Osteichthyes (Huxley, 1880).

Sarcopterygii (Romer, 1955).

Dipnomorpha (Ahlberg, 1991).

Porolepiforms (Berg, 1937).

Family Holoptychiidae (Owen, 1860).

Fig. 2 and Fig. 3A, B.

Remarks. We do not attempt to assign this tooth material to any existing species or genus of holoptychiid. This decision was made on the basis of the limited number of holoptychiid materials retrieved from the field, the incompleteness of the dental material, and the difficulty of identifying species or genera on the basis of incompletely-preserved, isolated remains.

Overall morphology—The tooth (UN-DG-PALV86) measures 2.1 cm in length (from the lowermost point of the exposed portion to the apex) and 0.7 cm in diameter at its base (Fig. 2A). It is slightly curved from the base to the apex. Although the tooth is isolated, it is most plausible that it was curved inwards (lingual curvature), a condition typical for piscine sarcopterygians and early tetrapods (Ahlberg and Clack, 1998 and Jarvik, 1972). There is no evidence of a reverse-curvature at the apex, a condition seen in rhizodontids (Jeffery, 2003), onychodontids (Andrews et al., 2006) and in the parasymphysial region of certain porolepiforms (e.g., Hamodus in Bystrow, 1939). The base of the tooth is not completely preserved, which is a frequent occurrence for isolated teeth, and so it is not possible to ascertain whether the tooth was also curved at its base. The presence of a swollen base cannot be confirmed either.

The external surface is gently striated by very thin, parallel striae. The striations extend continuously from the base to the apex, and are distributed along the entire surface of the tooth. In the uppermost part of the apex, the striae are more difficult to see but are still present. The striation pattern consists of both deep and more superficial plications, with one superficial thin stria generally located between two deep striae (Fig. 2B). There are about 20 deep striae, and accordingly, about another 20 superficial thin striae.

The tooth is rounded in cross-section, especially at its base, and becomes slightly flattened towards the apex. However, no carinae (or cutting edges) along the mesial and distal margins are present, contrary to the condition seen in the large teeth of holoptychiids such as Laccognathus or Glyptolepis (Bystrow, 1939 and Jarvik, 1972; pers. obs.). Due to its size, the tooth can be described as a fang, although it is not possible to determine whether it was located in the upper or lower jaw.

Histology—The tooth has been greatly re-crystallized (re-mineralized) during fossilization. Osteo- and orthodentine have been replaced by mineral elements (iron oxide) from the surrounding rock matrix and it is very difficult to identify the boundaries between them (Fig. 3A, B). However, the general branching pattern of the orthodentine matches closely the ‘fire-like’ branching of the typical dendrodont plicidentine of porolepiforms (Fig. 3C, D). The pulp cavity was probably filled with osteodentine and occupies about one-fourth (Fig. 3A) to one-half (Fig. 3B) of the inner surface of the tooth.

The inward dentine folds are regular and numerous (about 40) and penetrate deeply toward the pulp cavity. Beneath the enamel layer and extending into the folds (i.e., globular zone, see Warren and Turner, 2006), the dentine tubules show a radiating pattern. The outer enamel rim has completely disappeared (Fig. 3A, B). The external plications seen on the external surface of the elastomer cast appear to match those of the missing enamel layer and are directly correlated with the dentine folds, similar to rhizodontids and “osteolepiforms” (Jeffery, 2003). Because of the poor preservation issues and the weathering of the base of the tooth, it is not possible to determine whether the bone of attachment penetrates between the dentine folds or not.

Two scales (UN-DG-PALV50, 51; Janvier and Villarroel, 2000) previously figured by Janvier and Villarroel (1998), (fig. 11; 2000, pl.5, figs. 9, 10) and Janvier and Maisey (2010), (fig. 10E, F) are regarded as the best evidence for the occurrence of Holoptychius in the Cuche Formation (Fig. 4). They are incompletely preserved due to a fracture (Fig. 4A) or an incomplete exposure (Fig. 4B). They are about 3 cm long, rounded to elongate in shape, and devoid of cosmine. The external ornamentation consists of a series of spoon-shaped tubercles arranged in radiating rows, mostly located on the overlapped surface, and extending posteriorly through the boundary between the overlapped and exposed surfaces. These tubercles extend posteriorly into the exposed surface and anastomose, forming a series of antero-posteriorly elongate ridges that can vary in thickness.

