Convergence of Campylobacter Species: Implications for Bacterial Evolution

See allHide authors and affiliations

Science  11 Apr 2008:
Vol. 320, Issue 5873, pp. 237-239
DOI: 10.1126/science.1155532


The nature of species boundaries in bacteria remains controversial. In particular, the mechanisms of bacterial speciation and maintenance in the face of frequent genetic exchange are poorly understood. Here, we report patterns of genetic exchange that show two closely related zoonotic pathogenic species, Campylobacter jejuni and Campylobacter coli, are converging as a consequence of recent changes in gene flow. Population expansion into a novel ecological niche generated by human activity is the most probable explanation for the increase in genetic exchange between these species. Bacterial speciation can therefore occur by mechanisms analogous to those seen in metazoans, where genetic diversification and incipient speciation caused by ecological factors have been reported in several genera.

The dynamic nature of bacterial gene pools, especially the mobility of bacterial genes among unrelated groups (1), has complicated the development of a coherent species concept for these organisms (2). Biological speciation requires barriers to gene flow (3), and in the bacteria, this can occur as a consequence of a single clone expanding into a niche and becoming isolated from its parent population, for example, in the evolution of pathogens, such as Yersinia pestis (4) and Salmonella typhi (5), from ancestral hosts into human beings. In the absence of ecological isolation, we might expect that observed levels of genetic exchange should prevent the divergence of bacterial subpopulations (69). Although there are several ways in which gene flow can be limited (10), there is little evidence as to which of these have resulted in the evolution of particular bacterial species. Here, we demonstrate convergence of the zoonotic pathogens Campylobacter jejuni and C. coli, driven by changes in their ecology.

Campylobacter jejuni and C. coli, the most common bacterial causes of human gastroenteritis worldwide (11, 12), are recognized as distinct microbiological species. Their housekeeping genes share 86.5% nucleotide sequence identity (13), a level similar to that observed between the well-studied enteric bacteria Salmonella enterica and Escherichia coli, which are thought to have diverged 120 million years ago (14). Campylobacter species have a broad host range and are present in the gastrointestinal tracts and feces of many birds (15) and mammals (16). Both C. jejuni and C. coli can also be isolated from a wide variety of environmental sources, presumably as a consequence of fecal contamination (17), and there is evidence of genetic exchange within and between these two species (18, 19). As part of ongoing efforts to elucidate the epidemiology of human campylobacteriosis, isolates from diverse sources and multiple geographic locations have been characterized by the same seven-locus multi-locus sequence typing (MLST) scheme (19, 20). There is an extensive archive of these data available from the C. jejuni and C. coli PubMLST database (, which contains allelic profiles based on the nucleotide sequences of seven housekeeping gene fragments (MLSTalleles) (20). These profiles are assigned to sequence types (STs). Here, we analyze 2953 distinct STs either directly or as haplotypes of 3309 base pairs in length, generated by combining the seven allele sequences corresponding to each ST (21).

The distinct nature of C. jejuni and C. coli was demonstrated by the clustering of the haplotypes into two main groups corresponding to the microbiological species designations. Although some haplotypes did occupy intermediate positions in a neighbor-joining tree (Fig. 1A), these had alleles at the seven loci that could each be clearly assigned to one species or other, which indicated that the haplotypes were hybrids generated by interspecies gene exchange (19). A more formal species assignment was performed with the linkage model of the Bayesian clustering algorithm Structure (22, 23), with a threshold probability of 0.75 used as the cut-off for membership of a particular ST in each species. According to this criterion, 2221 (75.2%) of the STs were assigned to C. jejuni and 715 (24.2%) to C. coli; only 17 STs (0.6%, triangles in Fig. 1A) could not be assigned to a species in this way. A total of 79 (11%) of the C. coli STs exhibited C. jejuni ancestry, >1%, whereas only 15 (0.6%) of the C. jejuni STs exhibited any C. coli ancestry; these findings provide evidence for asymmetric gene flow between these populations.

Fig. 1.

