The adhesion disc of living remoras (Echeneoidea: Echeneidae) represents one of the most remarkable structural innovations within fishes. Although homology between the spinous dorsal fin of generalized acanthomorph fishes and the remora adhesion disc is widely accepted, the sequence of evolutionary—rather than developmental—transformations leading from one to the other has remained unclear. Here, we show that the early remora †Opisthomyzon (Echeneoidea: †Opisthomyzonidae), from the early Oligocene (Rupelian) of Switzerland, is a stem-group echeneid and provides unique insights into the evolutionary assembly of the unusual body plan characteristic of all living remoras. The adhesion disc of †Opisthomyzon retains ancestral features found in the spiny dorsal fins of remora outgroups, and corroborates developmental interpretations of the homology of individual skeletal components of the disc. †Opisthomyzon indicates that the adhesion disc originated in a postcranial position, and that other specializations (including the origin of pectination, subdivision of median fin spines into paired lamellae, increase in segment count and migration to a supracranial position) took place later in the evolutionary history of remoras. This phylogenetic sequence of transformation finds some parallels in the order of ontogenetic changes to the disc documented for living remoras.
From the orbital asymmetry of flatfishes  to the lures of anglerfishes , some of the most striking cranial innovations in vertebrates are found among acanthomorphs, a diverse clade of teleost fishes containing over 18 000 living species . One of the most bizarre cranial designs within acanthomorphs is found in Echeneidae (remoras or sharksuckers), a small clade of commensalists distinguished by a strongly depressed skull bearing a large, segmented adhesion disc that is used to fasten to hosts, including whales, turtles, sharks and other bony fishes [4,5].
The evolutionary roots of the remora suction disc have long been the subject of scientific debate. Developmental patterns [6–8], revealed in meticulous detail by recent ontogenetic work , combined with the segmented construction of the adhesion disc [9,10], indicate that this structure is derived from the anterior, spinous portion of the dorsal fin of other acanthomorphs. Modern outgroups of remoras bear spiny fins that differ considerably in their architecture, meristic counts and position from suction discs, leaving many questions about how this evolutionary transformation occurred. Because most fossil remoras are assigned to extant genera [11–13], they yield no clues beyond those already provided by living taxa. Exceptional among these palaeontological examples is the extinct genus †Opisthomyzon from the early Oligocene (Rupelian; approx. 30 Ma) of Switzerland. Although it was the first fossil remora to be described as such  and was initially embraced as a potentially pivotal taxon in understanding the origin of the group [6,9,10,15–17], †Opisthomyzon has subsequently been overlooked or dismissed as a probable relative of the living Phtheirichthys, and therefore an anatomically modern member of the crown clade [4,16].
Using a combined analysis of developmental, anatomical, fossil and molecular data, we find strong support for the placement of †Opisthomyzon outside the living radiation of remoras, refuting previous verbal arguments that this fossil genus nests within the echeneid crown clade [4,16]. More significantly, we show that the adhesion disc of this genus diverges substantially from that found in extant remoras, and instead displays striking positional, meristic and anatomical similarities to the spiny dorsal fins of outgroup taxa. This fossil evidence provides exceptional corroboration for hypotheses of homology drawn from painstaking study of recent anatomical and developmental datasets , implies a sequence of character transformation across phylogeny paralleling that occurring during ontogeny, and highlights the complementary roles of palaeontological and developmental data in documenting the evolutionary assembly of specialized anatomical structures.
2. Material and methods
(a) Comparative materials
Our analysis includes all living members of Echeneoidei, which contains the eight species of Echeneidae, two species of Coryphaenidae (dolphinfishes) and single species of Rachycentridae (cobia) . We also include two fossil taxa: the Oligocene remora †Opisthomyzon and the Eocene percomorph †Ductor, which has been associated with Echeneoidei . More distant relatives of Echeneoidei are represented by three carangiforms : two members of Carangidae (jacks) and the single species of Nematistiidae (roosterfish). All inferred phylogenies were rooted using the outgroup Pomatomus saltatrix, a non-carangiform and the only member of Pomatomidae (bluefish), which is argued to branch outside Carangiformes (sensu ) on the basis of both morphological  and molecular [3,22] evidence.
