Machaeracanthus goujeti n. sp. (Acanthodii) from the Lower Devonian of Spain and northwest France, with special reference to spine histology

ABSTRACT We describe here a new machaeracanthid acanthodian (Machaeracanthus goujeti n. sp.), based on isolated spines, scales and scapulocoracoids from the Lower Devonian (Lochkovian-Pragian) of the Nogueras Formation, Celtiberia, Spain. The new taxon also includes a fragmentary spine and isolated scales from the Lower Devonian of northern Spain (Palencia and Cantabrian Mountains) and western France (Saint-Céneré) originally assigned to Machaeracanthus sp. The spines of M. goujeti n. sp. comprise two morphotypes in agreement with the morphofunctional model of a pair of pectoral spines articulating with the pectoral girdle already indicated for M. hunsrueckianum Südkamp & Burrow, 2007, M. longaevus Eastman, 1907, and M. sulcatus Newberry, 1857. The morphology and size of the spines distinguish M. goujeti n. sp. from the coeval species M. bohemicus Barrande, 1872; the new species most closely resembles the younger species M. peracutus Newberry, 1857. The spines of M. goujeti n. sp. consist of trabecular and lamellar dentine layers which form the wall of the central axis (pierced by a longitudinal pulp cavity) and lateral expansions. The most superficial layer of dentine is centrifugally deposited in the complete spine; this condition is found in fin spines of some chondrichthyans and contrasts with that observed in typical acanthodian fin spines where the exserted portion is ornamented with ribs of centripetally growing dentine. Very small spines and scapulocoracoids of M. goujeti n. sp. described here, are the first report of juvenile specimens of a species of Machaeracanthus Newberry, 1857. The distal part of the juvenile spine lacks lateral expansions (keel and wing) and demonstrates the first stage in the development of the spine.

The affinities of Machaeracanthus are uncertain. The genus has traditionally been assigned to Acanthodii (Denison 1979) and the Order Ischnacanthiformes (Zidek 1975(Zidek , 1981Young 1986;Webers et al. 1992;Maisey et al. 2002;Maisey & Melo 2005;Südkamp & Burrow 2007). However, the systematic position of Machaeracanthus has become more uncertain in recent years in the context of the rejection of acanthodian monophyly by several workers (Janvier 1996;Brazeau 2009). Burrow & Young (2005) and Burrow et al. (2010b) considered Machaeracanthus as the type genus of the acanthodian family Machaeracanthidae, tentatively included in Ischnacanthiformes or in an Order incertae sedis. Finally, Hanke & Wilson (2010: 179) suggested that asymmetrical fin spines, including those of Machaeracanthus, "may actually represent stem chondrichthyans or stem members of the teleostome plus chondrichthyan clade".
In the present work we describe new disarticulated material, consisting of spines, scales and five scapulocoracoids assigned to the acanthodian genus Machaeracanthus, that occur recurrently together in several levels of the Lower Devonian sediments of the Nogueras Fm. (Celtiberia, Spain) and propose including all of them in a unique natural assemblage, Machaeracanthus goujeti n. sp. New data provided here increase our knowledge on the evolutionary history and taxonomic diversity the group reached during Early Devonian time. The chemical processing of sediment has led to recovery of very well preserved microremains, mainly scales and juvenile spines, which allow study for the first time of the early ontogenetic stages in spines of a Machaeracanthus species.

