A primitive actinopterygian braincase from the Tournaisian of Nova Scotia

The vertebrate fossil record of the earliest Carboniferous is notoriously poorly sampled, obscuring a critical interval in vertebrate evolution and diversity. Recent studies of diversity across the Devonian–Carboniferous boundary have proposed a vertebrate mass extinction at the end-Devonian, and recent phylogenies suggest that the origin of the actinopterygian crown may have occurred in the earliest Carboniferous, as part of a broader recovery fauna. However, the data necessary to test this are limited. Here, we describe a partial actinopterygian skull, including diagnostic elements of the posterior braincase, from the Tournaisian Horton Bluff Formation of Blue Beach, Nova Scotia. The braincase surprisingly shows a confluence of characters common in Devonian taxa but absent in Mississippian forms, such as an open spiracular groove; lateral dorsal aortae that pass through open broadly separated, parallel grooves in the ventral otoccipital region, posterior to the articulation of the first infrapharyngobranchial and an intertemporal–supratemporal complex. Phylogenetic analysis places it deep within the actinopterygian stem, among Devonian moythomasiids and mimiids, suggesting more phylogenetically inclusive survivorship of stem group actinopterygians across the end-Devonian mass extinction. With a high lineage survivorship in tetrapods and lungfish across the Devonian–Carboniferous boundary and high vertebrate diversity at Tournaisian localities, this hints at a more gradual turnover between Devonian and Carboniferous vertebrate faunas.

The vertebrate fossil record of the earliest Carboniferous is notoriously poorly sampled, obscuring a critical interval in vertebrate evolution and diversity. Recent studies of diversity across the Devonian-Carboniferous boundary have proposed a vertebrate mass extinction at the end-Devonian, and recent phylogenies suggest that the origin of the actinopterygian crown may have occurred in the earliest Carboniferous, as part of a broader recovery fauna. However, the data necessary to test this are limited. Here, we describe a partial actinopterygian skull, including diagnostic elements of the posterior braincase, from the Tournaisian Horton Bluff Formation of Blue Beach, Nova Scotia. The braincase surprisingly shows a confluence of characters common in Devonian taxa but absent in Mississippian forms, such as an open spiracular groove; lateral dorsal aortae that pass through open broadly separated, parallel grooves in the ventral otoccipital region, posterior to the articulation of the first infrapharyngobranchial and an intertemporal-supratemporal complex. Phylogenetic analysis places it deep within the actinopterygian stem, among Devonian moythomasiids and mimiids, suggesting more phylogenetically inclusive survivorship of stem group actinopterygians across the end-Devonian mass extinction. With a high lineage survivorship in tetrapods and lungfish across the Devonian-Carboniferous boundary and high vertebrate diversity at Tournaisian localities, this hints at a more gradual turnover between Devonian and Carboniferous vertebrate faunas.

Introduction
The Palaeozoic is marked by several biodiversity crises and major climatic events, including mass extinctions and longerterm transitions between climate states [1].

Phylogenetic analysis
We added the new taxon and Lambeia pectinata to the character matrix of Giles et al. [18]. We removed Tegeolepis clarki, as it was acting as a wildcard taxon following Anderson [19], and Dicksonosteus from the outgroup. We also added a character for the trajectory of the lateral dorsal aortae, which we coded from the literature (electronic supplementary material, S1). Our new specimen was coded directly from the specimen and from µCT, whereas L. pectinata was coded from the recent description of Mickle [20]. The resulting matrix contains 93 taxa and 266 characters (electronic supplementary material, S2).
We assessed phylogenetic relationships of actinopterygians using an equally weighted parsimony analysis in PAUP * 4.0a [21]. Ancestral states were calculated using DELTRAN and dashes were treated as gaps. Most parsimonious trees were found using a heuristic search, with 500 random addition sequences, five trees held at each step, nchuck = 10 000, chuckscore = 1 and the tree bisection and reconnection strategy enabled. Trees were rooted in a non-osteichthyan outgroup of Entelognathus, Acanthodes, Cladodoides and Ozarcus.

Etymology
Genus name for the Avon River, where Blue Beach is located. Species epithet for Chris Mansky, in honour of his many years of collecting, preparing, exhibiting and publicly presenting the fossils of Blue Beach.