Osteological features for fossil and living taxa were assessed from adult specimens. Owing to the rarity of appropriate material in systematic collections [5,16], larval and other developmental characters in modern taxa were scored from detailed, well-illustrated descriptions [5,23]. A complete account of our comparative sample is given in the electronic supplementary material.
(b) Assembly of phylogenetic dataset
The relationships of remoras and their immediate relatives were assessed through analysis of anatomical (taken here to include osteological, soft-tissue and developmental) features coded for fossil and extant species and molecular genetic characters assessed for living taxa. The anatomical dataset used here is substantially revised from that presented by O'Toole , with a complete account of corrections and modifications provided in the electronic supplementary material. Our matrix contains 108 osteological, two soft-tissue and nine developmental (larval) features, for a total of 119 anatomical characters. Molecular sequences were obtained from GenBank for mitochondrial 12S, 16S and ND2 gene regions, and the nuclear ITS-1 region (GenBank accession nos. FJ374786–FJ374798, AP004444). These correspond to the sequences used by Gray et al. , plus additional data for Carangoides.
(c) Phylogenetic analysis
Phylogenetic analyses were completed using maximum parsimony and Bayesian inference. Parsimony analyses were conducted in PAUP* v. 4.0b10  using the branch-and-bound search algorithm, with gaps in molecular sequence data treated as missing information. Support for inferred clades was assessed by character bootstrapping (10 000 pseudoreplicates).
Bayesian analyses were conducted using a metropolis-coupled Markov chain Monte Carlo strategy implemented in the program MrBayes v. 3.2.1 . A single partition was used for ITS-1, 12S and 16S rRNA gene regions, and three partitions were used for the mitochondrial ND2 gene based on codon position. Models of molecular evolution were determined using the Aikake information criterion (AIC) as implemented in MrModeltest v. 2.3  (see the electronic supplementary material, table S1). The optimal partitioning scheme for mitochondrial ND2 was selected from the log of the harmonic mean of the likelihood values sampled from the posterior distributions of the two compared MrBayes runs [26–28]. Morphological characters were treated with the standard Markov-variable (Mkv) model of Lewis , assuming gamma-shaped rate variation and unordered character state transitions. Posterior trees and model parameters were sampled from MrBayes runs of 2.0 × 107 generations. Burn-in was set at 2.0 × 106 generations, discarding all trees and parameter values sampled before the burn-in. Stationarity of the chains and convergence of the trees and parameter values were determined by plotting the likelihood score and all other model parameter values against the generation number using the Tracer v. 1.5 . Convergence of the runs was also assessed by monitoring the average standard deviation of the split frequencies between the two independent runs, assuming that stationarity of chains was achieved when this value was less than 0.005.
In order to test the sensitivity of our phylogenetic inferences to sampling disparate types of comparative data, we repeated our analyses with the following partitions excluded: developmental data (characters 113–119; coded from literature and unknown for fossils); molecular data (unknown for fossils); developmental and molecular data; fossil data (†Opisthomyzon and †Ductor); fossil and developmental data. Solutions arising from these pruned datasets are summarized below and in the electronic supplementary material, tables S1 and S2.
3. Systematic palaeontology
Acanthomorpha Rosen, 1973; Percomorpha Rosen, 1973; Carangiformes Jordan, 1923; Echeneoidei Bleeker, 1852; Echeneoidea Johnson, 1993; †Opisthomyzonidae Berg, 1940; †Opisthomyzon Cope, 1889; †Opisthomyzon glaronensis (Wettstein, 1886).
(a) Holotype and referred material
Holotype: Naturhistorisches Museum der Burgergemeinde Bern, Bern, Switzerland, NMBE 5016633, complete individual preserved in left-lateral view (counterpart NMBE 5017410). Referred material: Natural History Museum, London, UK, NHMUK PV P.1995, impression of right side of skeleton (specimen of †Uropteryx elongatus nomen nudum; see below); NHMUK PV P.4953, complete individual preserved in right-lateral view; Paläontologisches Institut und Museum, Zurich, PIMUZ AI 2110, complete individual preserved in left-lateral view; Sedgwick Museum of Geology, Cambridge, UK, CAMSM C 31451, disrupted impression of right side of skeleton.
