Mitochondrial phylogeny of the brittle star genus Ophioderma.

We reconstructed the mitochondrial phylogeny of the species of the brittle star genus Ophioderma, using sequences of the Cytochrome Oxidase I gene (COI) to address four questions: (i) Are the species of Ophioderma described on morphological evidence reflected in mitochondrial genealogy? (ii) Which species separated from which? (iii) When did speciation events occur? (iv) What is the rate of COI evolution in ophiuroids? We found that most of the 22 described species we sampled coincide with monophyletic clusters of COI sequences, but there are exceptions, most notably in the eastern Pacific, in which three undescribed species were indicated. The COI phylogeny lacks resolution in the deeper nodes, but it does show that there are four species pairs, the members of which are found on either side of the central American Isthmus. Two pairs with a genetic distance of ~ 4% between Atlantic and Pacific members were probably split during the final stages of Isthmus completion roughly 3 million years ago. The rate of divergence provided by these pairs allowed the calibration of a relaxed molecular clock. Estimated dates of divergence indicate that the lineages leading to extant species coalesce at times much older than congeneric species in other classes of echinoderms, suggesting that low extinction rates may be one of the reasons that ophiuroids are species-rich. The mean rate of COI substitution in Ophioderma is three times slower than that of echinoids. Conclusions of previous mitochondrial DNA studies of ophiuroids that relied on echinoid calibrations to determine divergence times need to be revised.

Introduction

The Ophiuroidea (brittle stars), with more than 2000 species, is one of the two most species-rich echinoderm classes[1]. They inhabit benthic environments in nearly all depths and latitudes[2]. Molecular data have been used to elucidate their higher level phylogeny[3-9], to delimit species borders[10-19], and to document intraspecific population genetic structure[20-26]. However, to our knowledge, there are only two published molecular phylogenies of brittle stars at the genus or family level, those of Macrophiothrix[27] and of ophiocomid brittle stars[28]. Phylogenies showing the order of splitting of congeneric species from each other, the time that these splits occurred, and sister species relationships are the first step towards determining the possible causes of speciation. This paper attempts to make the first strides towards this end through a mitochondrial phylogeny of the species of the genus Ophioderma.

The genus Ophioderma Müller & Troschel, 1840 encompasses 33 extant described species[1], though two, O. tonganum Lütken, 1872 and O. propinquum Koehler, 1895, both described from the Indo-Pacific, appear to be based on doubtful locality information and possibly misidentified[29-31]. Another species, O. besnardi Tommasi, 1970, described from Brazil[32] may be the juvenile form of O. cinereum[33]. Five of the 33 species were recently described. Stöhr et al.[34] split O. longicaudum (Bruzelius, 1805) into O. longicaudum, O. zibrowii Stöhr, Weber, Boissin & Chenuil, 2020, O. hybridum Stöhr, Weber, Boissin & Chenuil, 2020, and O. africanum Stöhr, Weber, Boissin & Chenuil, 2020 and also resurrected O. guineense Greeff, 1882, which Madsen[35] had placed into synonymy with O. longicaudum. Granja-Fernandez et al.[36] described Ophioderma hendleri Granja-Fernandez, Pineda-Enriquez, Solis-Marin & Laguarda-Figueras, 2020 and pointed out that a number of Ophioderma museum specimens from the eastern Pacific were erroneously identified as belonging to previously described species.

The species of Ophioderma are distributed on both sides of tropical and subtropical America, the Mediterranean, and the West African coast. O. anitae Hotchkiss, 1982, O. appressum (Say, 1825), O. brevicaudum Lütken, 1856, O. brevispinum (Say, 1825), O. cinereum Müller & Troschel, 1842, O. devaneyi, Hendler & Miller, 1984, O. elaps Lütken, 1856, O. ensiferum Hendler & Miller, 1984, O. guttatum Lütken, 1859, O. holmesii (Lyman, 1860), O. pallidum (Verrill, 1899), O. phoenium H.L. Clark, 1918, O. rubicundum Lütken, 1856, and O. squamosissimum Lütken, 1856 are found in the Caribbean, Bahamas, and Bermuda[37,38]. Of these, only O. brevispinum, O. cinereum and (perhaps) O. brevicaudum have a range extending south to Brazil[39,40], and only O. brevicaudum and O. brevispinum spread north to the Carolina Banks and to the Cape Cod respectively[38]. O. januarii Lütken, 1856 is common in Brazil[40-43]. A nominal Brazilian species, O. divae Tommasi, 1971was described from Baía de Santos in Sao Paulo State[44], but it has never been reported from any other location. O. longicaudum exists in the Eastern Atlantic from Bretagne to Macaronesia and widely in the Mediterranean[34,45]. O. zibrowii is found in the eastern Mediterranean. O. hybridum is only known from the coast of Tunisia. O. africanum is only known from Senegal. O. guineense is listed by Stöhr et al.[34] as extending from Senegal to the Gulf of Guinea, although one mitochondrial clade, designated as corresponding to this species, was found by Boissin et al.[14] to also be present in the Mediterranean. O. wahlbergii Müller & Troschel, 1842 is found on the Atlantic coast of South Africa from Namibia to Danger point[46,47]. Depth records of most Atlantic species range from the littoral to ca 200 m, but O. pallidum has only been reported from 198 to 360 m off Cuba[38].

