Detailed measurements of the investigated specimens are listed in the electronic supplementary material (Additional file 2: Table S1).
Composite phylogeny of extant and extinct Anguilliformes
The extinct anguilliforms, †Anguillavus and †Hayenchelys were identified as stem group members in a phylogenetic analysis [28] and are placed accordingly at the base of the tree. Extinct taxa such as †Abisaadia and †Libanchelys, that have not been included in any phylogenetic analysis up to now, were inserted manually by us at the conservative position of minimum assumption based on published information [26, 29, 30, 39, 50, 51]. The interrelationships of most stem-group representatives or even crown anguilliforms (e.g., Anguilla) consequently are unresolved displaying extensive polytomies. This, however, is not disadvantageous since we intend to identify evolutionary trends across clades rather than within smaller monophyletic units. The resulting tree was calibrated using the oldest occurrence of taxa and/or clades as extension ranges (Fig. 2). In doing this, we did not distinguish between actual stratigraphic ranges based on fossil occurrences and ghost lineages, since this is of no relevance here and does not influence our results. Nevertheless, origination and divergence estimates used here represent hard minimum rather than soft maximum age constraints.
The approach provides some new information about the interrelationships of extinct and extant anguilliforms even though it does not follow strict cladistics principles. Therefore, we provide a summary of the major results (Fig. 2), which are important for our evolutionary deductions. Node A depicts the sister group relationships between the most basal anguilliform, †Anguillavus from the Cenomanian (99-98 Ma) and all remaining anguilliforms. †Anguillavus is represented by two, probably synonymous, species [26]. Two other possible anguilliforms, †Enchelion and †Enchelurus, are known from the Cenomanian but are excluded here because of their unknown systematic status (see also [26]). The basal position of †Anguillavus is supported, inter alia, by the presence of reduced but still present pelvic fins and girdle [29, 51]. All other anguillforms lack the pelvic girdle. Divergence between †Anguillavus and all other remaining anguilliforms occurred at ca. 116 Ma [27] indicating that anguilliforms might have originated in the Aptian (Early Cretaceous).
Above node B an unresolved clade including †Abisaadia, †Luenchelys, †Hayenchelys, †Urenchelys, †Libanechelyidae, and crown-group anguilliforms (node D) form sister groups. Divergence is estimated here at ca. 116-99 Ma based on fossil occurrences. The clade comprising [†Abisaadia + †Luenchelys + †Hayenchelys + †Urenchelys] and †Libanechelyidae are thus supported here as stem-group anguilliforms. Nineteen characters support the monophyly of a clade above node C [51]. †Abisaadia might represent the most basal member of clade C because it still occasionally shows remains of the pelvic girdle and fins [29, 51]. †Libanechelyidae seemingly is the most advanced stem-group representative occupying an intermediate position between more basal stem group and crown-group members and is assumed to represent the sister taxon of crown anguilliforms [51].
Node E defines a monophyletic clade comprising all crown-group anguilliforms; divergence of crown-group anguilliforms is dated at ca. 99 Ma based on molecular data [27]. This correlates more or less with the oldest fossil occurrences of stem anguilliforms. The oldest fossil remains of clade E are known from the Santonian of Italy [29] and the Campanian-Maastrichtian of North America and Italy, respectively ranging from 84.7 – 74.5 Ma [52, 53]. This indicates that the origin of the total and crown clade as well as diversification of stem and basal crown members occurred between 116 and 74 Ma corresponding to a time when the climate was very warm and the supercontinent Pangea continued to break up resulting in the establishment of new near-shore habitats.
Node F represents the common ancestor of Anguilloidei (81 Ma). †Anguilloides from the Eocene of Italy (ca. 49 Ma) is a member of this clade but of uncertain systematic position. Within Anguilloidei, the family Anguillidae (node G) includes two fossil taxa, †Anguilla ignota from the Eocene (47.8 Ma) and †Anguilla elegans from the Miocene (ca. 13 Ma) of Germany, which can be considered as stem-group members of this clade. Nevertheless, all three terminal taxa are arranged in a polytomy here lacking detailed morphological trait analyses. †Anguilla ignota is known from maar deposits indicating that freshwater adaptation occurred early in the evolutionary history of anguillids. The origin of the total and crown groups is dated at ca. 60 Ma.
Congroid anguilliforms represent the sister group to muraenoids (node H) and a diverse group above node I. Their origin as inferred from bracketing divergence dates is estimated at 92 Ma. The genera †Bolcyrus and †Voltaconger, both known from the Eocene of Italy (ca. 49 Ma) are considered here to represent members of congroids with an uncertain relationship to other members. Most likely, they represent stem group members pending further phylogenetic analyses. The stem age for congroids is 86 Ma, which is in good accordance with the oldest known congroid, †Nardoechelys, from the Campanian-Maastrichtian of Italy [51].
The Eocene anguilliform †Paranguilla resembles that of muraenoids and places this taxon on the stem lineage of Muraenoidei [50]. The origin of the total-group Muraenoidei (node K) inferred from the fossil record dates back to 76 Ma, which indicates a major gap in our knowledge about their evolutionary history since divergence between congroids and muraenoids dates at ca. 92 Ma.
