I. — Introduction

Vertebrates are very rare in the Lower Devonian of western France. The first specimens to be recorded from mid-Palaeozoic deposits in the region were Machaeracanthus spines, assigned to two species and described briefly by Rouault (1858). According to Burrow et al. (2010, p. 59): “Machaerius archiaci Rouault, 1858 and Machaerius larteti Rouault, 1858 were erected the year after Newberry's first description, for small fragments of spines from the Lower Devonian of northwestern France. Unfortunately, these were poorly described, not figured, and have been lost, so Zidek (1981) in his review of Machaeracanthus spp. listed them as nomina vana.” In fact the fragments, based on their width (Bézier, 1913, table), were from quite large spines. Their provenance is unclear, as Rouault (1858) only recorded that they were from western France. According to a letter written by M. de Verneuil to M. d'Archiac in 1858 (Anonymous, 1858), Rouault's locality was reported to be Saint-Léonhard in north Sarthe department (= Saint-Léonard-des-Bois). However, on investigation (Tromelin in Guillier, 1971) the only rocks found in this area were Ordovician. Tromelin in Guillier (1871) and Tromelin & Lebesconte (1876) recorded that Rouault's specimens were from Les Courtoisières, near Brûlon, Sarthe, but gave no proof for this assertion. All other Machaeracanthus spines reported from western France (Mayenne, Maine-et-Loire, Manche, Sarthe departments; Fig. 1A) have been considered to belong to Machaeracanthus bohemicus (Barrande, 1872), or Machaeracanthus cf. bohemicus (Bézier, 1913). (Bézier [1913] listed the genus as Ctenacanthus as originally designated by Barrande [1872], not Machaeracanthus, even though Kayser [1883] had long since reassigned the species to the latter genus). The description given by Rouault (1858) for his genus Machaerius comprises generic characters of Machaeracanthus Newberry, 1857 and specific characters of M. bohemicus. Unfortunately the spines described by Rouault (1858) have still not been found.

Figures 1a and 1b

As well as spines, scales assigned to Machaeracanthus sp. have also been recorded in western France. Goujet (1976, 1980) described and figured scales from the Saint-Céneré and l'Armorique formations respectively. Botella et al. (2012) referred the scales from the Saint-Céneré Formation that Goujet (1976) originally assigned to Machaeracanthus sp. to a new species M. goujeti, based on their similarity in morphology and stratigraphic occurrence (Lochkovian-Pragian) with the type material of this species from Celtiberia, Spain.

Bézier (1913) published the only detailed description, including photographs, of two Machaeracanthus spines from western France. His specimens, collected at Saint-Ger- main-le-Fouilloux and originally described as “Ctenacanthus cf. bohemicus”, were until recently, like those of Rouault, considered to be lost. DG (technical curator in the Rennes University Geological Museum collections) located the specimens in his institute's collections in 2015: two were donated by G. de Vaucenay (IGR 132670 and IGR 132671), part and counterpart, and there is another spine (IGR 10869) from the Musée de Rennes collection. As such specimens are so rare, we consider it appropriate to refigure them and reassess their taxonomic affinity.

Figures 2a to 2m

II. — Geological Setting

Several localities in the Laval Synclinorium in the eastern part of the Massif Armoricain of western France (Fig. 1B), including Saint-Germain-le-Fouilloux village about 8 km N of Laval, Mayenne department, have been famous since Oehlert's work in the late 19th century on invertebrate fossils from the region (e.g. Oehlert, 1884). The La Roussière quarry at Saint- Germain-le-Fouilloux (Fig. 1B, locality 1) is the type locality for many Lower Devonian (Pragian) fossils. The quarry exposes the shales and limestones of the middle part of the Early Devonian Saint-Céneré Formation (Frýda et al., 2008). According to Bézier (1913), other Machaeracanthus localities in the Laval Synclinorium are les Courtoisières, Sarthe (Fig. 1B, locality 2), and Saint-Roch quarry, Changé, Mayenne (Fig. 1B, locality 3). These are in Pragian strata of the Saint-Céneré Formation. Machaeracanthus spine localities in western France also include Néhou, Manche department (Fig. 1B, locality 4), presumably in the Pragian Nehou Formation of the Cotentin region (for stratigraphy see Morzadec et al., 1988). Another locality listed by Bézier (1913) is near Angers, Maine-et-Loire department (Fig. 1B, localities 5-7), in the ‘Calcaire d'Angers' of the Angers Synclinorium. It is unclear if the ‘Calcaire' denoted was the Angers Limestone (Pragian-Emsian) or the Vern Limestone (Pragian), the latter being the regional stratigraphic equivalent of the Saint-Céneré Formation. However, Péneau (1935) recorded that Machaeracanthus spines are found in the Vern Limestone. Le Maître (1934) also recorded Machaeracanthus spines from the Valet quarry in Chaudefonds (= Chaudefonds-sur-Layon village) in the ‘Calcaires de Chaudefonds' of Maine-et-Loire.