In the eastern Pacific there are 8 nominal species: Ophioderma panamense Lütken 1859, O. pentacanthum Clark 1917, O. teres (Lyman 1860), and O. variegatum Lütken 1856 range from southern California to Ecuador at depths of 0 to approximately 100 m[48] O. hendleri is spread from the Gulf of California to Colombia[36]. Other species have more restricted reported ranges. O. vansyoci Hendler 1996 is known from only two localities, one on each side of Baja California[49,50], O. peruanum Pineda-Enriquez, Solis-Marin, Hooker & Laguarda-Figueras, 2013 is only known from the coast of Peru[30], and O. sodipallaresi Caso, 1986 is only known from Mazatlán, Mexico[51]. O. elaps, known from the Caribbean[29,31,52-54], where it was originally collected and described, has also been reported from one locality in the Galapagos[29,48,55].

Fossils ascribed to extinct species of Ophioderma have been reported from the Permian[56], Triassic[56-59], Jurassic[60-62], Cretaceous[63], and the Miocene[64]. Chen and McNamara[57] have pointed out that many of these records are based on disarticulated plates or possess characters that do not place them in the genus. Aronson[65], on the other hand, described Jurassic fossils of Ophioderma at the British museum as “exceptionally well preserved”. The transcriptomic phylogeny of O'Hara, et al.[5], calibrated by multiple reliably identified fossils, suggests that the Ophiodermatidae did not originate until the Cretaceous, approximately 100 MYA (million years ago).

The majority of species of Ophioderma have lecithotrophic vitellaria larvae that in the laboratory settle approximately in eight days[37,66-70]. However, O. wahlbergii off South Africa broods its young[71,72], as do individuals previously considered as belonging to O. longicaudum but recently described as O. zibrowii from the eastern Mediterranean and O. hybridum from Tunisia[14,15,34]. Individuals of some species are estimated to be quite long-lived; O. brevispinum is thought to reach an age of 25–28 years, and O. longicaudum 30 years[69].

In this study we use partial sequences of the Cytochrome Oxidase I (COI) mitochondrial gene to address the following questions: (i) Are the species of Ophioderma described on morphological evidence reflected in mitochondrial genealogy? (ii) Which species separated from which? (iii) When did speciation events occur in this genus? (iv) What is the rate of COI evolution in ophiuroids?

Results and discussion

The COI phylogeny of Ophioderma (Figs. 1 and 2) showed little resolution in deeper nodes. Thus, little can be said about the relationships between species on COI alone. However, COI of most morphologically recognized species was shown to be monophyletic, which bolsters the case that COI evidence can be used to delimit species in this genus. Part of the doubt created by the Maximum Likelihood (ML) reconstruction is that the genus Ophioderma may not be monophyletic with respect to Ophiarachnella. The node joining these two genera was supported with a posterior of 1.00 in MrBayes, but with only 55% of the bootstrap reiterations of RAxML (Fig. 1). Conversely, the node joining these two genera with Ophiarachna received support of 1.00 in MrBayes and of 89% in RAxML. Such discrepancies between methods of phylogenetic reconstruction have made it necessary to collapse nodes that did not receive high support from at least one of the two methods.

