Segmentation—repetition of anteroposteriorly polarized units along the anteroposterior body axis (Hannibal and Patel, 2013)—has been a frequently featured concept in the narrative of vertebrate development and evolution. Serial or iterative homology (homology of parts within a single body) was initially conceived to describe segments and bilateral counterparts in a comparative context (Owen, 1848). In that scheme, the observed segmented pattern corresponded to segments in an idealized archetype (“general homology”), whereas each pair of corresponding segments between different taxa was designated as “special homology” (Owen, 1848). The developmental basis of homology has generated long, convoluted discussions since Owen's time, and numerous schemes and definitions have been proposed (reviewed by De Beer, 1971; Patterson, 1988; Wagner, 1989, 2014; Hall, 1994, 2003; Amundson, 2001; Kuratani, 2009; Faunes et al., 2015). These include historical and biological criteria of homology ranging from trait-specific regulatory networks (“character identity network”; Wagner, 2007, 2014) to the flexibly and neutrally defined “continuity of information” (Van Valen, 1982; Roth, 1994). Specifically in regards to serial homology, gene expression patterns that underlie morphologically repeated structures along the anteroposterior axis, such as collinear Hox expressions, have attained an iconic status in evolutionary developmental biology (Patel et al., 1989; Krumlauf, 1994; Holland and Garcia-Fernàndez, 1996; Gellon and McGinnis, 1998; Manzanares et al., 2000; Trainor and Krumlauf, 2001; Carroll, 2008; Parker et al., 2014). Nevertheless, a problem persists for serial homology: body parts generally considered as segments and thus serially homologous do not necessarily share historical continuity of descent from a common ancestor (Wagner, 1989, 1994; Roth, 1994). We will present notable examples where developmental patterns reveal such decoupling (Figure 1A). This terminological paradox has led to two overlapping criteria in parallel with other categories of homology: (a) identified on the basis of historical continuity of form (phenotype) between similar parts (“evolutionary” serial homology); and (b) identified on the basis of developmental correspondence between parts (“developmental” serial homology; Wagner, 1994; Reno et al., 2013).

In this paper, we recognize serial homology on the basis of historical continuity (“evolutionary” serial homology sensu Wagner, 1994), where the use of serial homology is restricted to those developmentally corresponding parts with shared evolutionary history (arising either simultaneously or via a duplication event). Under this view, serial homology requires phenotypic similarity between corresponding parts as an ancestral state because any congruence not due to such ancestral state then involves some type of phenotypic homoplasy such as convergence and parallelism (Gould, 2002; Hall, 2007). In other words, historical discontinuity in form between parts would violate the pervasive assumption that the serial units originated in a common ancestor simultaneously and in identical forms. Major components of the vertebrate body such as pharyngeal apparatus, axial elements, and paired appendages have each served as a popular example to illustrate how the hypothetical ancestral segmentation underwent differential specialization of its parts (Gegenbaur, 1859; Sewertzoff, 1899, 1931; Goodrich, 1930; De Beer, 1937; Jarvik, 1980). Consequently, these repeating patterns in the vertebrate body have often generated an evolutionary scenario from polyisomerism (a series of identical or similar segments) to anisomerism (a series of differentially specialized units). That is, a series of identical units in an archetypical ancestor (polyisomerism) gave rise to serially organized but distinctly specialized structures (Figure 1B). In this context, the jaw has been long considered as a modified branchial bar (Sewertzoff, 1911, 1913; De Beer, 1937; Jarvik, 1980; Mallatt, 1996), and pectoral and pelvic appendages as serial homologs due to a duplication event (reviewed by Diogo et al., 2013).

In this review, we question the assumption that serial patterns are normally the default ancestral state. To provide a test, this review explores whether two striking cases of serial similarity in vertebrates—within the pharyngeal apparatus and between the paired (pectoral and pelvic) appendages—were gained through assimilation of dissimilar units (serial similarity as a derived state) or through differential modification of originally identical segments (serial similarity due to an ancestral state). When an underlying segmented scheme is extrapolated from observed serial morphological patterns, two possible scenarios exist: (1) segments were differentially modified through secondary specialization (ancestrally polyisomeric state); or (2) distinctly specialized regions secondarily assimilated each other by acquiring serial patterning (ancestrally anisomeric state; Figure 1B). When contrasting these alternative hypotheses, it is critical to not confuse observed or extrapolated serial patterns across different developmental stages. For example, serial organization of the pharyngeal arches at pharyngula stages of vertebrate development should not be confused with serial musculoskeletal patterning of the arch derivatives at later stages of development. Both are hypotheses testable on the basis of phenotypes—such as morphology and gene expression profiles—at that particular stage of development. This is because the mechanisms responsible for serial pattern at an earlier stage (e.g., hindbrain and pharyngeal Hox codes that characterize distinct streams of cranial neural crest cells) can be independent from those that pattern differentiation of the anlagen (e.g., Dlx code that dorsoventrally pattern the crest cells into elements of the pharyngeal apparatus; Hunt et al., 1991a,b; Hunt et al., 1998; Couly et al., 2002; Depew et al., 2002, 2005; Depew and Compagnucci, 2008; Gillis et al., 2013).