MATERIAL AND METHODS
All the material studied here come from different outcrops of the Nogueras Formation (Lochkovian-Pragian, Lower Devonian) in two different areas of Celtiberia (Spain; see Fig. 1): 1) the Axial Depression of the Río Cámaras (ADRC; Carls 1988) more specifically from the localities Barranco Sur Santo Domingo, Poyales-East, Maripló and Viñas (see Carls 1988;Dojen 2005); and 2) the Axial Depression of Nigüella (NI; Valenzuela-Ríos 1989), sections Ni-2 and Ni-4 (see Valenzuela-Ríos & Botella 2000). The ADRC is located in the south-eastern part of the Iberian Chains (Teruel Province) and NI is situated in its north-eastern part (Zaragoza Province; see Fig. 1). The Nogueras Fm. consists of a 140-150 m thickness of shallow-marine deposits with bioclastics limestones, marls and arenaceous shales (Fig. 2). This formation includes the "Leitbank A" (bed A), located between the local boundary d2bα/d2bβ, a laterally continuous dark mudstone bed, 35-50 cm in thickness, which corresponds almost exactly to the Lochkovian/Pragian boundary. The most important biostratigraphic marker is the brachiopod Vandercammenina sollei Carls, 1986, which indicates the beginning of the Pragian in rhenish facies (Carls & Valenzuela-Ríos 2002). A little below bed A, chitinozoans of the Lochkovian/Pragian boundary occur (Morzadec et al. 1991). The upper 80 metres of the Nogueras Fm. (submembers d2cα to d2cβ) are Pragian in age. Remains assigned here to Machaeracanthus goujeti n. sp. occur in numerous levels around the Lochkhovian-Pragian boundary (submembers d2aα to d2cα, see Fig. 2). Most of the material comes from the dissolution of limestone rocks with formic acid (5-10%), although a number of the largest spines are preserved in marly sandstone slabs.
Specimens were photographed with a Leica MZ12 binocular microscope connected to a digital camera "Leica" DFC420 and with a Scanning Electron Microscope (Philips XL-30) hosted at Electron Microscopy Service of the University of Valencia. For the histological study the spines and scales were embedded in Canada balsam and polished subsequently along transverse or longitudinal planes. The material, once prepared, was photographed with a petrographic microscope (James Swift England) connected to a digital camera. All Machaeracanthus remains studied here are reposited in the Museum of Paleontology at the University of Zaragoza (MPZ) and at the Museum of Geology of the University of Valencia (MGUV).   holotype. -MPZ 2010/948 (Fig. 4A); a complete spine, c. 53 mm long, slightly damaged at distal end, preserved in a marly sandstone slab from the late Lochkovian (Nogueras Fm.) of the Locality Poyales-East.   diAgnosis. -Machaeracanthus species with relatively slender spines with a maximum width to length ratio c. 1:7; two morphotypes can be distinguished, both showing narrow lateral expansions (keel and wing) of similar width (upper surface view); morphotype 1 is represented by spines with longitudinal striation mainly on the proximal third, and showing, in cross section, a triangular to sub-triangular longitudinal ridge on the upper surface and a more rounded longitudinal ridge on the lower surface; morphotype 2 is represented by densely striated spines which present, in cross section, a subsquare-shaped longitudinal ridge on the upper surface and a rounded and broader longitudinal ridge on the lower surface. Scales with eight to twelve ridges which converge from the anterior part to the centre of the crown and diverge posteriorly; the ridges extending behind the upper part of the neck, never reach the base; neck is pronounced, presenting a slight narrowing; base moderately convex. Mesodentine forms most of the crown, and the base is formed by cellular bone with bone cell lacunae arranged parallel to growth lines.

Institutional abbreviations
description Spine morphology (Figs 3; 4A-K; 5) More than 50 spines, from nearly complete to very fragmentary specimens, were studied. The preserved material indicates a wide range of sizes suggesting the assemblage of remains belonging to individuals of different ontogenetic stages (see below). The smallest complete specimen (  fig. 2B) are about 1.5 cm in maximum width pointing to complete spines more than 10 cm long. All the spines are asymmetrical, curved posteriorly and have a characteristic saber shape with a thick central axis (the body of the spine) and two narrow lateral expansions, an anterior keel and a posterior wing. As shown in specimen MPZ 2010/950 (Fig. 3), the lateral expansions start close to the distal termination of the spine, gradually increase their width in proximal direction to reach the maximum and then decrease in width to the proximal termination of the spine. Upper surfaces of keel and wing exhibit a similar width with the exception of the most proximal part of the spine where the keel is narrower than the wing. Concerning the lower side, the keel is narrower than the wing along the whole spine. Two different morphotypes can be recognised according to the morphology in transverse section of the

Spine histology (Figs 5-8)
The histological structure of the spines changes from distal to proximal ends (Figs 5-8). A pulp cavity extends along the central axis of the spine (Fig. 5); the cavity is very narrow in the distal part of the spine, becomes wider proximally and opens at the most proximal end of the spine (Fig. 3A, C). At the proximal part of the spine, the pulp cavity is higher than wide (transverse section view) in morphotype 1 and nearly rounded to wider than high in the morphotype 2 (Fig. 5).
