Adipose fins are appendages found on the dorsal midline between the dorsal and caudal fins in more than 6000 living species of teleost fishes. It has been consistently argued that adipose fins evolved once and have been lost repeatedly across teleosts owing to limited function. Here, we demonstrate that adipose fins originated repeatedly by using phylogenetic and anatomical evidence. This suggests that adipose fins are adaptive, although their function remains undetermined. To test for generalities in the evolution of form in de novo vertebrate fins, we studied the skeletal anatomy of adipose fins across 620 species belonging to 186 genera and 55 families. Adipose fins have repeatedly evolved endoskeletal plates, anterior dermal spines and fin rays. The repeated evolution of fin rays in adipose fins suggests that these fins can evolve new tissue types and increased structural complexity by expressing fin-associated developmental modules in these new territories. Patterns of skeletal elaboration differ between the various occurrences of adipose fins and challenge prevailing hypotheses for vertebrate fin origin. Adipose fins represent a powerful and, thus far, barely studied model for exploring the evolution of vertebrate limbs and the roles of adaptation and generative biases in morphological evolution.
The fins and limbs of vertebrates are a classic model for studying both homoplasy and serial homology [1–4]. On the basis of position, vertebrate fins can be categorized as two kinds: paired and unpaired. Pectoral and pelvic fins are paired bilaterally and are homologous to tetrapod fore- and hindlimbs, respectively . Caudal, dorsal, anal and adipose fins are unpaired median fins, positioned on the body midline. In vertebrates, the evolution of new fins is rare . Fin number and skeletal anatomy are diagnostic characters for major vertebrate clades, suggesting that these characters are conserved in phylogeny and constrained in development. For instance, pectoral fins appear to have evolved only once, and pelvic fins are a synapomorphy of jawed vertebrates [6,7]. New fins can originate by either the subdivision of an existing fin domain—one module becoming two—as in the dorsal fins of some acanthopterygians , or by de novo origin—a new appendage developing in a location where there was not one previously. Pectoral fins are an example of de novo fin origin, the developmental programme of median fins having been recruited from paraxial to lateral plate mesoderm [9,10].
Fins that originate de novo are first observed in the palaeontological record as rudimentary outgrowths, which secondarily evolve greater anatomical and functional complexity . Dermal skeleton appears first and is followed by endoskeleton, the presumed first direct evidence of muscular attachment and active control in fins . The consistency of these patterns across fin systems suggests a common route for evolving new, complex appendicular systems. Although these patterns are discovered from study of the palaeontological record, their generality can be tested through analysis of extant diversity.
Adipose fins are appendages found in more than 6000 species of living teleosts, the major radiation of ray-finned (actinopterygian) fishes [11,12]. Adipose fins are defined by their position on the dorsal midline between the dorsal and caudal fins, and also by their anatomy, which appears less complex than the other fins of extant fishes . Generally, adipose fins lack the endoskeleton, dermal skeleton and associated musculature that characterize the other fins found in extant fishes [14,15]. Instead, adipose fins tend to be passive structures supported by rods of collagen oriented along the proximodistal axis of the fin called actinotrichia, which sandwich a subdermal space composed of either adipose tissue, as in some Siluriformes [15–17], or a non-adipose tissue, loosely termed ‘connective matrix’ [17–19]. Adipose fin diversity is not well characterized, and the evolution, development and function of these structures remain poorly understood.