During the development of vertebrate embryos, the pharyngeal arches (PAs)—anteroposteriorly arranged columnar structures of mesodermal (MMCs) and neural crest cells (NCCs) set apart by endodermal pouches toward bilateral sides of a pharynx—contain anlagen of a remarkable variety of complex structures (Figure 2; Graham et al., 2005). The pouches begin forming as (a) Tbx1 expression in the mesoderm promotes Wnt11 expression to destabilize the pharyngeal epithelium; and (b) Fgf8 expression zones in the mesoderm guide the destabilized epithelium into pouches (Piotrowski et al., 2003; Crump et al., 2004; Choe et al., 2013; Choe and Crump, 2014; Jandzik et al., 2014). The MMCs are surrounded by NCCs that migrate in position-specific streams corresponding to anteroposterior levels of the hindbrain rhombomeres (Figure 2; Kimmel et al., 2001; McCauley and Bronner-Fraser, 2003; Cerny et al., 2004b; Noden and Trainor, 2005; Hall, 2009). This pattern applies to both of the major living lineages of vertebrates, cyclostomes (hagfish and lampreys) and crown gnathostomes (chondrichthyans and osteichthyans; see Figure 1A for phylogenetic scheme).

Each PA receives a specific stream of NCCs, and innervations by the cranial nerves (trigeminal, facial, glossopharyngeal, or vagal) follow this arrangement (Figure 2). Importantly, adjacent streams of the migratory NCCs do not mix (Köntges and Lumsden, 1996). The pharyngeal Hox code reflects this specificity of the migratory NCCs to the PAs: the NCCs (ectomesenchyme) in each of the non-mandibular PAs expresses a specific combination of Hox genes (Figure 2; Hunt et al., 1991a,b; Trainor and Krumlauf, 2001; Takio et al., 2007; Minoux and Rijli, 2010). Within respective PAs, NCCs interact with the ectodermal placodes, endodermal epithelium, and the MMCs to form the entire pharyngeal apparatus (Piotrowski and Nüsslein-Volhard, 2000; Couly et al., 2002; Noden and Trainor, 2005; Noden and Francis-West, 2006; Minoux and Rijli, 2010; Frisdal and Trainor, 2014). The differentiated PA structures in a generalized crown gnathostome include the jaw skeleton and branchial bars, jaw and branchiomeric muscles, trunks of cranial nerves, sensory structures like paratympanic organ, pseudobranch and respiratory gills, afferent and efferent vessels, associated ganglia of the cranial nerves, and others (Frisdal and Trainor, 2014). With the exceptions of the paratympanic organ and pseudobranch, these structures are generally repeated from one PA to another as a plesiomorphic condition (see Goodrich, 1930; Hyman, 1992; Shone et al., 2016 for a review of reduction in the PA-derived structures or the PA themselves in various lineages of vertebrates).

By this stage, two critical properties of the PA development are: (a) the initial spatial confinement (compartmentalization) of the NCCs and MMCs to the PAs and (b) the subsequent serial patterning of differentiating tissues (among all the PAs in crown gnathostomes, and among all the PAs except for the PA I in cyclostomes). The NCCs and MMCs in the PAs are spatially delimited in many ways. They are set apart by the pharyngeal pouches (Crump et al., 2004). The NCCs are attracted, repulsed, or maintained competent by various signaling molecules from epithelial structures (such as BMP4, FGF8, and SHH) including the pouches and placodes (Shigetani et al., 2000; Haworth et al., 2004, 2007; Eberhart et al., 2006). The NCCs regulate the MMC differentiation within the same PA into musculature and other connective tissues (Rinon et al., 2007; Grenier et al., 2009; Heude et al., 2010). As a result, it is possible to identify original PA identities of pharyngeal structures even well after tissue differentiation.

On the basis of the development and anatomy, the vertebrate pharyngeal structures have long been interpreted in a serial pattern between the PAs (Rathke, 1827; Gegenbaur, 1859; Goodrich, 1930; Sewertzoff, 1931; De Beer, 1937). In particular, the jaw has generated more than a century of investigation into its evolutionary origin because its morphology and function clearly depart from those of other PA-derivatives that ancestrally supported gills.