The wall of the central axis (body of the spine) and lateral expansions (keel and wing) consist of trabecular and lamellar dentine.
Centripetally growing trabecular dentine surrounds the pulp cavity in the distal and middle parts of the spine (Figs 5A, B, D; 6A; 7A). Centrifugally growing trabecular dentine surrounds the pulp cavity proximally (Figs 5C; 8A), covers the centripetally growing trabecular dentine in the mid-and distal parts of the spine and extends laterally to form the internal and main part of the keel and wing (Figs 5; 6A; 7A; 8). There is no evidence of sharp structural discontinuity between centripetal and centrifugal trabecular dentines, but both hard tissues can be easily distinguished by the difference in the dimension of the intertrabecular spaces (see below). In addition, growth marks in the trabecular dentine are also less apparent than in the lamellar dentine.
Centrifugal lamellar dentine with clear growth marks covers the centrifugal trabecular dentine at least in the distal half of the spines (morphotypes 1 and 2) (Figs 6; 7A, B).
Vascular pattern of the centripetally growing trabecular dentine is extremely regular as observed in transverse section (Figs 5A, B, D; 6A; 7A). Rows of round cavities and fairly straight canals radiate from the pulp cavity and connect with the centrifugal trabecular dentine. The round cavities represent the transverse sections of longitudinal canals. In general view, the intertrabecular spaces are wider in the centripetal dentine than in the centrifugal one.
Denteons around the vascular canals and interdenteonal areas can be distinguished in sections with less taphonomic alteration (Fig. 7). Dentinal tubules radiating from the vacular canals form a dense network in the borders of denteons and in the interdenteonal areas (Fig. 7B, C). Numerous interglobular spaces appear in the interdenteonal areas (Figs 7C; 8B). The spaces are usually filled by opaque authigenic minerals and present evidence of severe alteration postmortem in some regions of the spines ( Fig. 8B; see discussion below).
The most superficial layer of dentine, centrifugally deposited, is pierced by dentinal tubules that exhibit their finer distal branches in centripetal direction (Fig. 6B, C). Consequently, there is no evidence of hypermineralized enameloid.
Scale morphology (Fig. 9) Scales are large, 1 to 2 mm long and wide, and 0.6 to 1.5 mm height. Largest specimens are up to 2.4 mm long. The crown is flat, nearly parallel to the interface between the base and neck and extends posteriorly beyond the base (Fig. 9B, G, H). The crown is ornamented with a variable number of ridges (8-12) which rise from the upper part of the neck, more or less parallel along the rostral margin and then converge to the center of the crown. On A the posterior crown, when it is preserved, ridges diverge as radial ridges towards the caudal edge of the crown (Fig. 9A, C, E, G, L), which is divided into eight to twelve long parallel denticulations, with each of these denticulations corresponding to the end of a ridge (Fig. 9H, J, L). However, as the posterior part of the crown is thin and delicate, denticulation at the caudal margin is broken or damaged in most of the specimens (Fig. 9A, C, E). The anterior margin of the crown is rounded (Fig. 9A, C, E, G). The neck is pronounced and presents a slight narrowing between the crown and the base (Figs. 9D, F, J). The base varies from low to moderately convex and protrudes rostrally in front of the anterior edge of the crown (Fig. 9A, C, E, L). The shape of the base is rhomboidal and smaller than the crown (Fig. 9B, H). Small vascular canal openings can be observed in the neck.
