Ancient vertebrate dermal armor evolved from trunk neural crest

Significance

The body of early vertebrates was covered from the head to tail with extensive dermal armor, comprised of dentin and bone. There has long been controversy over whether this armor arose from the neural crest, mesoderm, or both. Since odontoblasts (dentin-producing cells) are exclusively neural crest derived, we probed the developmental origin of the bony component of the dermal armor in the sterlet sturgeon, an early branching lineage of ray-finned fishes. Here, we show that trunk neural crest of the sterlet gives rise to osteoblasts, producing the bone of the scutes. Together, our results support a primitive skeletogenic role for the neural crest along the entire body axis, that was later progressively restricted to the cranial region during vertebrate evolution.

Abstract

Bone is an evolutionary novelty of vertebrates, likely to have first emerged as part of ancestral dermal armor that consisted of osteogenic and odontogenic components. Whether these early vertebrate structures arose from mesoderm or neural crest cells has been a matter of considerable debate. To examine the developmental origin of the bony part of the dermal armor, we have performed in vivo lineage tracing in the sterlet sturgeon, a representative of nonteleost ray-finned fish that has retained an extensive postcranial dermal skeleton. The results definitively show that sterlet trunk neural crest cells give rise to osteoblasts of the scutes. Transcriptional profiling further reveals neural crest gene signature in sterlet scutes as well as bichir scales. Finally, histological and microCT analyses of ray-finned fish dermal armor show that their scales and scutes are formed by bone, dentin, and hypermineralized covering tissues, in various combinations, that resemble those of the first armored vertebrates. Taken together, our results support a primitive skeletogenic role for the neural crest along the entire body axis, that was later progressively restricted to the cranial region during vertebrate evolution. Thus, the neural crest was a crucial evolutionary innovation driving the origin and diversification of dermal armor along the entire body axis.

The evolutionary success of vertebrates is intimately associated with acquisition of the neural crest (NC), a stem-cell population with remarkable migratory potential and the ability to form a vast array of derivatives (1, 2). NC cells arise in the developing embryo within the forming central nervous system, then migrate away throughout the periphery and form diverse cell types, ranging from ectomesenchymal elements of the craniofacial skeleton to most of the peripheral nervous system (3). However, there are differences in NC cells along the body axis. For example, in birds and mammals, only the cranial NC has the ability to form ectomesenchymal cartilage and bone in vivo. While trunk NC cells normally lack this ability, they can be reprogrammed to do so using cranial crest transcription factors (4) or other experimental manipulations (5, 6). Moreover, single-cell analysis of murine NC has revealed promesenchymal bias in trunk NC cells such that overexpression of Twist1 can shift them toward ectomesenchymal fates (7).

The entire body of most nontetrapod vertebrates is covered with various dermal mineralized structures (8). This extensive dermal armor consists of either bone (osteogenic unit), dentin (odontogenic unit), or both tissues. During the course of vertebrate evolution, differential combinations of loss and/or retention of dentinous and bony components led to an enormous diversity of scales, scutes, and denticles in vertebrate lineages (9, 10). The head-to-tail distribution of the dermal armor has prompted discussion regarding the intriguing possibility that skeletogenic potential may have been present in NC cells along the entire body axis of gnathostome ancestors (10). During the transition of tetrapods to land, the odontogenic and osteogenic dermal armor units were gradually reduced, including the loss of dental ornament and the replacement of the thick bony base by a weakly mineralized plywood-like structure in the sarcopterygian elasmoid scales (11) as in teleosts. Thus, the skeletogenic ability of the NC likely became gradually restricted to the head, by spatiotemporal changes in some gene regulatory network (GRN) components (12). To date, the developmental evidence regarding the skeletogenic ability of the trunk NC remains sparse. Lineage tracing in the little skate has suggested that trunk NC cells have the ability to give rise to dermal denticles (13), which are dentinous structures that cover the entire body. This and other evidence strongly suggest that the odontoblast program was ancestrally associated with the NC along the entire anterior-to-posterior (AP) body axis (14). However, whether trunk NC has the potential to form osteogenic components lacks hard evidence.

While the modern chondrichthyan exoskeleton only retains odontogenic components, the dermal armor of primitive vertebrates typically was either comprised of dentin units that rested on dermal bones, or solely consisted of osteogenic components as in some extinct jawless fishes (15, 16). In zebrafish and medaka scales (17–19), the plywood-like osteogenic layer has been shown to be derived from mesoderm, refuting earlier studies claiming an NC origin (20, 21). This has led to the suggestion that the odontogenic layer of the dermal exoskeleton comes from trunk NC, while the osteogenic layer arises from paraxial mesoderm (13). A complication is that teleost elasmoid scales are highly derived relative to the ancestral condition, such that zebrafish or medaka may not be appropriate model organisms for evaluating the broader evolutionary pattern of ectomesenchymal potential of the trunk NC.