(b) Horizon and locality
Engi Slates, Matt Formation, Canton Glarus, Switzerland. Early Oligocene (Rupelian), but younger than approximately 32 Ma based on K/Ar and 40Ar/39Ar radiometric dates for the underlying Taveyannaz formation .
(c) Emended diagnosis
Echeneoidei differing from other members of that group in the following combination of characters: adhesion disc present and located over anterior trunk; disc lamellae median (rather than paired) ossifications in adults; spinous projections absent from disc lamellae; dermal skull roof strongly ornamented; anteriorly extensive soft dorsal fin, inserting far in advance of anterior insertion of anal fin; caudal fin deeply forked.
(d) Taxonomic remarks
Friedman & Johanson  mistakenly indicated that †Opisthomyzon glaronensis is a junior synonym of †Uropteryx elongatus. The latter is in fact a nomen nudum, because it is not accompanied by a description, definition or indication, and therefore fails to satisfy Article 12 of the International Code on Zoological Nomenclature .
We maintain placement of †Opisthomyzon in †Opisthomyzonidae rather than Echeneidae owing to the major morphological differences between this extinct genus and modern remoras. We also adopt a modified taxonomic scheme for remoras and their closest relatives. Johnson  identified Echeneidae, Coryphaenidae and Rachycentridae as echeneoids, but did not specify whether the intended formal name was Echeneoidei (an existing suborder) or Echeneoidea (a new superfamily). Both have since appeared in the literature, where their meaning has been identical in terms of composition [4,5,18,34]. We restrict Echeneoidea to the remora total group, presently comprising Echeneoidei and †Opisthomyzonidae. Echenoidei contains these families plus Coryphaenidae and Rachycentridae, matching the most recent usage of the name .
4. Description and comparison
We focus on the neurocranium and adhesion disc of †Opisthomyzon, and their relevance to documenting the evolutionary assembly of remora cranial anatomy. Abbreviated descriptions of the remainder of the skeleton are given for completeness, and emphasize differences between †Opisthomyzon and crown-group remoras.
The skull roof of †Opisthomyzon is broad, like that of living remoras, but differs from modern examples in a series of critical features. First, the lateral ethmoid does not make a large contribution to the dorsal margin of the orbit (see the electronic supplementary material, figure S1), unlike the derived condition characteristic of modern remoras. Second, the dorsal surface of the skull is flat or slightly convex (figure 1a), unlike the concave surface found in all extant remoras. Third, the frontals and parietals of †Opisthomyzon bear an irregular ornamentation (figure 1a,b), similar to that found in Rachycentron but different from the smooth surface that characterizes modern remoras. Together, these two final features indicate that the skull roof was covered by only a thin layer of soft tissue, rather than buried deep within tissue under an anteriorly positioned adhesion disc.
(b) Adhesion disc
Two specimens of †Opisthomyzon clearly show the adhesion disc (NMBE 5016633, figure 1a,b; NHMUK PV P.4953, figure 1c), and their slightly different modes of preservation provide insight into the anatomy of this feature. We accept the classical interpretation , recently corroborated by exquisitely detailed ontogenetic and anatomical study , that the lamellae, intercalary bones and interneural rays of the remora adhesion disc are the homologues, respectively, of the fin spines, distal radials and proximal-middle radials of other percomorphs. Of these major components of the adhesion disc, the lamellae are the most superficial, making them the easiest to examine in fossil material.
