A Late Devonian coelacanth reconfigures actinistian phylogeny, disparity, and evolutionary dynamics

Abstract

The living coelacanth Latimeria (Sarcopterygii: Actinistia) is an iconic, so-called 'living fossil' within one of the most apparently morphologically conservative vertebrate groups. We describe a new, 3-D preserved coelacanth from the Late Devonian Gogo Formation in Western Australia. We assemble a comprehensive analysis of the group to assess the phylogeny, evolutionary rates, and morphological disparity of all coelacanths. We reveal a major shift in morphological disparity between Devonian and post-Devonian coelacanths. The newly described fossil fish fills a critical transitional stage in coelacanth disparity and evolution. Since the mid-Cretaceous, discrete character changes (representing major morphological innovations) have essentially ceased, while meristic and continuous characters have continued to evolve within coelacanths. Considering a range of putative environmental drivers, tectonic activity best explains variation in the rates of coelacanth evolution.

Fig. 1. Acid prepared material and skull reconstruction of Ngamugawi wirngarri gen. et sp. nov.

A, B ‘Part a’ of WAM 09.6.148 (holotype) shown in left dorsolateral view and skull close up in left lateral view. C ‘Part b’ of WAM 09.6.148 (holotype) showing all exposed elements; D partial braincase of NMV P231504 (paratype) shown in right lateral view; E cleithrum of NMV P231504 (paratype) in mesial and lateral view; F, G skull reconstruction in dorsal and left lateral view. Abbreviations: Ang angular, Cl cleithrum, Clv clavicle, Dt dentary, Exc extracleithrum, icj intracranial joint, ioc infraorbital canal, L.Gu lateral gular, Lj lachrymojugal, mc mandibular canal, L.Ex lateral extrascapular, Op operculum, Par Parietal, Po postorbital, Pop preoperculum, Pp postparietal, Pmx premaxilla, Psym parasymphysial, Q quadrate, Ro.p1 anterior pore of the rostral organ, Ro.p2 antero-lateral pore of the rostral organ, Ro.p3 postero-lateral pore of the rostral organ, So supraorbitals, soc supraorbital canal, Sop Suboperculum, Spl splenial, Sq squamosal.

Fig. 2. Volume renderings of high-resolution µCT scans showing 3D anatomy of select elements of Ngamugawi wirngarri gen. et sp. nov.

A–D ‘Part a’ of WAM 09.6.148 (holotype) shown in two views; E–G cranial endocast reconstruction from WAM 09.6.148 in dorsal, ventral and left lateral view; H–J partial braincase of NMV P231504 (paratype) shown in left lateral, posterior and ventral views; K–N left mandible of WAM 09.6.148 (holotype) shown in left lateral, medial/mesial, dorsal and ventral views. Abbreviations: Ang Angular, ant.scc anterior semicircular canal, Bb basibranchial, Bpt Basipterygoid process, Cb ceratobranchial Cor coronoids, Dt dentary, end canals for endolymphatic ducts, Enpt entopterygoid EthSph ethmosphenoid, hf hypophysis, ica canal for internal carotid artery, lat.scc lateral semicircular canal, L.Gu lateral gular, Lj lachrymojugal, n.I-VII cranial nerves I to VII, Mand mandible, nc nasal capsule, nc notochordal canal, Op operculum, OtOcc otico-occipital, Pa parietal, Po postorbital, Pop preoperculum, Pp postparietal, post.scc posterior semicircular canal, PQ palatoquadrate, pr.a antotic process, pr.c processes connectens, Psph parasphenoid, Psym parasymphysial, pv canal for pituitary vein, ro rostral organ, sacc sacculus, Spl Splenial, Sq squamosal, soph canal for superficial ophthalmic nerve. All CT data and models are available via: www.morphosource.org/projects/000485769?locale=en.

Fig. 3. Phylogenetic relationships and divergence dates within coelacanths, based on tip-dated Bayesian inference.

Ngamugawi wirngarri gen. et sp. nov., shown in enlarged black text. Each branch is coloured according to median rate of morphological evolution of discrete characters under the uncorrelated lognormal relaxed clock (details in Fig. 4). Numbers at branches refer to posterior probability. Identification of, and credit for, the coelacanth silhouettes is provided in Supplementary Fig. 4.

Fig. 4. Results based on Bayesian total-evidence tip dating under the uncorrelated lognormal clock; there is a sharp slowdown in discrete character evolution in the most recent time bin, but no such slowdown for meristic or continuous characters.

Rates of evolution through time for A discrete, B meristic, and C continuous characters; each plot shows the duration and rate for every branch in the MCMC tree sample. Median rates for each branch for D discrete, E meristic, and F continuous characters; taxon labels for trees are as in Fig. 3. Due to constraints in BEAST2, absolute rates of change are shown for discrete characters, but relative rates (weighted average =1) shown for meristic and continuous characters. In A–C, fuzzy plots depict branch rates in 1000 MCMC tree samples (8000 post-burnin trees, thinned by factor of 8); box and whisker plots depict interquartile (50%) and range for the average rate in each sample for the relevant geological period. In D–F, branch rates are median rates for 8000 post-burnin trees.

Fig. 5. Temporal trends in environmental drivers of coelacanth evolutionary rates, showing that, of the variables measured, subduction flux had the greatest influence on coelacanth evolution overall.

A Scaled and centred rate of evolution of all coelacanths, B subduction flux; C continental flooded area (incorporating % shallow seas), D sea surface temperature; E atmospheric CO₂ concentration, F dissolved (marine) O₂, and G relative influence of each environmental driver on rate of coelacanth evolution. x-axis error bars in A-F are standard errors derived from the source literature, y-axis error bars are 95% confidence limits; error bars in G represent 95% confidence limits and are derived from 1000 iterations of the boosted regression tree analysis. Green = Palaeozoic, blue = Mesozoic, yellow = Cenozoic.

Fig. 6. Discrete and morphometric disparity of coelacanths.

A Principal coordinates analysis plot showing principal coordinate 1 versus principal coordinate 2, disparity based on discrete characters for all coelacanth taxa (n = 82 species). B–D Principal component analysis plot showing principal component 1 versus principal component 2 disparity based on 2-D geometric morphometrics for B body shape (n = 35 species); C cheek bones (n = 34 species); and D lower jaw shape (n = 38 species). Time-binned morphospaces per geological period compare the distinctiveness, similarity, and temporal evolution of the morphological disparity of coelacanths over a period of 410 Ma. Larger morphospaces indicate greater morphological disparity; overlapping morphospaces indicate similar body plans. There are distinct morphospaces for Devonian and post-Devonian coelacanths. Latimeria falls close to its closest, Mesozoic relatives for discrete characters (A) and cheek shape (C) but is quite distantly separated from Mesozoic forms for overall body shape (B) and lower jaw shape (D). All individual species data points plotted and identified in Supplementary Fig. 8. Identification of coelacanth silhouettes is provided in Supplementary Fig. 4 (complete body) and Fig. 11 (cheek region and lower jaw).