Relations among haplotypes and alleles derived from STs from the C. jejuni and C. coli pubMLST database. (A) Neighbor-joining tree of the 2953 C. jejuni and C. coli haplotypes (concatenated MLST alleles) rooted with a C. fetus sequence. Isolates belonging to C. jejuni are shown in gray and those belonging to C. coli in black. Haplotypes that could not be assigned to a species are shown as open triangles. (B) ClonalFrame genealogy generated from the C. coli haplotypes. Clade 1 haplotypes are indicated in blue, clade 2 in yellow, and clade 3 in red. (C) Neighbor-joining trees of individual gene sequences. The species and clade to which the haplotype containing each allele is assigned are indicated by color: gray, C. jejuni; blue, C. coli clade 1; yellow, C. coli clade 2; and red, C. coli clade 3. Open circles represent alleles that are found in C. coli and C. jejuni. Alleles found in more than one C. coli clade are indicated by the following colors: clades 1 and 2, black; clades 2 and 3, orange; and clades 1 and 3, purple. The colored circles show the boundaries used to assign alleles to a likely origin. For each tree, the scale bar represents a genetic distance of 0.02.

Structuring within the C. coli population, not apparent from the cluster analysis (Fig. 1A), was evident from a genealogical analysis of the derived C. coli haplotypes with ClonalFrame (24) (Fig. 1B), a model-based approach that estimates clonal relations while accounting for the fact that a single import can change multiple nucleotides at once. This indicated the division of C. coli into three subclades, which was supported by cluster analysis of the seven housekeeping genes individually (Fig. 1C). All but 26 (3.6%) of the 715 C. coli haplotypes could be assigned to one of the three clades (clades 1 to 3) in the genealogy, of which clade 1 was the largest and most diverse in this data set (Fig. 1B). Combining these data enabled the assignment of each ST to a given clade and the assignment of individual alleles to a clade of origin, independent of the clade assignment of the corresponding ST (Fig. 1, B and C).

Asymmetric gene flow was apparent among C. jejuni and C. coli clades 1, 2, and 3 (Table 1 and table S1); most gene flow involved C. coli clade 1, which had imported C. jejuni alleles at every locus on multiple occasions. A high proportion (18.6%, 54 out of 290) of the alleles in hybrids assigned to C. coli clade 1 originated in C. jejuni. Clade 1 also had the highest frequency of alleles originating in the other two C. coli clades, although the numbers were small. There was also evidence that clade 1 was a more frequent donor, at least to C. jejuni, with 25 of the 27 imported C. coli alleles arising from clade 1, which indicated a substantially elevated, although still low, rate of import compared with that observed from the other C. coli clades (Table 1 and table S1).

Table 1.

The predicted origin of alleles is given for haplotypes assigned to C. jejuni and C. coli clades 1 to 3. Assignment of haplotypes to C. jejuni or C. coli was on the basis of the Structure analysis, and the three clades of C. coli were defined by the ClonalFrame genealogy. Individual alleles were assigned to a predicted source on the basis of the neighbor-joining trees for each allele (Fig. 1).

View this table:

The rate of import of C. jejuni alleles by C. coli must have changed, subsequent to the divergence of the three clades from their shared common ancestor, with several lines of evidence indicating that the low rates seen for C. coli clades 2 and 3 were closer to historical levels than the high rates estimated for clade 1. First, because clade 1 and clade 2 are sister taxa (Fig. 1B), a change in rates of import for clade 1 is a more parsimonious explanation than changes for both clades 2 and 3. Second, there were short but similar branch lengths observed for each of the seven MLST loci between the predominant sequence clusters of the three clades (Fig. 1C). This pattern was consistent, with divergence occurring progressively by point mutation rather than abruptly by genetic import from a foreign source. Third, most (49 out of 50) of the alleles assigned a C. jejuni origin, but found in C. coli, were identical to alleles present in C. jejuni, which showed that they have not mutated since being acquired by C. coli. Finally, an ongoing process of recombination would be expected to fragment imported DNA progressively and to lead to the creation of hybrid alleles. Hybrid alleles were much less common than C. jejuni alleles in C. coli clade 1; they were observed at only three of the seven loci, namely, tkt, aspA, and glyA (Fig. 1C). In conclusion, the high rates of import into C. coli clade 1 represent a recent change from historical patterns of gene flow.