Some scales have a long and narrow crown, with a long neck; in these scales the ridges are strongly  marked and do not follow the pattern of convergence to the center, but run subparallel along the crown (Fig. 9I). These forms are similar to some figured by Goujet (1976: pl. 61, figs 15-17). Others have an asymmetrical rounded crown (Fig. 9K), where the ridges are smooth and short and are arranged throughout the crown. These morphologies are not present in the material figured by Goujet (1976) and are very similar to the forms "verwachsene" and "schamale" figured by Wang (1993: pl. 15, figs 5, 6).
Scale histology (Fig. 10) The crown is formed by apposed growth layers of mesodentine in the posterior part of the crown, and superposed growth layers in the anterior (Fig. 10A, H). A dense network of dentine tubules and lacunae occupies the internal zones of growth layers, although in zones of the outermost part of every growth layer (especially in areas that correspond with ridges) the network of dentine tubules is less dense, with large sinuous branched tubules, and no lacunae, so that the tissue resembles orthodentine (Fig. 10G, H). No vascular canals can be identified either in the base or in the crown, although it may be due to fossilisation problems. The base shows numerous concentric growth lines consisting of successive alternating of dark and light layers (Fig. 10A, C, F). It is formed by cellular bone with some bone cell lacunae aligned with the different growth lines (Fig. 10F) and is pierced by numerous Sharpey's fibers, thick and arranged radially and obliquely (Fig. 10F). The apex of the base, immediately above the center of the crown, is often crystallized and occupied by calcite that hides the inner structure (Fig. 10C).

Scapulocoracoid (Fig. 4L-O)
Three right and two left perichondrally ossified scapulocoracoids of typical Machaeracanthus morphology (see Burrow et al. 2010b) occur associated with scales and spines of M. goujeti n. sp. All the scapulocoracoids found are of small size, the preserved height (dorso-ventrally) of the largest specimen MPZ 2010/952 (Fig. 4L-O) is 13 mm (estimated not more than 18 mm if it was complete). Unfortunately, none of the elements is entirely preserved, missing the dorsal end of the scapular shaft and, to a greater or lesser extent, the ventral areas of the scapulocoracoid blades. The preserved specimens show an elongate constricted scapular shaft, which is subcircular in cross section (about 0.25 cm in dimension anteroposterior), and a triangular scapulocoracoid portion (Fig. 4L, M). The transition between the scapular shaft and the triangular scapulocoracoidal areas is not abrupt but gradual, although the scapular shaft slightly narrows and bends anteriorly. The better preserved scapulocoracoid (MPZ 2010/952, Fig. 4L-O, broken during photography) shows a flat medial face on the blade while the preserved part of the lateral face slightly flares out in its most ventral part, consistent with the presence of a ventrolateral expansion of the scapulocoracoid.

DISCUSSION
tAxonoMic AssessMent All the machaeracanthid elements found in Celtiberia (i.e. spines, scales and scapulocoracoids) appear repeatedly associated (but not articulated) in the same levels and all of them present a similar stratigraphic distribution (see Fig. 2), consequently we propose their inclusion in a single assemblage, all belonging to Machaeracanthus goujeti n. sp. This association is congruent with previous Machaeracanthus assemblages from other localities (see Burrow et al. 2010b). Mader (1986) and Wang (1993) figured and briefly described fragmentary spines from Lebanza Fm. (Palencia) and Nogueras Fm. (Celtiberia). The comparison with our material indicates that the specimen illustrated by Mader (1986: pl. 1, fig. 8) corresponds to the distal part (about 0.9 cm long, 4 mm maximum width) of a spine of morphotype 1 with the characteristic triangular upper longitudinal ridge.
The presence of spines of two different morphotypes within Celtiberian material is consistent with the presence of a pair of pectoral fin spines of different morphology on each side of the body as is known to occur in machaeracanthids (Burrow et al. 2010b).
