Acanthodes, Agassiz, 1833, Agassiz, 1833

RESULTS View in CoL

All three specimens have collapsed laterally, but the 3D shape of individual pharyngeal elements is retained and is consistent on either side of specimens. Between the three specimens most of the known endoskeleton of Acanthodes is preserved ( Figs 2–4 View Figure 2 View Figure 3 View Figure 4 ) including the jaws, hyoid arch, branchial skeleton, braincase, and shoulder girdle. Here we concentrate on the visceral skeleton: the braincase and shoulder girdle will be the focus of future descriptions. Generally speaking the details of the visceral skeleton of Acanthodes matches the detailed accounts of Miles (1964, 1965, 1968, 1973; Figs 5–10 View Figure 5 View Figure 6 View Figure 7 View Figure 8 View Figure 9 View Figure 10 ). In this account we generally follow Miles’ terminology; a comparison of the terminology we use to that used by previous authors is given in Supporting Information, Table S1. A drawing of our pharyngeal reconstruction is given in Figure 11 View Figure 11 .

All endoskeletal elements in A. confusus comprise a heavily mineralized outer shell surrounding an internal space ( Fig. 5 View Figure 5 ) that was presumably filled with unmineralized cartilage in life. Our scan data is of insufficient high resolution to show the histology of the outer tissue, but Ørvig (1951) interpreted the same tissue as perichondral bone in thin sections of acanthodians including Acanthodes specimens from Lebach on the basis of the tissue having fusiform cell spaces with canaliculi and Sharpey’s fibre attachments, albeit lacking lamination or vascular canals. This perichondral bony tissue is distinct from the varied cartilage tissues that have been described in other stemchondrichthyans ( Burrow et al. 2018, 2020, Maisey et al. 2020, Burrow and Blaauwen 2021). The thickness of this tissue is variable through the skeleton, for example being thicker in the mandibular arches than in the branchial arches, particularly around the mandibular articulation ( Fig. 5C View Figure 5 ). This probably reflects the function of different elements analogously to varying thicknesses in the prismatic tesselate calcified cartilage in extant and extinct chondrichthyans ( Maisey et al. 2020).

Mandibular arch

The mandibular arch ( Fig. 6 View Figure 6 ) comprises a palatoquadrate and Meckel’s cartilage, the morphology of which largely confirm the description of Miles (1968, 1973). The three ossifications of the palatoquadrates are unfinished and have open margins where they would have been joined by uncalcified cartilage. When reconstructed they conform with Miles’ account of their articulations with the neurocranium ( Miles 1968). The anterior edge of the autopalatine is finished, with no evidence for an anteriorly extending palatine commissure ( Jarvik 1977). Like the palatoquadrate ossifications, the two ossifications of the Meckel’s cartilage are unfinished towards the centre of the element. The relationship between the Meckel’s cartilage and the dermal mandibular splint matches the description of Dearden and Giles (2021). The cup-like anterior tip to the Meckel’s cartilages suggests a well-developed but mobile connective-tissue attachment between them, and may be a character uniting a subset of stem-group chondrichthyans ( Dearden and Giles 2021). The articulation between the palatoquadrate and Meckel’s cartilage confirm the description of Miles (1973), although the preglenoid process is notably more rounded than in that reconstruction. There is no pronounced retroarticular flange as in Gogoselachus and Tristychius ( Long et al. 2015, Coates et al. 2019, Frey et al. 2020). When closed the mouth is tilted dorsally at its anterior end, as reconstructed by Brazeau and de Winter (2015) and Davis et al. (2012), to some extent by Miles (1968, 1973) but not by Watson (1937) or Jarvik (1977).

