The phylogenetic relationships of Caenorhabditis and other rhabditids^*

A preliminary phylogenetic hypothesis for representative rhabditid species and four species from the outgroup is shown in Figure 1. The analysis is based on DNA sequences from nearly complete small and large subunit (SSU and LSU) ribosomal RNA (rRNA) genes and a portion of the RNA Pol II (RNAP2) gene (KK and DF, unpublished; original data and details of the analysis will be published elsewhere).

Two major clades of rhabditids can be distinguished, called here "Pleiorhabditis" and "Eurhabditis" (Figure 1). Although these two groups were combined in a single monophyletic group (supported by 87% bootstrap value) in an earlier analysis using only SSU rRNA genes (Sudhaus and Fitch, 2001), such a clade does not appear in the present jackknife analysis with additional genes. Pleiorhabditis comprises several genera of species with a posterior vulva (e.g., Mesorhabditis, Teratorhabditis) and also the genus Pelodera which includes Pelodera strongyloides, reported to be a facultative parasite of mammals (Sudhaus and Schulte, 1988; Jones et al., 1991). Eurhabditis includes almost all the other rhabditid species, including Caenorhabditis, and thus differs from the Eurhabditis originally proposed by Sudhaus on the basis of a different view of relationships (see Sudhaus and Fitch, 2001). Additionally, Eurhabditis includes the "Protorhabditis group", named after its most prominent members, which is the sister clade to Caenorhabditis. Eurhabditis includes a monophyletic genus Oscheius that comprises two separate clades, one of leptoderan species, the Insectivora group, and another of peloderan species, the Dolichura group, as proposed by Sudhaus and Hooper (1994; see also Figure 4). One member of this latter group, Oscheius tipulae (represented here by strain CEW1) is used as a satellite model organism for comparative studies of development (see Evolution of development in nematodes related to C. elegans; Sommer, 2000) and molecular biology (e.g. Evans et al., 1997). The strongylid parasites of vertebrates and the Heterorhabditis pathogens of insects have also clearly evolved from within the Eurhabditis clade (see A quick tour of nematode diversity and the backbone of nematode phylogeny).

Relationships are now well-resolved among the Caenorhabditis species which are represented by live strains (Figure 2; Kiontke et al., 2004; Cho et al., 2004). Of these species, C. briggsae and C. remanei are closest relatives. No single species is most closely related to C. elegans, which is the sister species of a clade that includes C. briggsae, C. remanei, and Caenorhabditis n. sp. (represented here by strain CB5161). The closest species to these four Elegans group species is C. japonica, which was chosen for a genome sequencing project in order to provide an outgroup comparison for the four Elegans group species. Another undescribed new species (represented here by strain PS1010), which has been used in some comparisons (e.g. Baldwin et al., 1997), diverged earlier than C. japonica. Uncertainty still exists for some Caenorhabditis relationships; instead of being related in the manner depicted in Figure 2, C. sp. PS1010 could be most closely related to C. drosophilae and Caenorhabditis n. sp. (DF5070) (Kiontke et al., 2004). For more about Caenorhabditis biodiversity and a phylogenetic hypothesis incorporating other Caenorhabditis species, see Ecology of Caenorhabditis species (also Sudhaus and Kiontke, 1996).

Rhabditids, as a monophyletic group, include two taxa that have been previously classified separately. First, the strongylids are a group of vertebrate parasites traditionally considered outside of rhabditids. Molecular analyses have placed strongylids clearly within the Eurhabditis group of rhabditids and as a sister group of the insect pathogenic Heterorhabditis (Fitch and Thomas, 1997; Blaxter et al., 1998; Sudhaus and Fitch, 2001; De Ley and Blaxter, 2002).

A second group of species traditionally classified outside of rhabditids is the group designated here informally as diplogastrids (called Diplogastrina, Diplogastridae, or Diplogasteromorpha in formal classification systems), which includes the satellite model system, Pristionchus pacificus (see Evolution of development in nematodes related to C. elegans). Specifically, there appears to be a relationship between diplogastrids and Rhabditoides inermis, a rhabditid (see also Sudhaus and Fitch, 2001). According to this phylogeny, R. inermis is more closely related to diplogastrids than to its traditional congener, Rhabditoides inermiformis, despite their morphological similarities. (Another species, Rhabditoides regina, resides within Pleiorhabditis, thus making "Rhabditoides" polyphyletic.) Furthermore, our phylogeny suggests that Poikilolaimus could be in the most anciently diverged lineage of rhabditids, placing diplogastrids distinctly within rhabditids, although the jackknife support (58%) is not strong for this conclusion. Together, these data suggest that diplogastrids are not a clade separate from rhabditids, but are instead derived from within rhabditids. These data contradict a recent proposal based on morphology that bunonematids are the sister taxon of diplogastrids (Fürst von Lieven, 2002).

Using the molecular phylogeny, we have traced the evolution of gonochorism (males and females must cross to propagate), self-fertile hermaphroditism (see Sex-determination in the germ line), heterogony (alternating generations of gonochorism and hermaphroditism), and parthenogenesis (sperm, if they exist, do not contribute genetic material; Figure 3). Among the taxa represented in Figure 3, we find that gonochorism is ancestral and that hermaphroditism has arisen at least ten times independently in all the major rhabditid clades, with only one possible reversal back to gonochorism occurring within the Dolichura species group of Oscheius. Such massive convergence provides an excellent opportunity to study how significantly developmental constraints limit or channel evolution.