To definitively examine whether the osteogenic component of the ancient dermal armor was trunk NC- or mesoderm-derived, here, we perform histological analysis on the dermal armor of the major phylogenetic lineages of ray-finned fishes (cladistians: bichir; chondrosteans: sterlet; holosteans: gar; and teleosts: armored catfish). Among them, sturgeons are a deep branch of ray-finned fishes, appearing in the fossil record from the Cretaceous (22, 23). They have a characteristic morphology with a long rostrum, reduced ossification of the endoskeleton, and an extensive dermal exoskeleton consisting of several types of scutes (24, 25). While sturgeon scutes likely evolved from ganoid scales (8), they have lost both ganoin (a multilayered ename) (26) and dentin. This makes them particularly useful for investigating potential NC involvement in the formation of trunk dermal bones. Therefore, we perform lineage tracing and RNA-seq analyses of the sterlet sturgeon (Acipenser ruthenus Linnaeus, 1758) to reveal whether trunk NC plays a role in the development of the bony part of the dermal armor.

Results and Discussion

Dermal Armor of Extant Ray-Finned Fishes.

The ray-finned fishes are remarkably diversified, representing about half of all extant vertebrates. Because several lineages retain dermal armor with clearly identifiable dentinous and osteogenic components (8, 15), ray-finned fishes are poised to provide important insights into the evolutionary origin of dermal armor and the NC’s ectomesenchymal potential. Particularly relevant species are nonteleosteans, such as bichirs, sturgeons, and gars, which represent early diverging groups of ray-finned fishes that may reflect ancestral conditions of bony fishes (including tetrapods). To this end, we compared the dermal exoskeleton in representatives of all major ray-finned fish lineages by examining the bony scutes of sterlet sturgeon (Chondrostei), the ganoid scales of Senegal bichir (Cladistia) and the spotted gar (Holostei), and the dermal armor of the bristlenose catfish (Teleostei) (Fig. 1A). We first used Alizarin Red for histological detection of mineralized structures (Fig. 1 B–I ⁱ and SI Appendix, Fig. S1). For closer examination of individual scales and scutes, we used 3D histology based on microCT (Fig. 1 J–M) and resin-based histology (Fig. 1 N–U).

Fig. 1.

Ray-finned fishes exhibit a remarkably diversified dermal exoskeleton. ( A) Simplified phylogeny of ray-finned fishes highlighting studied species. (B–Iⁱ) Alizarin Red showing the first developed scales and scutes (B–Eⁱ), and the general morphology of the dermal exoskeleton in bichir (B–F ⁱ), sterlet (C–G ⁱ), gar (D–Hⁱ ), and armored catfish (E–Iⁱ). (J–M) microCT reconstructions of single scale of bichir (  J), gar (L), scute of sterlet (K ), and catfish (M). The green corresponds to vascularization (vs), magenta to osteocytes, yellow to ganoin (g), and gold to acrodin (ac). bs, bony scute; dent, denticle; llc, lateral line canal; pc, pulp cavity. (N–U ) Transverse sections reveal details of internal morphology of individual scales and scutes. Black arrowheads mark the ornamented outer layer of the scale. White arrowheads mark osteocytes. bp, bony plate; de, dentin; ep, epidermis; lb, lamellar bone; m, muscle; odt, odontoblast; os, osteoid; ost, osteoblast; pd, pedicle; pde, predentin; pg, preganoin; sc, stratum compactum; I., anteriorly located scute to the posterior one (II.). [Scale bars: 1 mm (B–E ), 250 μm (B ⁱ), 125 μm (Cⁱ–E ⁱ ), 2 mm (F–I ), 150 μm (  J–M ), 50 μm (N–O), 200 μm (P ), 100 μm (Q), 20 μm (R–U ).]

For bichir scales, our data show that they resemble the ancestral rhomboid scales of early osteichthyans with a bony plate, dentin, and a superficial layer of ganoin (Fig. 1 J and N–O), consistent with previous studies (27–29). The entire scale is interlaced with a dense vascularization system (Fig. 1J, SI Appendix, Fig. S2, and Movie S1). Our findings show that bichir scales are formed at least in part by trunk NC cells, which are clearly involved in dentin formation; whether NC is also involved in bony base formation is less clear (Fig. 1 N–O and SI Appendix, Fig. S2).

Gar scales are composed of a thick bony plate covered by a ganoin layer without dentin in between (Fig. 1 L–R). Nevertheless, dentin is involved in the morphogenesis of the gar exoskeleton, in the form of small acrodin-capped denticles (27, 30), arising on the scale surface prior to full mineralization of ganoin (Fig. 1 H ⁱ, L, and R, SI Appendix, Fig. S3, and Movie S4). Interestingly, we found similar denticles on bichir ganoid scales near the anal fin (SI Appendix, Fig. S1D), and Hertwig (27) described them in the region of bichir pectoral fins. Although gar ganoid scales lack the dentin layer, the presence of denticles agrees with the hypothesis that trunk NC cells participate in the formation of gar dermal armor.