Only the external surface of the scales is preserved and so it is not possible to observe the internal surface or verify whether they possess an inner boss. The absence of internal surface ornamentation (i.e., articular ridges or bosses) in rounded, cosmine-free scales occurs in holoptychiids, derived dipnoans, actinistians, and onychodontids (Mondéjar-Fernández and Clément, 2012 and Ørvig, 1957), whereas the presence of a wedge-shaped inner boss is characteristic of the convergently acquired rounded scales of tetrapodomorphs (such as rhizodontids and tristichopterid ‘osteolepiforms’) (Jarvik, 1980).

The diversity of the tooth structure in vertebrates, and especially in sarcopterygians, has been well studied and documented since the pioneering works of Bystrow, 1938 and Bystrow, 1939. Bystrow's observations on the teeth of ‘crossopterygians’ were summarized and emphasized by Schultze, 1969 and Schultze, 1970. Schultze (1969) used the term plicidentine, first coined by Owen (1841), which was then formalised by Tomes (1878), as “a tissue with true dentinal tubules, which is derived from the calcification of a pulp, the odontoblast-carrying surface of which has been rendered complicated by foldings of its surface” (Tomes in Warren and Turner, 2006: 125). The structure and arrangement of the plicidentine has proven to be a key character for the study of tooth histological diversity in osteichthyans (Janvier, 1996).

The presence of plicidentine around the pulp cavity at the base of a tooth was considered a diagnostic feature of rhipidistians; i.e. sarcopterygians crownward to onychodontids and actinistians such as dipnomorphs and tetrapodomorphs (Vorobyeva, 1977). However, folded dentine has also been described in the stem-sarcopterygian Psarolepis from the Early Devonian of China (Yu, 1998 and Zhu et al., 1999). Therefore, the presence of dentine folding can no longer be considered a synapomorphy of rhipidistians (Cloutier and Ahlberg, 1996) but rather the plesiomorphic state for sarcopterygians.

Onychodontids and actinistians have simple, non-plicated teeth, which could be considered a secondary and probably convergent loss, whereas dipnomorphs and tetrapodomorphs show different dentine-folding morphotypes. Based on the degree and regularity of the dentine folding, Schultze, 1969 and Schultze, 1970 identified three principal different plicidentine morphologies: •

Polyplocodont: the pulp cavity is free from osteodentine; the orthodentine is folded simply and irregularly with branches of first or second degree; and the bone of attachment extends between the folds (the labyrinthodont folding of early tetrapods is a variant of the polyplocodont one in which the branches of the folds are apparently lost and the bone of attachment does not penetrate into the folds);

The different types of dentine folding have been usually regarded as organisational grades without a strong phylogenetic signal (Jeffery, 2003 and Warren and Turner, 2006). Polyplocodont plicidentine is primitively present in piscine sarcopterygians and early tetrapods; and is thus phylogenetically uninformative. However, the labyrinthodont folding (present in several Carboniferous tetrapods) and the eusthenodont folding (characteristic of certain derived tristichopterids such as Eusthenodon and Platycephalichthys) can be considered as autapomorphic traits in these taxa. By contrast, dendrodont folding seems to be restricted to the Porolepiforms, and constitutes a well-established synapomorphy of the clade (Panchen and Smithson, 1987 and Schultze, 1969). Vorobyeva (1977) cautioned that the dendrodont teeth could also have evolved in parallel among sarcopterygians from a primitive polyplocodont folding, by analogy with the eusthenodont condition. However, dendrodont plicidentine has not been found in any other sarcopterygian group so far. This suggests that the dendrodont plicidentine may be a reliable character of the Porolepiforms.