There are several conspicuous differences between the disc of †Opisthomyzon and that of all crown remoras. First, it lies posterior to the skull roof in both specimens where it is preserved (figure 1), confirming positional inferences based on the anatomy of the skull roof. Second, only six lamellae appear to be present, corroborating the low counts provided by previous workers . Third, the lamellae bear no pectinations along their posterior margin with the exception of the median spinule. Fourth, the lamellae are single, bilaterally symmetrical structures rather than paired elements. Fifth, the median spinules are not autogenous and are instead united with their associated lamella. Sixth, the disc is short, measuring slightly more than 10 per cent of standard length (SL), whereas the disc in modern examples ranges from 18–28% of SL in Phtheirichthys  to over 40% of SL in some species of Remora .
For the first five characters, states observed in †Opisthomyzon correspond to primitive conditions of the adhesion disc as predicted from dorsal fin structure in outgroups of remoras and developmental patterns in living echeneids (figures 1b,c and 2c). The position of the adhesion disc posterior to the skull in †Opisthomyzon disagrees with the supracranial position of living remoras, but corresponds with the arrangement found in generalized percomorphs and, to a lesser degree, that in early developmental stages in extant remoras, where the disc initially develops posterior to the orbits only to later extend to the anterior tip of the snout [5,7,8,36]. Living remoras  and other fossil examples [11,12] generally have discs bearing anywhere from 15 to 28 lamellae. Phtheirichthys is unusual among modern remoras in having a count of 9–11 lamellae, but this small number appears to be secondary based on the nesting of this genus high within the crown (figure 2a) . By contrast, the low count of lamellae in the adhesion disc of †Opisthomyzon corresponds closely to the number of dorsal spines in the closest relatives of remoras that bear distinct spinous and soft portions of the dorsal fin (7–9 in Rachycentron; 7 in †Ductor [19,37]). The absence of posterior pectinations along individual lamellae in †Opisthomyzon disagrees with the condition in extant remoras [4,5] and all other fossil representatives of the group with well-preserved adhesion discs [11,12]. However, the arrangement seen in †Opisthomyzon is in accord with the morphology of dorsal fin spines in outgroups of remoras, where trailing-edge pectination is absent [5,37]. The disc lamellae of †Opisthomyzon do not consist of paired right and left ossifications, as is the case in adult crown-group remoras , but instead consist of a single median bone. Additionally, the median spinule is united with the lamellae (figures 1b,c and 2c), rather than appearing as a separate ossification of the sort found in modern adult remoras (figure 2c). These last two features of the lamellae in †Opisthomyzon draw immediate comparisons to the unpaired dorsal fin spines of generalized percomorphs and the condition found in early developmental stages in extant remoras, where the disc lamellae are joined across the midline and are united with the median spinule .
(c) Cheek, jaws, suspensorium and opercular series
The jaws of †Opisthomyzon form a short, beak-like snout. The premaxilla bears a well-developed ascending process like that of generalized carangiforms, but which is absent in crown remoras (see the electronic supplementary material, figure S1). As in living remoras, the maxilla is slender and splint-like. The ventral process of the dentary is long, comparable with generalized carangiform conditions, but differing from the short process characteristic of extant remoras. Oral teeth in †Opisthomyzon are small, appearing to consist of a pavement of denticles. This contrasts with the condition in crown remoras, where long, slender teeth are present . The circumorbital series of †Opisthomyzon closely resembles that of living remoras, but obscures details of the suspensorium. It is clear, however, that the articular surface of the quadrate is directed anteriorly, matching the derived condition of extant remoras (see the electronic supplementary material, figure S1). The opercular series of †Opisthomyzon does not deviate radically from that found in modern remoras.
(d) Postcranial skeleton
The postcranium shows several proportional differences in comparison with living remoras (figures 1a and 2b; electronic supplementary material, figure S2). Principal among these is the shape of the body, which is relatively deep and fusiform, unlike the comparatively long and slender postcrania of extant remoras. This is complemented by the structure of the median fins. The anterior insertion of the dorsal fin lies far anterior to that of the anal fin in †Opisthomyzon, whereas insertion bases are effectively symmetrical in crown remoras. †Opisthomyzon also bears a caudal fin with a deeply notched posterior margin (figure 1a), which differs from the gently concave to convex profile that characterizes the caudal fins of modern remoras. The vertebral column of †Opisthomyzon comprises 10 abdominal and 13 caudal centra. Each centrum is anteroposteriorly elongate. †Opisthomyzon lacks the derived, laterally directed parapophyses common to living remoras  and all other fossil examples [11,12]. The body of †Opisthomyzon is covered with small, diamond-shaped cycloid scales (figure 1a; electronic supplementary material, figures S1–S3).
