotyping scheme from humans (n = 35), unknown (n = 11) and a goat (n = 1) were also included. These strains were of non-Danish origin and first isolated in the early- to mid-1980s.

Strains were grown on 5% horse blood agar plates under microaerobic conditions for 2–3 days at 37°C (for DNA extraction) or 1–2 days at 42°C (for antigen preparation), respectively.

DNA extraction and AFLP profiling

DNA extraction and AFLP profiling were performed as described previously (Kokotovic and On 1999), with modifications for use with an ABI 377 automated sequencer (Amersham Biosciences, Uppsala, Sweden) (On and Harrington 2000). Gel-loading conditions were further amended to obtain compatible profiles between the ABI 377 36-well configuration and an upgraded ABI 377 96-well system. Consequently, 1 μl aliquots of samples containing 2·0 μl deionized formamide, 0·5 μl Genescan ROX-500 marker (Amersham Biosciences), 0·5 μl loading dye and 1 μl PCR product were used for those samples run on the 96-well configuration.

Numerical analysis of AFLP profiles was performed using the Pearson product–moment correlation coefficient and UPGMA clustering (program BioNumerics v. 2·5; Applied Maths, Kortrijk, Belgium), with the best possible match among profiles obtained by use of a 0·1% optimization coefficient. Profiles containing fragments in the size range of ca 53–434 bp were used in the analysis.


Serotyping by passive haemagglutination of heat-stable antigens was performed according to the Penner serotyping scheme (Penner et al. 1983) as described previously (Nielsen et al. 1997), using the complete set of 66 antisera for C. jejuni and C. coli.


General characteristics and repeatability of AFLP profiling

The BglII-Csp6I-A derived AFLP profiles of most C. jejuni isolates typically contained among ca 60 to 70 labelled fragments, as described previously (Kokotovic and On 1999). Eight isolates yielded profiles containing fewer fragments (ca 44–57) in a size distribution somewhat different to those of most profiles observed. The two field isolates were of serotypes 36 and 35, and the remaining six isolates were the references of serotypes 22, 23, 32, 35, 53 and 64.

Amplified fragment length polymorphism assay repeatability was assessed by cluster analysis of 37 strains examined at separate occasions and including data from 36- and 96-well configurations. The mean similarity among duplicates was 92·6 %. This concurred with the clustering at a minimum similarity of 92·3% of the five outbreak strains examined (data not shown). A cut-off value of 92·0% was therefore used as a conservative measure to identify clones, i.e. groups of strains sharing exceptional genetic relatedness. Manual inspection of the profiles was also performed to confirm strain relatedness.

Genetic relatedness among C. jejuni serotypes

A cluster analysis of the AFLP profiles from each of 254 isolates is presented in Fig. 1 and strain relationships were examined to determine any correlation between clustering and serotype designation. In general, few distinct clusters containing all or most strains of a given serotype were observed. At the 77·6% similarity (S)- level, 54 phenons were formed, of which only five showed an association with a single serotype (Fig. 1). All strains (including the appropriate archetypal serotype reference) of serotype 11, 13/14 serotype 12 strains and 67/70 serotype 2 (along with a single serotype 4-complex isolate) strains formed discrete clusters. Field strains assigned to the complex serotype 6/7 (i.e. showing serological cross-reactions to both sera) were distributed between two clusters. One cluster comprised the reference strain for serotype 7 and seven field isolates of human, chicken, cattle and food origin. The second cluster contained the reference strains for serotypes 6 and 21, and 11 isolates from humans, ostriches, sheep and food.

Figure 1.