Material
NSM016GF025.001 (Nova Scotia Museum), comprising the posterior portion of the skull, collected in the intertidal debris in the cove below the lighthouse, one of the most fossiliferous areas, by J.S.A. in 2013. While this is the most likely source of the specimen, given the large daily tides in the basin (approximately 40 feet every 12 hours), it is possible that the specimen was transported into the area. If this were the case, then the fossil would have originated in the slightly older Blue Beach Member of the Horton Group, but this is still Tournasian, so there is no substantial impact on our results.

Locality
Hurd Creek Member of the Horton Bluff Formation, near Hantsport, Nova Scotia. NSM016GF025.001 was recovered from undifferentiated, eroded material on the western shore of the Avon River.

Diagnosis
Actinopterygian uniquely possessing: bifurcation of dorsal aorta enclosed in canal with widely divergent, long subparallel grooves for the lateral dorsal aortae medial and posterior to the articulation for the first infrapharyngobranchial; relatively anterior spiracular embankment of intertemporal; massively ossified, horizontal quadrate; groove for the facial branch of the hyomandibular nerve and foramen for the internal mandibular branch of the facial nerve relatively anteriorly situated. Shares with Moythomasia: relatively long intertemporal. Differs from Moythomasia: no serrations or barbs on vermiform ornamentation.
Elements are disarticulated such that ventral structures, including the base of the parasphenoid and the posterior otic-occipital, are displaced right-laterally, dorsally and anteriorly relative to more dorsally situated structures.
Overall, the skull roof bones are rectangular and well ornamented with elevated ridges of glossy tissue that run subparallel to the long anteroposterior axis and form vermiform ornamentation typical of early actinopterygians (figures 1 and 2).
We interpret separate supratemporal and intertemporal ossifications, rather than a single dermopterotic ossification, based on a partial suture on the right side of the skull roof; it forms an acute 'V' with an anterior point (figures 1 and 2). The elements are damaged just posterior to this point, so the full extent of the suture is unknown. The supratemporal has a long anteroposterior axis, with its apex posterior to the suture between the frontal and the parietal. The posterior margin of this element shows an occipital flange devoid of ornamentation. Medially, the occipital flange is narrow along its anteroposterior axis, but flares out posteriorly towards its right lateral-most margin (figures 1 and 2). This flare is less pronounced than in taxa such as Mimipiscis toombsi and Moythomasia durgaringa. The relative length of the intertemporal is difficult to assess, as it is damaged and overlain by the displaced dermosphenotic anteriorly. Whereas the preserved extent is shorter than the supratemporal, it is long relative to the intertemporal in other actinopterygian genera (e.g. Raynerius and Woodichthys) and similar in relative length to the intertemporal of Moythomasia [24][25][26]. There is a shallow embayment on the lateral margin of the intertemporal that probably allowed the passage of the spiracular. The intertemporal is slightly eroded on its dorso-lateral extent, which has an exposed part of the infraorbital canal (figure 2). The intertemporal is similar in length to the supratemporal, but it is hard to determine the total extent as it is overlain by the displaced and rotated dermosphenotic. The dermosphenotic is damaged, leaving its posterior extent unknown; the anterior limb of this element extends below the matrix.
The right parietal is partially preserved. The displacement of the supratemporal and intertemporal dorsally and medially obscures the suture between those elements and the parietal (figures 1 and 2). It is overlain posteriorly by an indeterminate extrascapular, which marks the posterior end of the specimen. The parietal appears to be relatively long in comparison with the preserved extent of the frontal. The suture with the frontal is incompletely known, but appears to run anterolaterally. The frontal itself is poorly preserved, being obliterated laterally and anteriorly, and is overlain laterally by the intertemporal (figures 1 and 2). Much of the medial edge of the frontal is preserved-the right side of the midline suture. This appears straight and without significant interdigitation. The medial margin of the frontal is laterally indented anteriorly. This may be the right-posterior corner of the pineal foramen.
The right suspensorium is displaced such that the long axis of the fused dermohyal and hyomandibula lies parallel to the skull roof (figures 2 and 3). The dermohyal displays prominent dermal texturing. The hyomandibula is displaced posteriorly from its point of articulation with the skull roof and post-orbital process. Ventrally, the posterior-most portion of the quadrate is preserved-it is large and robust and is tightly sutured to a densely denticulate quadrate ramus of the pterygoid. The groove for the facial branch of the hyomandibular nerve runs obliquely along the lateral surface of the quadrate and is pierced at the midpoint by a foramen serving the internal mandibular branch of the facial nerve (figure 3). The arrangement of this groove and foramen is similar to the condition described in Mimipiscis by Gardiner [27], but is more horizontal in orientation.