Figure 1

Part of the phylogeny of COI haplotypes of Ophioderma. Second part is shown in Fig. 2. Reconstruction is based on Maximum Likelihood (ML), using RAxML[109], and Bayesian Information (BI) using MrBayes[108]. The tree is rooted on Ophiocomella pumila, but Ophiarachna incrassata and Ophiarachnella petersi were also included in the phylogeny to verify monophyly of Ophioderma. All nodes that did not receive support of at least 80% in ML or 0.9 in BI have been collapsed. Numbers next to nodes indicate ML and BI support (in this order). Asterisks indicate support of 100%. Codes indicate individuals in which a haplotype was present.

Figure 2

Second part of the phylogeny of COI haplotypes of Ophioderma, continued from Fig. 1. All nodes that did not receive support of at least 80% in ML or 0.9 in BI have been collapsed. Numbers next to nodes indicate ML and BI support (in this order). Codes indicate individuals in which a haplotype was present. Asterisks indicate support of 100%. Haplotypes obtained from the study of Boissin, et al.[14] are identified by codes that start with either “JN” or “FJ” (GenBank numbers). Clade labels starting with “L” are those designated by Boissin, et al.[14] , labels starting with C are from Weber et al.[101].

Species delimitation

COI sequences of most morphologically defined species were monophyletic, but there were some inconsistences between morphology and DNA. In the Atlantic, two COI haplotypes, one from French Guiana and one from Suriname, which were morphologically identified as belonging to O. januarii, were nested within haplotypes of O. brevispinum. That this was not the result of low phylogenetic resolution is evident from their being reciprocally monophyletic from haplotypes of O. brevispinum from Belize, while the Belize and South American clades were sister to O. brevispinum from Massachusetts (Fig. 2). O. januarii and O. brevispinum are morphologically similar[74], but the arms of the former are broader and more carinate near the disk and more tapered distantly, and the arm spines are longer and more tapered[73,75]. It remains to be determined whether they are, indeed, separate species. Another possibility is that Caribbean O. brevispinum and South American O. januarii belong to the same species, whereas the Massachusetts “O. brevispinum” is a separate species. This hypothesis would be consistent with a strict interpretation of relative genetic distances. The Maximum Composite Likelihood Genetic Distance between Caribbean O. brevispinum and O. januarii (6.80%) is smaller than the genetic distance between Massachusetts and Caribbean O. brevispinum (9.53%) or between Massachusetts O. brevispinum and O. januarii (8.05%). It is also possible that all three clades are separate species. Although molecular sampling in Brazil is necessary to choose between these alternatives, it would not be surprising if North American populations ascribed to O. brevispinum, occurring at higher latitudes than any other species of Ophioderma, were a distinct species. These northern populations had, in fact, been described as belonging to O. olivaceum by Ayres in 1852[76], but Lyman[77], without providing an explanation, synonymized O. olivaceum with “Ophiura brevispina”.

In the eastern Pacific there were two distinct and distantly related COI clades, the morphology of which does not fit the species description of any species from this ocean. One clade, designated in Fig. 1 as O.sp.1 and found in Anacapa, San Clemente and Santa Barbara islands off California, has a genetic distance of 14.31% from a clade found in California and Mexico that is designated as O. sp. 2, which it resembles in gross morphology (Supplementary Table S1). The latter clade, is sister to O. panamense, but separated by a genetic distance of 10.53% from it. Given the magnitude of these genetic distances, these two clades in all probability represent undescribed species. The morphology of the specimens from which the sequences were obtained clearly indicates that they do not belong to O. pentacanthum, O. vansyoci, O. peruanum, or O. sodipallaresi, nominal eastern Pacific species which we were unable to sequence. Two haplotypes of Ophioderma from Clipperton are reciprocally monophyletic with those of Atlantic O. elaps (Fig. 2). Given that their genetic distance from O. elaps is 4.95%, slightly higher than the distance between O. holmesii and O. teres (Supplementary Table S1), and given that O. elaps has also been reported from the eastern Pacific[29], they might have been expected to belong to this species. However, they are morphologically very different from O. elaps and much more similar to O. pentacanthum. They differ from O. pentacanthum in that their arm spines are longer and more tapered, their dorsal arm plates are trapeziform and relatively narrow, their ventral arm plates are W-shaped, rather than rhomboid or quadrangular, and separated by large gaps. These specimens must also belong to an undescribed species to which we refer here as O. aff. pentacanthum.