According to the most parsimonious reconstruction, and assuming reversibility and equal chances of changes between character states, C. briggsae and C. elegans evolved hermaphroditism independently (Kiontke et al., 2004). If we consider that changes from gonochorism to hermaphroditism have occurred ten times more frequently than the reverse in rhabditids, it becomes even more plausible that hermaphroditism evolved independently in these two species. In such a case, there are two possibilities for evolutionary change: parallelism (in which independent changes affect the same mechanism) or convergence (in which the changes affect different mechanisms but yield similar results). In the case of C. elegans and C. briggsae, hermaphroditism appears to be convergent because there is a fundamental difference between the two species in the identity of FOG-2, for which no single ortholog exists in C. briggsae, and developmental role of GLD-1, which inhibits female fate in the germline of C. elegans but promotes female fate in C. briggsae (Nayak et al., 2005; see Sex-determination in the germ line).

We also find that heterogony evolved twice (in Heterorhabditis and an undescribed species represented by strain SB347) and parthenogenesis evolved once (within the Protorhabditis group).

As an example of morphological evolution, we have traced the history of changes in morphogenesis of the male tail tip. In C. elegans, the shapes of the hypodermal cells in the male tail tip change dramatically as the cells fuse during L4 male morphogenesis and the conically pointed tail tip of the larva retracts to form the rounded, "peloderan" tail of the adult male (Nguyen et al., 1999). Tail morphogenesis does not occur in hermaphrodites or in females of most other rhabditid species. Although males of most rhabditid species do undergo tail morphogenesis, the tip of the tail does not always retract. In such cases, the tail tip remains pointed, as in larvae, usually sticking out beyond the posterior edge of the fan or bursa, if a fan exists. This tail form is called "leptoderan".

There have been at least seven independent changes in the shape of the male tail tip during rhabditid evolution (Figure 4). Because some branches are still unresolved, the directionality of most of these changes is unclear. However, the present phylogeny allows us to conclude that at least two changes occurred from peloderan to leptoderan. Some of the changes in male tail tips must be convergent because we know that peloderan tail tips can be constructed in different ways (e.g., with or without fusions of the hypodermal cells), as can leptoderan tail tips (Nguyen et al., 1999; Fitch, 1997; Fitch, 2000). Because of these convergences, it is clear that there is no monophyletic clade of leptoderan species or of peloderan species (cf. Peloderinae of Andrássy, 1983). We have found that several other morphological characters (such as lack of glottoid apparatus) show similar homoplasy at gross morphological levels (e.g., Sudhaus and Fitch, 2001). Thus, rhabditids provide fertile ground for investigating the role of developmental plasticity in evolution.

Using the well-resolved Caenorhabditis phylogeny, we inferred the most parsimonious history of intron gains and losses in a 1,860-bp portion of the gene coding for the largest subunit of RNAP2 (Kiontke et al., 2004). This phylogeny of closely related species allows finer resolution of such events than do comparisons among distantly related model organisms, which have been the main staple of intron evolution fare (e.g. Rogozin et al., 2003). We find that intron evolution is extraordinarily dynamic (Figure 5). In this one region, the number of introns present in a species can range between 1 and 14. Considering only homologous intron positions at which a single most parsimonious reconstruction could be made ("unequivocal" positions) and assuming that intron gains and losses are equally likely, introns were lost 12 times and gained 4 times (Kiontke et al., 2004). If one considers the possibility that losses are more likely than gains and that it might be highly improbable for an intron to be regained in exactly the same spot, it is possible that as many as 27 losses and at most 3 gains (but at least 1 gain) occurred. Most of these intron losses occurred in the stem species of the Caenorhabditis clade, and several losses occurred later as the clade evolved (Figure 5). One intron in C. sp. CB5161 is unique among all rhabditids we have characterized so far, and is thus a clear gain (Figure 5). Characterization of the homologous region in other rhabditids suggests that intron evolution is very dynamic with periods of losses and gains (KK and DF, unpub.). Frequent intron loss in other Caenorhabditis genes has also been reported (Cho et al., 2004). Recent intron gains have also been documented, some of which have sequence similarity to other introns and could therefore have originated by reverse splicing (Coghlan and Wolfe, 2004).

Genome-level comparisons of molecular divergence show that C. elegans and C. briggsae are a little bit less similar at the molecular level than are human and mouse (Stein et al., 2003). We find the same relative differences when comparing these species with RNAP2 and rRNA genes alone (Kiontke et al., 2004). Because many taxa are represented by SSU rRNA sequences, we can use this molecule to compare the degree of molecular divergences among other vertebrate taxa with the divergences among rhabditids. The divergence between the least closely related Caenorhabditis species (C. briggsae versus C. sp. SB341) is comparable to the divergence between mouse and zebrafish, encompassing a broad range of vertebrate lineages (Figure 6). Furthermore, species within the Eurhabditis clade are at least as different at the molecular level as mouse and sea urchin, encompassing a broad range of deuterostome lineages. It is important to note that the Eurhabditis clade is but a small fraction of Nematoda. Thus, the nematodes diverged from other animals very anciently, rates of molecular evolution are much faster in nematodes than in deuterostomes, or both. Because there is no accurate method for calibrating a "molecular clock" for nematodes, estimates for dates of divergence will be extremely unreliable. Complicating this issue, there is considerable rate heterogeneity among different lineages for some genes (e.g., SSU rRNA; see Kiontke et al., 2004). The best measure of taxonomic difference in rhabditids is therefore simply the relative amount of sequence difference (as in Figure 6), not some date of divergence which could be very inaccurate.