Kansasiella eatoni (Poplin, 1974)

Kansasiella eatoni

The ethmoid region is incompletely preserved in the present specimens; only the ventralmost part could be recovered. There is a small midline longitudinal ridge (l.r; Fig. 7C, D) which terminates anteriorly, presumably allowing for an unpaired anterior ventral myodome (a. v.myo ; Fig. 7C, D). The articulation for the palatoquadrate (eth.p.a; Figs 7C, 9C) is less pronounced than that of Phoebeannaia , though positioned similarly with respect to the midline longitudinal ridge and postnasal wall. The probable lack of endochondral bone in the central ethmoid region is consistent with observations in other early actinopterygians (e.g. Moythomasia , Mimipiscis Choo, 2011 , Lawrenciella ) ( Gardiner, 1984; Hamel & Poplin, 2008; Pradel et al., 2016).

PHYLOGENETIC RESULTS

Specimens NHMUK PV P 10419a & b, FMNH UF464 and FMNH UF462 were coded into a morphological matrix and analysed using parsimony and Bayesian inference. The unconstrained parsimony analysis retrieved 60 equally parsimonious trees of 163 steps. The strict consensus ( Fig. 12A) recovered Phoebeannaia in a polytomy with Coccocephalus wildi , Luederia kempi , Lawrenciella , a branch leading to Kansasiella eatoni and Pteronisculus magnus , and a clade consisting of the neopterygian crown plus Boreosomus piveteaui Stensiö, 1921 and Australosomus kochi Stensiö, 1932 . This fairly unresolved node has a Bremer Index of two, and is supported by seven character state changes, four of which are unambiguous and none of which have a Consistency Index (CI) of one. The strict consensus has a tree length of 173, CI of 0.3526, Homoplasy Index (HI) of 0.6474, Retention Index (RI) of 0.7248 and Rescaled Consistency Index (RC) of 0.2556. The topology of the strict consensus tree contradicts known morphological and molecular evidence regarding the branching order of crown Actinopterygii, thus alternative phylogenetic placements under parsimony were investigated using a constrained backbone search.

The backbone-constrained parsimony analysis retrieved 12 equally most parsimonious trees of 165 steps, and the strict consensus ( Fig. 12B) recovered Phoebeannaia in the same position, supported by the same characters, with a Bremer Index of two. The strict consensus has a tree length of 172, CI of 0.3547, HI of 0.6453, RI of 0.7273 and RC of 0.2579. While the recovered topology is no longer contradictory, the insertion for Phoebeannaia is still relatively unresolved and supported by low CI characters. The backbone-constrained analysis was therefore re-run after character reweighting by RC.

The reweighted and backbone constrained parsimony analysis found three equally most parsimonious trees, and the strict consensus ( Fig. 13A) is nearly fully resolved; Phoebeannaia is sister to Kansasiella and Pteronisculus magnus (Bremer Index = 2, Bootstrap = 66.52), with the clade comprising the Neopterygian crown plus Coccocephalus wildi , Luederia kempi , Lawrenciella , Boreosomus piveteaui and Australosomus kochi as their sister group and Pteronisculus stensiöi as the nearest outgroup. The strict consensus has a higher CI (0.5347), RI (0.8513) and RC (0.4552), and a lower HI (0.4653) than the previous analyses, as expected, which could be interpreted as a partial correction for high levels of homoplasy in the dataset. The clade of Phoebeannaia , Kansasiella and Pteronisculus magnus is supported by one character, the presence of a ‘buttressed’ basisphenoid pillar (CI = 1). This feature manifests via the lateral expansion of the basisphenoid pillar with excavation of paired recesses through to the posterior myodome, as seen in Figure 10. It might have allowed for larger ocular muscles while simultaneously bracing the basipterygoid processes to withstand increased stress from palatoquadrate motion. Kansasiella and Pteronisculus magnus are further held together by three unambiguous character state changes, and five characters (four unambiguous) bind Phoebeannaia , Kansasiella and Pteronisculus magnus in place. The short branch lengths calculated in the most parsimonious trees (one example shown in Fig. 13B) emphasize the similarity between these taxa.

The unconstrained Bayesian analysis ( Fig. 14A) retrieved Phoebeannaia as sister to Kansasiella and Pteronisculus magnus (Posterior Probability 0.59). That clade was placed in a large polytomy with Coccocephalus wildi , Luederia kempi , Lawrenciella and the group of Boreosomus piveteaui , Australosomus kochi and crown Neopterygii, with Pteronisculus stensioi as the nearest outgroup. The placement of polypterids and chondrosteans contradicting modern consensus justified re-running the analysis with an imposed backbone topology ( Fig. 14B), resulting in the addition of the Chondrostei plus Birgeria groenlandica Stensiö 1932 and Saurichthys sp. ( Argyriou et al., 2018) to the large polytomy.