As well as these records of spines, Machaeracanthus sp. scales were collected and described from Saint-Céneré, Mayenne in the Saint-Céneré Formation (Goujet, 1976; Fig. 1B, near locality 3), and from la Pointe de l'Armorique, Finistère, in the L'Armorique Formation (lower Pragian), Châteaulin Synclinorium (Goujet, 1980; Fig. 1B, locality 8).

III. — Systematic description

†Class ACANTHODII Owen, 1846
Order indet.
Family MACHAERACANTHIDAE Burrow & Young, 2005
Genus Machaeracanthus Newberry, 1857

Type species. Machaeracanthus peracutus Newberry, 1857.

Machaeracanthus bezieri nov. sp.
(Figs. 2A-K, 3A-F)

Synonymy.

? 1858 Machaerius Larteti Rouault. p. 102

? 1858 Machaerius Archiaci Rouault. p. 103

? 1858 Machaerius. Anonymous, p. 370

? 1868 Machaerius Larteti et Archiaci. D'Archiac, p. 32

in part 1872 Ctenacanthus bohemicus Barrande. p. 641, pl. 28 figs. 8-10, 16, 20, 24

1878 Ctenacanthus bohemicus Barr. Kayser, p. 4, pl. 35 figs. 12, 12a, 12b

? 1879 Ctenacanthus bohemicus, Barr. Oehlert & Davoust, p. 699

1883 Machaeracanthus. Kayser, pl. 4 fig. 2

1889 Ctenacanthus bohemicus. Barrois, p. 333

? 1890 Ctenacanthus bohemicus, Barr. Oehlert, p. 755

1913 Ctenacanthus cf. bohemicus Barr., “probably Ctenacanthus bohemicus Barr.”. Bézier, p. 67, plate figs. 1-2

1924 Machaerius Larteiti. Havel, p. 16

1928 Machaeracanthus cf. bohemicus (BARRANDE). Péneau, p. 132-133

Machaeracanthus bohemicus. Schmidt, p. 240, fig. 6a

Machaeracanthus bohemicus BARRANDE. Le Maître, p. 106, pl. 17 fig. 8

Machaeracanthus cf. bohemicus (BARRANDE). Péneau, p. 43-44

1953 Machaeracanthus cf. bohemicus. Pillet, p. 16

1959 Machaeracanthus sp. n. Gross, p. 30

? 1976 Machaeracanthus sp. Goujet, p. 313, fig. 54 A-E, pl. 61 figs. 3-17, pl. 63 fig. 1a,b

in part 1979 M. bohemicus (Barrande) 1872A. Denison, p. 52

? 1979 Machaeracanthus sp. Morzadec et al., p. 187

? 1980 Machaeracanthus sp. Goujet, p. 311, pl. 42 figs. 1-5

? in part 1993 Machaeracanthus Goujet, p. 28

in part 2010 Machaeracanthus bohemicus. Burrow et al. p. 76, fig. 8

? in part 2012 Machaeracanthus goujeti Botella et al., p. 765

Type material. (All from Saint-Germain-le-Fouilloux) Holotype IGR 132670; Paratypes IGR 132671a,b part and counterpart, and ML-PAL-03006.

Etymology. Acknowledging Toussaint Bézier for his important contributions as curator of the Natural History Museum of Rennes, and his contribution to the study of Machaeracanthus remains from France.

Material examined. From Saint-Germain-le-Fouilloux: type specimens: holotype IGR 132670 (Fig. 2A–C), paratype 132671a,b (Fig. 2D–G) in the Rennes University Geological Museum collection (IGR 132670 collected by de Vaucenay possibly in 1909; IGR 132671 collected in 1914); paratype spine ML-PAL-03006 Fig. 2H-J), Collection Oehlert, Laval Museum. From an unrecorded locality (matrix indicates Saint-Germain- le-Fouilloux): ML-PAL-29822 (Fig. 2K), Collection Oehlert, Laval Museum.

Type horizon. Saint-Céneré Formation (Pragian).

Type locality. La Roussière quarry, Saint-Germain-le- Fouilloux, Mayenne.

Geographical and stratigraphic distribution. Western France: Saint-Céneré Formation (Pragian); western Germany: Taunus Quartzite (Pragian); Czech Republic: Dvorce-prokop Formation (Pragian).