Relationships between species

Despite the low resolution of the COI phylogenetic reconstruction, evidence of shared common ancestors among some species is recorded in this mitochondrial marker (Figs. 1 and 2). The common ancestor of O. holmesii and O. teres split from the common ancestor of O. phoenium and O. cinereum. That O. phoenium and O. cinereum are sister to each other was also shown by Bribiesca-Contreras et al.[8], based on sequence from1462 exons and from COI, and by Christodoulou et al.[9] based on the same data, plus sequence from 28S. These studies did not sample O. holmesii but reported that O. peruanum, which we did not sample, is sister to O. teres. It is, therefore, possible that the eastern Pacific O. teres and O. peruanum are closely related and their ancestor was separated from the Atlantic O. holmesi by the rise of the Isthmus of Panama. Contrary to the unsupported speculation of Madsen[35], and in accordance with the conclusions of Tortonese[45], which were based on several reliable morphological features, O. cinereum and the O. longicaudum complex, with an average genetic distance of 12.43% from each other (Supplementary Table S1) are not related. O. guttatum is an outgroup of O. appressum and O. hendleri. In the phylogeny of Bribiesca-Contreras et al.[8] (which does not include O. hendleri), O. guttatum is an outgroup to a clade composed of O. squamosissimum and O. vansyoci, but the lower resolution of COI in our data did not provide support for the node joining it with O. squamosissimum. O. appressum and O. hendleri are another amphi-isthmian pair, but probably separated earlier than the completion of the Isthmus (see below). The members of the majority of geminate species in a variety of organisms were either separated by the protracted emergence of the central American Isthmus before the final closure, or else appear as if they have done so because of the extinction of true geminates[78,79]. Bribiesca-Contreras et al.[8] show O. variegatum, rather than O. hendleri, as the sister species of O. appressum, but, as O. hendleri had not yet been described at the time of their publication[36], it is not unexpected that they would assign the specimen they sequenced to O. variegatum. As Granja-Fernandez et al.[36] point out, O. hendleri and O. variegatum can easily be confused, because they both have radial shields covered with granules and naked adoral shields. The Isthmus also appears to have separated O. variegatum (proper) from the common ancestor of O. brevispinum and O. januarii, and also O. aff. pentacanthum at Clipperton from Atlantic O. elaps (Fig. 2). O. wahlbergii in both the Bribiesca-Contreras et al.[8] and the Christodoulou et al.[9] phylogenies appears as an outgroup of nearly all other species of Ophioderma, which is not incompatible with our COI phylogeny that lacks support for deep nodes. The COI phylogeny, however, shows O. wahlbergii to be in a clade that includes O. devaneyi and O. elaps and O. aff. pentacanthum, species that were not included in the Bribiesca-Contreras[8] et al. and the Christodoulou et al.[9] phylogenies.

We have included in our set of data sequences of O. longicaudum from a study by Boissin, et al.[14] primarily to determine the affinities of our sample from the Kassandra peninsula in the northern Aegean to the six clades they discovered. Stöhr, et al.[15], Boissin, et al.[14] and Weber et al.[16,80] found that individuals with brooded young belonged to mitochondrial clades labeled L2, L3 and L4 (as defined in ref.[14]), all found in warm, oligotrophic waters of the Mediterranean east of Peloponnese, whereas lineages L1, L5, and L6 contained broadcast spawning individuals. They suggested that the brooding lineages are separate species, and they were subsequently described as such[34]. In our phylogeny (Fig. 2) the Kassandra specimens were nested in lineage L1 along with specimens from the eastern Atlantic, Cyprus, Croatia, and Marseille. Their average genetic distance from the rest of clade L1 is 0.62%, whereas from the other clades of the O. longicaudum complex their average distance is 4.39% (Supplementary Table S1). Clade L1is predominantly found in the western Mediterranean but also in the Saronic Gulf, east of the Peloponnese[14]. Its presence at the Kassandra peninsula in the northern Aegean may be related to the colder waters of this area, because this clade is adapted to lower temperatures[80].

