Phylogeny of Myzostomida
Additional file 1: Table S1 lists the nominal species and terminals used to assess myzostomid phylogeny, including the genes sequenced for this study and those obtained from GenBank. The maximum parsimony (MP) and maximum likelihood (ML) results for each gene partition are summarized in Additional file 1: Table S2. Results from individual genes were basically congruent with each other, and with the combined data analyses (MP, ML, and Bayesian), and are not shown here. The aligned sequence data from the four genes (COI, 16S, 18S, H3) yielded a concatenated `complete’ dataset of 3019 characters with 1187 parsimony informative sites and 252 variable, but uninformative, sites. The `restricted’ dataset (with 16S, 18S, and H3 Gblocked and third positions of COI excluded) had 2501 characters, with 908 parsimony informative sites and 182 variable, but uninformative, sites.
The ML and Bayesian analyses of the complete (ML = -ln 38806.84) and restricted (ML = -ln 26899.17) datasets gave the same tree topology (Figure 2). Endomyzostoma Perrier, 1897 (living in galls or cysts on crinoids) formed a well-supported clade sister to all other myzostomids. Within this latter clade, Asteromyzostomum Jägersten, 1940 (associated with a seastar) was found with high support as sister to Protomyzostomum Fedetov, 1912 (living in ophiuroids) and this clade was then sister to the remaining myzostomids found on crinoids. Myzostoma cirripedium Graff, 1885, a free-living form on a stalked crinoid (previously found within Endomyzostoma and referred to this genus [18]), was sister to the well-supported clade of Pulvinomyzostomum Jägersten, 1940 and the remaining myzostomids. The placement of Myzostoma cirripedium requires that it be placed in a new genus, Eenymeenymyzostoma n. gen. (see Taxonomy section). Pulvinomyzostomum, with type species Pulvinomyzostomum pulvinar (Graff, 1884), formed a clade (through likelihood, but not parsimony analyses) with two other undescribed myzostomid species. Pulvinomyzostomum inaki nomen nudum (Summers and Rouse in press) has the same lifestyle as P. pulvinar (in the mouth of the crinoid), while Pulvinomyzostomum messingi nomen nudum (Summers and Rouse in press) is a free-living form on its stalked hyocrinid host.
The MP analysis of the complete dataset yielded 100 most parsimonious trees, length 9386. The MP analysis of the restricted dataset yielded 56 most parsimonious trees with a tree length of 5451. The MP results were similar to each other and largely congruent with the ML and Bayesian analyses, but the grade of taxa shown in Figure 2 with respect to Myzostomatidae, was instead a grade with respect to Endomyzostoma (Additional file 1: Figure S1). Also, the three species forming the Pulvinomyzostomum clade in the ML and Bayesian analyses were not monophyletic in the MP results. Two basic topologies were considered for the interpretation of the transformations, the first ML- and Bayesian- based (Figure 2) and the second MP-based (Additional file 1: Figure 1).
With regard to the relationships within the well-supported Myzostomatidae, all MP, ML, and Bayesian analyses showed Myzostoma Leuckart, 1827 as paraphyletic with the genera Contramyzostoma Eeckhaut & Jangoux, 1995, Hypomyzostoma Perrier, 1897, Mesomyzostoma Remscheid, 1918 and Notopharyngoides Uchida, 1992 nested inside (Figure 2, Additional file 1: Figure S1). Relationships in the MP analyses varied among the species of Myzostoma in the various shortest trees for the complete and restricted analyses and between these analyses and the ML and Bayesian analyses. This is reflected in the poor support for many of the nodes across Myzostomatidae (Figure 2). What all analyses did show was that Myzostoma australe Rouse, 2003 was the sister group to all other Myzostomatidae.
The ML and Bayesian analyses presented a monophyletic Hypomyzostoma as a poorly supported sister-group to a well-supported Mesomyzostoma clade (Figure 2), as did some MP trees, though with poor support. Hypomyzostoma has the type species H. folium (Graff, 1884), which was not available for this study. However, based on morphology and following Lanterbecq et al. [18], we apply the name Hypomyzostoma to the clade comprising four nominal species (H. jonathoni nomen nudum (Summers & Rouse, in press.), H. jasoni nomen nudum (Summers & Rouse, in press), H. nanseni (Graff, 1887), and H. fasciatum (Remscheid, 1918)).