Denticles are also a key component of the armored catfish dermal exoskeleton (Fig. 1 E–U). They form sequentially at the posterior edge of the scutes (SI Appendix, Fig. S4, Oa-Od), composed of a cone of dentin covered by a cap of acrodin, and anchored to the underlying bone by a pedicle and ligament (Fig. 1 M–U, SI Appendix, Fig. S4, and Movie S2). Given the tooth-like composition and similar development to odontogenesis (31), this is also consistent with the idea that trunk NC cells participate in the development of armored catfish dermal exoskeleton. It is worth noting that while catfishes lost their scales during evolution (32), armored catfishes secondarily developed an extensive dermal exoskeleton (33), perhaps due to the reactivation of an ancestral developmental program within the trunk NC.

Collectively, these findings suggest that trunk NC has at least odontogenic potential across the ray-finned fishes and participates in the formation of remarkably diverse dermal armor. However, the embryonic origin of osteogenic unit of dermal armor remains unexplored.

The sterlet dermal exoskeleton is comprised of several types of scutes varying in size and shape (24, 25). First to arise are the dorsal scutes which extend from the head to far into the trunk region. Subsequently, four rows of lateral and ventral scutes appear with scattered tiny scutes between them (Fig. 1 C–G ⁱ). The dorsal scutes are comprised of only cellular bone, forming a longitudinal crest which is higher in its hind part (Fig. 1 K–P ⁱⁱ and SI Appendix, Fig. S5), with vascularization confined to one central canal (Fig. 1K, SI Appendix, Fig. S5, and Movie S3). These scutes lack dentin or hypermineralized layers, which makes them particularly useful for investigating potential NC involvement in the formation of trunk dermal bones.

Trunk NC Gives Rise to Osteoblasts in Sturgeon.

As a first step in examining a possible NC contribution to scutes, we analyzed the spatiotemporal expression of the NC specifier gene foxd3 in the sterlet embryo to determine the time course of its appearance. The first emigrating trunk NC cells were observed at stage 25 undergoing an epithelial-to-mesenchymal transition in the dorsal part of the neural tube (Fig. 2 A–Aⁱⁱ). Thus, we chose stage 24 for focal microinjection of the fluorescent dye CM-DiI into the lumen and dorsal midline of the neural tube to label emigrating trunk NC cells and test their ability to contribute to the dorsal bony scutes of sterlet (Fig. 2B and SI Appendix, Fig. S6 A–Aⁱ). One day postinjection, we observed labeled trunk NC cells migrating around the somites similar to other vertebrates (Fig. 2 C–Cⁱ and SI Appendix, Fig. S6 B–B   ⁱ  ). To examine the long-term fate of these trunk NC cells, embryos were allowed to develop to 22 to 23 mm (Fig. 2 D and Dⁱ), by which mineralization of the dorsal scutes had initiated. Histological examination revealed CM-DiI-positive cells located in the dorsal root ganglia (Fig. 2Dⁱⁱ), an expected trunk NC derivative. Importantly, dye-labeled cells were observed around the bone forming in the scute (n = 44) (Fig. 2Dⁱⁱⁱ and SI Appendix, Fig. S6). While the majority of CM-DiI-labeled cells were located in the ventral part of the scute, there was also label in the dorsal part tightly associated with the bone (Fig. 2 Dⁱⁱⁱ–F  ⁱ and SI Appendix, Fig. S6 F–Iⁱ).

Fig. 2.

Trunk neural crest gives rise to osteoblasts of sterlet sturgeon scutes. (A) Expression of NC specifier foxd3 showing premigratory trunk NC cells (Aⁱⁱ) at stage 25, dorsal view. Transverse sections as shown in a through the trunk region (Aⁱ and Aⁱⁱ). (B) Illustration of CM-DiI microinjections at stage 24 into the lumen (light blue) of the neural tube. Small Inset shows an embryo at stage 24 immediately after microinjection. (C) CM-DiI-positive trunk NC cells migrating from the dorsal part of the neural tube (NT) at 1 day postinjection (dpi). (Cⁱ) Transverse section showing migrating trunk NC cells (white arrowheads) around the somite. (D and Dⁱ) At 60 dpi, CM-DiI-positive trunk NC cells can be seen in the dorsal scutes (marked by asterisks), dorsal (D) and lateral (Dⁱ) views. Transverse sections reveal CM-DiI-positive cells within expected NC derivative such as dorsal root ganglia (Dⁱⁱ) (DRG; yellow dotted line). (Dⁱⁱⁱ–Fⁱ) Transverse sections show CM-DiI-positive trunk NC cells (white arrowhead) around the bone (b) forming the sterlet scute. Insets show higher magnification of CM-DiI-positive cells. (G–Hⁱⁱ) HCR against runx2, col1a1, osx, and postn (genes associated with osteoblastic differentiation) reveals expression corresponding to the distribution of CM-DiI-positive cells (compare Fⁱ and G–Hⁱⁱ). (I–Iⁱⁱ) Transverse section of sterlet scute shows CM-DiI signals (magenta) overlapping with HCR expression of col1a1 (early marker of osteoblast; blue). (J) microCT reconstruction (for details, see Material and Methods and SI Appendix, Fig. S5) of osteocytes (magenta) and vascularization (green) within the scute, lateral view. (K) Schematic representations suggesting that trunk NC cells give rise to the osteoblasts (ost) and osteocytes (ostc) of the scute. Every letter represents a different experimental animal. h.r., head region; not, notochord. [Scale bars: 500 μm (A and C), 25 μm (Aⁱ, Aⁱⁱ, Cⁱ, and G–I), 50 μm (Dⁱⁱ–F), 150 μm (J).]