Diagnosis. Machaeracanthus species with adult spines c. 16 cm long and a length to maximum width ratio c. 6:1; smooth surfaced central shaft with semicircular cross sections of different diameters above and below the plane of the side expansions; inner lateral expansion (wing) and outer lateral expansion (keel) of similar widths; one morphotype has wing with one surface convex (side to side) bearing thin longitudinal ridges, other surface smooth, and keel with both surfaces smooth and flat; second morphotype with: keel with one surface convex and bearing thin longitudinal ridges, and wing with both surfaces smooth and flat.

Figures 3a to 3g

Remarks. Bézier (1913) did not describe the morphology of the spines from Saint-Germain-le-Fouilloux, he merely referred to Barrande's (1872) original description and plates of Machaeracanthus bohemicus, Rouault's (1858) descriptions of Machaerius laterti and M. archiaci, and Oehlert & Davoust's (1879) description of Ctenacanthus cf. bohemicus, noting that these descriptions fitted the Saint-Germain-le-Fouilloux spines as well.

Description. The Saint-Germain-le-Fouilloux spine IGR 132670 (Fig. 2A-C), nominated as the holotype, lacks the proximal end, and is 13 cm long and 2.5 cm at its widest. As exposed by the natural cross-section terminating the specimen, the central shaft has a semicircular upper and lower profile. The surfaces of the central shaft and the keel show no ornament (although very little of the shaft surface is preserved intact), whereas the wing has a convex exposed surface with longitudinal ridges and grooves (six visible near the fractured end). Paratype IGR 132671a (Fig. 2D-F) is the distal half of a spine, lacking the tip, 7 cm long and 2 cm at its widest, with a cross-sectional profile similar to that of IGR 132670. The exposed surface is smooth, including the wing, although the surface appears abraded with possible traces of longitudinal grooves on the slightly convex wing. The counterpart IGR 132671b preserves the distal end of the spine that is missing in the part, and shows impressions of longitudinal ridging on the wing (Fig. 2G: r). The two specimens IGR 132670 and 132671 are morphotype one spines.

Two of the morphotype two spine forms are identified. Paratype spine ML-PAL-03006 (Fig. 2H-J) is 10 cm long, lacking the distal and proximal ends. On this spine, a short impression of the side originally buried in the matrix (at the proximal end of the preserved spine fragment) shows that the keel on this spine has longitudinal ridges (Fig. 2H: r). ML-PAL-29822 (Fig. 2K) is 12 cm long and almost complete, only missing the distal tip. The keel is partly preserved on the distal half of the spine, and is ornamented with at least three longitudinal ridges. The exposed surfaces of the wing and shaft are smooth.

Discussion and comparison. It seems probable that surfaces exposed in the two type specimens IGR 132670 and 132671a are the upper and lower surfaces (although we cannot confidently nominate which is which), as the narrowest and shallowest side of the shaft is exposed in IGR 132670 and the widest and deepest in IGR 132671a. If this is the case, it is also probable that only one or other of the upper and lower surfaces of the wing or keel is ridged. We thus recognize two morphotypes (Fig. 3A, B), one with a ridged wing (IGR 132670, 132671) and the other with a ridged keel (ML-PAL-03006, 29822). We propose that the two morphotypes were contiguous in life, as a pair of spines articulating via the ridging on the wing and keel respectively (Fig. 3C). It is not possible to distinguish the upper and lower (i.e. dorsal and ventral, in life) surfaces, as neither surface of the central shaft is worn smoother than the other (cf. Burrow et al. 2010, for a discussion of this feature in other species).

Bézier (1913) provided a table comparing the lengths and widths of the Czech Republic and western France spines, which showed that the dimensions of the latter compared closely with those of Machaeracanthus bohemicus. The Armorican spines resemble most closely some of the Pragian syntype spines assigned to M. bohemicus (e.g. Barrande, 1872, pl. 28, figs. 8–10, 16, 20, 24; Burrow et al., 2010, fig. 8B; Fig. 3D here) rather than the Emsian lectotype and syntype spines of M. bohemicus (Barrande, 1872, pl. 28, figs. 4, 6). The Armorican and Pragian Bohemian spines have one convex, longitudinally ridged surface on either the inner (concave edged) wing or the outer (convex edged) keel. This feature is also visible on other Pragian spines previously assigned to M. bohemicus from the Taunus Quartzite, Germany (Kayser, 1878, pl. 35, figs. 12, 12a, 12b; Fig. 3E here; Schmidt, 1933, fig. 6a; Fig. 3F here). This feature is not found in any other Machaeracanthus species, including M. goujeti, which is distinguished from other species in having adult spines with narrow equal width keel and wing, and a wide shaft with dense longitudinal striations on the upper surface on one of the two morphotypes. The other Machaeracanthus species from the Taunus Quartzite, M. kayseri Kegel, 1913, is easily distinguishable from M. bezieri nov. sp. by the robust sharp- crested longitudinal ribbing (Kegel, 1913) on the lower surface of the axial shaft (Fig. 3G).