Dendrogram of the cluster analysis of amplified fragment length polymorphism profiles of 254 Campylobacter jejuni strains correlated with serotype and source. The suffix ‘P’ to a serotype denotes the position of the appropriate ‘Penner’ serotype reference strain; n, number of field isolates of a given serotype; NT, nontypable by ‘Penner’ serotyping; ND, serotype not determined. Clonally significant groups of greater than two strains formed at or above the 77·6% similarity level are triangulated. For details see text

Identification and composition of clonal groups

The cluster analysis of the AFLP profiles (shown in condensed form in Fig. 1) was re-examined using the 92% S-level (as defined above) to identify groups of highly related strains, and revealed 160 phenons. Of these, 117 comprised single strains (data not shown). The remaining 43 phenons contained two or more strains, as shown in Table 1 alongside their serotype and source of isolation. Thirteen phenons contained strains from a single source only, whilst one phenon contained one strain of human and unknown origin respectively. Twenty-nine phenons (accounting for 99 strains) comprised strains from diverse origin, of which 13 (accounting for 51 strains) contained strains of human origin and at least one nonhuman source (namely ostriches, sheep, poultry, cattle, dogs and foods derived from chicken products). Of the 43 highly related phenons, 40 comprised isolates assigned to the same serotype. Two of the three remaining phenons contained strains of different serovars, although the serotypes included are known to give cross-reactions as members of the so-called serogroup 4-complex, and one phenon contained strains of serotypes 1,44 (n = 6) and 12,17 (n = 1).

Table 1.  Groups of Campylobacter jejuni strains sharing or exceding the 92% similarity threshold value in a comparison of 254 amplified fragment length polymorphism profiles
  1. CCUG = Culture collection of the University of Göteborg; SVS = Danish Veterinary Institute.

 1SVS 7819Ostrich42
CCUG 12782Human42
 2SVS 7793Ostrich60
SVS 7278179Turkey60
SVS 7338Dog60
SVS 7332Dog60
 3SVS 7814Ostrich42
SVS 7816Ostrich42
 4SVS 7829Food42
SVS 7847Food42
SVS 5803Human42
 5SVS 5826Human27
SVS 5833Human27
 6SVS 5786Human 6,7
SVS 4244Sheep 6,7
SVS 7930Ostrich 6,7
 7SVS 7795Ostrich 6,7
SVS 7792Ostrich 6,7
SVS 7827Food 6,7
 8CCUG 17625Human12
SVS 7849Food12
 9SVS 1431Chicken12
SVS 1430Chicken12
SVS 1433Chicken12
SVS 7341Dog12
10SVS 787-657Chicken 6,7
SVS 5791Human 6,7
SVS 4039Cattle 6,7
11SVS 7834Food 6,7
SVS 7857Food 6,7
12SVS 1436Chicken 6,7
SVS 1445Chicken 6,7
13SVS 7270118Turkey29
SVS 4273Cattle29
14CCUG 10966Human36
SVS 4238Cattle23,36
15SVS 7632Ostrich23,36
SVS 7772Ostrich23,36
SVS 7770Ostrich23,36
SVS 7784Ostrich23,36
SVS 5832Human23,36
16SVS 1452Chicken1,44
SVS 1434Chicken1,44
SVS 1437Chicken1,44
17SVS 1421Chicken4c
SVS 1461Chicken4c
18CCUG 10398Human4
CCUG 10945Human13
19SVS 7781Ostrich37
SVS 7780Ostrich37
20SVS 7274172TurkeyNT
SVS 7279301Turkey4c
SVS 7261449Turkey4c
21SVS 4232Cattle2
SVS 4236Cattle2
22SVS 1455Chicken2
SVS 4258Sheep2
23SVS 1429Chicken2
SVS 4254Sheep2
SVS 4247Sheep2
24SVS 4067Cattle2
SVS 1117Chicken2
SVS 5104Outbreak2
SVS 1096Chicken2
SVS 4017Cattle2
SVS 5003Outbreak2
SVS 1104Chicken2
SVS 5002Human2
SVS 5001Outbreak2
SVS 5014Outbreak2
SVS 5141Water2
SVS 5004Human2
25SVS 7333Dog2
SVS 7929Ostrich2
SVS 3141Pig2
26SVS 4266Cattle2
SVS 4229Cattle2
27SVS 1127Chicken2
SVS 4030Cattle2
28SVS 4014Cattle2
SVS 4022Cattle2
SVS 4009Cattle2
SVS 1093Chicken2
SVS 1044Chicken2
SVS 1103Chicken2
SVS 1101Chicken2
SVS 1042Chicken2
SVS 4026Cattle2
SVS 1072Chicken2
SVS 690Chicken2
29SVS 7329Dog2
SVS 5005Human2
SVS 4016Cattle2
30SVS 7822Food2
SVS 5040Human2
31SVS 5829Human2
SVS 5784Human2
SVS 5813Human2
SVS 1462Chicken2
SVS 5808Human2
32SVS 7789Ostrich2
SVS 7821Ostrich2
SVS 7783Ostrich2
SVS 7272122Turkey2
33SVS 7775Ostrich1,44
SVS 7855Food1,44
34CCUG 10935Human1
SVS 7864Food1,44
35SVS 4264Cattle1,44
SVS 4265Cattle1,44
SVS 5795Human1,44
SVS 7831Food12,17
SVS 104-733Chicken1,44
SVS 4260Cattle1,44
SVS 7344Dog1,44
36SVS 1463Chicken1,44
SVS 1425Chicken1,44
SVS 1456Chicken1,44
SVS 1442Chicken1,44
37SVS 7832Food1,44
SVS 7853Food1,44
38SVS 7279002Turkey31
SVS 7275990Turkey31
SVS 7277238Turkey31
SVS 4252Sheep31
39SVS 7356Dog4c
SVS 7262285Turkey4c
40CCUG 12790Human50
CCUG 24869NR65
SVS 4231Cattle4c
41SVS 5836Human11
SVS 5838Human11
SVS 5830Human11
42SVS 1443Chicken11
SVS 7833Food11
43SVS 7263001Turkey3,17
SVS 7264077Turkey17