All results recovered Contramyzostoma as polyphyletic, with C. sphaera Eeckhaut, Grygier, & Deheyn, 1998 as the sistergroup to Myzostoma cf. viride, and C. bialatum Eeckhaut & Jangoux, 1995, the type species of the genus, as sister to another clade of Myzostoma species. Notopharyngoides was also consistently polyphyletic, with N. aruensis (Remscheid, 1918) and N. platypus (Graff, 1887) well separated and sister groups to different clades of Myzostoma.
With regards to family-ranked taxa, the concatenated ML and Bayesian analyses (see Figure 2) returned four of the current families as clades: Asteromyzostomatidae (monotypic), Protomyzostomatidae, Pulvinomyzostomatidae, and Mesomyzostomatidae (Table 1-left column), though the latter taxon rendered Myzostomatidae paraphyletic. The MP analyses did not recover what is delineated as Pulvinomyzostomatidae in Figure 2 as a clade (see Additional file 1: Figure S1). The concatenated ML, Bayesian, and MP analyses recovered Endomyzostomatidae as polyphyletic (two genera sequenced). Endomyzostoma formed a clade that was sister to all other myzostomid terminals (Figure 2) and is here retained as Endomyzostomatidae, while the two terminals of Contramyzostoma were nested within clades composed primarily of Myzostoma (Figure 2, Additional file 1: Figure S1). Myzostomatidae was recovered as paraphyletic (all three genera sequenced), owing to the placement of Contramyzostoma (Endomyzostomatidae) and Mesomyzostoma (Mesomyzostomatidae) among the various Myzostomatidae terminals, and these taxa are now referred to this family. Figure 2 and Table 1 (right column) indicate the revised family taxon memberships and names.
Lifestyle transformations
Most myzostomids live externally on their host following a period of development in a cyst (i.e., cyst-to-free-living). This condition evidently arose once in both the ML/Bayesian and MP topologies, and is found in Eenymeenymyzostoma n. gen. and Myzostomatidae (Figure 3). Gall-forming myzostomids were recovered as a single clade within Endomyzostoma. Adult cyst-dwelling myzostomids were recovered in both Endomyzostoma and Myzostomatidae. In Endomyzostoma, these cyst-forming taxa were a clade. Three other adult cyst-dwellers (Contramyzostoma sphaera, Contramyzostoma bialatum, and Notopharyngoides platypus) were distributed across Myzostomatidae, each sharing a most recent common ancestor with a form that transitioned from a cyst to an external adult life. In all of these cases, this ancestor was estimated to have been cyst to free-living (prop. likelihood 1.00). Two taxa found in the mouth or digestive tube were in Pulvinomyzostomum and one in Myzostomatidae (Notopharyngoides aruensis), each arguably arising from a cyst to free-living ancestor (Figure 3). Host-eating forms were within Asteromyzostomum (found on/in asteroids), Protomyzostomum (living in ophiuroids), and Mesomyzostoma (living within crinoids). For both ML/Bayesian and MP topologies, internal host-eating was most likely the plesiomorphic condition for Mesomyzostoma (prop. likelihood 0.94) and possibly Asteromyzostomum + Protomyzostomum clades (prop. likelihood 0.58 and 0.52 respectively).
Patterns of host use and specificity
Figure 4A shows transformations for general host type. Association with asteroids and ophiuroids is restricted to Asteromyzostomum and Protomyzostomum, respectively. Eenymeenymyzostoma n. gen. resides exclusively on stalked crinoids, and the sequenced Myzostomatidae only inhabit feather stars. Endomyzostoma and Pulvinomyzostomum infect both stalked and feather star crinoids. The ML/Bayesian topology suggests a stalked crinoid was the ancestral host for Myzostomida (prop. likelihood 0.87), with one transition to asteroids (Asteromyzostomum), one to ophiuroids (Protomyzostomum), and three transitions to feather stars (in Endomyzostoma and the clade of Pulvinomyzostomatidae + Myzostomatidae, with a reversal back to a stalked crinoid for Pulvinomyzostomum messingi nomen nudum). Transformations on the MP topology suggest a feather star crinoid as the ancestral host (prop. likelihood 0.87), with one transition to asteroids, one to ophiuroids, and two or three transitions to stalked crinoids (for Pulvinomyzostomum messingi nomen nudum, Eenymeenyzostoma cirripedium, and in Endomyzostoma, or Pulvinomyzostomum messingi nomen nudum and the ancestor of E. cirripedium + Endomyzostoma, with a reversal to stalked crinoids in Endomyzostoma). Six most parsimonious reconstructions (MPRs) were found for host type on both ML/Bayesian and MP topologies, the variation due to unresolved nodes for Asteromyzostomum and Protomyzostomum.