Chronology of branching and rate of evolution

In the COI phylogeny (Figs. 1 and 2), there were four clades separated by the Central American Isthmus: (a) Sequences of O. appressum were nested in those of O. hendleri (Fig. 1); (b) O. variegatum was reciprocally monophyletic with the O. brevispinum-O. januarii clade (Fig. 2); (c) Sequences of O. holmesii were nested in those of O. teres (Fig. 1); and (d) Sequences of O. elaps were sister to sequences of O. aff. pentacanthum (Fig. 2). The genetic distance between members of pair (a) was 13.03%, the distance between members of pair (b) was 12.99%, whereas that between members of pair (c) was 4.02% and between members of pair (d) was 4.95% (Supplementary Table S1). We, therefore, assumed that the divergence between the members of the last two pairs was more likely to reflect separation caused by the final stages of the emergence of the Central American Isthmus roughly 3 million years (MY)[79]. Thus, the split of O. holmesii from O. teres and the split of O. elaps from O. aff. pentacanthum were given an offset of 3 MY, and the resultant rate of COI divergence in Ophioderma was used by BEAST to estimate dates of clade separation by a log normal relaxed clock (Fig. 3). To estimate divergence times, it is necessary to preserve information on branch length. Nodes, if collapsed because of low support, would produce a false impression of antiquity. For this reason, all nodes in the fully resolved RAxML tree needed to be included in the BEAST analysis, even when they might have received low support. Figure 3, therefore, constitutes a chronogram. Given the wide 95% Highest Posterior Density (HPD) intervals of each node, the chronogram is compatible with the collapsed phylogenetic tree (Figs. 1 and 2).

Figure 3

Chronogram of Ophioderma generated by BEAST v. 1.10.4[111], showing median and 95% Highest Posterior Density (HPD) times of divergence between the clades. A log-normal uncorrelated relaxed clock was calibrated with an exponential prior that assumes that the minimum time of splitting between O. holmesii and O. teres and also between O. elaps and O. aff. pentacanthum (marked with green bars) was the completion of the Central American Isthmus 3 million years ago. Numbers next to nodes indicate median estimates of divergence times.

The general impression provided by the BEAST chronology (Fig. 3) is that in comparison to echinoid and asteroid molecular phylogenies, clades leading to extant species of Ophioderma are quite old. In fact, they may be much older than is estimated by median ages in COI, because codon positions of mitochondrial DNA (mtDNA) that are free to vary are likely to become saturated with time. That COI sequences underestimate the antiquity of the older clades is revealed by a comparison of the age of deeper nodes of Ophioderma involving the outgroups, with the estimated ages for the separation of ophiuroid families obtained by O'Hara, et al.[5] from 285 kb of sequence from 1552 exons and 11 fossil-based reliable date calibrations. By COI, the Ophiocomoidea (represented by Ophiocomella) and the Ophiodermatoidea (represented by Ophiarachna, Ophiarachnella and Ophioderma) were estimated by the COI chronogram to have branched from each other between the Paleocene and the Oligocene (median: 49.2 MYA, 95% HPD: 86.1–25.5 MYA). The Ophiomyxidae (Ophiarachna) were estimated as having split from the Ophiodermatidae (Ophiarachnella, Ophioderma) at 37.6 MYA (95% HPD: 64.8–18.9 MYA). According to the more reliable analysis by O’Hara et al. these separations between superfamilies and families occurred much earlier, between the Triassic and the Cretaceous 250–100 MYA.

More recent dates estimated on the basis of COI, which may be more accurate as they involve fewer multiple hits, suggest that Ophioderma clades that lead to extant species are older than those of extant congeneric species in other echinoderm classes. In Ophioderma, lineages terminating in extant species coalesce in the Oligocene; much of the cladogenic activity has occurred 27–10 MYA between the middle-Oligocene and early Miocene (Fig. 3). In most phylogenies of extant echinoids, COI lineages of extant species within genera coalesce more recently: ~ 0.5 MYA in Paracentrotus[81], ~ 3 MYA in Tripneustes[82] and also in Lytechinus[83], ~ 4 MYA in Echinometra[84], Heliocidaris[85] and in Arbacia[86], ~ 5 MYA in Eucidaris[87], ~ 5.5 MYA in Mellita[88] (but ~ 14 MYA in Diadema[89], and ~ 15 MYA in Encope[90]). The limited number of dated phylogenies of asteroid genera also suggest that lineages within a genus coalesce more recently than they do in Ophioderma. In Asterias they only go back to ~ 3.5 MYA[91] and in Leptasterias ~ 8 MYA[92]. In holothurians the most distant species of Stichopus go back to 4.6–8.8 MYA[93], but those of the immense, paraphyletic genus Holothuria are quite old, coalescing in the Triassic, some 240 MYA[94]. The case of Holothuria illustrates that taxonomic decisions as to the size of a genus will influence the appearance of antiquity of the species it contains. However, it is not likely that the species of Ophioderma appear to be old only because the genus has been too broadly defined, because there are no obvious morphological or molecular discontinuities that would suggest that it should be subdivided.