Hyoid arch

The hyoid arch comprises paired hyomandibulae, interhyals, and ceratohyals, which join directly to a median basibranchial ( Figs 2–4 View Figure 2 View Figure 3 View Figure 4 , 7 View Figure 7 , 8 View Figure 8 ). The hyomandibula comprises a posterior and poorly mineralized anterior ossification ( Fig. 8A, B View Figure 8 ; Miles 1968, 1973). The end of the hyomandibula contacting the neurocranium is laterally compressed in cross-section, meaning that overall the hyomandibula grades into a circular cross-section posteriorly, and the posterior part of the medial face of the hyomandibula is split by a marked, longitudinal ridge, which separates it into dorsomesial and ventromesial faces ( Fig. 7F View Figure 7 ; Miles 1968). Miles (1968) drew a comparison between this shape and the hyomandibula of hexanchiform sharks. Amongst Palaeozoic forms the laterally flattened, curved, and posteriorly tapered shape of the hyomandibula is more comparable to that of the actinopterygian Mimipiscis ( Gardiner 1984) , the symmoriiforms Ozarcus ( Pradel et al. 2014) and Ferromirum ( Frey et al. 2020) , and cladoselachians ( Maisey 1989) than to the stem-group chondrichthyan Gladbachus ( Coates et al. 2018) or to the stouter hyomandibulae of crown-group chondrichthyans such as xenacanths or hybodonts ( Hotton 1952, Maisey 1987).

An interhyal is present between the hyomandibula and ceratohyal of Acanthodes ( Figs 4 View Figure 4 , 7 View Figure 7 , 8A, B, G, H View Figure 8 ), confirming the account of Miles (1973). The interhyal is subrectangular and laterally flattened, with gently convex dorsal and ventral surfaces. Of the three specimens described here, interhyals are only preserved in MNHN-F-SAA24 ( Fig. 4 View Figure 4 ) and before that were only known from a single mouldic specimen, MFN MB 23, now lost but from which casts are preserved (e.g. NHMUK PV P 49990, Miles 1973, plate 7). The proportions of the interhyal in MNHN-F-SAA24 relative to the hyomandibula approximately match those of the interhyal in NHM PV P 49990, which comparison suggests probably preserves a postero-lateral view of the element. As Miles (1973) highlighted it is unclear whether this ‘interhyal’ is homologous to the interhyal, stylohyal, or symplectic in osteichthyans ( Patterson 1982, Véran 1988). Whether or not these are homologous, additional elements of the hyoid skeleton are not found in this position in Gladbachus ( Coates et al. 2018) , or in other articulated Palaeozoic chondrichthyan hyoid arches ( Pradel et al. 2014, Frey et al. 2020, Klug et al. 2023).

The ceratohyal is ossified in two parts with no hypohyals present ( Figs 2–4 View Figure 2 View Figure 3 View Figure 4 , 7 View Figure 7 , 8A–E, I, J View Figure 8 ; Miles 1965, 1968, 1973, Gardiner 1984) and its morphology confirms the account of Miles (1968). The posterior end is laterally flattened and lacks the lateral fossa seen in some early chondrichthyans ( Coates et al. 2018). It also lacks the sharp dorsal angle at the posterior end of the ceratohyals of Ferromirum ( Frey et al. 2020) , Phoebodus ( Frey et al. 2019) , and Maghriboselache ( Klug et al. 2023). The anterior end of the ceratohyal is not spatulate anteriorly like some Palaeozoic chondrichthyans such as Phoebodus ( Frey et al. 2019) , instead pinching in and expanding out to form the articulation with the basibranchial (the expanded knob of Miles 1968). The longitudinal groove runs along the mesial surface of the element which may be homologous to a similar groove on the ceratohyal of Gydoselache ( Maisey et al. 2019) and which Miles (1968) suggested was for the insertion of muscles including the anterior interhyoideus musculature.