In all specimens, the pectoral girdle is largely concealed by opercular bones. However, details of the pelvic girdle are apparent. The girdle in †Opisthomyzon is narrow in comparison with its length, unlike the broad pelvic girdle common to extant remoras. However, the presence of a medial anterior arm of the girdle represents a derived feature linking †Opisthomyzon to modern remoras (see the electronic supplementary material, figure S3).
5. Evolutionary relationships
All phylogenetic analyses agree in the placement of †Opisthomyzon on the remora stem as the immediate sister group of crown echeneids (posterior probability = 1 in Bayesian analyses of complete datasets and all pruned datasets that include fossils; clade recovered in 100% of bootstrap replicates in parsimony analyses of both complete dataset and all pruned datasets that include fossils; figure 2a; electronic supplementary material, figures S4–S6 and tables S1 and S2). This genus shares with living remoras a series of unambiguous morphological synapomorphies, the most obvious of which is the adhesion disc. However, †Opisthomyzon retains several primitive characters not found in living remoras. Most striking are the fusion of disc lamellae along the midline (i.e. the presence of true dorsal fin spines), low number of disc lamellae comparable with the number of fin spines in living outgroups of remoras and absence of pectination, which are joined by a series of generalized features found elsewhere in the skeleton, including: placement of the spinous component of the dorsal fin posterior to the skull; lateral ethmoid not contributing to dorsal margin of orbit; dorsally convex, ornamented frontals bearing a sensory-line canal on the body of the bone rather than the margin; a small median ethmoid; an ascending process of the premaxilla; dorsal and ventral processes of the dentary of equal length; weakly developed anterior wing of the preopercle; presence of two postcleithra; a long pelvic girdle; and lack of greatly expanded horizontal parapophyses (see the electronic supplementary material, figures S5 and S6). The monophyly of crown-group remoras to the exclusion of †Opisthomyzon is strongly supported (posterior probability = 1 in all Bayesian analyses including fossils; clade recovered in 100% of bootstrap replicates in all parsimony analyses including fossils; figure 2a; electronic supplementary material, figure S4 and tables S1 and S2).
Significantly, most of our analyses provide strong support for the hypothesis that remoras form the sister group to a clade comprising Coryphaena and Rachycentron, consistent with published molecular phylogenies  and arguments made on the basis of larval development . This contrasts with recent cladistic treatments of anatomical data  and classical schemes , both of which have aligned Rachycentron with remoras to the exclusion of Coryphaena. The historical association of Rachycentron and remoras reflects conspicuous phenotypic similarities shared between these groups, but we argue that many of these traits are more general in their distribution. Most remoras and Rachycentron share a broadly similar external appearance, with a relatively slender body and darkly pigmented brown-to-black dorsal surface [36,37] that strongly contrasts with the vibrant coloration and distinctive body outline characteristic of Coryphaena. The early Eocene †Ductor, from the Lagerstätte at Bolca, Italy, shares with remoras and Rachycentron a slender postcranium with a darkly pigmented dorsal surface (e.g. NHMUK PV P.1987; electronic supplementary material, figure S7). Placement of †Ductor on the stem of Echeneoidei or as sister to Rachycentridae + Coryphaenidae in all of our analyses suggests that these characteristics are simply symplesiomorphies of crown Echeneoidei, and therefore have no bearing on the relationships between remoras and Rachycentron. We only recover a sister-group relationship between remoras and Rachycentron when both larval and molecular characters are ignored in maximum-parsimony analyses, or when molecular data, or these in combination with larval characters, are excluded from Bayesian analyses (see the electronic supplementary material, tables S1 and S2).
