INTRODUCTION

Interpretations of the anatomy of Brochoadmones milesi have evolved with discovery of increasingly well-preserved fossils. The species was described by Bernacsek and Dineley (1977) based on two poorly preserved specimens collected from the Devonian fish layer at the MOTH locality, Mackenzie Mountains, N.W.T., Canada. The holotype (NMC 22710) includes a few head scales, body scales, tooth whorls, and fragments of dorsal, prepelvic, pelvic, and anal fin spines, but much of the fish could not be reconstructed accurately (Bernacsek and Dineley, 1977:text-fig. 14; Fig. 1A). The pectoral fin spines were missing from the first two known specimens, so Bernacsek and Dineley reconstructed an elongate pectoral spine with the expectation that it would be similar to that of other acanthodians. Few scales were preserved on the holotype, and therefore, Bernacsek and Dineley (1977) could not provide a detailed account of the squa-mation.

Bernacsek and Dineley (1977) erected the new family Brocho-admonidae for their new species, although they were unsure whether Brochoadmones milesi represented an ischnacanthiform or a climatiiform. Denison (1979) found no reason to assign Brochoadmones milesi to a new family, and instead, placed Brochoadmones in the family Climatiidae, based on the presence of multiple prepelvic (then called ‘intermediate’) spines, multiple tooth whorls, and ‘cranial tesserae.’ The ‘cranial tesserae’ mentioned by Bernacsek and Dineley (1977) likely were fragments of displaced tooth whorls, because the labial side of the tooth whorls of B. milesi have a complex, tuberculated surface (see Gagnier and Wilson, 1996a:figs. 7g–j). Denison thought that the fin spines, which are long, narrow, and have smooth ribs, represented a derived feature relative to other climatiiform taxa.

Gagnier and Wilson (1996a) added to the description of Brochoadmones milesi based on 14 new specimens collected between 1983 and 1990. They recognized unique features of B. milesi and resurrected the original family assignment provided by Bernacsek and Dineley (1977); they also erected a suborder Brochoadmonoidei to emphasize the distinctiveness of B. milesi, and suggested a relationship between B. milesi and another taxon from MOTH, Kathemacanthus rosulentus Gagnier and Wilson, 1996a. Gagnier and Wilson (1996a) were able to define the shape of the head and trunk, the cranial squamation, dentition, sensory lines of the head and trunk, branchial openings, and details of the external ornament on the prepelvic and dorsal fin spines. In addition, Gagnier and Wilson (1996a) described the unique structure of the pectoral fin spine of B. milesi, and showed that dentigerous bones are absent. Unfortunately, the anal and caudal fins were not preserved on any of the specimens available to Gagnier and Wilson (1996a), who suggested that B. milesi had an elongate and laterally compressed body (Gagnier and Wilson, 1996a:fig. 2; Fig. 1B).

Until now, Brochoadmones milesi has received little attention in phylogenetic analyses of acanthodians, probably because it was imperfectly known. Brochoadmones was excluded from cla-distic analyses performed by Long (1986), Maisey (1986), and Janvier (1996). In addition, Gagnier and Wilson (1996a, b) did not test their proposed classification of B. milesi with a cladistic analysis.

The Climatiiformes as currently defined may be paraphyletic (Gagnier and Wilson, 1996b; Janvier, 1996; Hanke, 2001a; Hanke and Wilson, 2004). Hanke and Wilson (2000, 2004) and Hanke (2001a) presented results from the first cladistic analysis to include Brochoadmones, which was found to be derived relative to Lupopsyrus pygmaeus and other typical ‘climatiiforms,’ and, in a strict consensus analysis, contributed to a polytomy including diplacanthids, ischnacanthiforms, and acanthodiforms. In contrast to the classification proposed by Denison (1979), Brochoadmones milesi possesses no synapomorphies to indicate relationship with members of the Climatiidae (e.g., Brachyacanthus and Climatius spp.); in this respect, the recent cladistic analyses support the family-level distinction proposed by Gagnier and Wilson (1996a). Furthermore, these recent cladistic analyses do not support the monophyly of the Climatiiformes as a whole.

In 1996, an extraordinary fossil slab was collected from the MOTH locality (Fig. 2A). When collected, only a small area of bone was visible, but subsequent preparation revealed well-preserved fossils of eight individual vertebrates representing seven species. Now known informally as the ‘Wonder Block,’ the slab includes, as its most prominent features, two nearly complete specimens of Brochoadmones milesi, one with its right side exposed and the other with its left side exposed (Fig. 2). These two specimens, supplemented by corroborating evidence from other material, provide the basis for the present paper, leading to the third, and, we hope, final reconstruction of the anatomy of this species (Fig. 1C). The new specimens provide data on the pectoral, dorsal, anal, and caudal fins and several other unique features of Brochoadmones, and have implications for its phylo-genetic position as well as its possible mode of life.