The component forming the ventral floor of the pharyngeal skeleton is composed of two mineralized parts ( Figs 2–4 View Figure 2 View Figure 3 View Figure 4 , 7 View Figure 7 , 8A–E, K–M View Figure 8 ). This has variously been termed a basibranchial or basihyal; here we use the former due to its extending considerably further posteriorly than most elasmobranch basihyals. The anterior mineralization has a hammerhead-shaped anterior end with ventrally-oriented articulation surfaces for the ceratohyals. In its posterior half this becomes taller with deep ventral attachment surfaces for the coracohyoid and coracobranchial musculature, with an unfinished posterior face (i.e. not covered by perichondral bone). The posterior basibranchial mineralization isflatwithaposteriortail, itsunfinishedanteriorfacesuggeststhat it was joined to the anterior mineralization of the basibranchial by cartilage. There are no obvious articulation surfaces on the posterior part of the basibranchial, but based on their preserved position it seems likely to have articulated with the first and second branchial arch as reconstructed by Miles (1973). This tall, narrow basibranchial is dissimilar from the broad flat basihyals of other known stem-group chondrichthyans ( Brazeau 2012, Coates et al. 2018); a possible exception is in Halimacanthodes , where an element identified as an?autopalatine (or possibly a basibranchial) could be the front of the basibranchial in lateral profile ( Burrow et al. 2012). A tapering posterior end is common in basibranchial copulae in extant and extinct chondrichthyans ( Coates et al. 2018, Dearden et al. 2021), although otherwise these are dissimilar from the basibranchial in Acanthodes in that copulae extend considerably beyond the posterior-most branchial arch (e.g. Dearden et al. 2021). We find no evidence for a chain of basibranchial elements as given in some reconstructions ( Fig. 1A–C View Figure1 ): this aspect of the reconstruction appears to be based on the holotype of A. confusus ( Heidtke 2011: fig. 8) in which preserved hypobranchials give the impression of there having been a chain of basibranchials.

Branchial arches

Five branchial arches are preserved which extend well-posterior to the braincase ( Figs 2–4 View Figure 2 View Figure 3 View Figure 4 , 7 View Figure 7 , 9 View Figure 9 ). It remains unclear whether they articulated with the basioccipital, as reconstructed by Miles (1973). In MNHN-F-SAA21 they overlap with the underside of the braincase anteriorly( Fig.2 View Figure 2 ) but this is not the case in MNHN-F-SAA24 and MNHN-F-SAA20 [ Figs 3 View Figure 3 , 4 View Figure 4 ; see discussion of this character state in Dearden et al. (2019), Supplementary material, Character 50, of Frey et al. (2020)]. The branchial arches comprise hypo-, cerato-, epi-, and pharyngobranchials. We find no evidence for accessory elements as in some osteichthyans and in the symmoriiform Ozarcus ( Pradel et al. 2014) .

Four pairs of ceratobranchials are preserved, a fifth ceratobranchial is absent although rakers on the ventral part of the fifth arch in MNHN-F-SAA20 ( Figs 3 View Figure 3 , 7C View Figure 7 ) indicate that there was a ventral component to the fifth branchial arch. The ceratobranchials all have a similar overall structure, with a ventral groove for the afferent branchial artery and plinths on their pharynx-facing surfaces for the branchial rakers ( Fig. 7 View Figure 7 ). Ceratobranchial I and II are segmented into anterior and posterior sections, and ceratobranchial I has a pronounced articular facet on its anterior end which must have articulated with the basibranchial, although there is no corresponding facet on that element. More posterior ceratobranchials become progressively shorter and more pronouncedly curved until the fourth pair is almost as broad as they are long ( Fig. 7E, F View Figure 7 ). The posteriormost ceratobranchial is not enlarged as in Gladbachus and the stem-group gnathostome Paraplesiobatis ( Brazeau et al. 2017, Coates et al. 2018), yet is more flattened relative to the others as observed in these taxa.

Hypobranchials are present on at least the second and third branchial arches ( Fig. 7 View Figure 7 ). No hypobranchial IV is visible in our datasets ( Fig. 7 View Figure 7 ), although a cast of a now lost specimen (NHMUK PV P 49990; Miles 1973, plate 7) shows a hypobranchial in this position quite clearly that conforms with the anatomy of the more anterior hypobranchials so it may be either lost or unmineralized in the specimens studied here. The hypobranchials are short and curved, with a lateral groove, and are oriented anteriorly. Ceratobranchial I has a well-developed condyle at its anterior tip and is preserved in close association with the basibranchial in MNHN-F-SAA21 ( Fig. 8C, D View Figure 8 ), suggesting a direct connection between the two without an intervening hypobranchial. We find no evidence for a fifth hypobranchial.