To address the possibility that some CM-DiI-positive cells are osteoblasts, we performed hybridization chain reaction (HCR), a highly sensitive in situ hybridization technique, to examine the expression of runx2 and other genes associated with bone formation such as osx, col1a1, and postn in the scute. The results show that the expression pattern of these skeletal genes corresponds (cf. Fig. 2 F, F ⁱ, and G–H  ⁱⁱ) and overlaps with the distribution of CM-DiI-positive cells (Fig. 2 I–Iⁱⁱ and SI Appendix, Fig. S6 J–K ⁱ ), consistent with the possibility that trunk NC cells give rise to osteoblasts and osteocytes of the scute (Fig. 2 J and K and SI Appendix, Fig. S6). As osteoblasts in teleost scales are exclusively derived from the mesoderm (17–19), we also tested the possible contribution of mesodermal cells to scute development. To this end, we microinjected CM-DiI into the paraxial mesoderm (SI Appendix, Fig. S7 A–B ⁱ). This resulted in labeling cells between the epidermis and the layer of osteoblasts, but not tightly associated with the bone in any of the scutes (n = 19; SI Appendix, Fig. S7 C ⁱⁱⁱ–H). These findings demonstrate that only the trunk NC gives rise to the osteoblasts that form the bony dermal armor of sterlet sturgeon, as well as a significant contribution to the mesenchymal cells, suggesting an essential role of the trunk NC in the formation of dermal exoskeleton.

NC GRN Gene Signature in Scutes and Scales.

To further explore the molecular characteristics of scutes and scales across different fish species, we performed a hierarchical clustering analysis (12) of genes identified as being part of NC GRN (34) in sterlet scutes, bichir ganoid scales, in comparison with zebrafish scales (Fig. 3A and SI Appendix, Fig. S8 A–C). Consistent with our lineage tracing of sterlet trunk NC cells, differential gene expression (DGE) analysis of three developmental stages of sterlet scutes displayed significant enrichment of NC-GRN genes including sox10, tfap2a, and tfap2b (Fig. 3A and SI Appendix, Fig. S8 I–Kⁱ ). Furthermore, HCR in situ hybridizations support this finding by demonstration of NC-like gene expressions in sterlet scutes (Fig. 3 B–D ⁱ and SI Appendix, Fig. S8 D–F   ⁱ ). Similarly, DGE analysis of ganoid scales of the Senegal bichir, which has osteogenic and odontogenic components, identified significantly enriched NC-GRN genes (Fig. 3A and SI Appendix, Fig. S8) compared to the gene signature analysis of mesoderm-derived zebrafish scales (17–19). Our results indicate that trunk NC possesses ectomesenchymal potential at least in two early branching ray-finned fishes and participates in the development of their dermal exoskeleton. These findings are consistent with the hypothesis that the trunk NC contributed to both odontogenic and osteogenic components of the dermal armor of some vertebrates (10).

Fig. 3.

Scale RNA sequencing comparisons between sterlet, bichir, and zebrafish reveal neural crest GRN gene signature in bichir scales and sterlet scutes. (A) Hierarchical clustering analysis focused on genes identified as being part of NC GRN reveals significant enrichment of NC-like genes in sterlet scutes and bichir scales. (B-D) Multiplexed fluorescent mRNA in situ hybridizations by HCR reveals expressions of NC-like genes (sox10, sox9a, tfpa2a, and tfap2b) and osteoblast markers (osx and runx2) in sterlet scutes. [Scale bars: 50 μm (B–D).]