A prevailing idea has been that the jaw evolved as a specialization within the mandibular arch (PA I). Previous jaw-origin hypotheses explicitly or implicitly assume such differential specializations. Some postulate the jaw as a modified branchial bar (Gegenbaur, 1859; Sewertzoff, 1911, 1913; Goodrich, 1930; De Beer, 1937; Mallatt, 1996, 2008). Some premise on correlated shift of character in all of the PAs (Kimmel et al., 2001; Cohn, 2002; Cerny et al., 2004b) and others focus on phenotypic changes required for specialization of the PA I derivatives into a jaw (Forey, 1995; Janvier, 1996; Cerny et al., 2010; Medeiros and Crump, 2012). One hypothesis—the Heterotopy Hypothesis—is free from this assumption (Kuratani et al., 2001, 2013; Shigetani et al., 2002; Kuratani, 2012), but this is discussed in the context of test of the most recently proposed hypothesis. Despite the fact that none of them provides a comprehensive account of all character transitions from the jawless to the jawed vertebrates, it is difficult to reconcile any pair of these hypotheses with each other under the assumption that the vertebrate pharyngeal apparatus has always been serially patterned. Not only is the assumption non-parsimonious on a phylogenetic tree (i.e., taking more character changes than minimally required), it appears to burden each hypothesis with character changes that are either implausible or clearly decoupled from the origin of the jaw (reviewed by Miyashita, 2015). Whether they are based on anatomical or gene expression patterns, the previous hypotheses capture conditions necessary for a jaw to develop in a crown gnathostome, but the proposed phenotypic changes alone are clearly not sufficient for a jaw to evolve.

This incompatibility questions that the patterning of PA I shares its evolutionary history with that of other PAs prior to the origin of the jaw. The question is two-fold. Were the NCCs and MMCs spatially confined (compartmentalized) in all of the PAs in the last common ancestor of all living vertebrates? Were the PA derivatives (differentiation of the NCCs and MMCs) serially patterned across the pharynx in that ancestor? Miyashita (2015) explored these questions through a review of gene expression patterns, functional analyses, fate-mapping experiments, and adult and embryonic morphology in living jawless and jawed vertebrates (cyclostomes and crown gnathostomes) and through a review of adult morphology in extinct jawless vertebrates (stem gnathostomes; Figure 1A for phylogenetic scheme). In brief, the NCCs and MMCs were likely not spatially confined or delimited in PA I before the jaw evolved. Neither were the PA I derivatives patterned in series with the derivatives of the other PAs in that part of the phylogenetic tree. However, both spatial confinement and serial patterning appear to govern the non-mandibular PAs across all known vertebrates. On the basis of cyclostome development (Figures 3, 4) and morphology of stem gnathostomes, only at the origin of the jaw did the PA I likely acquire both spatial confinement of the NCCs and MMCs and serial patterning of the derivatives. The Mandibular Confinement Hypothesis was generated to explain these predicted character transitions. It proposes: (a) the NCCs and MMCs of the PA I are delineated along interfaces with those of the premandibular, hyoid/hyomandibular (pharyngeal pouch I), and hypobranchial domains in jawed vertebrates; and (b) confinement along these interfaces were acquired in steps as the jaw evolved (Figure 4; Miyashita, 2015).

If such scenario is correct, then the jaw evolved through assimilation of an otherwise distinctly patterned region with the rest of the pharyngeal series, and not through specialization of a metameric unit. Many lines of evidence corroborate the scenario of mandibular confinement. This review focuses on those particularly relevant to the developmental aspect of the Mandibular Confinement Hypothesis.

A striking insight from the Mandibular Confinement Hypothesis is that both the jawless cyclostomes and the jawed gnathostomes—two major lineages of living vertebrates—retain distinct features of PA I patterning from the rest of the PAs, which cannot be easily attributed to secondary modification. Serial patterns between the PA I derivatives and the derivatives of other PAs only occur in jawed gnathostomes, whereas cyclostomes show no such patterns. This contrast is apparent in the morphology of the chondrocranium and pharyngeal muscles (reviewed by Miyashita, 2015; also summarized in Table 1), but is also manifest in embryos.

The two conserved gene expression patterns—the pharyngeal Hox and Dlx codes—serve as an illustrating example (Table 1, Figures 2, 3). Both in cyclostomes and gnathostomes, the non-mandibular PAs collinearly express a specific combination of Hox genes, but the PA I stands out in lacking Hox expression in its NCCs (Figure 2; Hunt et al., 1991a,b; Takio et al., 2007). Within crown gnathostomes, the absence of Hox expression in the PA I ectomesenchyme is required for proper patterning of a jaw (Pasqualetti et al., 2000; Couly et al., 2002; Creuzet et al., 2002; Kitazawa et al., 2015). In reverse, Hoxa-deficient gnathostome embryos develop ectopic jaw-like cartilages in posterior PAs (Rijli et al., 1993; Baltzinger et al., 2005; Minoux et al., 2009). These results suggest that expression of the Hox code likely governs serial pattern of the branchial skeleton (PA II and more posterior arches). Because the PA I-derived skeleton requires the absence of the Hox expression, some other mechanism should explain the serial pattern observed between the jaw and the branchial skeleton in gnathostomes.