Hypotheses of adipose fin evolution are rooted in questions of homology and whether adipose fins are adaptive. Garstang  postulated that adipose fins were vestigial and homologous to the second dorsal fin that is plesiomorphic in gnathostomes, seen today in Latimeria and many chondrichthyans. Bridge  similarly concluded from anatomical study that adipose fins were vestigial. But, recognizing a single dorsal fin to be the primitive condition of ray-finned fishes, Bridge argued that adipose fins represented the degeneration of a second dorsal fin, which itself originated by the subdivision of the actinopterygian dorsal fin. Fossils, however, do not support either of these hypotheses. Within ray-finned fishes, structurally elaborated second dorsal fins (i.e. with dermal skeleton, endoskeleton and associated musculature) are found almost exclusively within the Acanthopterygii, which lack adipose fins . Study of the fossil record highlights two lineages of non-acanthopterygian fishes that have evolved two dorsal fins: macrosemiids (e.g. †Agoultichthys chattertoni ), an extinct clade of ginglymodian fishes ; and †Placidichthys, an extinct genus of Halecomorphi [24,25]. Thus, most who have considered adipose fin evolution have concluded that the rudimentary appearance of these fins is plesiomorphic [15,19,26–30]. It is generally thought that adipose fins evolved once within the Actinopterygii, usually on the basis of their apparently limited function [19,26,28,29,31,32]. Some have proposed that their phylogenetic distribution might be best explained by homoplasy [15,27]; however, these hypotheses have never been tested.
Recent large-scale phylogenetic analyses are transforming our understanding of the evolution of teleost fishes [33–35], and hypotheses of character evolution must be revisited in the light of these new phylogenies. Here, we test competing hypotheses of adipose-fin origin and conclude that adipose fins have evolved repeatedly. We also characterize adipose fin skeletal anatomy for a diversity of fishes to test for similar patterns of structural elaboration among independent derivations of adipose fins and compare these patterns with what is known from the fossil record of other fins. Adipose fin skeletal evolution is highly homoplastic, differs significantly between instances of adipose fin origination and exhibits patterns of elaboration not observed in the fossil record of other fins. Finally, we discuss how fins, as serial homologues, originate and evolve.
(a) Ancestral-state reconstruction
To test how many times and in which groups adipose fins originated, we reconstructed a phylogeny of ray-finned fishes using the published data from Near et al. , which is comprised of 7587 base pairs from multiple nuclear genes for 232 actinopterygian fishes and 36 fossil age constraints. Using BEAST  with the same strategy as the original study, we recovered a pool of post-burn-in trees with a consensus topology that was congruent and similar in support of Near et al.  (electronic supplementary material, figure S1). The resulting trees are ultrametric, with branch lengths proportional to time. To account for evolutionary uncertainty, models of ancestral-state reconstruction were run across a distribution of 200 trees sampled randomly from the posterior distribution of post-burn-in trees generated by the Bayesian analysis.
Models of ancestral-state reconstruction were run treating the adipose fin as a binary character. Each of the 232 terminal taxa were coded for the presence or absence of an adipose fin with reference to specimens at the Field Museum of Natural History (Chicago, IL) and the literature (electronic supplementary material, figure S1). Maximum-likelihood and maximum-parsimony models of ancestral-state reconstruction were run in Mesquite . Two likelihood models were considered: a Markov k-state one-parameter (Mk1) probability model, and a Markov k-state two-parameter (Mk2) probability model. The Mk1 model defines the rate of transition between states as equal, whereas the Mk2 model allows the rate of transition between two states to differ, allowing for bias in the direction of character state transformation. A likelihood ratio test was used to determine whether MK1 or MK2 models produced a better fit of the data for each of the 200 sampled topologies. The two-rate (Mk2) model of character evolution fit the data significantly better than the one-rate (Mk1) model over all trees (likelihood ratio test: p < 0.001, d.f. = 1), and so only results from the Mk2 model are reported. The MK2 model was run both by treating the root state frequencies as equal and root state frequencies equal to equilibrium. Results were consistent for each method, and results are shown only for the model in which root state frequencies are equal to equilibrium.
