The history of studies on Palaeospondylus [1–10] constitutes a heuristic search for homologous body parts, a task that we suggest may be completed with the present study, in which it was finally compared with the hagfish embryo. Previous studies comparing Palaeospondylus with hagfishes relied solely on adult morphology [8]. Importantly, it has recently been shown that adult hagfish species possess vestigial vertebral elements [21]. The presence of a vertebral column composed of cylindrical vertebrae in Palaeospondylus (Additional file 1: Figure S1), which simply sheathed the notochord [5, 8], thus does not necessarily preclude its affinity with the modern hagfish. Our comparative analysis of Palaeospondylus and the hagfish embryo, revealing previously unrecognized topographical relationships between skeletal elements specific to these taxa, was made possible by the introduction of modern techniques to the study of hagfish embryonic development in recent years [20–23, 29].
According to comparative developmental biology two mutually exclusive systems of craniofacial anatomical configuration are present in vertebrates [22, 24]. In both, during embryonic development, the rostral portion of the cephalic crest-derived ectomesenchyme (trigeminal crest cells) initially follows a common distribution pattern, comprising the pre- and postoptic crest cells and mandibular arch crest cells (up to early pharyngula stage) [30]. However, the difference between the cyclostomes and crown gnathostomes later becomes conspicuous (Additional file 3: Table S1) in association with differences in placode distribution [22, 31]. In cyclostomes, the craniofacial morphology develops from the mandibular arch (MA), post-hypophyseal process (PHP), and anterior nasal process (ANP) [24, 32]. In crown gnathostomes, the embryonic head consists of lateral and medial nasal prominences (LNP, MNP), postoptic ectomesenchyme (PO), maxillary and mandibular processes (MX, MN) [24, 32]. Given that the craniofacial pattern of stem gnathostomes (osteostracans and galeaspids) conforms with the cyclostome pattern [22], the cyclostome pattern (involving MA, PHP, and ANP) thus represents the ancestral developmental pattern for the total-group vertebrates, whereas the crown gnathostome pattern (involving LNP, MNP, PO, MX, and MN) represents a derived condition. However, the craniofacial development of crown gnathostomes does not, even transiently, recapitulate the cyclostome pattern at any developmental stage [22, 24].
The Palaeospondylus cranium is congruent with the cyclostome craniofacial pattern described above (Fig. 3). In this comparison, the nasal capsule, subnasal rostral cranium (palatine bar, and rostral part of the dorsal longitudinal bar and trabecula), and dental and lingual plates of the Palaeospondylus are reconstructed to develop from the ANP, PHP and MA of the cyclostome pattern, respectively (Fig. 3c,d; Additional file 4: Table S2). The inferred positions of the nasal epithelium and adenohypophysis in Palaeospondylus are consistent with this scheme (Fig. 3d). Contrastingly the Palaeospondylus cranium cannot be derived from the crown gnathostome pattern, which involves a dorsoventrally bifurcated mandibular arch, separate paired nostrils, and adenohypophysis. A previous hypothesis that suggested a Devonian lungfish affinity of Palaeospondylus [10] was flawed in that it did not account for the neurocranial element developed medially to MX.
More detailed homology relationship of skeletal elements can be established between Palaeospondylus and the hagfish embryo at specific developmental stages, but not with the lamprey skeletal elements at any developmental stage. The cage-shaped nasal capsule, two separated longitudinal bars (the dorsal longitudinal bar and trabecula), velar bar, and large lingual plates (Fig. 3a, b) are seen exclusively in the hagfish embryo and Palaeospondylus, and the topographical arrangement of these homologous elements is consistent. Based on these synapomorphies, it is parsimonious that Palaeospondylus was related to hagfishes.
A comparison of the rostral portion of the neurocranium among vertebrates further underscores the congruence between Palaeospondylus and the hagfish. There is a morphological disparity among the crown gnathostomes, lamprey and hagfish, which results from the difference in ‘trabeculae’ within embryonic heads; in fact, the term ‘trabecula’ does not designate the same body part among the crown gnathostomes, lamprey and hagfish. The neurocranium of the crown gnathostomes is heterogeneous in cell lineage [30, 33]; the rostral part, or trabecula, develops from the cephalic neural crest cells [34–38], while the caudal part, or parachordals, from the mesoderm [33]. In the cyclostomes, a large part of the rostral portion of the neurocranium develops from mesodermal cells [23, 39–43], and only the nasal capsule and oropharyngeal skeleton develop from the cephalic (ANP and PHP, respectively) crest cells [24, 44]. In the lamprey, the trabecula (‘lamprey trabecula’) consists of a mesodermal neurocranial wall secondarily elongated rostrally [39–43]. In the hagfish, the corresponding prechondrogenic precursor later splits dorsoventrally into two bars: the dorsal longitudinal bar and ‘hagfish trabecula’ [23, 24]. Of these, the configuration of rostral elements of Palaeospondylus neurocranium agrees only with the hagfish-type, as the dorsal longitudinal bar and ‘hagfish trabecula’ occupy the corresponding position in this taxon.