Asymmetric spines of two different morphologies occur in the material from the Nogueras Fm.  fig. 2B), maximum width to length ratio c. 1:7 (adult individuals), and in both, the keel and the wing are narrow and of approximately equal width; they differ in the cross sectional shape of the upper longitudinal ridge and in the degree of longitudinal striation. As we have still not found associated pairs of spines in a single slab in Celtiberia, we cannot rule out definitely the possibility that the two morphotypes belong to two different species. Nevertheless, the association of these two morphologies in a single species is in agreement with the model proposed by Südkamp & Burrow 2007 (see also Burrow et al. 2010b), for Machaeracanthus with two unequal spines articulating on the In addition to the scales from Celtiberia, we assign to M. goujeti n. sp. scales of Machaeracanthus sp. from Saint-Céneré Fm., northwest France (Goujet 1976), and scales of Machaeracanthus sp. A from Lebanza Fm. and La Vid Fm. in Northern Spain (Mader 1986).
Rouault (1858) erected two species, Machaerius archiaci and Machaerius larteti, for fragments of spines coming from the Mayenne department (Brittany, France), the same region of the scales studied by Goujet (1976). Unfortunately, the exact origin (locality and age) of the material was not indicated. These spines, not illustrated and poorly described, have been lost and the two species have been considered nomina vana (Zidek 1981).
Mader (1986) was the first author who pointed out the similarities between the Spanish and French material. Interestingly, Mader suggested that the complete assemblage could be a new species differing from M. bohemicus and M. stonehousensis, the latter being a species also identified in the Lower Devonian of Spain. Later, Wang (1993) distinguished between scales of Machaeracanthus sp. A (including the original material described by Mader and part of the material studied by Goujet) and Machaeracanthus sp. B (including here material from Celtiberia and one element of the French material); however, this variation probably corresponds to different topological scale types within a single taxon rather than a distinct species. It can also correspond to ontogenetic variation in the development of squamation; for example the small scales with a simple crown morphology (reduced number of ridges) figured by Goujet (1976: pl. 61, figs 7, 11) could belong to juvenile individuals.
As well as the distinctive arrangement of the growing dentine layers (see above), scales of M. goujeti n. sp. also show the general morphology present in the scales of other Machaeracanthus species (i.e. which tend to converge toward the centre of the scale and posteriorly diverge and the denticulation at the caudal margin of the crown (see Figure 9). Scales of M. goujeti n. sp. resemble those of M. bohemicus, but as Goujet (1976) pointed out, they can be clearly distinguished by the lower protrusion of the base present in the latter species (see Gross 1973: pl 28, figs 21c, 22b;pl. 29, figs 6-8) and from the lateral aspect of the crown ribs (see Mader 1986: 30). Additionally, the posterior crown margin is markedly pointed in M. bohemicus (Gross 1973: pl. 29, figs 1-5) but not in the new species. The base of M. goujeti n. sp. scales is of cellular bone and mesodentine forms most of the crown. This histological structure differs from M. bohemicus and M. pectinatus, where orthodentine forms most of the crown and no cell lacunae are present in the base (Burrow & Young 2005). Otherwise, M. stonehousensis presents "stranggewebe"-like tissue (Mader 1986;Vergossen 1995Vergossen , 2000 and radial rows of crown pores (Vergoossen 1995;, features absent in other described Machaeracanthus scales. Because of these pore rows, Vergoossen (1995Vergoossen ( , 2000 did not include M. stonehousensis within the genus and considered this species to be a poracanthodid ischnacanthiform. Nevertheless, Burrow et al. (2010b) based on an unpublished manuscript by Denison supported its assignment to Machaeracanthus.