(b) The evolution of form in adipose fins
To characterize the evolution of form in adipose fins, skeletal anatomy was studied making reference to (i) cleared and stained specimens from the Fishes Collection at the Field Museum of Natural History (Chicago, IL), (ii) photographs of X-rays uploaded to the electronic collections of the California Academy of Sciences (San Francisco, CA) and Le Muséum National d'Histoire Naturelle (Paris, France) and (iii) the literature. A total of 746 specimens belonging to 620 species, 186 genera and 55 families were studied (table 1 and electronic supplementary material, table S1). Additionally, the adipose fin skeleton of Mochokus niloticus was characterized by making reference to micro-CT scans, which were made available to us by Dr John Friel, Curator of Fishes at Cornell University's Museum of Vertebrates. The specimen CUMV91386 was methanol-preserved, treated with iodine solution (I2M) following the methods from Metscher , and portions of it were CT scanned at 1.2, 3, 12 and 41.5 µM resolution. Mimics (Materialise Inc.) was used to process the data and to generate three-dimensional reconstructions of skeletal anatomy.
(a) Adipose fins have evolved repeatedly
All models of ancestral-state reconstruction found support for multiple origins of adipose fins within Teleostei. The Mk2 likelihood model finds support for an origin of adipose fins in the group that includes Siluriformes and Characiformes, and again independently in the Euteleostei excluding Lepidogalaxiidae (figure 1). Specifically, adipose fins were likely absent in the lineage that diverged from the Euteleostei and led to the Otophysi (the clade composed of Siluriformes, Characiformes and Gymnotiformes; figure 1, nodes f,h,i,k,n). Although absence is recovered as unambiguously optimal in only a small fraction of the trees, proportional probabilities consistently show support for adipose fins having been absent. The Mk2 likelihood model also found equivocal support for the hypothesis that adipose fins originated a third time, in the Percopsidae (figure 1, nodes t–v). The parsimony model unambiguously found support for the same three adipose fins origination events (electronic supplementary material, figure S2).
(b) Adipose fins have repeatedly evolved endoskeleton and dermal skeleton
Endoskeleton is found in the adipose fins of three taxonomically distinct groups: Horabagrus brachysoma (Siluriformes), Stomiiformes and Myctophiformes (figure 2). In each case, the adipose-fin endoskeleton was composed of a cartilaginous plate, undifferentiated and generally positioned at the posterior of the adipose fin base (figure 2a) [14,15]. The presence of stained cartilage varied within genera (i.e. Neoscopelus and Spirinchus) and even within a species (i.e. Salangichthys microdon).
Dermal adipose-fin spines formed from modified midline scutes are found in three taxonomically distinct groups of Siluriformes: a clade within Loricarioidei, which includes Callichthyidae, Astroblepidae and Loricariidae; Sisor spp. (Sisoridae); and some Amphilidae (observed in Phractura, described in Andersonia  and Trachyglanis ; figure 2b–d). In each case, spines are anterior to the adipose fin membrane, posteriorly projecting and anteriorly bounded by additional midline scutes.
Fin rays are found in the adipose fins of at least four taxonomically distinct groups of fishes: three catfish groups and at least one characiform (figure 2e–k). Siluriformes with adipose fin rays include Clarotes spp. (Bagridae) [42,43], Phractocephalus hemioliopterus (Pimelodidae)  and Mochokus spp. (Mochokidae) . Within Characiformes, adipose fin rays were observed in Colossoma macropomum (Serrasalminae) and have also been described for C. brachypomum, and Pygocentrus piraya [44,45]. In all lineages except Mochokus, adipose fin rays are described as developing only after fishes have achieved a large adult size [42,44,45]. This may explain why some species previously reported with a rayed-fin phenotype, were not observed with adipose fin rays in our study (e.g. Clarotes laticeps, FMNH 50304; Pygocentrus piraya, FMNH 69988). Pygocenrus is the only genus previously described in the literature with a rayed adipose fin phenotype that we were unable to confirm. If P. piraya does indeed exhibit this phenotype, it likely represents a fifth unique origination of adipose fin rays, Colossoma and Pygocentrus being phylogenetically distinct [46–48].