The composition of clonal groups was largely supported by visual inspection of strain AFLP profiles. However, in some cases, profiles belonging to neighbouring phenons were seen to closely resemble those in adjacent phenons that were distinguished at the 92% S-level. Each of these examples belonged to clusters defined at the 77·6% S-level that were dominated by a single serotype (see above). Examples of AFLP profiles of genetically highly related strains are shown in Fig. 2.

Figure 2.

Representative amplified fragment length polymorphism profiles of highly related strains identified in the study. The 92% similarity level used to delineate these genotypes is denoted by a stippled line. Strain number, ‘Penner’ serotype designation and source information are annotated to the right. CCUG, Culture Collection of the University of Göteborg; SVS, Danish Veterinary Institute


In examining strains from human, animal and environmental sources with high-resolution genotyping- and serotyping methods, this study represents a comprehensive molecular epidemiological analysis of potential sources of sporadic human campylobacteriosis in Denmark. The discriminatory potential and efficacy of AFLP for identifying highly related strains has been demonstrated before (Duim et al. 1999; On and Harrington 2000) Profiles clustering at similarity values >92% have been shown to identify groups of strains with epidemiological or clonal significance (On and Harrington 2000). The results with replicated analyses and outbreak strains in this study also indicated the 92% similarity cut-off value to be a pragmatic means by which highly related strains could be identified. We defined 43 groups of strains that clustered together at or above this similarity value (Table 1). Thirteen such groups contained only isolates from a single source; such strains may represent host-specific genotypes of relatively low risk to humans, as suggested previously (On et al. 1998; Dingle et al. 2002). However, three strain genotypes were found only in humans, but these are unlikely to represent ‘human-specific’ clones as interpersonal transmission is considered a rare event in developed countries (Friedman et al. 2000). In our study, isolates from certain sources were under-represented (e.g. pigs, water) or not included (e.g. wild birds, nonpoultry foods) and this sampling bias may have contributed to our failure to indicate possible infectious sources for the aforementioned three genotypes. Furthermore, our criteria for the identification of highly related clones could be considered to be overly conservative. Visual analysis generally supported our application of the 92% S-level to define clonal groups, but several phenons contained profiles that closely resembled those observed in neighbouring phenons (see Fig. 2 for examples). Such profiles were invariably encountered in apparently clonal lineages dominated by a single serotype. Although the same cluster topology was invariably obtained when strains belonging to these clonal serotypic groups were compared in distinct analyses (data not shown), the overall homogeneity of profiles within such groups certainly suggested them to be highly related.