The remora adhesion disc is one of the most striking anatomical specializations of living vertebrates, but the consistent anatomy of this structure among modern species does not provide a clear picture of its evolutionary history. Detailed anatomical and developmental study has delivered compelling evidence for the homology between components of generalized percomorph dorsal fins and remora discs , but does not provide direct evidence for the sequence of evolutionary events by which one was transformed into the other. This problem is not unique to the remora adhesion disc. Many groups of teleosts are characterized by elaborate anatomical innovations absent in their closest living relatives. Fossils play a particularly important role in such cases and can deliver unique insights into the evolutionary origin of extreme morphological specializations [1,38–40].
The conclusion that †Opisthomyzon is a stem-group remora, combined with novel anatomical information arising from the discovery and preparation of additional specimens, provides the first clues about the sequence of evolutionary changes that led to the assembly of the adhesion disc. Mapping character states on our robustly supported phylogenies provides two-stage resolution relating to the evolution of this remarkable structure. It is clear that many key transformations, most notably modification of fin spines into laterally expanded lamellae, took place while the disc occupied a postcranial position. At this stage, lamellae were still joined along the midline, comparable with the condition of dorsal fin spines in generalized percomorphs. The second stage was characterized by anterior migration of the disc, the separation of lamellae into paired ossifications, the development of pectination along the posterior margins of the lamellae and an increase in the number of segments in the disc. Even at this coarse level of resolution, is it apparent that there are clear parallels between the ontogenetic origin of the adhesion disc and its sequence of phylogenetic transformation, the clearest of which are the migration of the disc on to the cranium, delayed division of individual lamellae and the late origin of spinules in both sequences . However, disagreements in the timing of specific events between these two trajectories, such as the expansion of the most posterior laminae and intercalary bones only after migration of the disc on to the skull during the development of living forms , indicate that one cannot be read as a surrogate for the other.
Many striking anatomical innovations of acanthomorphs—from the asymmetry of flatfishes to the bills of swordfishes to the lures of anglerfishes—appear to have arisen as part of a large-scale morphological radiation in the latest Late Cretaceous and earliest Palaeogene [3,41]. By contrast, the remora adhesion disc would seem to be a comparatively modern innovation. The early Oligocene †Opisthomyzon, along with roughly coeval material provisionally assigned to the extant genus Echeneis [11,12], represent the earliest known remoras. Rachycentrids and coryphaenids have no fossil record, and no remoras of Eocene or earlier age are known. The earliest representative of Echeneoidei is the early Eocene †Ductor, and the oldest carangiforms are jacks from the late Palaeocene . This palaeontological timescale agrees with that of recent molecular clock analyses that place the origin of the remora total group—and therefore the earliest possible date for the origin of the adhesion disc—in the mid-late Eocene . This interval has historically been poorly sampled for articulated marine acanthomorphs , but excavation of the few known deposits of this age continue to yield new taxa , raising the possibility that future work might uncover deeper branches of the remora stem group that will add further detail to our understanding of the evolution of the remarkable adhesion disc.
Financial support was provided by the John Fell Fund, St Hugh's College and NERC grant no. NE/J022632/1 to M.F. and Z.J.
We thank R. Arrindell, B. Brown (American Museum of Natural History), H. López-Fernández (Royal Ontario Museum), P. Campbell (NHMUK) and S. Raredon (National Museum of Natural History, Smithsonian) for providing access to Recent comparative material, H. Furrer (PIMZ), U. Menkveld (NMBE) and M. Riley (CAMSM) for access to fossil specimens, J. McDowell (Virgina Institute of Marine Science) for providing previously published molecular sequence alignments, and R. Britz (NHMUK) for conversations about the structure, development and homology of the remora adhesion disc. H. Taylor (NHMUK) took photographs of fossil specimens. M. V. H. Wilson (University of Alberta) and an anonymous reviewer provided insightful comments that improved this contribution, while C. Kammerer (Museum für Naturkunde) clarified the taxonomic status of the remora from the Engi Slates.
- Received May 13, 2013.
- Accepted June 20, 2013.
- © 2013 The Author(s) Published by the Royal Society. All rights reserved.