Adipose fin rays are similar in their morphology to the lepidotrichia of other actinopterygian fins. They are composed of segmented and bilaterally paired hemitrichia, but are distinguished by their lack of associated musculature (figure 2e–k). In P. hemioliopterus, lepidotrichia segments appear to fuse to form non-segmented dermal rods, though segmentation remains visible in the smallest rays (figure 2g,h). In some species, adipose fin rays are separated by a thin fin membrane (e.g. C. macropomum and Mochokus spp.), whereas in other species, adipose fin rays are embedded in a thick fleshy fin (i.e. Clarotes spp. and P. hemioliopterus). In addition to flexible fin rays, Clarotes spp., also develop an anterior spine in large individuals (figure 2f). Conflicting accounts have been given as to how this spine develops: either through the elaboration of an anterior fin ray  or by the fusion of ‘fulcra-like scutes’ . Because of this developmental ambiguity, we have not included Clarotes spp., among the list of taxa with dermal skeleton derived from dermal scutes. The adipose fin rays of M. niloticus are significantly differentiated along the anteroposterior axis. The two anterior-most fin rays are unsegmented, and anterior fin rays have flared bases (figure 2i,j).
Homoplasy is defined as independently derived similarity . It has been studied, thus far, in biological systems as a pattern that arises between lineages and can reflect the operation of developmental constraints and natural selection [50,51]. Parts of an organism evolve with degrees of independence and can similarly exhibit derived similarity. For example, similarity between the fore- and hindlimbs of crown group tetrapods is a derived condition [7,52]. By studying anatomies that have evolved repeatedly, both between lineages and within an organism, we might discover common trajectories and logic in the routes by which adaptive landscapes are navigated.
(a) Adipose fins have evolved repeatedly
New hypotheses of actinopterygian phylogeny provide consensus on the relationships of major actinopterygian clades [33–35]. Our result, that adipose fins have evolved repeatedly, is based on the phylogenetic hypothesis of Near et al.  and appears robust relative to minor topological changes among euteleosts . A few analyses have found Characiformes to be paraphyletic, with Siluriformes nested within Characiformes [53,54]. Although our analysis does not include trees of this topology, characiform paraphyly is unlikely to change our result. The analysis presented here includes trees of three alternative topologies of the Otophysi, the clade composed of Gymnotiformes, Characiformes and Siluriformes. Each of the three topologies recover the same result that adipose fins originated in a common ancestor of Characiformes and Siluriformes.
There is equivocal evidence that Gymnotiformes might have primitively possessed adipose fins. A fossil gymnotiform was interpreted as having an adipose fin ; however, this interpretation has been challenged [56,57]. It has also been proposed that the ‘dorsal filament’ , an appendage found on the dorsal midline of some apteronotid gymnotiformes, which lacks a skeleton and musculature, is a transformed adipose fin. Assessment of this hypothesis yielded inconclusive results . Regardless, when the entire Otophysi clade is reconstructed as primitively possessing adipose fins, our reconstructions still support the hypothesis that adipose fins in this group evolved independently from those in the Euteleosti.
Although adipose fins usually are composed of soft tissue and preserve poorly, these fins have been identified in a number of fossils as body outlines. In each case, they are found on fishes belonging to families known already to possess adipose fins [60–62]. If adipose fins originated only once in teleost fishes, we would expect to find evidence of their loss in multiple lineages; however, this is not the case. For example, fossil Gonorynchiformes, which are abundant and often exquisitely preserved , are never observed with an adipose fin. Thus, although the occurrence of adipose fins in fossils was not coded in our analysis of ancestral-state reconstruction, we argue that treating these data independently is very likely to impede our ancestral-state reconstruction, because the distribution of fossil fishes with adipose fins is congruent with what is predicted by our model. Finally, the only non-acanthopterygian actinopterygian fishes known to have elaborated second dorsal fins are the macrosemiids  and Placidichthys [24,25]. Both these Mesozoic holostean clades are nested deeply within groups otherwise characterized by the presence of a single dorsal fin. This suggests further instances of distinct fin origination, but in these examples perhaps by subdivision of the primary dorsal outgrowth.