The post-orbital process descends from the dorsal braincase anterior to the suspensorium. The ventral extent of this element cannot be discerned, as it reaches the plane of shear. The posterior-most portion of the interorbital septum is displaced and rotated such that its ventral extent is lateral to its dorsal extent. The surface of the interorbital septum preserves a foramen-this probably served the oculomotor nerve ( figure 3).
The braincase is missing anterior to the apex of the base of the parasphenoid-the anterior-most preserved surface of the parasphenoid preserves an indentation which may represent part of the buccohypophyseal canal, but the complete canal is not preserved. The posterior plate of the paraphenoid is broad, extensive and anteriorly restricted (figures 4 and 5). The posterior palatal surface has broken, so the extent of denticulation cannot be fully determined, but circular denticle bases are identifiable on the anterior-most portion of the ventral surface. The erosion of the parasphenoid has exposed the dorsal margin of the parasphenoid in ventral view, permitting a determination of its full posterior extent. There is an indentation in this margin for the attachment of the subcephalic muscles. No parotic toothplates are preserved in this specimen, and the damage renders it impossible to evaluate if there is a point of articulation for these elements (figures 4 and 5). The spiracular groove is preserved on the lateral surface of the base of the parasphenoid (figure 6); it is shallow and positioned posterodorsally. The spiracular groove appears to be bounded dorsally by a strut on the anterior ascending process of the parasphenoid, although it is weathered, so the full morphology of the strut remains unknown. The ascending process is reduced in extent in comparison to the ascending process of Howqualepis [28,29] of the ventral otic fissure (figure 6). Vascular canals through the basisphenoid-parasphenoid complex can be seen using µCT (figure 7). The internal carotids enter the parasphenoid posteriorly, within the ventral cranial fissure, and begin to converge anteriorly. A second pair of canals are visible in µCT and run medially and anteriorly towards the internal carotids; these probably served the pseudobranchials ( figure 7). There is no foramen for these canals in the preserved extent of the parasphenoid. Part of the posterior otic-occipital is preserved. The specimen is broken just anterior to the posterior end of the braincase, producing an oblique transverse section of the canals for the dorsal aorta and notochord, and the foramen magnum. The foramen for the occipital nerve is preserved on the right lateral otic-occipital, posterior to the confluence of the grooves of the lateral dorsa aortae into the aortic canal (figures 4 and 5). This foramen is overlain by the craniospinal process. The craniospinal process forms the posteroventral margin of the otoccipital fissure, which runs posterodorsally. The otic-occipital fissure widens for the passage of the vagus nerve. There is another foramen dorsal to this fissure, probably serving the glossopharyngeal nerve. The otic-occipital fissure appears to be continuous with the vestibular fontanelle, but the shape and extent of the vestibular fontanelle is obscured by the overall crushed nature of the otic region (figures 4 and 5).
The   polytomy including moythomasiids and Raynerius. This clade is, in turn, the sister group to mimiids. This represents a further indication of widespread survivorship of Devonian actinopterygian lineages into the Mississippian [26]. The position of A. manskyi is supported by the dorsal aortae open well posterior to the articulation for the first infrapharyngobranchial, the large open spiracular groove, a triangular anteriorly restricted parasphenoid that does not cross the ventral cranial fissure and the welldeveloped persistent ventral cranial fissure. This phylogenetic placement makes A. manskyi the earliest diverging actinopterygian lineage currently known to have survived the Hangenberg extinction. What is less clear is the phylogenetic relationship between A. manskyi and the broader diversity of Carboniferous actinopterygians from the Tournaisian and later. Several other actinopterygians are known from isolated remains from Blue Beach, including a second, larger, actinopterygian neurocranium (BWC 536), a partial skeleton of a deep-bodied form (BWC 1029) possibly attributable to the Platysomoidea and a mandible from a very large actinopterygian (CM 9827), as well as scales and isolated remains possibly attributable to several Palaeozoic actinopterygian 'wastebasket' taxa, including Paleoniscum, Rhadinichthys and Acrolepis [15]. A more extensive actinopterygian fauna is present from another Horton Group locality in New Brunswick [20], with an overall similar list of taxa: one or several species attributed to Rhadinichthys, a species probably incorrectly attributed to Palaeoniscum, a species attributed to Elonichthys, a possible Canobius and the possible elonichthyid L. pectinata [20]. Although we have not explicitly treated these taxa in our analysis here, rhadinichthyids (Wendyichthys) and elonichthyids (Mansfieldiscus) are both deeply nested within a more derived radiation of non-neopterygian actinopterygians, suggesting that much of the Palaeozoic actinopterygian radiation may have occurred by the earliest Carboniferous.