Consistent with this, the pharyngeal Dlx code—dorsoventrally nested expressions of the Dlx family genes—spans across the PAs in crown gnathostomes. In that clade, the NCCs of all of the PAs (including the PA I) express Dlx1 and Dlx2 at all dorsoventral levels, Dlx5 and Dlx6 from intermediate to ventral levels, and Dlx3 and Dlx4 at ventral levels (Figure 2; Depew et al., 2002, 2005; Medeiros and Crump, 2012; Gillis et al., 2013). Slight differences emerge in the ventral-most Hand2 expression zone both among taxa and between the PA I and other PAs. In zebrafish, dlx1 and dlx2 are expressed in this zone in all of the PAs, but dlx5 expression is also detected for the PA I (Akimenko et al., 1994; Ellies et al., 1997; Kimmel et al., 2003; Talbot et al., 2010). In mice, no Dlx expression is apparent in the equivalent ventral-most zone in all of the PAs (Depew et al., 2002; Ozeki et al., 2004). Despite these minor differences, the Dlx code is an evolutionarily conserved serial patterning mechanism in the pharynx among crown gnathostomes (Compagnucci et al., 2013; Gillis et al., 2013). The pharyngeal Dlx pathway has upstream regulatory components such as the endothelin signaling (upregurating Dlx) and Hand2 (downregulating Dlx) that are expressed ventrally (reviewed by Clouthier and Schilling, 2004; Medeiros and Crump, 2012).

In cyclostomes, however, the Dlx expressions in the NCCs are neither exactly serial across nor restricted to the PAs. Notably, the expression patterns differ between the PA I and others. In the sea lamprey Petromyzon marinus, DlxA has dorsal and ventral expression domains in the PA I (without expression in the intermediate region), whereas its expression in other PAs is restricted to the ventral domain (Figure 3A; Cerny et al., 2010). DlxC and DlxD are expressed throughout in the PA I, but its expression domain in other PAs does not extend to the ventral region (Cerny et al., 2010). Although, initially reported as uniform patterns (Myojin et al., 2001; Shigetani et al., 2002), the Japanese lamprey Lethenteron camtschaticum appears to have differential expressions of Dlx cognates as well. Assuming that the published figures (Kuraku et al., 2010) show representative phenotypes (and assessing them with respect to the description from Cerny et al., 2010), LjDlxA is expressed throughout in the PA I and in the intermediate domain in other PAs; LjDlxB has a weak dorsal expression domain in the PA I but has clear dorsal expression in other PAs; LjDlxC and LjDlxD are expressed throughout in the PA I but clear expression does not extend to dorsal domain in other PAs; LjDlxE has strong expression domain from dorsal to ventral regions of the PA I, but is not expressed so strongly ventrally in other PAs; LjDlxF lacks its dorsal expression domain in the PA I (Kuraku et al., 2010). Whether the Dlx expression patterns in L. camtschaticum are truly nested or uniform, however, interspecific variations in Dlx expression patterns exist among cyclostomes. The analysis of the patterns is further complicated by unclear orthology of Dlx cognates among cyclostomes and with respect to crown gnathostome Dlx family (Kuraku et al., 2010; Fujimoto et al., 2013; Takechi et al., 2013). In addition to this variation, the cyclostome Dlx cognates are expressed in the NCCs of the premandibular region outside the PAs in both hagfish and lampreys (Myojin et al., 2001; Shigetani et al., 2002; Kuraku et al., 2010; Fujimoto et al., 2013). This premandibular expression does not exist in crown gnathostomes. Therefore, Dlx cognates in cyclostomes have dorsoventrally patterned expressions like crown gnathostomes (Cerny et al., 2010; Medeiros and Crump, 2012), but their expression patterns differ (a) from crown gnathostomes, (b) between the PA I and the rest of the pharyngeal series, and (c) from species to species.

These observations are consistent with the scenario of assimilation predicted by the Mandibular Confinement Hypothesis (Miyashita, 2015). That is, the distinct evolutionary history of the PA I is consistent with the exclusion of the PA I from the Hox code in both jawless cyclostomes and jawed gnathostomes. Similarly, the Dlx expressions and the musculoskeletal morphology ancestrally differed between the PA I and the rest of the pharyngeal series, as seen in cyclostomes. Only at the origin of the jaw, the PA I derivatives (jaw) assimilated the pattern of the branchial skeleton from the rest of the pharyngeal series. This event can be interpreted as assimilation of the Dlx expression patterns between the PA I and other PAs—in line with the pan-pharyngeal expression of the Dlx code across crown gnathostomes (reviewed by Miyashita, 2015). This scenario of assimilation better explains the serial pattern of the jaw, hyoid, and branchial skeletons in crown gnathostomes than the scenario of differential specialization from the elusive archetypical ancestor with identical skeletons in all the PAs.