In addition to the number of originations recovered by ancestral-state reconstruction, detailed analysis of adipose-fin variation within major actinopterygian clades suggests that adipose fins might have originated twice more, in Cypriniformes and in Clariidae (Siluriformes). The cypriniform Paracobitis rhadinaeus is described with a ‘soft fin’ between the dorsal and caudal fins [18,64]. Originally described in the appropriately named genus Adiposia, the anatomy of its posterior fin is reminiscent of the adipose fin of the madtom catfish, Noturus exilis, and its structure is similar in its histology to other adipose fins: a superficial epidermal layer covers fibrous rods, likely actinotrichia, which run proximo-distally and posteriorly and sandwich a core of ‘connective tissue’ . Related species have been described with similar, but less prominent, extensions of the midline anterior to the caudal fin on both the dorsal and ventral midline . Paracobitis rhadinaeus is nested deep within the Cypriniformes , a clade that includes over 4000 species otherwise lacking adipose fins , and this structure thus originated within Cypriniformes. Additionally, within the Clariidae, a group of air-breathing catfishes, there is phylogenetic and anatomical evidence that adipose fins may have been lost and subsequently re-evolved. Only three of the 14 clariid genera possess adipose fins: Heterobranchus, Dinotopterus and Encheloclarias . Initially, it was proposed that these fishes represented the primitive Clariidae condition, the possession of an adipose fin and that the rest of the family had lost the fin . However, a recent phylogeny of clariid catfishes recovers Heterobranchus and Dinotopterus nested well within a clade of clariids lacking adipose fins . The adipose fins of Heterobranchus and Dinotopterus are anatomically unique when compared with all other adipose fins. Remarkably, they are supported by neural spines, which extend to the distal margin of the fin [67,70]. Whether this distinct anatomy reflects the independent derivation of an adipose fin of unique morphology or secondary modification of a more typical adipose fin is unclear. Regrettably, Encheloclarias is not included in recent molecular phylogenetic analyses, and it remains to be seen how its placement might affect interpretations of adipose fin evolution within the group.
(b) Are adipose fins adaptive?
The repeated evolution of adipose fins raises the question—what, if any, is the adaptive function of adipose fins? Homoplasy is not unto itself evidence of adaptation, as it can reflect non-adaptive processes [51,71]. However, natural selection readily purges fishes of their fins. For example, pelvic fins have been lost at least 80 times in teleosts alone . The maintenance of adipose fins for millions of years and the discovery of unanticipated structural complexity in some fins [15,19] implies functionality. Attempts have been made to identify adipose fin function through an eco-morphological approach , but fishes with adipose fins may be too varied in their ecologies for such an approach to be broadly informative. Adipose fins are found on fishes ranging in size from centimetres to metres in length; on fishes from diverse ecosystems, ranging from montane rivers to the deep sea and on fishes occupying nearly every trophic level that fishes hold, ranging from planktivores, to molluscivores and piscivores. Discovering the adaptive function of adipose fins might be best achieved through biomechanical modelling and experimental functional approaches, as in Reimche & Temple . Accumulating evidence suggests that some adipose fins function to direct pre-caudal flow or serve as a flow sensor (reviewed in reference ). These hypotheses are not mutually exclusive and, given the diversity of adipose fins in their size, positioning and composition, may not capture the full repertoire of adipose fin functions.
(c) Homoplasy in the adipose fin skeletal evolution
Adipose fins have evolved several kinds of skeleton, and each has evolved repeatedly. Endoskeleton appears to have evolved in adipose fins three times. This study highlights what is known about adipose fin endoskeletal elements, but likely underestimates the occurrence and therefore, too, the diversity of endoskeletal supports. It is notable that X-rays did not detect any adipose-fin endoskeleton, even in species expected to have them, for example, those species belonging to a genus in which endoskeleton had been previously described (electronic supplementary material, table S1). It is also noteworthy that Alcian blue does not always stain the acidic polysaccharides (e.g. glycosaminoglycans) in cartilages. For example, the adipose fin endoskeleton of H. brachysoma was identified by analysis of histological sections and does not stain with Alcian blue .