Actinopterygian survivorship across the Hangenberg extinction
The Devonian-Carboniferous transition is marked by at least two marine mass extinction events, the Kellwasser event at the Frasnian-Famennian boundary and the Hangenberg event at the Famennian-Tournaisian boundary. Of these two, the Kellwasser has a greater impact on marine invertebrate diversity [2,13]. However, the vertebrate communities of the Upper Devonian differ substantially from those of the Lower Carboniferous, and it has been argued that the Hangenberg represents a major extinction of vertebrate lineages, even if the impact on invertebrates is more moderate [13]. The presence of a classic Devonian actinopterygian lineage in the Horton Bluff Formation hints that post-Hangenberg recovery fauna may retain a more phylogenetically inclusive sample of Famennian actinopterygian diversity than previously appreciated. The only Devonian actinopterygian lineages which appear to go extinct before the Carboniferous are cheirolepids, which do not appear after the Frasnian [30], howqualepids, restricted to the Givetian [31], and basal-most forms such as Meemannia, Ligulalepis and Dialipina, which are restricted to the earliest Devonian [32][33][34]. The relative rarity of articulated actinopterygian fossils makes it difficult to determine whether these lineages reached the end of the Devonian, or whether they went extinct prior to the Hangenberg. One possibility is that the replacement of 'archaic' Devonian-type actinopterygian lineages by 'advanced' Carboniferous-type clades in the aftermath of the Hangenberg event was a function of differential diversification in the aftermath of the extinction rather than a function of clade selection within the mass extinction itself. Future approaches to the impact of the end-Devonian mass extinction on actinopterygian diversity should distinguish between the extent and selectivity of the mass extinction itself versus diversification patterns within the Lower Carboniferous recovery fauna.
This secondary question, whether actinopterygians underwent an adaptive radiation in the earliest Carboniferous, is more difficult to assess, as the phylogenetic relationships of the vast majority of Carboniferous actinopterygians remain unresolved. Although it is certainly the case that actinopterygian communities achieved high diversity within the Mississippian (e.g. [24,35,36]), it is unclear whether this diversity appeared suddenly or whether it accrued gradually over the Late Devonian and Early Carboniferous. Further phylogenetic work incorporating broader character samples (e.g. [37]) and more fully sampling Carboniferous and Permian actinopterygian diversity is necessary [18].
More inclusive survivorship of actinopterygians across the Hangenberg opens up the possibility that actinopterygian diversity may more closely align with the emerging pattern of mass survivorship among tetrapods [8,9] and dipnoans [10,12]. An extinction of archaic tetrapods has been previously proposed based on the apparent disappearance of ichthyostegalid-grade tetrapods [6], but recent work has shown that Devonian ichthyostegalid-grade tetrapods persist into the Tournaisian [8,9] and later [38], and that Carboniferous-type whatcheeriid-grade tetrapods are present in Devonian assemblages [39,40]. Similarly, recent work on dipnoans also shows a high diversity of this group in the Tournaisian [10], and several recent studies [11,12] have shown that many major post-Devonian lungfish lineages may have diverged as early as the Frasnian, well before the Hangenberg extinction. This suggests that