M. bohemicus, M. pectinatus and M. stonehousensis): a crown ornamentation consisting of simple ridges
Scapulocoracoids themselves are probably the least diagnostic machaeracanthid elements, showing a morphology similar to those of ischnacanthiforms but with an extra ventrolateral expansion (Burrow et al. 2010b). Scapulocoracoids from Celtiberia present a slender scapular shaft which broadens out to a triangular blade, similar to those of other Machaeracanthus species (compare with Burrow et al. 2010b: fig. 5). Unfortunately, all specimens from Celtiberia are incomplete, lacking the ventral areas of the scapulocoracoid blades, so that we cannot definitively assert the presence of a ventrolateral expansion, although the most complete specimen suggests it (see above).
size And ontogeny Fritsch (1893) and Gross (1965) speculated about the absolute size of individuals of Machaeracanthus species and estimated a total length of about 200 cm for Machaeracanthus bohemicus and 140-170 cm for Machaeracanthus sp. (originally identified as Gemuendolepis muelleri Gross, 1973) from the Hunsrück Slate. Still the fossil record of Machaeracanthus is composed of isolated elements, mainly spines and scales (see Introduction), so that the allometric relationship between the skeletal elements and the body of the fish is unknown, hampering estimation of the length of the individuals belonging to different species. However, we can compare the dimensions of the dermal spines of the recorded species and assume that the size of the Machaeracanthus spines is an appropriate indicator of the complete size of the individual, in similar fashion to most acanthodians known from a more complete fossil record (e.g. Upeniece 1996;Zajíc 1998Zajíc , 2005. In this respect, M. goujeti n. sp. appears to be a small species Several very small spines of M. goujeti n. sp., less than 0.5 cm in width, probably belong to juvenile individuals. Specimen MPZ 2010/950 (Fig. 3), the smallest nearly complete spine of M. goujeti n. sp. is the best example, found by acid preparation of the fossiliferous limestones. In addition to the extremely small size, the specimen differs in several morphological features from the other, larger spines form the new species and the other Machaeracanthus spines described until now (see review in Burrow et al. 2010b). The spine in its distal part (about one fourth of the total length) lacks lateral expansions and presents a diamond-shaped transverse section. Although the actual spine is relatively robust with a ratio of maximum width to length of 1:4, it could originally be slender with a longer region lacking lateral expansions as suggested by the distal worn surface. Interestingly, an unpublished small specimen MB f. 14194 (Machaeracanthus sp.) (27 mm GEODIVERSITAS • 2012 • 34 (4) long, lacking the proximal end) from the Eifelian (?) of Müllerberg near Nettersheim (Germany) exhibits a distal half without lateral expansions ranging 0.5-1 mm in width; the proximal half of the specimen, 4 mm in maximum width, presents the usual Machaeracanthus morphology with well developed keel and wing, lateral to the central axis. The distal part of the juvenile spine lacking lateral expansions, document the first stage in the development of the spine.
The presence of juveniles of M. goujeti n. sp. in the Spanish localities is also suggested by the size of the scapulocoracoids associated to the spines. The elements from Celtiberia are significantly smaller than that from M. sulcatus. The scapulocoracoids of M. goujeti n. sp. are not more than 18 mm in height (inferred height) whereas those from M. sulcatus reach 43 mm. The scapulocoracoids of M. major and M. bohemicus are even larger than those from M. sulcatus, in agreement with their enormous spines. As a good example a very large complete scapulocoracoid of M. major is 300 mm high (Burrow et al. 2010b).
