Speciation is such a contentious topic [see: chlamydial taxonomy], even for sexually reproducing populations (Mayr and Provine, 1998), that I approach the divergence of Chlamydiaceae into species with some trepidation. Divergence results when a population radiates into different habitats in which subpopulations become reproductively isolated and follow unique pathways to eventually form new species. This is the basis of allopatric speciation in sexually reproducing populations. In a very general way, the same concepts can be applied to divergence in Chlamydiaceae, with the last common ancestor (LCA) of the family forming the original population and different hosts serving as the isolating habitats. The extreme discontinuity of the chlamydial habitat favors isolation. Whole genome comparisons give a good idea of what the LCA was like, but there are no comparable hints as to the nature of the host or hosts for the LCA. Until recently, it would have been assumed that the LCA lived in an ancient mammal or bird because Chlamydiaceae were thought to be restricted to warm-blood vertebrates, but now that they have been isolated from lower vertebrates and even from invertebrates (Part I; see also Moulder, 1989 and Everett et al., 1999), this assumption must be discarded.

Radical adaptation to specific hosts was probably not required, a conclusion supported by the great phenotypic and genotypic resemblance among the species of Chlamydiaceae. From phylogenetic analyses of five different genes in several Chlamydiaceae spp., Bush and Everett (2001) concluded that virulence phenotype is more strongly influenced by specific adaptation to new habitats than by ancestry. Genetic differences among species, differences that determine unique behavior, are of intense interest in medical and veterinary circles but are less interesting from an evolutionary point of view: too short a segment of evolutionary time and too few fundamental changes in character of either hosts or parasites.

Imagine a clonal chlamydial population separated into discrete subpopulations in identical hosts. Over time each subpopulation will go its separate way because of founder effects, bottlenecks, genetic drift, mutations, recombinations, and deletions. When the hosts also change, different selective pressures will shape each population. There will be opportunities for coevolution. Chlamydiae will evolve to more efficiently exploit their hosts, and hosts will evolve to resist exploitation. However, coevolution cannot be expected to be as potent a force in intracellular parasitism as it is in endosymbiosis (Bell, 1997; Funk et al., 2000). In maternally transmitted endosymbiosis, the lineages of host and parasite are tightly linked. While there is no such obligatory linkage in intracellular parasitism, the chances are still good that progeny will be transmitted to another host of the same species. Adaptation to one host will be marked by increased efficiency of reproduction in that host but decreased efficiency in others. As adaptation proceeds, host range narrows and populations are isolated in different hosts where numerous evolutionary pressures will operate to increase divergence.

This explains the origin of chlamydiae which have a single host, but how did species with multiple hosts arise? One way would be chance infection of a novel host, adaptation to that new host, and the emergence of a variant still recognizable as belonging to the parent species but with a preference for its new host. Another way would be for a selection of traits that facilitate promiscuous crossing of species barriers without special adaptation to any one host. It is axiomatic in evolutionary thought that specialists prevail over generalists (Bell, 1997) so what advantage do chlamydiae gain by being able to live in different hosts? Despite the preponderance of multihost pathogens in nature (Woolhouse et al., 2001), the evolutionary advantages of broad host range are not well understood. It may be that it frees an obligate intracellular parasite from utter dependence on just one host species. This suggests that there will be pressure for single-host pathogens to evolve multihost capabilities.

Although truly extensive data are lacking, it appears that in cell culture Chlamydiaceae behave as multihost pathogens. They multiply in a variety of cells from different species and tissues, albeit with varying efficiencies. Woolhouse et al., (2001) have defined true hosts as ones that are infected in nature, that support multiplication of the pathogen and that release its progeny to infect more individuals of the same host species. Chlamydiaceae may have one or several true hosts, although the "true host range" is always narrower than the "cell culture host range" (Table 1). Chlamydia spp. tend to be single-host pathogens, whereas Chlamydophila spp. may be either single-host (as with C. felis and C. caviae) or multihost (as with C. psittaci and C. pecorum). The discrepancy between true host range and cell - culture host range is well-known, as are the many differences between cell cultures and true hosts. It could be that extant Chlamydiaceae species infect such a broad range of cells in culture because all that is required is the basic genetic toolbox inherited from the familial LCA, whereas the much more demanding requirements for indefinite propagation in true hosts demand special tools acquired during radiation. If this is true, then many of the genes responsible for host range and other traits that define the individuality of chlamydial species will be expressed only in true hosts and not in cell culture. In a somewhat comparable way, it is believed that many virulence genes of facultative intracellular bacteria are expressed only inside host cells (Chiang et al., 1999).