Five pairs of epibranchials are present, one in each arch ( Figs 7 View Figure 7 , 9 View Figure 9 ). The epibranchials are gently curved, with a dorsolateral groove for the efferent branchial artery ( Fig. 9A, B View Figure 9 ). This groove is bordered medially by a ridge corresponding to the posterior flange ( Coates et al. 2018). At the distal end of the ceratobranchial this develops into a dorsal process ( Miles 1968). Like the ceratobranchials the surface facing into the branchial chamber is pitted forming a series of plinths to carry the gill rakers ( Fig. 10E View Figure 10 ). The proximal end of the epibranchials has an articular surface for the ceratobranchial ( Fig. 9C View Figure 9 ). The epibranchials have the same overall form, but become progressively shorter posteriorly ( Figs 7 View Figure 7 , 9H View Figure 9 ). The posteriormost (fifth) epibranchials are different in shape, with broad heads ( Figs 7 View Figure 7 , 9E View Figure 9 ). Although the epibranchials in MNHN-F-SAA20 and MNHN-F-SAA21 appear to be segmented, in MNHN-F-SAA24 they are ossified into a single structure ( Figs 4 View Figure 4 , 9 View Figure 9 ).

There are four pairs of pharyngobranchials, best preserved in MNHN-F-SAA24 ( Figs 4 View Figure 4 , 7 View Figure 7 , 9 View Figure 9 ) where all four pairs are preserved in articulation with the epibranchials. The individual anatomy of these pharyngobranchials is consistent with the description of Miles and that observed in casts of mouldic specimens (e.g. NHMUK PV P 59959, Miles 1973: pl. 2), and there is no evidence for separate supra- and infrapharyngobranchials as in Ozarcus and osteichthyans ( Gardiner 1984, Pradel et al. 2014). The morphology of each successive pair of pharyngobranchials is serially consistent, although the more posterior pharyngobranchials are slightly smaller. Unlike living chondrichthyans, but like Gladbachus and Ptomacanthus ( Coates et al. 2018, Dearden et al. 2019) the posteriormost epibranchials and pharyngobranchials are not fused into a single complex element. The only other acanthodiform in which pharyngobranchials have been figured is Halimacanthodes , in which they appear broadly similar in shape ( Burrow et al. 2012: fig. 2b, ph. br.).

The articulation between each pharyngobranchial and the head of each epibranchial confirms the account of Miles (1973), with two cotyli on the anterior end of each pharyngobranchial fitting with two condyli on the head of each ceratobranchial. This articulation is preserved in situ in the right, second branchial arch of MNHN-F-SAA24 ( Fig. 9G–J View Figure 9 ), and the positions of the other pharyngobranchials in this specimen are consistent with the same articulation. Nelson (1968) outlined two alternative arrangements of the pharyngobranchials in Acanthodes ( Fig. 1C View Figure1 ): one where each pharyngobranchial was oriented anteriorly from the more posterior arch to meet the one in front, and another where pharyngobranchials were oriented posteromedially from the top of each arch. Reconstructing the articulation between this pharyngobranchial and epibranchial leads to an angle of about 90 ° between the two, suggesting that the latter of Nelson’s reconstructions is accurate ( Figs 9I, J View Figure 9 , 12 View Figure 12 ). Moreover there is no visible articulatory surface on the posterior end of each pharyngobranchial ( Fig. 9K, N View Figure 9 ), with them instead being narrow and unfinished ( Miles 1973). Based on this we interpret these new data as confirming that Acanthodes had postero-medially oriented pharyngobranchials. The effect of the posterior process and dorsal ridge on each pharyngobranchial ( Fig. 12 View Figure 12 ) is thus to form a concave surface on the dorsal face each serving as anchoring points for the m. interpharyngobranchialis ( Jarvik 1977).