Whether this longevity of lineages is characteristic of ophiuroids in general or of Ophioderma in particular remains to be determined. The phylogeny of Ophiocomidae by O'Hara, et al.[28] reveals that their genera are also old. In this family, the genus that evolved most recently dates back to the Paleogene, 30 MYA. The persistence of ophiuroid lineages terminating to the present time suggests that this class of echinoderms may suffer a lower rate of extinction than echinoids. This may be part of the reason that they contain more than double the number of extant species than echinoids.

Contrary to conclusions from previous studies that the ophiuroid mitochondrial mitogenome in general and the COI gene in particular evolve faster than that of other classes of echinoderms[95,96], COI substitution rate in Ophioderma, as calibrated from the age of the completion of the Isthmus of Panama, is three times slower than the average rate of similarly dated echinoids. The separation between the geminate species O. teres and O. holmesii dated from BEAST, occurred 3.5 MYA, and the genetic distance between the two species was 4.02%. The separation of O. elaps from O. aff. pentacanthum was also dated at 3.5 MYA and the genetic distance was 4.95%. Thus, Ophioderma divergence rate in COI has proceeded at approximately 1.15–1.41% per MY. Roy and Sponer[97] estimated the rate of COI divergence in Ophiactis to be 0.87% per MY. In echinoids, the average rate of COI divergence in six genera summarized by Lessios[78] is 3.66% (range 2.90–4.50%) per MY. Since that publication, a higher rate of 7.85% per MY has been found in Mellita[88] and a much slower rate of 0.23% in Encope[90], illustrating that even within a single family rates of substitution can vary by an order of magnitude, but preserving the echinoid mean at 3.76% per MY. With the exception of Roy and Sponer [97] and of Richards, et al.[25], previous studies of evolution in ophiuroids based on COI[12,14,17,18,98] relied on rate calibrations from echinoids, as the echinoderm group in which the calibrations were most extensively determined at the time that the studies were conducted. The dates in these studies are in need of revision, as are some of the conclusions based on them. Given the substitution rate of COI in Ophioderma, the clades (now different species) of O. longicaudum did not begin separating at the time of Pleistocene glaciations after 2.4 MYA as Boissin, et al.[14] estimated, but more likely at about 7 MYA, before the Messinian crisis[99]. Assuming that class-specific calibrations provide better estimates than phylum-based estimates, and applying the Ophioderma calibration to studies of other ophiuroid genera, the sister clades of Ophiarachnella, Ophiopeza and Ophiolepis discovered by Hoareau, et al.[18] in the southern Indian Ocean did not separate between 1.6 and 3.9 MYA, but rather between 4.2 and 11.7 MYA. The divergence between intertidal and subtidal populations of Acrocnida brachiata occurred closer to 10 MYA instead of 3.5[98] or 5 MYA[12], adding evidence in favor of recognizing the intertidal form as a separate species[100]. Similarly, divergence between lineages of European Ophiothrix occurred 14–22 MYA, rather than the Miocene–Pliocene transition 4.8–7.5 MYA[10].

Conclusions

The COI phylogeny of the species of Ophioderma is far from the last word on the reconstruction of the relationships between its species, but it does illustrate that several undescribed species may be present, and that, dated with calibrations specific to this genus, lineages coalesce farther back in time than those of the studied genera of echinoids, asteroids and holothuroids.