Tooth—The absence of a curved apex clearly distinguishes our isolated tooth from the ‘rhizodontid-like’ teeth already described in the Cuche Formation (Janvier and Villarroel, 2000), which are tentatively attributed to Strepsodus (but see Jeffery, 2003 for an alternative interpretation). Further, the rhizodontid teeth with folded enamel have striation patterns that differ from our specimen. Rhizodus, for example, has deep plications that are restricted to the base of the crown (Jeffery, 2003, fig. 6). However, other rhizodontids, such as Strepsodus, have striations that extend across the entire surface of the crown, but which do not follow the contour of the tooth (Jeffery, 2003, fig. 11c–d). In our specimen, the striae run parallel to each other, merge at the apex, and follow the contour of the tooth. This pattern is identical to that of porolepiforms, such as Glyptolepis, Hamodus and Holoptychius (Bystrow, 1939).

In porolepiforms, the size and recurvature of the teeth is variable, not only between parasymphysial teeth and coronoid fangs, but also among coronoid fangs, depending on their location in the jaw (Ahlberg, 1991, Ahlberg, 1992a and Jarvik, 1972). In the lower jaw, the parasymphysial fangs are slightly recurved at their apex but do not exceed in height the coronoid fangs, except in the case of Duffichthys (Ahlberg, 1992a).

In our specimen, the absence of recurvature at the apex, the uniform striation pattern, and the absence of well-developed carinate margins are consistent with the dental characters in holoptychiid porolepiforms such as Holoptychius, Glyptolepis and Hamodus (Bystrow, 1939). However, unlike Hamodus, UN-DG-PALV86 lacks a hooked apex and sigmoid aspect. The absence of carinate margins is common in several sarcopterygian groups (e.g., porolepiforms and “osteolepiforms”) and it is considered to be another example of tooth variability related to their position along the jaws, especially in the lower jaw (Ahlberg, 1992a and Jarvik, 1972).

When histology is considered, Schultze, 1969 and Schultze, 1970 noticed that in the “porolepidid” Porolepis the osteodentine of the pulp cavity and the orthodentine are separated by a clear boundary. In holoptychiids, both ortho- and osteodentine grow closely connected, and no trace of a well-delimited boundary between these tissues is visible. Such a boundary is not visible in any of our histological sections (Fig. 2), thus supporting its assignation to the Holoptychiidae. However, we should note that this boundary between ortho- and osteodentine is also absent in the “porolepidid” Heimenia from the Early-Middle Devonian of Spitsbergen (pers. obs.), thus adding a new feature suggesting the intermediate condition of Heimenia between Porolepis and holoptychiids (Mondéjar-Fernández and Clément, 2012).

Scales—As described above, the external exposed surface of our scale material bears a series of tubercles and stout ridges. Among porolepiforms, the presence of such ridges has been identified in holoptychiids such as Holoptychius, Quebecius and Glyptolepis (Cloutier and Schultze, 1996, Ørvig, 1957 and Schultze and Arsenault, 1987). The ridges in Glyptolepis and Quebecius are made of dentine and are narrower and more numerous than in Holoptychius. This primitive condition is present in several other sarcopterygian groups (e.g., actinistians, holodipterid dipnoans, and to a lesser extent onychodontids). Laccognathus shows a unique ornamentation composed of rounded dentine tubercles or small ridges capped with enamel (Downs et al., 2011 and Ørvig, 1957), probably derived from narrow dentine-made ridges as seen in Glyptolepis.

The condition of Holoptychius is also derived relative to the condition seen in Glyptolepis; the ridges are solely made of bone and look thicker than in any other holoptychiid. However, their morphology and ornamentation varies throughout ontogeny, even at different points along the body (Ørvig, 1957). For example, the scales of the ventral region can exhibit stout bony tubercles that are arranged in rows, whereas dorsal and flank scales display antero-posteriorly oriented bony ridges (Cloutier and Schultze, 1996). Moreover, the numerous nominal species of Holoptychius have been mostly diagnosed on the basis of disarticulated material and isolated scales (Cloutier and Schultze, 1996 and Downs et al., 2013). The assignation of isolated scales to different species of Holoptychius can thus be considered suspect and poorly informative when such assignations rely exclusively on external descriptions of scale remains (Downs et al., 2013 and Miller and Brazeau, 2007).