Accepting that elements of the PA I did not ancestrally share serial patterning program that operated in other PAs, what distinct features did the ancestral PA I exhibit? The living cyclostomes provide insights into this question. In cyclostomes, the NCCs and MMCs of the PA I are not spatially confined as in those of other PAs in the same animals or those of the PA I in crown gnathostomes. The boundaries observable in crown gnathostomes between the NCC and MMC populations of the PA I is shifted in position or has different attributes in cyclostomes (Miyashita, 2015). These features are linked to the adult morphology.

The anterior interface for the PA I in crown gnathostomes sits between the premandibular region and the PA I (Table 1, Figures 2, 5). There is no pharyngeal pouch that separates these two domains as between PAs. Instead, the skeleton around the hypophyseal pit mark this boundary (Kuratani et al., 2004, 2013). Earlier during development, the trigeminal NCCs migrate in three distinct subpopulations: pre-optic, post-optic, and mandibular (Kuratani, 2012). The pre-optic and post-optic NCCs fill in the premandibular region and proliferate around the tripartite olfactory and adenohypophyseal placodes to form the trabecula cranii, interorbital septum, and nasal capsule, whereas the mandibular NCCs migrate to the PA I and interact with the mandibular MMCs to form a jaw apparatus (Kuratani et al., 2001; Eberhart et al., 2006; Szabo-Rogers et al., 2009; Wada et al., 2011). In some crown gnathostomes such as axolotls, the subpopulations initially occupy slightly different positions, but the fates are still distinct from one another (Cerny et al., 2004a). The trigeminal NCCs of the mandibular subpopulation are Dlx-positive, whereas those of the pre-optic and post-optic are Dlx-negative (Depew et al., 2002; Shigetani et al., 2002). So in crown gnathostomes, the premandibular-PA I boundary can be recognized by Dlx expression, and the trigeminal NCCs of the mandibular subpopulation give rise to the skeletal derivatives of the PA I.

In cyclostomes, however, the premandibular-PA I boundary cannot be clearly delineated (Table 1, Figures 3, 5). Not only is it challenging to delineate three separate subpopulations from each other in histological observations, it is difficult to map identities and fates of the trigeminal NCCs with precision. This is (a) because all three subpopulations of the trigeminal NCCs express Dlx cognates in both lampreys and hagfish (Myojin et al., 2001; Neidert et al., 2001; Cerny et al., 2010; Kuraku et al., 2010) and (b) because the trigeminal NCCs of the post-optic position migrate to what corresponds to the “premandibular” region in crown gnathostomes, and integrate with the MMCs to form a muscular upper lip (Horigome et al., 1999; Kuratani et al., 1999, 2004; McCauley and Bronner-Fraser, 2003; Kuratani, 2012; Oisi et al., 2013b). Medial to this region, the cyclostome counterpart to the olfactory and adenohypophyseal placodes of crown gnathostomes is a single nasohypophyseal placode (Janvier, 1996; Oisi et al., 2013b). Therefore, the skeletogenic proliferation of the pre-optic and post-optic trigeminal NCCs does not occur between the tripartite placodes as in crown gnathostomes, but only surrounds the single placode in cyclostomes (Figure 5).

These features blur a premandibular-PA I boundary in cyclostomes. What would otherwise characterize a PA I in a crown gnathostome—the Dlx-positive NCCs and the MMCs—have different distributions in the cyclostome head. In crown gnathostomes, the PA I derivatives secondarily extend anteriorly later in development, but at the stage of the NCC migration, positions of the NCCs are delimited from each other and tightly linked to distinct skeletal fates (Cerny et al., 2004a; Wada et al., 2011; Kuratani et al., 2013). The maintenance of this boundary is critical in crown gnathostomes. When manipulation disrupts molecular signals expressed in structures that sit at the boundary and thereby delimit the PA I elements in these taxa (e.g., hypophyseal pit), the resulting phenotypes often include jaw defects and even cyclostome-like distribution of the differentiating PA I structures. The shh⁻ and noc⁻ mutant zebrafish show severe reduction of chondrogenic proliferation of the pre-optic and post-optic NCCs (Neuhauss et al., 1996; Eberhart et al., 2006). They also exhibit conditions that parallel cyclostome pattern: fusion of left and right palatoquadrate anterior to the notochord (shh⁻) and lower labial cartilage upturned to overlap the premandibular region from lateral side (noc⁻) (Figure 4E; reviewed by Miyashita, 2015).