Conclusions

It has long been argued that the origin of vertebrate dermal armor is linked to the evolution of neural crest skeletogenic potential (10, 35). The first identifiable dermal armor is attributed to fragments of an enigmatic vertebrate Anatolepis from the Late Cambrian to the Early Ordovician (~490 Mya to 450 Mya) (36). It is comprised of tubercles with a central pulp cavity capped by dentin-like tissue; the individual tubercles are connected with a lamellar tissue that may represent a precursor of dermal bone (37) (Fig. 4, Bottom). Later ostracoderms and placoderms (jawless and jawed members of the gnathostome stem group) show a definite odontoosteogenic arrangement of dermal armor with reduction or elaboration varying in individual lineages (9, 38, 39) (Fig. 4). Based on lineage tracing of dermal denticles of cartilaginous fishes and highly derived elasmoid scales of teleost fishes, it has been suggested that the odontogenic component of the ancient dermal armor is trunk NC derived, whereas the osteogenic component is mesoderm derived (13).

Fig. 4.

Trunk NC has the ability to form ectomesenchymal progeny across gnathostomes. Simplified phylogenetic tree of vertebrates, with highlighted extinct lineages (dashed lines) showing composition of dermal armor (rectangle with three units; from top to bottom: hypermineralized, odontogenic, and osteogenic unit) in individual groups of vertebrates. I and II indicate two possible phylogenic positions of anaspids, (I) as a deeply branching ostracoderms (40) or (II) anaspids as stem cyclostomes (41, 42). Based on this, the dermal armor was absent or secondarily lost in the cyclostome lineage (lampreys and hagfishes). Given that the oral and postcranial odontodes in extant gnathostomes seem to be NC derived (13, 43), this strongly implies that the odontogenic unit of dermal armor is trunk NC derived in fossil crown gnathostomes (such as Andreolepis) and likely in stem gnathostomes as well. During the rise of tetrapods (lungfishes, porolepiforms, tetrapodomophs) (11), the odontogenic unit was gradually reduced. Indeed, loss of both dermal armor units is apparent in Amniotes (half-filled unites of rectangle). Note trunk NC (magenta)/mesoderm (green) polarity in osteogenic unit contrasting with the exclusively NC-derived odontogenic system. The half-colored box in gar indicates the presence of dentin in denticles. The scheme is based on experimental evidence from zebrafish, little skate, and sterlet sturgeon, for which developmental evidence exist. b-l, bone-like tissue; de-l, dentin-like tissue; od?, possible odontode; pc?, possible pulp cavity.

To understand the evolutionary origin of the ancient dermal armor, we analyzed the dermal exoskeleton of the representatives of all major phylogenetic lineages of ray-finned fishes, including nonteleost groups, such as sturgeons. To directly test the embryonic origin of the osteogenic component of the dermal armor, we took advantage of the accessibility of the sterlet sturgeon, whose scutes completely lack the dentin component. Our data show that trunk NC cells migrate to the developing dorsal scutes where they give rise to osteoblasts comprising the bony elements of the scutes (Fig. 2). This is an unambiguous demonstration of trunk NC forming dermal bone, posterior to the pectoral girdle (44). This is also supported by indirect evidence of gene expression studies in several amniotes (45–47), which inferred that the osteoderms of tetrapod integumentary skeleton may be trunk NC derived as well. For instance, preliminary lineage tracing of late emerging trunk NC in the hard shell turtle suggested the NC origin of the plastron bones (45, 46).

While our results suggest that the postcranial bones can form from trunk NC (cf. Figs. 2 and 4), data from teleost elasmoid scales suggest that the origin of the osteogenic unit was mesodermal (17–19). Three opposing scenarios may explain the origin of sterlet scutes from ancestral ganoid scales (Fig. 4): 1) Trunk NC gave rise to both units, but after reduction of the dentinous portion, trunk NC gave rise to the bony part; or 2) Trunk NC was involved only in the formation of the odontogenic unit; when this unit disappeared, the trunk NC was coopted to make dermal bone; 3) Both NC and mesoderm are equally competent to form bone, such that teleost scales versus sterlet scutes reflect shifts in the contribution of one or the other of these mesenchymal populations in the trunk; a similar situation occurs in the bones of skull vault (48) or pharyngeal endoskeleton (49). We also speculate that the NC/mesoderm origin may correlate with the presence of either a rigid bony plate as in bichir or gar, or an incompletely mineralized elasmodin (plywood-like tissue) as in most teleost scales, respectively. These two types of skeletal tissue (bony versus plywood like) are considered distinct by some authors (50). The origin of the elasmoid scales is still unclear, but it is possible that they have completely lost the thick bony plate typical of ancestral rhomboid scales (8) and evolved a novel tissue produced by a different cell population expressing bone matrix genes (51). This suggests that a mesodermal origin observed in some teleosts may reflect a highly derived condition for elasmoid scales. Further insights into the potential mesodermal origin of plywood-like structures may come from exploring Latimeria and lungfish scales. While sterlet scutes are derived as well, the presence of bone is likely a retained primitive character. Thus, we propose that the osteogenic unit of the ancient dermal armor was likely derived from the trunk NC. Together, our results suggest that the trunk NC may have had dentoosteogenic potential along the entire body axis across gnathostomes (Fig. 4), and this ability became restricted to the cranial region only later in tetrapod evolution by spatiotemporal expression changes of some GRN components (12).