Most (29/43) highly related genotypes were not host specific and occurred in at least two of the following sources: humans, chickens (and food products thereof), turkeys, cattle, sheep, water, dogs and ostriches (Table  ). Several other studies using pulsed-field gel electrophoresis, fla-PCR analysis, AFLP and multi-locus sequence typing (MLST) analyses have included isolates from three to five nonhuman reservoirs (On et al. 1998; Dingle et al. 2001; Fitzgerald et al. 2001; Wolfs et al. 2001; Dingle et al. 2002; Gillespie et al. 2002) and together have implicated most of the above as potential sources of human infection. We believe our study to be the first to suggest, via high-resolution genotyping, that ostriches are potential sources of sporadic human infection. These data suggest that many C. jejuni strains are well adapted to colonizing multiple animal hosts that, in turn, can act as sources of human infection. Genotyping data cannot quantify the relative risk of infection from these sources, but should be integrated into risk assessments to do so. Until this is performed, it seems reasonable to recommend that all foods of animal origin (not solely chicken), and dogs, be considered as sources of human infection. Suitable hygienic measures of prevention should thus be applied equally in efforts to reduce the current burden of disease.

The genetic diversity of C. jejuni is well recognised and is attributed to a number of distinct phenomena, including genomic rearrangements and horizontal gene transfer (Wassenaar et al. 2000). The latter has been suggested by MLST to play an important role in the generation of genomic heterogeneity in C. jejuni and results in a ‘weakly clonal’ population structure (Dingle et al. 2001; Suerbaum et al. 2001; Dingle et al. 2002). This model assumes that C. jejuni comprises several distinct lineages (i.e. groups of organisms sharing a recent common ancestor), with phylogenetic relationships among lineages obscured through frequent lateral gene transfer events. The correspondence between AFLP and MLST analyses of C. jejuni has been noted previously in a study of isolates from human infections, poultry and cattle, with 116 AFLP types, and 117 MLS types defined from 184 isolates (Schouls et al. 2003). Given the numbers of strains examined, the number of AFLP types determined in our study (160 types from 254 strains) is in proportion with those obtained by these authors (Schouls et al. 2003). Similarly, our study indicated that Penner serotypes were infrequently associated with discrete AFLP clusters, supporting the results of MLST analyses that suggested serotype to be a poor marker of clonality in C. jejuni (Dingle et al. 2002). This is perhaps not surprising considering that the genetic basis of the Penner serotyping scheme appears to be related to a single genetic locus (Karlyshev et al. 2000), whereas both AFLP and MLST methods analyse a broader range of genomic polymorphisms. Nonetheless, our study identified several clusters of strains assigned to a single serotype only. This result resembles observations of Desai et al. 2001, where a few serotypes based upon an adaptation of the classical ‘Penner’ serotyping system used here also formed homogeneous clusters in a numerical analysis of AFLP profiles. Overall, these data confirm the extensive genetic heterogeneity of C. jejuni and of the existence of several clonal lineages, a few of which share a common serotype. As each cluster of Penner serotypes 2, 6, 7, 11, and 12 in our study also included the archetypal reference strain isolated over 20 years previously in a different country, they may represent ‘epidemic clones’ (Tibayrenc 1996) that are genetically relatively stable (On and Harrington 2000; Manning et al. 2001) Penner serotypes 2, 6, 7 and 11 have been quite frequently reported in human disease (Patton and Wachsmuth 1992), perhaps suggesting that relatively few epidemic clones may account for a considerable burden of human disease and may pose a higher risk to human health than others. In summary, our findings support the need for research aimed at identification of virulence factors that possibly differentiate strains of higher risk to human health than others, but also the need for awareness regarding the probable multiplicity of sources of human infection, to improve public health protection measures.


We thank Birte Hald (Danish Veterinary Institute) for canine isolates, and Penelope Jordan, Katja Kristensen, Kenn Kristiansen, Nette Larsen, Sidsel Boelsen and Sussi Kristoffersen for excellent technical help. This work was supported by a grant from the Directorate for Food, Fisheries and Agri Business, no. FøSI00-SVS-4.