METHODS

The new specimens of Brochoadmones milesi occur on the ‘Wonder Block’ (Fig. 2A) along with remains of one osteostra-can (as yet undescribed), two fork-tailed thelodonts (Drepanolepis maerssae Wilson and Caldwell, 1998, and Furcacauda heintzae Wilson and Caldwell, 1998), two acanthodians (Lupopsyrus pygmaeus Bernacsek and Dineley, 1977, and Ischnacanthus sp., similar to material identified as Ischnacanthus cf. I. gracilis by Bernacsek and Dineley, 1977), and the holotype of a problematic, putatively chondrichthyan taxon, Obtusacanthus corroconis Hanke and Wilson, 2004. Only a small area of the first dorsal spine of one of the specimens (UALVP 41495) of B. milesi was visible in the field and consequently shows some weathering (Fig. 2A, C); the remainder of the slab was covered by matrix. A thin layer of bone visible on the edges of the slab suggested the presence of additional fossils, as subsequent preparation confirmed. The slab was prepared by repeatedly immersing it in dilute acetic acid followed by rinsing in fresh water; silt-sized siliciclastic residues remaining after each acetic-acid treatment were removed while wet using soft paint brushes. After preparation, the slab was dried slowly and stabilized using a 5% solution of Glyptal™ cement. For most of the photographic illustrations (except Fig. 2), ammonium-chloride sublimate was used to whiten specimens.

For histological study, small groups of scales were removed from other, imperfect specimens of B. milesi, embedded in Lu-minate 83 HA-4 epoxy, polished to expose histological structure using 600- and 1000-grit wet-dry sandpaper, and given a final polish using moistened alumina powder on a glass plate. High-magnification images were taken using a Nikon coolpix 990 digital camera attached to a Nikon SMZ 1500 dissecting microscope. Line drawings were made with the same dissecting microscope and its camera lucida attachment.

SYSTEMATIC PALEONTOLOGY

Class ACANTHODII Owen, 1846

Order incertae sedis

Suborder BROCHOADMONOIDEI

Gagnier and Wilson, 1996a

DESCRIPTION

Brochoadmones milesi is a deep-bodied acanthodian with a prominent ‘nuchal’ hump in its dorsal profile immediately posterior to its head (Figs. 2–4). The ratio of body length to depth is 0.28, with the deepest part of the body at the level of the origin of the anterior dorsal-fin spine. The caudal peduncle is short and deep. The axial skeleton is unossified.

The shape of the head is unusual for an acanthodian, tapering to an acutely pointed snout anteriorly, with large orbits (Figs. 2–4). In all well-preserved specimens, the posterior extent of the braincase is indicated by two masses of sandy material interpreted as otic statoconia (Sahney and Wilson, 2001; Figs. 3–5). The structure and shape of the head are indicated only by the arrangement of scales and teeth, because the braincase and visceral arches are unossified.

The orbits are located above the anterior half of the mouth gape, and are ringed by a series of rectangular circumorbital scales (Fig. 4B). These circumorbital scales and other specialized sensory-line scales on the head cradle the cranial branches of the sensory canals in a groove between raised margins of the crown (Figs. 4, 5). There are no sclerotic plates within the orbits of any specimens.

No enlarged cranial tesserae or compound cranial plates are seen in Brochoadmones milesi. Scales of the head have a simple, flat crown, and a flat to slightly concave basal surface (Figs. 5B, 7A). The polygonal crowns of head scales fit closely together to form a ‘cobble-stone’ covering. There is no indication of multiple odontodes on any head scales as exist for some climatiid head scales (Denison, 1979:fig. 16). We were unable to prepare informative thin sections of head scales so it has not been possible to determine the extent of basal tissue or the microstructure of the crown tissue.

The terminal mouth is large. At least 10-11 multicusped, adsymphyseal tooth whorls per side are positioned in a continuous battery along the margins of both the upper and lower jaws (Gagnier and Wilson, 1996a; Figs. 4, 5A). Individual whorls were described in detail by Gagnier and Wilson (1996a). Each whorl is oriented labio-lingually, with the larger cusps of each tooth whorl on its lingual side, suggesting that the whorls rotated la-bially during the growth of the fish as the fish required larger teeth or as teeth were worn and replaced. The tooth whorls are smallest near the jaw angle and increase in size anteriorly, a pattern reminiscent of the growth of teeth on ischnacanthid dentigerous jaws and constrained by basic jaw mechanics.