Quantitative analysis of these recent events showed that the genetic changes leading to the convergence of C. coli clade 1 with C. jejuni occurred at least four times more frequently than those leading to their divergence (21). If maintained over time, these relative rates would lead to progressive genetic convergence unless countered by strong genome-wide natural selection against C. jejuni nucleotide sequence in C. coli.

There are several ways in which barriers to genetic exchange among bacterial populations, sufficient to cause speciation, could arise. These can be divided into three categories: mechanistic, ecological, and adaptive. Mechanistic barriers could be imposed by dependence on homologous recombination (10) or by other factors promoting DNA specificity (25), such as restriction and/or modification systems. Ecological barriers are a consequence of physical separation of bacterial populations, resulting from the occupation of distinct niches, whereas adaptive mechanisms imply selection against hybrid genotypes (26). We have shown that there has been a recent, bidirectional increase in the rate of recombination between C. jejuni and C. coli and a concomitant reversal of the speciation process. One of the three clades of C. coli is more affected by this increase in recombination than the others. This asymmetry cannot be easily explained by a mechanistic barrier but could be the result of the proliferation of this clade in an environment where C. jejuni dominates numerically. C. jejuni and C. coli display some host specificity and dominate in, for example, wild birds (15) and pigs (16), respectively; however, there is appreciable niche overlap. Both species are frequently isolated from chickens and cattle, in which they are provided opportunities for genetic exchange (17). Taken together, these observations suggest that ecological, and possibly adaptive, factors historically prevented gene flow between the two species and that an ecological change has disrupted this barrier and has led to a process whereby the two species are merging or “despeciating” (Fig. 2).

Fig. 2.

An illustration of the genetic divergence and a scenario for despeciation of C. jejuni (gray) and C. coli clades 1 (blue), 2 (yellow), and 3 (red). Between time t–2 and t–1, campylobacter split into two separate species, C. jejuni and C. coli. Between t–1 and the present (t0), C. coli further separated into three distinct lineages representing incipient species. At t0 C. coli clade 1 starts to accumulate genetic material imported from C. jejuni, owing to expansion into a novel agricultural niche. By t1, recombination has been sufficient to make strains with a clade 1 clonal origin indistinguishable from C. jejuni. A more speculative projection would be that the change in environmental conditions could also have a substantial effect on clade 2 and clade 3 strains, with an elevated rate of exchange with the cosmopolitan and numerous C. jejuni clade 1 bacteria, which could lead to complete despeciation at the nucleotide level by time t3.

C. jejuni isolated from the intestinal tracts of different animal species have differentiated gene pools of MLST alleles that allow probabilistic assignment of alleles to host (27). We found that alleles imported from C. jejuni by C. coli were indistinguishable from C. jejuni alleles taken from agricultural sources (ruminants and poultry), but distinct from those found either in the environment or wild birds (21). Agriculture represents a highly plausible novel environment within which the two species could have been brought together. However, wider sampling, particularly of C. coli from natural environments, will be needed to confirm this hypothesis.

The interaction of two distinct bacterial species, C. coli and C. jejuni, has provided an opportunity to observe evolution by hybridization as it is occurring. The despeciation process that we have identified appears to be a consequence of ecological factors. This mechanism is analogous to that observed in sexual eukaryotic populations, such as Darwin's Finches, where incipient species are associated with distinct niches (28), and this process can be reversed through hybridization when selection pressures change (29). Humans continue to cause rapid and extensive environmental change. Preexisting environments have been degraded, and novel ones, such as intensive farms or acid mine drainage areas, have been created that provide a new adaptive landscape and opportunities for hybrid forms to evolve (2931). By understanding the mechanisms of microbial evolution, we can hope to mitigate some of the harmful consequences of both environmental change and the biotic response to it.

Supporting Online Material

Materials and Methods

Fig. S1

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article