Figure 4B shows the maximum likelihood transformations for major host clades, which mostly correspond to family-level classification of crinoids and also include Asteroidea de Blainville, 1830 and Ophiuroidea Gray, 1840. Isocrinid stalked crinoids are hosts to gall-Endomyzostoma and cyst-to-free-living Eenymeenymyzostoma n. gen., while Pulvinomyzostomum messingi nomen nudum is found on a hyocrinid stalked crinoid. Among feather stars, species of Antedonidae Norman, 1865 host two Pulvinomyzostomum taxa and two Myzostomatidae taxa. Myzostoma australe is found on a Ptilometridae AH Clark, 1914 and M. divisor Grygier, 1989 and Myzostoma josefinae nomen nudum (Summers & Rouse, in press.) on Zenometridae AH Clark, 1909. Comatulidae Fleming, 1828, Mariametroidea AH Clark, 1909, and Tropiometridae AH Clark, 1908 all host taxa within Myzostomatidae. The large number of MPRs for both ML/Bayesian and MP topologies (132 and 66 respectively) suggest multiple scenarios regarding switches among major host clades, especially at the family-level for myzostomids. Within Myzostomatidae, both topologies suggest two transitions to association with Comatulidae, three to four switches to Mariametroidea, and one or two inhabitations of Tropiometridae.
Additional file 1: Figure S2 presents the distribution of host specificity on the ML/Bayesian topology. Occurrence on only one host is most common, and is present in all families except Eenymeenymyzostomatidae n. fam. The 12 MPRs suggest that infestation of more than one host arose multiple times independently.
Comparison of myzostomid and host phylogenies
Figure 5 shows associations among myzostomids and their hosts (88 links in total). The host was known for 69 of 75 myzostomid terminals (Additional file 1: Table S1). Additional file 1: Table S3 lists the nominal species and 53 terminals used to estimate the host phylogeny. The MP and ML results for the concatenated datasets and each gene partition are provided in Additional file 1: Table S2. The overall phylogeny of the 69-myzostomid phylogeny was congruent with that recovered for all included terminals. For hosts, there were four major well-supported clades: stalked crinoids (Isocrinida and Hyocrinida), asteroids and ophiuroids, one family of feather stars (Comatulidae), and all other feather stars (Figure 5).
Most clades of myzostomids were associated with specific clades of hosts (branches shown in the same color - Figure 5). Exceptions include Pulvinomyzostomum (the two included terminals associate with a stalked and feather star crinoid, respectively), Notopharyngoides aruensis, Mesomyzostoma (taxa are found within two clades of feather stars), and Endomyzostoma (specimens infest stalked crinoids and two clades of feather stars). Endomyzostoma species are known from all the crinoid groups referenced here (as in Figure 4B) and along with Pulvinomyzostomum and Mesomyzostoma, are likely considerably undersampled.
Parafit analysis of myzostomids and their hosts rejected the null hypothesis of random association (ParaFitGlobal = 2439.93156; P = 0.00050). Three of the 88 individual links were significant (for terminals of Asteromyzostomum and Protomyzostomum) (prob1 and prob2 < 0.016). JANE analyses under default settings led to 4297 reconstructions (cost = 139) for the ML/Bayesian-rooted topology (e.g., Figure 5). Duplication with host switch was most used (40-41), followed by 29-30 losses, 19 failures to diverge, 18 co-speciation events, and 9-10 duplications. The mean costs were estimated as 377 (STD 30) for random associations and 331 (STD 25) for random parasite topology.