Nonetheless, among holoptychiids, the combination of coarse bony ridges in the exposed surface associated with a fan of spoon-shaped tubercles in the overlapped surface diagnoses Holoptychius. The occurrence of these scale characters in our isolated scale material supports the previous attribution of these scales to Holoptychius sp. indet. (Janvier and Villarroel, 2000).

Reconstructing the biogeographic history of Devonian fish is made difficult by the patchy distribution of data (Ahlberg, 1992b). Much data from Africa, the Middle East, and South America remain largely unknown, but these regions will most certainly reveal new taxa and unexpected biogeographic patterns in the coming years. Nonetheless, the discovery and description of new fossil sites from Venezuela, Colombia, Brazil, Bolivia, and the Falkland Islands have greatly improved our knowledge of vertebrate diversity in the Devonian of South America (Janvier and Maisey, 2010).

Placoderm- and osteichthyan-dominated assemblages from the Devonian ‘intertropical belt’ characterize the Eifelian–Frasnian vertebrate assemblages from Venezuela and Colombia (Janvier, 2007 and Janvier and Maisey, 2010). The Floresta Formation and the overlying Cuche Formation (Late Frasnian) of Colombia are thought to represent a shallow, low energy, marine influenced depositional environment (Janvier and Villarroel, 1998 and Janvier and Villarroel, 2000). This is consistent with the classical environments associated with Euramerican holoptychiid porolepiforms (Cloutier and Schultze, 1996).

Porolepiforms are still considered a rare component of Gondwanan faunas, despite their reported presence in Iran (Lelièvre et al., 1993 and Schultze, 1973), Australia (Johanson and Ritchie, 2000, Johanson et al., 2013 and Young et al., 2010), and Antarctica (Young et al., 1992). Porolepiforms, especially holoptychiids, have been well studied and more commonly found in Euramerican localities, from North America to Russia (Fig. 5). This lead to the assumption that holoptychiids might be considered Euramerican endemics in the Middle Devonian with a worldwide distribution of certain genera, such as Holoptychius, by the Late Devonian. However, this misinterpretation is likely explained by the poor sampling and preservation of Gondwanan material.

Holoptychiids occur in Euramerica by the late Lower Devonian (Emsian) (Nasogaluakus in Schultze, 2000), with Holoptychius first undisputed occurrences by the Frasnian (P. Ahlberg pers. comm.). Holoptychius presents a more ubiquitous mode of life and a greater potential of dispersion than any other sarcopterygian taxon (Cloutier and Schultze, 1996). This potential for a great dispersal ability appears to have favored its expansion between Euramerica and Gondwana, probably through both eastern and western routes (Middle East and South America, respectively), and likely as soon as the Frasnian. Similarly, “porolepidids” like Porolepis are known from both marine and freshwater deposits from Euramerica, and recently from eastern Gondwana (Johanson et al., 2013). By contrast, holoptychiids appear to have been somehow restricted to freshwater and possibly estuarine or lagoon environments (Ahlberg, 1992b). However, both families are widely distributed throughout the Devonian, which suggests a great adaptability and dispersal potential of porolepiforms as a whole.

Holoptychius has been reported from the Frasnian of Iran (northern margin of Gondwana) (Lelièvre et al., 1993) and is well represented in numerous sites from Euramerica (e.g., Cloutier and Schultze, 1996). The presence of Holoptychius in the Frasnian of Colombia thus represents one of the earliest occurrences of the genus, both in Euramerica and Gondwana. South America may have functioned as a bridge between northern Gondwana and southern Euramerica (Fig. 5), which allowed fish faunas to disperse more easily from one continent to the other. This suggests that the earliest range of Holoptychius extends from the southern margin of Euramerica to the northern margin of Gondwana. The rich Middle Devonian holoptychiid record of Euramerica might point to a Euramerican origin of Holoptychius, but this remains to be tested.

Finally, the co-occurrence of Holoptychius and Asterolepis in the Cuche Formation of Colombia suggests the dispersion of fish faunas between Euramerica and Gondwana through South America by the beginning of the Late Devonian.