Two kinds of dermal skeleton have evolved in adipose fins. First, anterior spines derived from modified scutes have evolved in adipose fins three times, each time in the Siluriformes. Only one siluriform group, the Doradidae, has evolved dermal plates, but not an adipose fin spine . Prominent spines at anterior boundary of fins are common in early gnathostomes , which often exhibit similarly enlarged scales and scutes immediately anterior to the fin spine (e.g. Guiyu oneiros ). Second, fin rays have evolved in adipose fins at least four times, but only in characiforms and siluriforms. Adipose fin rays may evolve as a saltatory change. Two abnormal individuals, both siluriformes, have been reported with adipose fins possessing lepidotrichia [28,75]. These apparently spontaneous transformations imply that simple developmental or genetic changes can generate rayed-fin morphology in some adipose fins.
Patterns of dermal skeletal evolution differ dramatically between adipose fin originations. Dermal skeleton has evolved independently in adipose fins at least seven times, but it has only evolved within the clade that includes characiforms and siluriforms, never within euteleosts. Different patterns of dermal skeletal evolution might reflect differences between these two groups in how adipose fins develop. Regrettably, studies of adipose fin development are limited. Bender & Moritz  characterized the timing of adipose fin development in several species and, intriguingly, identified two major developmental modes, which they call characiform and salmoniform types. Development of the characiform type is characterized by adipose fin outgrowth after the complete reduction of the larval fin fold. By contrast, salmoniform-type development is characterized by adipose fins apparently developing as a remnant of the larval fin fold. If these different developmental modes are broadly representative of adipose fins in the Otophysi and Euteleostei, then it would be further evidence of their independent origins and might explain differences in their propensity to evolve dermal skeleton.
(d) The origin of vertebrate fins and limbs
Vertebrates seem to be constrained in their ability to evolve new fins and limbs. When new fins do evolve, they originate either by the subdivision of an existing fin domain or de novo. Fully elaborated de novo fins are observed in malformed fishes with some frequency [77–80]; however, this seems not to be a realized route of evolutionary innovation. Rather, fins that originate de novo are primitively rudimentary in form and function, secondarily evolving greater structural and functional complexity .
Primitively, adipose fins lack endoskeleton, dermal skeleton and associated musculature. Each of these components has evolved in the adipose fins of some lineages (this study and ). Adipose fins can evolve new tissue types and increased structural complexity by expressing fin-associated developmental modules in this new territory. This is best exemplified by the repeated evolution of fin rays in adipose fins, and has been proposed for adipose fin endoskeleton  and musculature . Broadly, high degree of developmental similarity, and thus serial homology, between fins may be a derived feature, general outgrowths converging on similar forms owing to common functional demands and biases in the generation of morphological variation.
Studies of the palaeontological record have identified a common pattern of elaboration in de novo vertebrate fins, with dermal skeleton evolving first and endoskeleton evolving second . Our analysis challenges the general nature of this pattern. In adipose fins, endoskeleton repeatedly evolves in the absence of dermal skeleton. Dermal and endoskeleton, therefore, need not evolve in a hierarchical pattern; the capacity to evolve one is not dependent on the presence of the other. Multiple routes of elaboration are possible.
Research is supported in part by the NSF-GRFP and University of Chicago Division of Biological Sciences.
The authors thank B. Winger, M. Nelson and J. Mitchell for helpful conversations on ancestral-state reconstruction methods, T. Sosa for illustrations and J. Friel for graciously sharing micro-CT scans. Additional thanks to K. Criswell, B. Aiello, H. Stinnett, A. Olsen, M. Hale and R. Ho for feedback on the manuscript.
- Received November 27, 2013.
- Accepted February 5, 2014.
- © 2014 The Author(s) Published by the Royal Society. All rights reserved.