Analysis of body shape
Two morphospaces occupied by the investigated families of extant and extinct anguilliforms are evident (Fig. 4). The correlation analysis, which compares body length (Laxis1) with the total number of vertebrae (n) revealed two distinct morphospaces, which are characterized by members of extant and extinct anguilliforms, respectively. The morphospace of extant taxa is larger than the one of fossil taxa with an overlapping region comprising †Anguilloides branchiostegalis and Anguilla rostrata (Fig. 4). Throughout all investigated clades, extinct taxa have fewer vertebrae with shorter vertebral centra than extant ones. The highest number of vertebrae is found in the muraenoid Rhinomuraena quaesita, the congroid Saurenchelys fierasfer, and the anguilloidei Scolenchelys breviceps. The longest species of all investigated specimens are represented by †Anguilla elegans and Anguilla rostrata.
In the second analysis, the ‘vertebral shape index’ (VSI) is calculated to quantify body shapes in vertebrates [13]. Extinct and extant taxa occupy two different morphospaces with extant taxa having a larger morphospace than extinct species (Fig. 5). Additionally, an overlapping area of the morphospaces is identifiable containing the investigated specimens of †Anguilloides branchiostegalis, †Paranguilla tigrina, Anguilla rostrata, Scuticaria tigrina, and Gavialiceps taeniola. The precise loadings of the components of the principal component analyses are provided in the electronic supplementary material (Additional file 2: Table S1). The first component (PC1) correlates positively with the length and width of the body, the length of the skull, and the ratio between skull length and skull width. It correlates negatively with the number of precaudal and caudal vertebrae, and the ratio of length and width of the related vertebrae. The second principal component (PC2) correlates positively with the number of precaudal and caudal vertebrae, the length of the head in relation to the length of the individual vertebral centra, the ratio of the length and width of caudal vertebrae, and the ratio of head length and head width. This component correlates negatively with the length and width of the body, and the ratio of the width and length of the precaudal vertebrae. The third principal component (PC3) correlates positively with the total body length, the number of precaudal vertebrae, the ratio between the length and width of precaudal and caudal vertebrae, and the ratio of the width and length of the head.
Anatomy of the musculotendinous system in extinct anguilliforms
The ‘W’-shape of the myosepta in extinct anguilliform fishes is identified by the fossilized epineural bone (ENB) and epipleural bone (EPB), and, in one specimen, by the dorsally positioned myorhabdoid tendon (MT) (Figs. 1 and 3). In all investigated species, ENB and EPB are identified at different positions of the body, inclined posterodorsally and posteroventrally in the backward flexure of the corresponding myoseptum, and mostly can be traced to the posterior-most caudal region (Figs. 1 and 3, Additional file 2: Table S1). The length of ENB and EPB indicate the attachment line of the myosepta on the vertebral axis. The length of the lateral tendon, which represents the length of the myosepta, is inferred from the number of traversed vertebrae by ENB or ENP, and by adding one vertebra, which is spanned by DAC and VAC as seen in the extant specimens of Anguilla rostrata [7]. Attachment lines and length of myosepta of all investigated specimens are found in the supplementary material (Additional file 2: Table S1).
In stem group anguilliforms, represented here by †Luenchelys minimus, the attachment line of the myosepta traverses two vertebrae (N + 1). Adding the length of one additional vertebra corresponding to the anterior dorsal and anterior ventral cone, respectively, the length of the lateral tendon can be assumed to have traversed three vertebrae (Additional file 2: Table S1).
In †Anguilloides branchiostegalis, the attachment line of the myosepta attaches on the dorsal margin of the vertebral centrum N and crosses posterodorsal and posteroventral two subsequent vertebrae (N + 2). Adding the additional traversed vertebra of the lateral tendon, four vertebrae represent the length of the myosepta. Dorsally arranged MTs are preserved between 19–46 %TL. The ossified tendons of the horizontal septum can be identified as posterior oblique tendons (POTs), whereas epicentral bones (ECB) are not preserved.
In crown anguilloids represented here by †Anguilla elegans, ENB and ENP are traversing one to two vertebrae (N + 1/2). Adding the additional crossed vertebra, the length of the lateral tendon is represented by four vertebrae.
In the investigated fossil congroids, †Bolcyrus formosissimus and †Voltaconger latispinus, ENB attaches epaxially on the dorsal margin of the vertebral centrum N and continues posterodorsally across three subsequent vertebrae (N + 2) disappearing in the posterior body region between 89–97 %TL. Hypaxially, the EPB proceeds posteroventrally and at least traverses two additional vertebrae (N +1) but not more than three segments (N + 2) and disappears in the last 3–10 %TL. It is possible to infer a total length of four vertebrae hypaxially based on the epaxially and hypaxially symmetrical anatomy of the musculotendinous system and the additionally traversed vertebra of the lateral tendon.
The length of ENB and EPB decreases due to reduced length in posterior vertebrae. The dorsally positioned MT appears between 56–64 %TL and disappears at 85%TL. The hypaxial MT is only preserved in a single investigated specimen.
†Paranguilla tigrina, which is assumed to represent a basal muraenoid and which also displays preserved soft tissues in the posterior body portion, displays the longest lateral tendons of all investigated extinct taxa (Additional file 2: Table S1). The total length of the lateral tendon of this extinct taxon equals 4.5 vertebrae as indicated by the epaxial myosepta, which traverses N + 2.5 plus the additionally included vertebrae. Thus, basal muraenoids seemingly have the longest lateral tendon of anguilliform fishes. Hypaxially, no bony tendons are preserved.