Meckel's and palatoquadrate cartilages were unossified, but the posteriorly descending course of the mandibular sensory canal suggests that Meckel's cartilage was deeper posteriorly and more slender and slightly hooked or curved upwardly near the symphysis (Gagnier and Wilson, 1996a; Figs. 4, 5A). A few isolated, tiny, whorl-like denticles are visible in the pharynx of UALVP 41495 (also see Gagnier and Wilson, 1996a:figs. 7g–h), but given the rarity of these denticles, it is possible that they represent small, posteriorly displaced tooth whorls rather than pharyngeal teeth.

The deep branchial chamber is approximately as long as the braincase (Figs. 2, 3, 4A). No branchial ossifications are evident and we could not find any branchial denticles to indicate the position of gill arches. Five external gill slits are present (Gagnier and Wilson, 1996a). The posterior two or three slits are long dorsoventrally, but the anterior two or three slits appear to have been shorter (Figs. 3, 4A, 7E). The best-preserved slits (Figs. 4A, 7E) are separated by an external covering of overlapping scales similar to, but smaller than, the normal flank scales on the body. Where this covering has been broken away, one can see tiny, well-spaced scales with rose-shaped crowns. These may have lined the anterior face of the inter-branchial septum separating adjacent gill slits (Fig. 7E).

The pectoral endoskeleton is unossified and neither dermal plates nor prepectoral spines are present (Gagnier and Wilson, 1996a). The pectoral spine (termed the ‘flank spine’ by Gagnier and Wilson, 1996a, because of uncertainty about its homology) resembles the valve of a brachiopod or bivalve and was at first mistaken for one early in the study of the material eventually described by Gagnier and Wilson (1996a). Finding a similar paired structure on every relevant specimen showed that it was in fact a pair of pectoral spines. Each spine is small, short, and nearly flat, with a wide, shallow basal opening (Figs. 2C, 3, 6A), and presumably did not protrude far from the flank. The spine is ornamented with smooth ribs that converge at the apex of the spine (Gagnier and Wilson, 1996a; Fig. 6A). The pectoral spine is in an elevated position on the flank, immediately posterior to the branchial chamber and well above the level of the prepelvic paired spines (Figs. 2, 3).

The pectoral fins were not preserved on the specimens available to Gagnier and Wilson (1996a) and Bernacsek and Dineley (1977); however, a long, slender pectoral fin is present on one specimen (UALVP 41494; Fig. 6A), and a broken line of attachment for the fin is present on the other (Fig. 3). The detailed shape of the pectoral fin of UALVP 41494 cannot be determined because it appears to have been folded during decay and burial of the carcass.

The pelvic series includes six closely spaced pairs of prepelvic spines (Figs. 1–3, 6B). Gagnier and Wilson (1996a) identified seven spine pairs; however, it is now possible to see that the posterior-most pair in their interpretation corresponds to the pelvic spines, which are very similar. The prepelvic spines are positioned in two diverging rows (Figs. 2B, C, 3, 6A, B), each row beginning anteriorly beneath the anterior half of the branchial chamber, extending posteriorly to end just before the pelvic-fin spines.

Prepelvic spines are ornamented with smooth ribs, which converge on the leading edge of the spine (Figs. 6A, B). Each spine has a large, hollow basal cavity. The lack of a prominent insertion area for each spine is not surprising given their proximity to the coelomic cavity. The prepelvic spines appear to have been held at a low angle to the body margin; their shallow insertion may have allowed some limited mobility. The anterior-most spines are the smallest and the posterior few pairs are the largest in the series (Gagnier and Wilson 1996a; Figs. 2, 3, 6B).

The two specimens of Brochoadmones milesi described here provide the first evidence of prepelvic finlets in any vertebrate. Each prepelvic spine forms the leading edge for a small web of scale-covered skin (Figs. 3B, 6B). The paired finlets have scales on both inner and outer faces and were clearly ventrolateral projections from the body wall. The combination of spines plus scale-covered webs resembles theoretical expectations of multiple paired finlets in primitive gnathostomes (e.g., Zug, 1979:fig. 7.2). The prepelvic finlets increase in size posteriorly in concert with their respective spines.