As noted by Burrow et al. (2010, p. 77), “the stratigraphic range of material in the National Museum in Prague which has been assigned to M. bohemicus is Lochkovian to Eifelian”, and several authors (Perner, 1918; Gross, 1959; Burrow et al., 2010) have indicated that there is more than one species in the Barrandian sequence. Based on our observations of the spines from western France described here and Barrande's material, we have assigned the Pragian examples with the distinctive longitudinal ridges on the wing or keel, from France, the Czech Republic, and Germany, to the new species M. bezieri.

Machaeracanthus sp.
(Fig. 2 L-M)

Material examined. From Saint-Germain-le-Fouilloux: spine fragment IGR 10869. From Changé: one spine fragment IGR 144560 (Fig. 2L) found by DG during the construction of the railroad (LGV) in 2014. From Vern-d'Anjou (now Erdre- en-Anjou): spine fragment UCO-71631, Collection Davy, Université Catholique de l'Ouest, Angers. From Saint-Malo Quarry, Saint-Barthélémy-d'Anjou near Angers: spine fragment UCO-71929 (Fig. 2M), Collection Jouitteau, Université Catholique de l'Ouest, Angers.

Description. Specimen IGR 10869 (not figured) from Saint- Germain-le-Fouilloux is probably the proximal end of a spine with the wider side of the shaft exposed, but it is very poorly preserved with only a small area of hard tissue present. The spine fragment from Changé (Fig. 2L; IGR 144560) is very poorly preserved; it retains the characteristic cross-sectional shape of Machaeracanthus, but shows no specific characters. The surface of the hollow central pulp cavity shows narrow parallel ridges near the fracture surface; these are probably exposed vascular canals of the osteodentine. The specimens from Angers (UCO- 71629, 71631; Fig. 2M) are short midspine fragments which are poorly preserved. They do not show evidence of ridging on the side expansions (wing, keel), possibly because these are on the sides buried in the matrix. They are thus only assigned to Mach- aeracanthus sp. rather than to M. bezieri.

IV. — Taxonomic Discussion

Unfortunately we have been unable to examine the well preserved Machaeracanthus spine from Sablé-sur-Sarthe mentioned by Goujet (1993) and Janvier (1996), which should be in the collection of the Muséum National d'Histoire Naturelle, Paris, as it appears to have been misplaced and no images of it are available.

It is possible that the new species M. bezieri is synonymous with Rouault's (1858) species, but as his original descriptions make no mention of the distinguishing feature we recognize in M. bezieri nov. sp., and his specimens were not figured and are still lost, M. larteti and M. archiaci must remain nomina vana. We question whether the scales from western France, originally assigned by Goujet (1976, 1980) to Machaeracanthus sp., can reliably be assigned to M. goujeti as proposed by Botella et al. (2012) given that we have not seen spines like those of the latter species from western France. Although it is only known from limited material, the Pragian M. bezieri nov. sp. is characterised by the distinctive longitudinal ridging on the inner wing or outer keel of the spines, a feature that is not present on the type material of any other Machaeracanthus species. For this reason, we feel justified in assigning both the morphotypes showing complementary ridging on wing and keel to the same biological species rather than to two form species, despite not finding them as associated pairs. All of the type specimens are, however, from the same quarry. It now seems highly likely that most, if not all, Machaeracanthus fish had paired pairs of spines, as two morphotypes have now been identified in M. hunsrueckianum, M. longaevus, M. peracutus, M. sulcatus, and M. polonicus (Burrow et al. 2010; Burrow & Srek, 2017). The geographical distribution of M. bezieri includes France, Germany and the Czech Republic.

Acknowledgments. — We wish to thank G. Beaulieu (UCO), D. Goujet and P. Janvier (MNHN), Jérôme Tréguier and Valentin Prugneaux (ML), Fritz Geller-Grimm (Wiesbaden Natural History Museum) for images and information on specimens in their collections, and A. Blieck (Université des Sciences et Technologies de Lille) for instigating our collaboration. We also thank T. Wright and coauthors for allowing us to use the base map for Figure 1B. CJB acknowledges the support of the Queensland Museum. We thank Hector Botella (University of Valencia, Spain) and Alexander Ivanov (St Petersburg University, Russia) for their reviews of our manuscript.