One previous study argued against the cyclostome affinity of Palaeospondylus, citing the presence of a paired fin located caudally distant from the cranium [9]. The extant hagfishes, on the other hand, possess a cartilage at the origin of the lingual muscle (the ‘cyclostome tongue’ [45]), or the perpendicular muscle cartilage, which is located far caudal to the cranium [46] (Fig. 4). The perpendicular muscle cartilage of the extant hagfish is not decay-prone [47], and thus can be preserved separately from the cranial skeleton in fossils. Here, we suggest that the ‘paired fin’ of Palaeospondylus may in fact represent such a skeletal element at the origin of the lingual muscle.
Unique among known cyclostomes, Palaeospondylus possessed distinct features, including the biomineralization of cartilages and the well-developed vertebral column (notochordal centrum). These features, however, may have evolved from cyclostome-like conditions. The calcified cartilage of Palaeospondylus lacked perichondral ossification [48], which is specific to the gnathostomes [38], and it is conceivable that the calcified cartilage of Palaeospondylus evolved either through loss of perichondral ossification in gnathostomes, or through acquisition of biomineralization in cyclostomes. The latter possibility appears more plausible, as the hypertrophied cell lacunae in the cartilages of Palaeospondylus [11, 48] are reminiscent of cyclostome cartilages in their thin layers of extracellular matrix [21, 49]. A recent study also suggested that the vertebral element is synapomorphic to the vertebrates [21, 50], and the cylindrical vertebral column in Palaeospondylus (Additional file 1: Figure S1) may have evolved secondarily from arcualia in basal cyclostomes, through the invasion of cartilaginous cells into the fibrous sheath within the elastica externa [51]. The absence of arcualia in Palaeospondylus fossils does not preclude this scenario, since calcification often occurs differentially among vertebral elements, as seen in elasmobranchs [52].
There are nonetheless minor differences between Palaeospondylus and the known hagfish species, including the absence of some skeletal elements in Palaeospondylus. We suggest that this is likely due to a taphonomic bias. In the extant cyclostomes, there are two types of cartilage, hard and soft, which differ in the amount of extracellular matrix [49, 53, 54]. Soft cartilage is less resistant to decay than hard cartilage [47]. A similar heterogeneity of cartilage composition may have been present in Palaeospondylus, given that the branchial skeleton has not been identified in Palaeospondylus. The absence of branchial skeleton in Palaeospondylus fossils may indicate that cartilages of the branchial basket were not calcified, as in extant cyclostomes, for the functional reason that the branchial basket changes shape during ventilation. Among fossil cyclostomes, the Late Devonian putative stem-lamprey Euphanerops is a unique taxon that possessed calcified branchial basket [55]. The gill basket of Euphanerops was, however, unique in extending caudally across the half of the trunk, suggesting a functional requirement different from that in other cyclostomes.
In the extant hagfish chondrocranium, the rostralmost skeletal element made of hard cartilage is the subnasal cartilage (Fig. 2e). The subnasal cartilage is expected to be found rostral to the nasal capsule [47], but is apparently missing in Palaeospondylus. Since the proportion of the Palaeospondylus cranium resembles the embryonic, rather than the adult, cranium in known hagfish species (Fig. 2e), the absence of overt subnasal cartilage in Palaeospondylus fossils may reflect a very short snout containing a delicate subnasal cartilage. Cartilages supporting the nasal duct and tentacles in the extant hagfish are composed of soft cartilage [54], thus they are expected to decay rapidly.
Given the potential extensive taphonomic bias and the lack of data about soft tissue anatomy unlike circumstances of other species [56–58], it is difficult to build a character matrix to conduct a cladistic analysis of Palaeospondylus. Nevertheless, the above comparison strongly suggests that Palaeospondylus and the hagfish share a cranial skeletal configuration that is distinguishable from those of the lamprey and crown gnathostomes. On the other hand, some features seen in adult individuals of the known hagfish species are less conspicuous in Palaeospondylus. In particular, in the extant hagfishes, as well as in the Late Carboniferous hagfish Myxinikela siroka [59], the position of the nasal capsule, which initially develops at the rostral end of the cranium becomes relatively caudal in the cranium of adult individuals, whereas in Palaeospondylus, the nasal capsule remained at the rostral end. Based on this synapomorphy between extant hagfishes and Myxinikela, we suggest that the phylogenetic position of Palaeospondylus is best explained as a stem hagfish lineage basal to Myxinikela (Fig. 5).
The hagfish affinity of Palaeospondylus is important for the time of divergence between hagfish and lamprey lineages. The oldest reliable fossil record of the divergence time is the fossil lamprey Priscomyzon riniensis from the Upper Devonian (Famennian: 372–359 Ma in [60]) [61, 62]. An estimate using nucleotide and amino acid sequences, however, dated the divergence time to 470–390 Ma [63] (by the Middle Devonian), leaving a gap between the molecular-based estimate and the fossil record. Our classification of Palaeospondylus into the hagfishes pushes the oldest fossil record of the divergence time back to the Middle Devonian (393–383 Ma, Fig. 5), filling the gap between the molecular estimate and the fossil record, as well as depicting a hagfish species before the end-Devonian mass extinction that wiped out many vertebrate groups, including placoderms [64, 65].
Early cyclostomes may thus have been more morphologically diverse than previously recognized. At present, the early evolution of the cyclostomes has been less clear than that of the gnathostomes [62, 65]. Further analysis of Palaeospondylus, as well as other putative cyclostomes in the fossil record [55, 58, 66], may shed new light on the evolution of this lineage.