Pelvic spines are slightly shorter than the posterior-most pre-pelvic spines and approximately half the length of the anal fin spine. The straight pelvic spines have a broad base with an enlarged basal opening (Gagnier and Wilson 1996a; Figs. 2, 3, 6B). Pelvic spines are not deeply inserted in the abdomen. The ribs of the pelvic spines are smooth, of nearly equal thickness, and converge on the leading edge of the spine (Fig. 6B).

Each pelvic spine supports a large, broad-based pelvic fin web. The margin of the pelvic fin web is convex ventrally and posteriorly (Figs. 2, 3, 6C) and slightly overlaps the origin of the anal fin spine, implying that the anus must have been located between the bases of the pelvic fins and the anal spine origin. The scales on the pelvic fin web are aligned in rows and decrease in size towards the fin margin. There is no evidence of endoskeletal support in the pelvic fin base.

The anal-fin spine is long, straight, and has a short base with a prominent insertion area (Figs. 2, 3, 6C). The basal cavity is open along the basal third of the spine. The ribs of the anal spine parallel the leading edge of the spine to within a centimeter of the spine tip. There are no denticles on the posterior face of the anal spine.

The anal fin web is attached to the anal spine for most of the spine's length (Figs. 3B, 6C). In addition, nearly one-third of the trailing edge of the anal fin is continuous with the leading edge of the hypochordal lobe of the caudal fin; in this way the anal spine reinforces the leading edge of the anal + caudal fin complex. The anal fin web has a very short base and is the same length as, or slightly longer than, the anal fin spine (Figs. 3, 6C).

The deep caudal peduncle supports a large, epicercal caudal fin. The large surface area of the caudal fin means that it had a low aspect ratio, the biological significance of which is discussed below. The convex trailing margin of the caudal fin extends about as far posteroventrally as the anal fin and anal spine (Figs. 2, 3, 6C). The transition from typical scales on the caudal peduncle to the scales of the caudal fin web is gradual. Scales on the distal parts of the caudal fin web are aligned in rows (Fig. 6C). The low, lobate, epichordal portion of the caudal fin is covered with small, thickened scales, along the anterior two-thirds of the dorsal midline of the caudal axis (Fig. 7F). The axis of the caudal fin extends about 2 cm posterior to the epichordal part of the fin, and 1 cm beyond the hypochordal portion of the fin (Figs. 2, 3).

Two dorsal fins are present; each dorsal fin is slightly shorter than, and is attached to the basal half of, its fin spine (Figs. 2, 3). The anterior dorsal fin has a convex trailing margin and is shorter anteroposteriorly than the posterior dorsal fin; the posterior dorsal fin is larger, longer based, and more triangular (Figs. 2, 3). The dorsal fin bases are shorter than their respective fin webs; this condition is especially obvious in the second dorsal fin, the base of which is less than half as long as its fin web, a condition similar to that of many modern sharks and unlike that of bony fishes. The scales on the dorsal fin webs are aligned in rows, and the transition from body to dorsal fin scales is gradual.

The dorsal fin spines are long, slender, tapered, and straight, and each has a narrow base with a well-developed insertion area (Figs. 2, 3). The basal cavity of each dorsal spine is open along the basal one-third of the spine (Gagnier and Wilson 1996a; Figs. 2, 3). The dorsal spines have smooth, closely spaced, longitudinal ribs, which are approximately equal in thickness and parallel the leading edge of the spine to within one centimeter of the spine tip (Gagnier and Wilson, 1996a). The posterolateral ribs on each dorsal spine do not converge on the leading edge of the fin spine; instead, they terminate in a paired series of denticles along the posterior face of each spine (Gagnier and Wilson, 1996a).

The transition from head to body scales occurs anterior to the nuchal hump (Fig. 7A). The scales adjacent to the main lateral-line sensory canal just posterior to the head are somewhat thicker than other body scales and show concentric growth rings (Gagnier and Wilson, 1996a; Figs. 7B, C), but otherwise are shaped like typical body scales. These thickened scales commonly are among the few scales remaining on poorly preserved specimens. Body scales are thin and flat (Fig. 7D), and as a result, usually are poorly preserved (Gagnier and Wilson, 1996a). There is no evidence for the typical inflated mass of basal tissue that would be expected on the underside of a typical acanthodian scale. The crowns of body scales are teardrop shaped in crown view and extend posteriorly to overlap the scales more posteriorly by about half their length, thus forming a shingle-like armor (Gagnier and Wilson, 1996a; Figs. 7B, D). There is little difference in scale size over the entire body, although slightly larger body scales are found around the base of the posterior dorsal fin. Smaller scales are found posterior to the head and between the gill slits (Fig. 7E). There is a gradual transition from larger body scales to the smaller scales on the fins, and scales decrease in size towards the trailing margins of all fin webs.