The PA I is delimited posteriorly from the PA II by the hyomandibular pouch (hmp in Figure 2). The pouch sits between the PAs I and II and below the geniculate ganglion of the facial nerve (a branchiomeric nerve for the PA II). The pouch epithelium plays a critical role in patterning the Hox-negative trigeminal NCCs of the PA I (Couly et al., 2002) and often persists into adulthood in crown gnathostomes as the spiracle or paratympanic cavity. A variety of sensory structures develop in the hyomandibular pouch of crown gnathostomes, including the pseudobranch (folded epithelial structure with sensory/thermoregulatory functions) and the spiracular organ (paratympanic organ) (O'Neill et al., 2012). These structures are associated with the PA II. They are innervated by the facial nerve and/or irrigated by the hyoidean arteries (reviewed by Miyashita, 2015).

Cyclostome development shows no evidence for these hyomandibular structures (Figure 3). The hyomandibular pouch never opens externally as in crown gnathostomes. Instead, the trigeminal NCCs and mandibular MMCs of the PA I occupy this position to form an outpocket into the pharynx, which later becomes a ventilation structure called velum (reviewed by Oisi et al., 2013b; Miyashita, 2015). Although, differentiated from elements of the PA I, and although innervated by the trigeminal nerve, the velar skeleton abuts against the skeleton of the PA II in both hagfish and lampreys (Figure 3B; Janvier, 1996).

As such, the hyomandibular pouch persists as a barrier for the NCCs and MMCs of the PA I only among vertebrates with jaws, probably as a functional requirement for the PA II structures that the pouch hosts. Despite considerable spatial overlap between the PA I- and PA II-derived muscles later in ontogeny, the anlagen are spatially confined in crown gnathostomes. The NCCs and MMCs of the PA I do not observe such clear boundary in cyclostomes. To illustrate the importance of the spatial confinement at the hyomandibular pouch in crown gnathostomes, induced reduction of the pouch (e.g., fgf8⁻ and fras1⁻ mutant zebrafish) is linked to severe defective phenotypes in the jaw skeleton (Figure 6B; Crump et al., 2004; Talbot et al., 2012). Conversely, pax1^−∕− medaka fish retain the hyomandibular pouch but fail to form segmented pouches for the rest of the pharynx (Okada et al., 2016). Although these pax1 mutants develop jaws without pronounced phenotypic effect, the branchial skeleton is reduced to a severe defective phenotype, and the hyoid elements become fused to each other (Okada et al., 2016).

In crown gnathostomes, the NCCs and MMCs of the PA I are delimited along a mid-ventral boundary with the hypobranchial musculature, which functions as jaw depressors and/or a tongue (Figures 2, 4G; Sambasivan et al., 2011). The hypobranchial muscle precursors originate in the trunk (they are not part of the head branchiomeric musculature), migrate along a circumpharyngeal path, and extend anteriorly beneath the pharyngeal floor (Lours-Calet et al., 2014). The differentiated muscles become a deep mid-ventral muscle mainly connecting the pectoral girdle and the jaw. Several phenotypes with hypobranchial defects demonstrate confinement of the NCCs and MMCs of the PA I by the hypobranchial muscle precursors. These phenotypes can be induced in knockout/knockdown of the ventrally expressed elements of the endothelin and Dlx pathways (e.g., EdnRA). As the hypobranchial muscle precursors are reduced, the PA I-derived ectopic tissues extend mid-ventrally and develop in the space that would otherwise be occupied by the hypobranchial structures (Figures 4H,I; Abe et al., 2007; Ruest and Clouthier, 2009; Heude et al., 2010; Barron et al., 2011). The hypobranchial defects are linked with jaw defects in edn1 and myoD phenotypes of zebrafish (Miller et al., 2000; Hinits et al., 2011).

On the other hand, cyclostomes appear to have no such boundary between the PA I and the hypobranchial musculature (Figures 3, 4B). In both hagfish and lampreys, the NCCs and MMCs of the PA I extend posteriorly and mid-ventrally to the pharyngeal floor to form a lingual apparatus (Holmgren, 1946; Johnels, 1948; Oisi et al., 2013a,b). Contribution to the lingual apparatus from other PAs in cyclostomes would reject the prediction that the PA I-derived structure in midventral pharyngeal position only became spatially confined at the origin of the jaw. However, no such evidence exists. This cyclostome condition is strikingly similar to the EdnRA mutant phenotype of mice in which the ectopic PA I structures occupy the mid-ventral position (Figures 4H,I; reviewed by Miyashita, 2015). Furthermore, the cyclostome muscles that correspond to the crown gnathostome hypobranchial musculature form the most superficial muscle layer in the ventral region of the pharynx, instead of becoming a deep mid-ventral muscle (Kusakabe and Kuratani, 2005, 2007). These muscles draw their precursors broadly from somites, not just restricted to occipital somites as in crown gnathostomes, and continue to extend anteriorly into the head well after the NCCs occupied the PAs (Kusakabe et al., 2011; Oisi et al., 2015). Therefore, the PA I-hypobranchial interface as observed in crown gnathostomes does not exist in cyclostomes.