spine histology Several authors have studied the microstructure and histology of the spines of Machaeracanthus: Barrande (1872); Gross (1933); Wells (1944);Ørvig (1951);Denison (1979);Zidek (1975Zidek ( , 1981Young (1986);Webers et al. (1992); Anderson et al. (1999);Burrow & Young (2005);Südkamp & Burrow (2007), Fernández-Herrero et al. (2009) and Burrow et al. (2010b. In a preliminary note, Fernández-Herrero et al. (2009) gave the first description of the histology of the spines from Celtiberia. The authors described the wall of the spine as consisting mainly of "compact cellular bone highly vascularized" with the development of osteons and osteocytes. Fernández-Herrero et al. (2009: fig. 3C, D) also distinguished in some sections the presence of an "internal layer, in contact with the central cavity [pulp cavity] of a darker colour and with numerous cellular (?) rounded spaces". Our re-study of the original material indicates that the so-called bone and osteocytes correspond to altered interglobular dentine with interglobular spaces infilled by authigenic minerals resembling bone-cell lacunae (Soler-Gijón 1999: figs 11B;16B, C;Sansom et al. 2005: 380, fig. 2B, D). Dentine tubules, which were not recognized by Fernández-Herrero et al. (2009), still remain in a few areas of the sections despite harsh diagenetic alteration (see Fig. 8). Interglobular dentine, representing areas with weak mineralization during dentinogenesis, has been described in dermal spines and teeth from Palaeozoic to Recent, and from chondrichthyans to mammals (see Soler-Gijón 1999 and references therein). On the other hand, the darker internal layer of the spine wall surrounding the pulp cavity described by Fernandez-Herrero et al. (2009) corresponds to an authigenic cement (ferruginous or phosphatic) covering the trabecular dentine and infilling the vascular canals connected with the pulp cavity. Pulp cavity, vascular canals and interglobular spaces are optimal microenvironments for taphonomic alteration mediated by microorganisms (Kierdorf et al. 2009 fig. 3A). These comparative results suggest that the microstructure of M. goujeti n. sp. described here for distal, middle and proximal parts are representative of the general structure in the Machaeracanthus spine.
The spines of M. goujeti n. sp. consist of dentine only, and this was probably the case for the majority or even all of the species of the genus. Recently, Burrow et al. (2010b: fig. 3G), figured and described denteons, dentine tubules and interdenteonal areas lacking cell lacunae in osteodentine (called trabecular dentine in this paper) surrounding the pulp (central) cavity of the spine of M. peracutus. Machaeracanthus goujeti n. sp. and M. peracutus also share the absence of outer ortho-or mesodentine layers, a condition which clearly distinguishes Machaeracanthus from the other acanthodians (Burrow et al. 2010b: 65). In this respect, we have to note here that the smooth longitudinal carination ornamenting the Machaeracanthus spines develop by the regular deposition of centrifugal dentine in constrast to the longitudinal ridges of other acanthodians (e.g., climatiforms, acanthodiforms) which develop by centripetal deposition of orthoor mesodentine (Denison 1979;Beznosov 2009). Interestingly, longitudinal striation with centrifugal dentine appears in dorsal spines of xenacanth sharks (cf. Soler-Gijón 1999) which suggests developmental similarities between Machaeracanthus and some primitive chondrichthyans.
True cellular bone is described and figured by Burrow & Young (2005: fig. 8) in isolated small dermal elements tentatively identified as "Fin ray or spine elements" possibly from M. pectinatus (? late Emsian/early Eifelian of the Craven Peaks Beds, Georgina Basin, western Queensland, Australia). Although these dermal elements are associated with Machaeracanthus scales (see Burrow & Young 2005, fig. 7), they are deeply different from the spines of M. goujeti n. sp. and other species of the genus discussed above. The dermal elements of Craven Peaks Beds superficially resemble Machacaeracathus spines in transverse sectional shape. However they lack the central pulp cavity and the extremely regular vascular pattern, with denteons shown by the Machaeracanthus spines (see above). The general morphology (partially bifurcated) and histology (cellular bone and possibly Williamson's canals) of these elements correspond to fin rays of actinopterygians (e.g., Arratia 2008), but no other possible actinopterygian skeletal elements are found in the Cravens Peak Beds Our results from M. goujeti n. sp. are relevant to continue the discussion by Burrow & Young (2005) who tried to explain the possible origin of the particular morphology of Machaeracanthus spines, so different to the rest of the acanthodians (see Janvier 1996). Based on the morphology of the possible "fin spines/rays" of M. pectinatus, Burrow & Young (2005: 20-21) indicated that "If machaeracanthids derived from an ischnacanthiform ancestor, perhaps Machaeracanthus spines developed by enlargement of the fin basals after loss of the pectoral fin spines". Spines of M. goujeti n. sp. are different in morphology and histology from fin spines, fin rays and radials of teleostomes. In contrast, Machaeracanthus presents some features found in dentine spines of some chondrichthyans (e.g., xenacanths). A detailed study of the histology of the spines of Machaeracanthus, including serial cross sections, (in progress) will give information about their growth and development and could provide new information on the affinities of the genus.
