Resolution of the Early Jurassic Actinopterygian Fish Pachycormus and a Dispersal Hypothesis for Pachycormiformes

ABSTRACT 
 The Early Jurassic (Toarcian) actinopterygian Pachycormus is a basal taxon within Pachycormiformes, a Mesozoic marine neopterygian radiation that evolved an extreme ecomorph divergence between hyperspecialized ‘billfishlike’ macrocarnivores and gigantic suspension feeders, including the largest fish of all time. Current phylogenies place Pachycormus as an early member of the suspension-feeding lineage; however, species disparity renders character states uncertain and potential exists for considerable intraspecific variability. Given its importance for resolution of pachycormiform phylogenetic topology, we comprehensively reassessed Pachycormus fossils housed in collections across Europe and found that the proportional traits traditionally used to discriminate between species are actually consistent with an ontogenetic size morphocline. Our cladistic analyses further show that the monotypic senior synonym, Pachycormus macropterus, is a wildcard that manifests a mosaic of transitional states. This has significant implications for hypothesized Toarcian marine vertebrate provincialism because P. macropterus had a ubiquitous Boreal Tethyan distribution. Moreover, our tree-based palaeobiogeographical optimizations infer that the western Tethyan region was a pachycormiform dissemination center, with global dispersals occurring through transoceanic migration and invasion of epeiric basins.

bollensis (Quenstedt, 1858) forming a proportional morphocline (see historical assessments of Smith Woodward, 1895;Arambourg, 1935;Rayner, 1948;Lehman, 1949;Wenz, 1967;Mainwaring, 1978). Indeed, the diagnostic separation of these taxa has traditionally relied upon comparative skeletal dimensions (Mainwaring, 1978), leading to the dubious identification of many museum specimens (e.g., Hauff and Hauff, 1981). This has potentially significant implications for the established framework of European Toarcian marine faunal provinces (based on ichthyosaurians and plesiosaurians; Godefroit, 1994;Maisch and Ansorge, 2004;O'Keefe, 2004;Großmann, 2007;Smith and Vincent, 2010;Sachs et al., 2016) because P. macropterus and P. curtus were seemingly confined to the Yorkshire, Paris, and South German basins, whereas P. bollensis was restricted to the Posidonia Shale of Baden-W€ urttemburg in Germany. Furthermore, the Toarcian represents an important time frame for pachycormiform taxonomic appearances (Hauff and Hauff, 1981;Arratia and Schultze, 2013) and thus implicates the Boreal and western Tethyan region as a possible origination center for the clade. To test this hypothesis, we evaluated species disparity within Pachycormus as a model for assemblage discretization across Toarcian marine vertebrate provincial boundaries and investigated the concomitant dispersal history of Pachycormiformes using a phylogeny-based palaeobiogeographical optimization approach.

PROPORTIONAL VARIATION WITHIN PACHYCORMUS
The differential diagnoses of Pachycormus species have traditionally relied upon proportional characters. For example, (1) Smith Woodward (1895) defined P. macropterus as being »1 m long with the skull and mandible occupying 0.25£ the maximum body length. Pachycormus curtus, on the other hand, was reportedly only »500 mm long with the skull and mandible measuring >0.25 of the maximum body length. Wenz (1967) alternatively found no discrete size disparity between these species, although the specimens attributed to P. macropterus were thought to be generally larger, with a more elongate head and dentigerous elements. Mainwaring (1978) additionally noted dimensional change in (2) the length of the preorbital region relative to the entire cranium. Smith Woodward (1895) distinguished (3) the opercula as being almost or quite as broad as deep in P. macropterus, versus deeper than broad in P. curtus. Mainwaring (1978) later corrected this to the opercula being twice as broad as deep in both species, but added an amendment that the ventral angle of the opercula is positioned in the posterior half of the bone in P. macropterus, compared with the mid-length or anterior half in P. curtus. Smith Woodward (1895) further described a meristic trend towards more numerous branchiostegal rays (ranging from 40 to 50) and the occurrence of a few finely striated scales in P. macropterus (see Meunier and Brito, 2004), a feature also evident in other pachycormiforms (Schultze, 1996).
To test the discriminating dimensional features of Pachycormus for allometry, we first pooled our specimens to accommodate for taxonomic uncertainty and then screened them for diagenetic distortion and artificial restoration of the skeletal elements. This yielded final sample sizes of (1) n D 31, (2) n D 37, and (3) n D 9 (Tables S1-S3). The data were then log 10 -transformed and plotted in PAST 3.06 (Hammer et al., 2001) to visualize linear (isometric) correlations, which were significant (P < 0.001) for all parameters: (1) Pearson's r D 0.97516, r 2 D 0.95093, RMA (Reduced Major Axis) slope a D 0.99466 for skull/body length (Fig. 1D); (2) r D 0.9567, r 2 D 0.91528, a D 1.04 for preorbital/cranium length (Fig. 1E); and (3) r D 0.94198, r 2 D 0.88733, a D 0.8409 for operculum length/depth (Fig. 1F). This concurred with a principal component (PC) analysis ( Fig. 2A) that returned 72.665% of variance reified by the size-determined PC1 vector (McKinney, 1990), and 26.72% for PC2. More than 99%-94% of the correlated variable loading was also derived from (1)-(3), implying that increasing postcranial skeletal length was proportionate with that of the cranium and preorbital regions. Conversely, a discriminant analysis with Box's M test for homoscedasticity (raw data P < 0.18463) correctly classified 93.75% of the P. macropterus/P. bollensis specimens relative to P. curtus. Significant difference was also found in the multivariate mean (Hotelling's T 2 : P < 0.0001), inferring substantial size distance between these groups. We ascribe this to ontogeny, which was further visualized using a relative warps (RWs) analysis in tpsRelW version 1.46 and related software (Rohlf, 2008). We applied a set of fixed anatomically corresponding landmarks, selected to reflect general body shape and defined as follows: (1) apex of rostrum; (2) center of orbit; (3) anterior base of dorsal fin; (4) apex of dorsal caudal fin lobe; (5) point of angle between the upper and lower lobes; (6) apex of ventral caudal fin lobe; (7) anterior base of anal fin; (8) anterior base of pectoral fin; and (9) posterior-most point of cleithrum (Fig. S2). We used only the most skeletally complete examples of each species morph of P. macropterus (n D 3), P. bollensis (n D 6), and P. curtus (n D 5), together with an overall variance cutoff at >5% to reduce noise (Table S4). The results (Fig. 2B) showed that 63.71% of the cumulative variation occurs in RW1/RW2, with a successive morphospace overlap between P. curtus, P. macropterus, and P. bollensis. This demonstrates progressive increase in skull size, pectoral fin length, and caudal fin aspect ratio along RW1, with RW2 depicting elongation of the body outline, which was variable and might have been affected by postmortem distortion.