Biovars of C. pneumoniae have been isolated from humans, horses, koalas, and frogs. Complete genome sequences are available only for the human biovar, but the genomes of the other biovars have been mapped well enough to show that they are different (Hotzel et al., 2001). The biovars of C. pneumoniae may be examples of the acquisition of multiple host range by isolation in different hosts and the emergence of genetic variants. The common ancestor could have radiated into one of the present - day hosts, acquired species identity, and then reached other hosts where it adapted to them with minimal genetic change. What is needed to prove this idea is to show that the several biovars, under natural conditions of infection, have preference for the host of isolation. It would be satisfying to learn the genetic basis of these preferences. In which host did C. pneumoniae emerge as a distinct species? Its near-universal distribution in the human host and the near-identity of strains from diverse human populations argue for people being the first hosts. However, we were last on the evolutionary scene, so it may also be argued that another animal was first and that humans are a recent host in which C. pneumoniae has spread with unprecedented rapidity. How C. pneumoniae radiated among the four known biovars may become clear when their complete genomes are established.

Many difficulties lie in the way of unequivocally demonstrating that more than one family of animals are true hosts for a single Chlamydiaceae species. Unambiguous instances of the transmission of a single chlamydial strain among different host animals living under natural conditions are hard to find. The best example I know of is the transmission of an ornithosis strain (C. psittaci) from turkeys to sheep and then back to turkeys (Pierce et al., 1964). I cannot even guess what unique chlamydial trait might have permitted this strain to breach the species barrier with such impunity. For other pathogens with broad host range, such as rabies virus, there has also been no adequate explanation.

C. trachomatis offers a good example of different host preferences within a single host species. If true host preference is to be considered analogous with allopatric speciation, then tissue preference roughly resembles sympatric speciation, speciation within a single habitat. C. trachomatis may be separated into 19 serovars on the basis of differences in the four variable segments of the omp1 gene and its product the major outer membrane protein (Brunham, 1999). All serovars have a tropism for human mucosal epithelial cells, but serovars A, B, Ba, and C infect mainly the conjunctiva. Serovars D, Da, E, F, G, H, I, Ia, J and K are predominately isolated from the urogenital tract, while serovars L1, L2, L2a, and L3 are found in inguinal lymph nodes. These preferences in large part determine the manifestations of human disease due to C. trachomatis (Schachter, 1999). Thus, serovars A through C cause trachoma, serovars D through K are responsible for urethritis in men and cervicitis in women, while the more invasive serovars L1 through L3 cause lymphogranuloma venereum. The separation of ocular and genital serovars persists even in trachoma - endemic areas. Multiple variations in antigenic structure are usually attributed to immune pressure, but immune pressure alone does not explain separation of the serovars into these three functional groups. A simple explanation would be that the serovars in each group share some property that determines tissue tropism, but attempts to relate disease patterns to genetic differences among the serovars have yielded conflicting results.

The genetic variations in C. trachomatis that produced within a single host species the many serovars with their different tissue preferences must have occurred in reproductive isolation. How could this have occurred in a single host species? Differing modes of transmission and portals of entry might explain the separation of the ocular and urogenital serovars, but then how to separate the urethritis-cervicitis serovars from those of lymphogranuloma venereum? It is hard to even speculate on the genotypic and phenotypic nature of the ancestral C. trachomatis from which all the serovars have descended. The only published genomes for Chlamydia are those of C. trachomatis serovar D and C. muridarum. The close similarity of these two genomes suggests that differences among the serovars are small indeed. The tissue tropisms exhibited by the different serovars in human disease are frequently not maintained in experimental infections of laboratory animals, indicating that the basis of tissue preferences, like that of true host preferences, may not be found in cell culture.
[JWM]