Material and Methods

Animal Husbandry.

Sterlet sturgeon (Acipenser ruthenus Linnaeus, 1758) embryos and larvae were obtained from the Faculty of Fisheries and Protection of Waters, Research Institute of Fish Culture and Hydrobiology, University of South Bohemia in Ceske Budejovice, Czech Republic. Embryos were held in containers with dechlorinated water at 15 °C until the hatching stage (st. 35). Free-swimming larvae were kept in well-oxygenated containers at 15 to 17 °C until desired developmental stage, and they were euthanized using an overdose of tricaine (1 g/L) prior to fixation in 4% PFA. Sterlet development was staged as previously described by Dettlaff et al. (52). Senegal bichirs (Polypterus senegalus, Cuvier 1829) were obtained from the Department of Zoology, Faculty of Science, Charles University in Prague, the Czech Republic, and spotted gars (Lepisosteus oculatus, Winchell 1864) from the Department of Integrative Biology, Michigan State University, MI, USA. Fixed bristlenose catfishes (Ancistrus sp.) were a generous gift from the personal collection of Mr. Dominik Miler. Individuals were staged based on the total body length (TL). Sterlet embryology work and animal care were approved by the Ministry of Agriculture of the Czech Republic (MSMT-12550/2016-3), followed the principles of the European Union Harmonized Animal Welfare Act of the Czech Republic and Principles of Laboratory Animal Care and National Laws 246/1992 “Animal Welfare,” and were conducted in accordance with the Animal Research Committee of RIFCH. The authors of the study own the Certificate of professional competence for designing experiments and experimental projects under Section 15d(3) of the Czech Republic Act no. 246/1992 Coll. on the Protection of Animals against Cruelty. All animal work with bichir was approved by the institutional animal care and use committee of the Charles University in Prague, the Czech Republic. Gar animal work was approved by the IACUC at Michigan State University (protocols 10/16-179-00 and 201900309). Adult zebrafish were maintained in the Beckman Institute Zebrafish Facility at Caltech, and all work was compliant with the animal protocol No. 1764 approved by the IACUC at Caltech.

Cell Lineage Tracing.

The sterlet embryos at stage 24 (trunk neural crest) and stage 25 (paraxial mesoderm) were manually removed from the jelly egg membrane and positioned in suitable hollows in modeling clay, allowing proper orientation and stability. Cell tracker CM-DiI (1 mg in 1 mL of 100% ethanol; Thermo Scientific) was diluted (1:5) into 10% sucrose and microinjected through the chorion with injector using microcapillaries (Drummond Microcaps) prepared in a Narishige pc-10 puller (58 °C with two weight elements). After labeling, embryos were kept in the 48-well plates containing E2 zebrafish medium with Pen/Strep antibiotics (120 ng/mL of penicillin and 200 ng/mL of streptomycin) at 15 °C. After hatching, the individuals were maintained in well-oxygenated containers with dechlorinated water in incubators at 15 to 17 °C to the desired developmental stage (22 to 23 mm, approx. 60 d; depending on feeding frequency) and then euthanized using an overdose of tricaine (1 g/L) prior to fixation in 4% PFA. As a control of proper injection, a few embryos were fixed immediately after the injection and analyzed (T0; Fig. 2B and SI Appendix, Fig. S6 A and Aⁱ). Cell lineage tracing was performed in four independent rounds of injections (always different breeding pairs) during spring 2020 and 2022. To avoid staining of paraxial mesoderm, CM-DiI dye was injected into the lumen of the neural tube (Fig. 2B and SI Appendix, Fig. S6 A and Aⁱ); thus, only prospective trunk neural crest and dorsal midline of the neural tube were stained. Sixty-six embryos were successfully injected and fixed at stage 24 (n = 7), stage 25 to 26 (n = 10), stage 30 (n = 2), at 15 mm (n = 3), and at 22 to 23 mm (n = 44) (SI Appendix, Fig. S6 and Table S1). In the case of paraxial mesoderm fate mapping, the capillary was aimed from above into the paraxial mesoderm which is located laterally to the developing neural tube in sterlet sturgeon embryos; thus, only paraxial mesoderm and nonneural ectoderm were stained. In total, 49 embryos were successfully injected and fixed at stage 25 (n = 10), stage 26 to 27 (n = 20), and at 22 to 23 mm (n = 19) (SI Appendix, Fig. S7 and Table S1). To verify that trunk neural crest gives rise to the scute osteoblasts, several specimens with positive CM-DiI signals were used for HCR, but only col1a1 probe worked properly on 4% PFA fixed samples (trunk NC: n = 6; mesoderm: n = 4; Fig. 2I and SI Appendix, Fig. S6 J–Kⁱ). The samples for histological analysis were mounted into JB-4 resin or paraffin (SI Appendix, Fig. S7 F–G), sectioned (5 μm or 12 μm, respectively), and stained with Fluoroshield with DAPI (Sigma-Aldrich).