HYOID AND BRANCHIAL RAKERS

Branchial and hyoid arches carry rows of small rakers. Our data is insufficiently high resolution to show any histological detail, but these elements are separate from the endoskeletal arches, with distinct bases and crowns, and are ornamented ( Zidek 1985) so we consider it probable they are dermal rakers ( Miles 1968) rather than endoskeletal projections from the pharyngeal arches ( Ørvig 1973, Jarvik 1977). On the basis of MNHN-F-SAA21 and MNHN-F-SAA20 ( Figs 2 View Figure 2 , 3 View Figure 3 ) there is a single row of antero-medially directed rakers on each branchial arch, although more than one row on each arch has been reported, which may vary through ontogeny ( Reis 1896, Watson 1937, Miles 1968). Some small rakers are present on the ventral part of the hyomandibula in MNHN-F-SAA20 ( Reis 1896, Watson 1937, Miles 1968), and the ceratohyals carry a single row of rakers ( Figs 2 View Figure 2 , 3 View Figure 3 , 7 View Figure 7 ). Rakers are present on the dorsal and ventral parts of all five branchial arches, as can be seen in MNHN-F-SAA20 ( Fig. 7C, D View Figure 7 ). They are positioned on plinths along the length of the branchial elements ( Fig. 10 View Figure 10 ).

The morphology of the rakers is variable on different elements ( Fig. 10 View Figure 10 ). On the ceratohyal and hyomandibula the rakers comprise simple prongs with a broad, flat base ( Fig. 9I View Figure 9 ), and on the hyomandibula these are even simpler and smaller ( Fig. 10C View Figure 10 ). Contrastingly, on the epibranchials and ceratobranchials the blade of each raker is considerably longer and developed into a flattened leaf shape slightly inflected away from the midline, and a base that is quite rounded and concave. These branchial rakers are larger on more anterior arches, and also decrease in size towards the top and bottom of either arch ( Fig. 7C, D View Figure 7 ). More detailed views of raker morphology can be seen in casts, notably NHMUK PV P 49973 and 49990 ( Miles 1973: pls 6, 7).

HYOID/GULAR RAYS

The hyoid/gular rays are preserved in all three specimens, although the lateral collapse of the skeleton in each case has disarticulated them to some extent ( Figs 1–3 View Figure1 View Figure 2 View Figure 3 , 10A, B View Figure 10 ). They are short narrow elements, some of which having a slight sinusoidal inflection ( Dearden and Giles 2021). They are oriented posteromedially from the ventral margin of each mandible [this can be clearly seen in NHMUK PV P 49973 ( Miles 1973: pl. 6)], and appear to be absent from the first fifth of the mandible’s length and then be arranged into a row running just beyond the mandible’s posterior margin. In life they would have underlain the gular region.

FUNCTIONAL MORPHOLOGY

Our 3D reconstruction confirms that the reconstruction of Miles (1968) with three points of articulation with the neurocranium is plausible ( Fig. 12 View Figure 12 ) and that the effect of swinging the jaw laterally from these articulatory points is that the palatoquadrates swing laterally as he predicted. The double articulation of the Meckel’s cartilage and the palatoquadrate means that the mandible could only lower vertically relative to the palatoquadrate ( Miles 1968); as such the relative angles of the left and right mandibles change during jaw opening. This is perhaps the reason for the expanded symphyseal tip on the mandible, to accommodate the connective tissue (ligaments, cartilage) that allows this movement. Another effect of this movement is that the symphysis of the mandible moves anteriorly during jaw opening ( Fig. 12 View Figure 12 ). Our reconstruction suggests the proximal end of the hyomandibula would have been capable of staying close to the jaw articulation even with major abduction of the palatoquadrate. We note that we did not place any constraints on movement beyond the direction of rotation, so this reconstruction should be interpreted as a plausible rather than a maximum gape. Moreover, it is likely that in life the unmineralized cartilaginous midsections of the jaws and hyoid arch would have given them some degree of flexibility, something that is not incorporated into our model.