Materials and methods

Collection of specimens

We sampled a total of 185 individuals of 21 species of Ophioderma from 25 localities (Fig. 4) either collected by us, donated at our request, or available in the Natural History Museum of Los Angeles County. To the set of our data, we added 16 Cytochrome Oxidase I (COI) sequences of O. longicaudum from Boissin, et al.[14], 3 to 5 from each of the six clades of COI they identified. Their clade L1 (C3 in Weber et al[101]) corresponds to O. longicaudum[34], their clades L2, L3 and L4 (C2, C5 and C6 in Weber et al.[101]) correspond to O. zibrowii, clade L5 (C2) is present in O. africanum, and L6 (C1) in O. guineense, although the species name of Mediterranean specimens that share this clade is unclear (see introduction). Sequences of O. longicaudum from the Canary Islands and from Madeira (GenBank Accession numbers FJ716117, FJ716121, FJ716122, JN603483-JN603485, JN603517- JN603525, JN603556) that appeared in Boissin, et al.[14] had been obtained by us for the present study. Of the species of Ophioderma that are currently regarded as valid in the World Register of Marine Species[1], we were unable to either obtain specimens or to amplify DNA from O. besnardi, O. ensiferum, O. divae and O. pallidum from the western Atlantic and from O. pentacanthum, O. vansyoci, O. peruanum, and O. sodipallaresi from the eastern Pacific. As outgroups we included one specimen of Ophiarachnella petersi from the Bahamas, one of Ophiarachna incrassata from the Philippines, and one of Ophiocomella (previously Ophiocoma) pumila from the Atlantic coast of Panama. Samples were preserved in 95% ethanol or in high-salt DMSO buffer[102].

Figure 4

Collection localities of specimens of Ophioderma used in this study. Colors indicate the species as determined from their morphology; letters indicate localities, numbers indicate sample size of each species. D: Belize (Carrie Bow Cay). E: Panama (Bocas del Toro), D: Belize (Carrie Bow Cay). E: Panama (Bocas del Toro), D: Belize (Carrie Bow Key), W: Puerto Rico (off Isla Maguayez), Σ: Bahamas (San Salvador), X: US Virgin Islands (St. Croix). D: Belize (Norval Cay), Y: Massachusetts (Waquoit Bay, Cape Cod). : D: Belize (Carrie Bow Cay, Twin Cays), B: Saint Vincent, E: Panama (Portobelo), F: Salvador, Brazil. G: Gulf of Mexico (S. of Galveston), H: Florida (NE of Vero Beach). Σ: Bahamas (San Salvador, Eleuthera, Rum Cay), B: Saint Vincent, C: Barbados. : D: Belize (Carrie Bow Cay), B: Saint Vincent. Θ: Isla del Coco, R: Bay of Panama (Islas Perlas). : H: Florida (NE off Vero Beach), G: Gulf of Mexico. J: South Carolina (Edisto Island). : K: Guyana (Maroni), L: Suriname (Paramaribo). Γ: Southern California (Anacapa Island, San Clemente Island, Santa Barbara Island, Santa Catalina Island), Γ: South California (Corona del Mar, San Clemente Island, Santa Catalina Island), Δ: Isla Guadalupe, Mexico. R: Bay of Panama (Perlas Islands). : D: Belize (Carrie Bow Cay), E: Panama (Portobelo). E: Panama (Bocas del Toro, Portobelo), G: Gulf of Mexico, M: Curacao (Habitat Reef, W. Coast), D: Belize (Carrie Bow Cay). : N: Off El Salvador. P: Clipperton Island. : Q: Navassa Island, Σ: Bahamas (San Salvador). : R: Bay of Panama (Isla Taboguilla, Islas Perlas). : V: South Africa (False Bay, Cape of Good Hope). : Z: Gran Canaria, Ψ: Madeira, Ω: North Aegean (Kassandra Peninsula). Two-letter codes starting with “A” in O. longicaudum, O. zibrowii, O. africanum and O. guineense are localities of sequences obtained by Boissin, et al.[14] with species names designated by Stöhr et al.[34]. AA: Senegal (Dakar), AB: France (Marseille), AE: Lebanon, AF: Crete, AG: Dodekanese (Symi, Rhodes), AH: Cyprus, AI: Portugal (Algarve), AJ: Croatia, AK: African Spain (Ceuta), AL: Tunisia, AM: Italy (Naples). Detailed location data are listed in Supplementary Table S2. See text for explanation of taxa designated as O. sp,1, O. sp,2, and O. aff. pentacanthum. Map outline downloaded from http://woodshole.er.usgs.gov/mapit/.