Coupled with fossil and anatomical evidence (reviewed by Miyashita, 2015), the establishment of these interfaces of the PA I appears to have coincided with the origin of the jaw, because only jawed vertebrates exhibit full attributes of spatially confined NCCs and MMCs of the PA I (Figure 1A). The spatial confinement would have allowed the NCCs and MMCs of the PA I—which would otherwise have differentiated into peripheral structures as seen in cyclostomes—to acquire new developmental fates (Figure 4A). A likely scenario is that they co-opted patterning programs that operate in more posterior PAs, such as the pharyngeal Dlx code. Then serial similarities among the PA-derived structures—absent in cyclostomes but present in crown gnathostomes (e.g., jaws and branchial bars)—should reflect incorporation of the PA I into the serial patterning of the PA derivatives at the origin of the jaw. A full scope of this Mandibular Confinement Hypothesis and an evaluation of evidence and counterevidence are provided in the original paper (Miyashita, 2015). The present review looks to new evidence and future directions that were not considered in length in Miyashita (2015).

Spatial confinement is a difficult concept to test, as it ultimately requires precise fate mapping and specific functional analysis. Crucially, the spatial confinement proposed at the three interfaces concerns the undifferentiated NCCs and MMCs of the PA I (Miyashita, 2015). It does not necessarily mean that differentiated structures continue to occupy the same positions as in the embryonic PA I, or that the area occupied by the NCCs and MMCs of the PA I quantitatively decreased in proportion to schematic diagrams. Instead, the confinement implies that the PA I interfaces clearly observed in crown gnathostomes likely had different attributes in jawless ancestors. The NCCs and MMCs “confined” to the PA I as in other PAs—and the resulting serially similar patterns across the pharynx—are a feature specific to vertebrates with jaws. Fate mapping experiments in zebrafish, chick, and Xenopus (and other amphibian models) corroborate delimitation of the NCCs and MMCs of the PA I from three proposed interfaces (trabecula cranii, hyomandibular pouch, and hypobranchial muscles; e.g., Couly et al., 1993; Richman and Lee, 2003; Eberhart et al., 2006; Benouaiche et al., 2008; Gross and Hanken, 2008; Szabo-Rogers et al., 2009; Wada et al., 2011; Talbot et al., 2012; Lours-Calet et al., 2014). Such high-resolution fate mapping remains as a challenge for cyclostomes.

In zebrafish, for example, the premigratory trigeminal NCCs that would contribute to the trabecula cranii and other cartilages of the premandibular region originate more anteriorly than those that would contribute to the jaw cartilages of the PA I (Eberhart et al., 2006). Spatial segregation of the skeletogenic fates among the trigeminal NCCs is maintained through migration, and the “premandibular” trigeminal NCCs require the hedgehog signaling from the hypophyseal pit to chondrify (Eberhart et al., 2006). These observations document the predicted delimitation of the PA I-derived mesenchymal tissues from the premandibular region in crown gnathostomes, even though the elements that occupy this region in cyclostomes are derived from the position of the PA I (reviewed by Miyashita, 2015). Interestingly, the hedgehog-impaired shh⁻ mutant zebrafish have median fusion of the PA I-derived cartilages (Eberhart et al., 2006)—a condition reminiscent of cyclostomes (Figure 4E).

Overall, it waits for future tests to determine how robust predictions arising from the Mandibular Confinement Hypothesis are, but its implications are promising. Cumulative evidence suggests ancestral anisomerism in the gnathostome pharynx: the PA I derivatives were patterned differently from those of other PAs (with cyclostomes representing some examples of this), and the serially patterned state of the PA derivatives as observed in crown gnathostomes likely emerged with the origin of the jaw. Otherwise the PA I-specific features in cyclostomes and crown gnathostomes must have evolved independently within the respective lineages, and the cyclostome- or crown gnathostome-like traits identified in stem gnathostomes (such as galeaspids, osteostracans, and placoderms; Figure 1A) must have been misguided. There is no question that cyclostomes and crown gnathostomes have each undergone their own specializations, or that inferences for fossil phenotypes remain challenging. However, alternatives would be best offered through test of predictions arising from the Hypothesis. Currently, it requires far fewer character changes to postulate an anisomeric pharyngeal apparatus in the last common ancestor between cyclostomes and crown gnathostomes than to assume a polyisomeric pharyngeal apparatus.