IMPLICATIONS FOR PHYLOGENY AND PALAEOBIOGEOGRAPHY
The comprehensive pachycormiform phylogenies of Friedman et al. (2010) and Friedman (2012) scored Pachycormus macropterus and P. curtus as a generic chimaera. We therefore modified the data set of Friedman (2012) by coding these two species, along with P. bollensis as separate terminals, using our first-hand inspection of original fossils. Irrespective of this, the morphotypes of Pachycormus differed by only two out of 121 unordered and unweighted characters (parameters following Friedman, 2012): size of the premaxillary teeth, which are enlarged relative to the maxillary dentition in P. bollensis (60[1]), or homodont in P. macropterus and P. curtus (60[0]); and ossification of vertebral centra (arcocentra sensu Arratia and Schultze, 2013), which is absent in P. curtus (96[0]) and partial in P. bollensis and P. macropterus (96[1])-and additionally consistent with the variable   presence of centrum elements in other pachycormiforms (Mainwaring, 1978;Liston, 2007Liston, , 2010Arratia and Schultze, 2013;Liston et al., 2013aListon et al., , 2013b. Our heuristic searches in PAUP* version 4.0b10 (Swofford, 2002), with TBR (tree-bisection-reconnection) branch swapping and 100 random-addition replicates, returned two most parsimonious trees (length D 226, consistency index [CI] D 0.5442, rescaled consistency index [RC] D 4272) that unsurprisingly incorporated P. macropterus, P. bollensis, and P. curtus as a polytomy. The placement of Pachycormus also differed from that of Friedman (2012) in returning alternate generic resolution either as the sister to Saurostomus C Ohmdenia C Middle Jurassic and Cretaceous suspension-feeding pachycormids (Fig. 3A) or as a basal pachycormiform more derived than Euthynotus (Fig. 3B; character-state distributions are given in the differential diagnosis).

CONCLUSIONS
The classic Toarcian actinopterygian taxon Pachycormus represents a monospecific basal 'wildcard' within pachycormiform phylogenies (see strict consensus in Fig. 4A). We ascribe this to a unique character-state mosaic, which incorporates key traits present in both the macrocarnivorous and the suspension-feeding pachycormiform clades. In a paleobiogeographical context, such topological placement also infers an early distributional center within the Boreal Tethys (Fig. 4B), a region that seemingly hosted the nascent radiation of pachycormiforms during the late Early Jurassic (although age-compatible assemblages from Gondwana are admittedly sparse; Arratia, 2015;Cione, 2015a, 2015b). The historical diversity of Pachycormus species (e.g., Smith Woodward, 1895;Mainwaring, 1978) would thus further imply intrageneric vicariance amongst local European faunas. Concordant differentiation has been suggested between Toarcian marine vertebrates from the northwestern 'British province' of the Yorkshire Basin and those found in the southeastern 'Germanic province' of the South German Basin; this is usually attributed to segregation by the exposed landmass of the London-Brabant-Massif (Godefroit, 1994;Maisch, 1999;Maisch and Ansorge, 2004). In marked contrast, our analysis advocates taxonomic synonymy and the corresponding absence of an effective barrier to the dispersal of pelagic vertebrates across the central European seaway during the Toarcian. Finally, the proportional disparity evident amongst Pachycormus species concurs with the ontogenetic and sexual size differentiation observed in other actinopterygians (Parker, 1992;Mabee and Trendler, 1996;Micklich and Klappert, 2004;Liston, 2008;Liston et al., 2013b) and provides evidence for one of the most numerically prolific examples of intraspecific variability yet reported in the Mesozoic actinopterygian fossil record.