HCR.

HCR v3.0 was performed according to the protocol suggested by Molecular technologies (53) for zebrafish with several modifications. Briefly, methanol-fixed tissues were rehydrated by series of methanol/PBS-Tween solutions (2 × 100%, 75%, 50%, 25%; every step 15 min), washed in PBS-Tween (2 × 10 min), depigmented by a bleaching mix-solution (formamide, 20× SSC, 30% hydrogen peroxide, and distilled water) under the direct light, washed in PBS-Tween (10 min), treated with proteinase-K for 60 min (22 to 23 mm) or 70 min (26 mm) at room temperature, washed in PBS-Tween (2 × 10 min), postfixed in 4% PFA (10 min), washed in PBS-Tween (2 × 10 min), prehybridized in 30% probe hybridization buffer at 37 °C (60 min), and incubated with probes (usually 2 μL of 1 μM stock per probe mixture) in probe hybridization buffer at 37 °C overnight. HCR on sections (tfap2b and tfap2a) was performed as described by Criswell and Gillis (54), with the following modifications: cryosections were used instead of paraffin sections, and slides were hybridized overnight at 37 °C with 2 μL of 1 μM probe stock/100 μL of hybridization solution. All probe sets (sox10, sox9a, tfap2a, tfap2b, col1a1, osx, runx2), hairpins, and buffers were purchased from Molecular Technologies (https://www.molecularinstruments.com). Each HCR was repeated at least three times with a minimum of three sterlet scutes for each run of HCR (for HCR on sections (Fig. 3D and SI Appendix, Fig. S8D), four slides containing approx. 10 sections from two different specimens were used). HCR on CM-DiI-positive samples followed the same protocol described above, with the following modifications: The individual scute was dissected, depigmentation treatment was skipped, and hybridization was prolonged to 24 h. The samples for histological analysis were mounted into JB-4 resin (described below), sectioned (5 μm), and counterstained with Fluoroshield with DAPI (Sigma-Aldrich). All photographs were taken with Zeiss AxioImager.M2 equipped with an Apotome.2.

Cloning of Sterlet foxd3 Ortholog and In Situ Hybridization.

cDNA was synthesized using SuperScript VILO kit (Invitrogen) and used for the design of PCR amplification primers (foxd3: 5′-GAYGTGGAYATCGAYGTGGT-3′; 5′-CTSARRA​ARCTVCCGTTGTC-3′). The PCR product was cloned into pGEM-T-Easy vector (Promega) in DH5α-competent Escherichia coli. The in situ hybridization protocol is initially similar to HCR, but the samples are prehybridized in the hybridization buffer (50% formamide, 4× SSC, 0.1 mg/mL heparin, 1× Denhardt’s, 0.1% CHAPS, 0.2 mg/mL yeast RNA, 10 mM EDTA, 0.1% Tween-20) for at least 120 min. Hybridization was performed at 60 °C with a 1:1,000 dilution of DIG-RNA probe overnight. The next day, the samples were washed at least eight times (each step for 30 min) in wash solution (50% formamide, 4× SSC, 0.1% Tween-20) at 58 °C, then three times in MAB-T (100 mM maleic acid, 150 mM NaCl, 0.1% Tween-20), and blocked in blocking solution (2% blocking reagent, 20% sheep serum in MAB-T) for at least 2 h. Next, the samples were incubated at 4 °C in a blocking solution containing antidigoxigenin antibody (1:3,000; Roche) overnight. The third day, the samples were washed at least eight times (each step for 30 min) in MAB-T and equilibrated in NTMT (0.1 M Tris, 0.1 M NaCl, 0.05 M MgCl2, 0.1% Tween-20). The color reaction was developed in BM-Purple (Roche) until the signal developed. The hybridization was performed at least three times with two embryos per developmental stage, and selected embryos were prepared for vibratome histology as previously described by Stundl et al. (55).

Library Preparation and Sequencing.

Samples for RNA-sequencing were preserved in RNAlater (Sigma-Aldrich). For transcriptome sequencing, total RNA was isolated from scales/scutes: bichir (60 mm, 65 mm, 88 mm TL; 10 scales were dissected for each of n = 2 biological replicates); sterlet (22 mm, 26 mm, 68 mm TL; 2 scutes were dissected for each of n = 2 biological replicates); and zebrafish (~3 cm TL; 15 scales were dissected for each of n = 2 biological replicates) using the RNAqueous kit (Ambion) and assessed using the Agilent Bioanalyzer. The RNA-sequencing was performed at the Millard and Muriel Jacobs Genetics and Genomics Laboratory (California Institute of Technology, Pasadena, CA, USA) at 50 million, 50 bp, single-ended reads on two biological replicates for all the three species. The libraries were built according to Illumina Standard Protocols. SR50 sequencing was performed in an HiSeq2500 Illumina sequencer.