biostrAtigrAphy And pAlAeobiogeogrAphy The stratigraphic distribution of M. goujeti n. sp. in Celtiberia, shown in Figure 2, ranges from unit d2aα (middle-late Lochkovian) to unit d2cβ (early Pragian) being relatively continuous during the entire interval. Thus, the only coeval species of M. goujeti n. sp. is M. bohemicus from the Lochkovian of the Czech Republic, with other Machaeracanthus species known from younger strata (middle Pragian to Eifelian) with the exception of the Late Silurian species M. stonehousensis of eastern Canada, which is known only from scales. Remarkably, both species, M. goujeti n. sp. and M. bohemicus, are so far the only Machaeracanthus species known from associated spines, scales and scapulocoracoids.
The stratigraphical and palaeogeographical distribution of the oldest machaeracanthids (i.e. M. stonehousensis, M. bohemicus and M. goujeti n. sp.) suggest an origin for the genus in seas surrounding Gondwana-derived terranes (including Ibero-Armorican and Bohemian massif ) during the Late Silurian and then distributed worldwide (North America, Africa, Australia and Antarctica) during Pragian-Emsian times. The dispersion of Machaeracanthus and other organisms could be favoured by the development of large marginal marine areas. As commented above, the Nogueras Fm. consist of neritic sediments, with some episodes of high shallowing during the Late Lochkovian (unit d2aβ5) (see Carls 1999;Dojen 2005). These shallow near-coastal marine environments offer a number of restricted and protected areas from open seas which could have been a perfect place for development and growth of juvenile and adult Machaeracanthus, although the adult forms probably also inhabited deeper water environments, as usually occurs in recent fishes. The frequent changes in the subsidence of the Celtiberian Basin during Lochkovian and Pragian times, often associated with high shallowing events (Carls 1999), facilitated the dispersion (and interchange) of shallow-water faunas of Celtiberia, as has been recently documented for ostracods (Dojen 2005), primitive chondrichthyans (Martínez-Pérez et al. 2010), and placoderms (Dupret et al. 2011).

CONCLUSIONS
The study of numerous spines from the Lower Devonian of Celtiberia has confirmed the presence of a new species, M. goujeti n. sp., very different in morphology and size to the coeval species M. bohemicus. Machaeracanthus goujeti n. sp. comprises two morphotype spines in similar fashion to M. hunsrueckianum, M. longaevus and M. sulcatus. In this respect, the new material from Celtiberia supports the morphofunctional model for pectoral girdle-spines of Machaeracanthus proposed by Südkamp & Burrow (2007). The model, based on partial articulated/associated material indicates that a pair of spines articulated with the pectoral girdle (see Burrow et al. 2010b: fig. 5 I).
Juvenile spines and scapulocoracoids of M. goujeti n. sp. are the first record of juvenile individual of Machaeracanthus. The distal part of the juvenile spine lacks lateral expansions (keel and wing, typical in the "adult" Machaeracanthus morphology) documenting the first stage in the development of the spine.
The spine of M. goujeti n. sp. (morphotypes 1 and 2) exhibits an elongated pulp cavity which opens proximally. The wall of the central axis of the spine surrounding the pulp cavity, and the lateral expansions, consist of dentine. Centripetally growing trabecular dentine obliterates the pulp cavity in the distal and middle parts of the spine. Centrifugally growing dentine (trabecular and lamellar) forms the rest of the spine without development of an ornament with longitudinal ribs of ortho-or mesodentine on the exserted part.
The morphology and histology of Machaeracanthus, as seen in M. goujeti n. sp., differ from those of fin rays and radials of acanthodians and osteichthyans. The results presented here are in disagreement with Burrow & Young (2005) who proposed that the Machaeracanthus spine developed by enlargement of the fin rays or radials of some ischnacanthiform acanthodian.