DNA extraction and sequencing

Genomic DNA was extracted from pieces of the arms of the ophiuroids by proteinase K digestion[103], or using Qiagen DNeasy Blood and Tissue®, or Gentra Puregene Tissue® Kits, or Lucigen “Quick Extract” protocol®, or Promega Wizard Plus Purification System®. Amplifications of a 625 bp fragment of the COI region of mtDNA were carried out with primers CO1-f 5 CCTGCAGGAGGAGGAGAYCC or OphCOI-For 5' CAACAYYTATTYTGRTTYTTYGG in the forward direction, and CO1-a 5' AGTATAAGCGTCTGGGTAGTC or OphCOI-Rev 5' CCTARRAARTGTTGWGGGAARAA or CO1-TR1 5' GGCATTCCAGCTAGTCCTARAA in the reverse direction. PCR amplification was performed in 50 μL of PCR reaction mixture A (0.3 units of Promega Flexi Go Taq® , 2.5 μL of 5X colorless buffer, 0.625 μ1 of 10 μM of each primer, 1.25 μL of 8 mM dNTPs, 1.25 μL of 25 mM MgC12) or reaction mixture B (0.4 units of Invitrogen Platinum Taq® , 1.25 μL of 10X buffer, 0.625 μL of 10 μM of each primer, 2 μL of 0.8 mM dNTPs, 0.625 μL of 100% Dimethyl Sulfoxide, 0.75 μL of 50 mM MgC12). The samples were heated to 96 °C for 5 s, then cycled 39 times through 94° C for 30 s, 50° C for 45 s, 72° C for 60 s, followed by 5 m in 72 °C and 5 m in 10 °C. The PCR products were cleaned with the ExoSap-IT® kit (USBCorporation), then cycle sequenced in both directions using the amplification primers and electrophoresed in an ABI3130 or and ABI3500 automated sequencer. Attempts to amplify nuclear markers, the i51 intron[104] and an Actin-2 intron, produced unreliable amplifications and inconsistent results in different extractions; these data were not used.

Phylogenetic analyses

We eliminated redundant haplotypes and the outgroups from the set of data and then used Posada’s[105] jModelTest v. 0.1.1 program to determine the simplest model of mitochondrial DNA that produced the best fit of the data to the tree, based on the AIC criterion[106]. This was the General Time-Reversible model[107] with a gamma correction (α = 0.907) and a proportion of invariable sites (P = 0.5480). After adding the outgroups, we used this model and two algorithms to reconstruct the mitochondrial phylogeny of the species of Ophioderma. We performed phylogenetic reconstruction by Bayesian Inference (BI) in MrBayes v.3.2.2[108] and by Maximum Likelihood (ML) in RAxML v. 8.2.6[109] (without the invariable site correction). The ML analysis was run in the CIPRES Gateway[110]. We used the options for rapid bootstraps and automatic halting. Support values for the nodes were estimated from 504 bootstraps. In MrBayes we employed the models suggested by jModelTest, but let the program estimate the parameters. Mr. Bayes was run in 2 chains for 3 × 107 steps, which allowed the average standard deviation of split frequencies to fall below 0.01, and the potential scale reduction factor to be equal to 1.00. Convergence was also determined in multiple runs, which produced the same topology. Nodes that received < 80% support in ML and < 0.9 in BI were collapsed.

To estimate dates of divergence between major clades we used BEAST v. 1.10.4[111]. The program was given the fully resolved tree produced by RAxML, which was compatible with the MrBayes tree. Operators causing topology searches (“SubtreeSlide”, “Narrow Exchange”, “Wide exchange”, “WilsonBalding”) were turned off to force BEAST to place time estimates on the nodes of the ML tree, but BEAST was allowed to estimate branch lengths. To calibrate rate of divergence, the separation between Atlantic and Pacific haplotypes of two pairs of amphi-American sister species with the least divergence between their members, O. holmesii-O. teres and O. elaps-O. aff. pentacanthum (see results), was given an offset of 3 million years (MY) in a Lognormal Uncorrelated Relaxed clock. Three MY is the generally accepted approximate date of the completion of the Central American Isthmus[78,79,112,113]. However, as there are claims that there were intermittent closures starting at approximately 13 MYA[113,114], the priors for these calibration points were set with exponential distributions, ranging from 3 to 13 MY. Three separate runs of 107 steps each, recording every 1000th tree were performed. Logs from the three runs were combined in LogCombiner v. 1.10.4 after removing the first 103 trees from each run and viewed in Tracer v. 1.6 to verify that there were no trends and that effective sample size (ESS) values for all estimated parameters was > 231.

Maximum likelihood composite genetic distances[115], taking into account differences in composition bias[116], were calculated in MEGA v. 7.0.20[117] with gamma corrections as estimated by jModelTest.

Supplementary Information 1.