Proteome Alignment and Statistical Analysis of Sterlet, Bichir, and Zebrafish Scute and Scale RNA-seq Datasets.

To identify orthologous genes between sterlet sturgeon, Senegal bichir, and zebrafish, we followed the methods described by Martik et al. (12), with minor modifications. Sterlet sturgeon (genome assembly ASM1064508v1)/Senegal bichir (genome assembly ASM1683550v1) proteome obtained from NCBI was aligned to the zebrafish (genome assembly GRCz11) proteome using the BLAST alignment software available on the UCSC genome browser (56). In brief, every sterlet/bichir protein sequence was queried locally against the zebrafish proteome, following which regions with the longest alignment were matched to the respective zebrafish proteins. Using this alignment-based approach, proteins with highest alignment percentage score (Dataset S1; Dataset S2 for exact scores for each ortholog) were identified as orthologs. Sterlet/bichir scute/scale RNA sequencing libraries were aligned to the sterlet/bichir sequences, while the zebrafish scale RNA sequencing libraries were aligned to the zebrafish sequences using Bowtie2 (57). Transcript counts were calculated using featureCounts (58), and DGE analysis was performed using DESeq2 (59). Using zebrafish gene annotations as a reference, we added the transcript counts for duplicated orthologs found in the sterlet/bichir genome to calculate an “aggregated” transcript count for each gene as described by Martik et al. (12). These aggregated transcript counts were then normalized using the formula: Z_i = T_i − min(T)/max(T) − min(T) where Z_i is the normalized transcript count and T_i is the absolute transcript count. A subset of genes previously identified as being part of the neural crest gene regulatory network (34) was then isolated from the count matrix and plotted as a heatmap using ComplexHeatmap (60) package in Rstudio. The sterlet has a paralog retention rate of about 70% from its genome duplication; thus, all IDs of candidate genes were used for further analysis, and paralogs (SI Appendix, Table S2) were phylogenetically analyzed (amino acid tree made using the maximum likelihood method [PhyML 3.0 (61) http://www.atgc-montpellier.fr/phyml/] on a Clustal Omega 1.2.3 alignment, bootstrap score n = 100).

X-Ray Microtomography (microCT).

Specimens for microCT were dissected and mounted without contrast staining (62) and scanned with an MicroXCT-200 (Zeiss/Xradia, Germany) at the Department of Evolutionary Biology, Theoretical Biology Unit, University of Vienna. The scans were acquired from bichir scale (20 cm TL; 1.0 μm voxel size), sterlet dorsal scute (68 mm TL; 2.5 μm voxel size), gar scale (125 mm TL; 1.0 μm voxel size), and armored catfish scute (35 mm TL; 1.0 μm voxel size). 3D modeling from the virtual thin sections was segmented in the software VG Studio 3.4 at the Department of Organismal Biology of Uppsala University, Sweden.

Histology.

Visualization of mineralized tissues was carried out as described previously (63). All photographs were taken in numerous focal planes with Olympus MVX10 stereoscope with AxioCam and the final images were prepared with Helicon Focus Pro (HeliconSoft), allowing to form high-resolution images. Mineralized tissues for histological analysis were washed in distilled water and placed in Morse’s solution (10% sodium citrate and 22.5% formic acid) at room temperature for decalcification until the tissues were soft. Next, the samples were dehydrated in 100% ethanol and incubated in an infiltration solution of JB-4 resin (prepared according to the manufacturer’s instructions; Sigma-Aldrich) at room temperature overnight. The next day, the infiltration solution was replaced by an embedding solution (prepared according to the manufacturer’s instructions), placed into an embedding mold (PolyScience), and transferred to a vacuum chamber which accelerated the polymerization (~3 h). The resin block was sectioned at 5 μm, and sections were stained with Mayer’s hematoxylin (except SI Appendix, Fig. S2I) or Verde Luz-orange G-acid fuchsin (VOF) stain (SI Appendix, Fig. S2I).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (CSV)

Dataset S02 (CSV)

Movie S1.

3D reconstruction of bichir ganoid scale.

Movie S2.

3D reconstruction of gar ganoid scale.

Movie S3.

3D reconstruction of bristlenose catfish scute.

Movie S4.

3D reconstruction of sterlet sturgeon scute.

Data, Materials, and Software Availability

The RNA-seq data reported in this paper have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus database (accession no. GSE235280) (64).