Fossils indicate marine dispersal in osteoglossid fishes, a classic example of continental vicariance

Abstract

The separation of closely related terrestrial or freshwater species by vast marine barriers represents a biogeographical riddle. Such cases can provide evidence for vicariance, a process whereby ancient geological events like continental rifting divided ancestral geographical ranges. With an evolutionary history extending tens of millions of years, freshwater ecology, and distribution encompassing widely separated southern landmasses, osteoglossid bonytongue fishes are a textbook case of vicariance attributed to Mesozoic fragmentation of the Gondwanan supercontinent. Largely overlooked fossils complicate the clean narrative invoked for extant species by recording occurrences on additional continents and in marine settings. Here, we present a new total-evidence phylogenetic hypothesis for bonytongue fishes combined with quantitative models of range evolution and show that the last common ancestor of extant osteoglossids was likely marine, and that the group colonized freshwater settings at least four times when both extant and extinct lineages are considered. The correspondence between extant osteoglossid relationships and patterns of continental fragmentation therefore represents a striking example of biogeographical pseudocongruence. Contrary to arguments against vicariance hypotheses that rely only on temporal or phylogenetic evidence, these results provide direct palaeontological support for enhanced dispersal ability early in the history of a group with widely separated distributions in the modern day.

Keywords: FBD; biogeography; bonytongue fishes; dispersal; total-evidence; vicariance.

Figure 1.

Geographical distribution of extinct and extant osteoglossid bonytongues. Fossil occurrences are divided by preservation state (circles: fragmentary and disarticulated fossils; squares: articulated fossils) and palaeoenvironment (orange fill: freshwater deposits; light blue fill: marine deposits). Fossil occurrences with thicker borders indicate where the extinct osteoglossids included in the phylogenetic and biogeographical analyses have been found. The red area in the late Cenozoic map displays the current geographical distribution of extant osteoglossids. Palaeogeographic maps at 0, 50 and 85 Ma were generated in the R package mapast under the MULLER2016 model [30]. Fossil osteoglossid occurrences from [26,31,32]. Geographical distribution of extant osteoglossids from [33].

Figure 2.

Time-calibrated phylogeny of Osteoglossomorpha. The phylogeny plotted here is the ‘AllCompat’ summary consensus tree from the Bayesian total-evidence analysis. Tips are coloured according to geographical distribution. Taxa found in marine settings are highlighted in bold. Coloured triangles at internal nodes represent the most likely ancestral geographical area under the DEC+j model when it is at least 3.2 times more likely than the second most likely geographical area, indicating substantial strength of evidence under a Bayes factor framework. Coloured triangles at internal nodes for which all descendant tips and the immediately ancestral node inhabit the same area were masked to avoid figure cluttering. Circles at internal nodes indicate node support as Bayesian posterior probabilities of clades when equal to or larger than 0.50. White bars represent 95% highest posterior densities (HPDs) of node ages. Boxes encompass terminals belonging to total group (crown group + stem group) of named clades.

Figure 3.

Historical biogeography of bonytongue fishes integrated over phylogenetic uncertainty. Marginal probabilities and numbers of dispersal events shown in this figure have been integrated over a random sample of 200 phylogenies from the Bayesian posterior distribution. (a) Marginal probabilities of ancestral biogeographical area under the DEC + j model for crown Osteoglossidae when fossils are included in (left) or excluded from the analysis (right). (b) Average number of dispersal events into (immigration) and from (emigration) biogeographical areas, calculated under biogeographical stochastic mapping (BSM) and integrated over phylogenetic uncertainty. (c) Directionality of dispersal between biogeographical areas, calculated under BSM. Arrow thickness is proportional to the average number of dispersal events, indicated by the number next to the arrow. Arrows from one area to another are not shown when the average number of dispersal events is lower than 0.3.

Figure 4.

Ancestral habitat estimation for bonytongue fishes. (a) Marginal ancestral states under an all-rates-different (ARD) model on the Bayesian consensus tree of Osteoglossomorpha. Orange fill indicates freshwater environment, while light blue fill indicates marine environment. Tip identities are the same as in figure 2, with light grey box encompassing terminals belonging to total-group Osteoglossidae. (b) Maximum-likelihood (ML) estimates of transition rates between freshwater and marine environments, expressed per million years per lineage (Myr⁻¹ lin⁻¹). (c) Distribution of the inferred number of environmental transitions under stochastic character mapping (SCM) on the Bayesian consensus tree of Osteoglossomorpha. Freshwater-to-marine transitions are in dark blue and marine-to-freshwater transitions are in rosy brown.