Small, bead-like scales reinforce the dorsal midline of the axis of the caudal fin (zone 2 scales of Gagnier and Wilson, 1996a; Fig. 7F), and these scales reveal histological detail that previously was unknown. Microstructure of these scales is generally of the Nostolepis type. The basal tissue of each of these scales forms a thin, convex pad containing traces of Sharpey's fibers and cavities that may represent traces of cell lacunae (Fig. 8). Crowns are composed of mesodentine. The scale primordium is large, and relatively few, thick growth zones are present (Fig. 8). The short, large-diameter ascending canals (Fig. 8C) supply each odontode in the scale crown, but in the specific thin-sections prepared, there was no trace of vascular canals running through basal tissue.

The position of the lateral-line canal on the trunk is indicated by parallel rows of modified body scales, the canal presumably lying between the rows in life. Both of the new specimens (UALVP 41494 and 49495) reveal both left and right lateral-line canals, the canal on the exposed side of the fish being more prominent than that on the hidden side (Fig. 3). The canal extends in an arc dorsal to the branchial chamber, then continues, uninterrupted, posteriorly along the trunk and caudal peduncle, at about two-thirds the height of the body, until it reaches the caudal fin web. There, it turns posterodorsally to parallel the ventral edge of the main caudal lobe; about 1 cm from the pos-terodorsal end of the lobe, it turns posteroventrally and extends across the hypochordal fin web, nearly parallel to the scale rows, to terminate within millimeters of the margin of the fin web (Figs. 2B, C, 3).

DISCUSSION

CONCLUSIONS

The original and previous descriptions of Brochoadmones milesi were limited by the preservation of the available specimens. Two new, nearly complete specimens allow description of previously unknown anatomical details. Specimens of Brochoadmones milesi are deep bodied and compressed, have a small, tapered head anterior to a prominent nuchal hump, long median fin spines, large dorsal fins, a short-based, ribbon-like anal fin that is partly attached to the caudal fin, and small, plate-like pectoral fin spines. Ventrally, the body carries a paired series of prepelvic finlets, each consisting of a spine and a scale-covered web of skin. The confluent anal and caudal fins, the presence of fin webs associated with each prepelvic spine, and the reduced pectoral fin spine are unique features for an acanthodian. The upper and lower jaws of B. milesi are covered with a series of adsymphyseal tooth whorls that increase in size anteriorly. Scale microstructure is of the Nostolepis type, with few, thick growth zones, crown tissue of mesodentine, and cellular basal tissue with abundant traces of Sharpey's fibers.

The body scale microstructure, fin-spine complement, multiple adsymphyseal tooth whorls and gill slits, and the lack of endo-skeletal ossifications are retained primitive features of Brochoadmones milesi, but the structure of the anal fin, the small pectoral fin spine elevated on the flank, the lack of gill plates, the prepelvic finlet series, and the thin body scales are autapomor-phies of the species. The lack of plate-like pectoral armor, lack of prepectoral spines, the ornamentation of the median fin spines, and the structure of the head scales are features that suggest B. milesi is more closely related to Ischnacanthiformes or Acan-thodiformes than to typical climatiiform fishes. Kathemacanthus rosulentus Gagnier and Wilson, 1996b, is removed from the Bro-choadmonidoidei, which is left with the single genus and species Brochoadmones milesi Bernacsek and Dineley, 1977.

The confluent anal and caudal fins are consistent with the development of median fins from a single, continuous fin fold as seen in embryos of bony fishes. The reduced, dorsally situated pectoral fin and the long pelvic series (prepelvic finlets plus pelvic fin) that extends far anterior to the pectoral fin are evidence for a decoupling of the development of the two fin series, with the pelvic series maximally expressed and the pectoral series minimally expressed in Brochoadmones. The prepelvic finlets themselves are evidence of a long field of pelvic gene expression extending from beneath the branchial chamber to a point lateral to the anal opening.

The anatomy of Brochoadmones milesi resembles that of several ambush predators from Recent freshwater and marine habitats. While discussion of feeding behavior is largely speculative, the piscivorous nature of B. milesi is corroborated by the presence of fish (acanthodian) remains in some well-preserved specimens.