The idea that the structures of the forelimb (FL) and hindlimb (HL) are “serial homologues” was first proposed by Vicq-d'Azyr (1774), Oken (1843), and Owen (1849, p. 184). Oken and Owen's ideas were profoundly influenced by Johann Wolfgang Goethe, and so was their notion of polyisomeric archetypes. Interestingly, in their FL and HL comparisons—which were almost exclusively based on skeletal traits—the use of the term “serial homology” referred to what is currently viewed as parallelism (i.e., a subset of homoplasy) and not to true homology. Examples of striking similarity between the FL and HL in Owen (1849) mainly come from tetrapods with highly derived limbs (e.g., bats, horses, and plesiosaurs). Owen used the term parallelism more often than “serial homology” to describe the FL-HL similarity. When he discussed outgroup lineages to tetrapods (e.g., chondrichthyans), he considered that those taxa “confuse” the notion of “archetype” and “serial homology.” Thus, Owen intended homoplasy rather than the concept of serial homology as we understand it today (Diogo et al., 2013).

There is a conceptual difference between developmental biologists and evolutionary biologists over the perception of serial homology, including segmentation. The former tend to focus on the use of similar developmental mechanisms (“developmental serial homology” sensu Wagner, 1994), whereas the latter tend to consider both developmental similarity and evolutionary continuity (“historical serial homology” sensu (Wagner, 1994; Brigandt, 2003). Reno et al. (2013) recently proposed a broadly applicable criterion to identify evolutionary/historical serial homology: “confirmation of its evolutionary continuity throughout a lineage, such that the structure can be shown to have a shared ancestry with segmental duplicates within the same body (paralogues, usually named serial-homologs) or similar structures in other lineages (orthologues, usually named homologs)” (p. 217). By contrast, Goethe's romantic idea of a polyisomeric archetype did not refer to a true ancestral serial pattern, but instead to a theoretical, ideal serial pattern that could be fulfilled through vitalist forces within the body (Richards, 2004). This idea of the vertebrate body composed of segments implies that the pelvic appendage is nothing more than a second pectoral appendage, which was in turn viewed as a derivative of a posterior branchial arch (Gegenbaur, 1859).

Despite the decline of romantic and vitalistic ideas, authors continued to view the structures of the pectoral and pelvic appendages as serial homologs (Diogo et al., 2013). Efforts to work out the origins of the paired appendages have not explicitly questioned the serial homology until recently. One such example is a hypothesis that fins evolved from continuous stripes of competency for appendage formation located ventrally and laterally along the embryonic flank (Shubin et al., 1997; Don et al., 2013). An extension of this hypothesis proposes that the paired appendages evolved with a shift in the zone of competency to the lateral plate mesoderm in conjunction with the establishment of the lateral somitic frontier, thus allowing formation of the limb/fin buds with endoskeletons (Don et al., 2013). The idea that paired appendages are serial homologs assumes that: (a) these appendages were originally similar to each other; and (b) there was a subsequent functional/anatomical divergence between them (Figure 7). In line with these assumptions, Don et al. (2013) proposed: (a) the ancestral Tbx4/5 cluster of vertebrates underwent a duplication event; (b) Tbx4 is related to the HL and Tbx5 with the FL; and (c) pectoral fins evolved first and then were duplicated to form pelvic fins.

This supposed polyisomeric to anisomeric polarity links two distinct hypothetical phenomena: (a) duplication of the Tbx4/5 cluster and subsequent co-option associated with the ontogeny of the different paired appendages; and (b) morphological duplication of the appendages at the levels of individual musculoskeletal elements. As we will argue below, (a) is true but (b) is not. Gene expressions may facilitate an outgrowth that gives rise to different limbs. However, these limbs do not necessarily represent serial homologs. An alternative explanation is that some genes and regulatory cascades/networks have been recruited due to homoplasy between FLs and HLs to coordinate limb development (Willmer, 2003). At least some of these genes/networks have also been recruited to coordinate the development of other appendages that are clearly not serial homologs with pectoral or pelvic appendages (sexual organs and head barbels; Archambeault et al., 2014).

If the paired appendages represent evolutionary (historical) serial homology, four testable predictions must be met:

Recent anatomical, developmental, and functional studies raise intriguing questions about the serial homology concept. Among them are marked differences in the development of muscles between the pectoral and pelvic appendage in both tetrapods and non-tetrapod gnathostomes (Cole et al., 2011; Don et al., 2013). These differences are clearly more marked in phylogenetically older structures (e.g., girdles and pectoral-arm and pelvic-thigh muscles attached to them) than in structures that are derived. The hypothetical pectoral-pelvic duplication event in early gnathostomes cannot explain this pattern. In no known extant tetrapod does topological correspondence exist between the muscles attached to the pectoral girdle (i.e., pectoral and arm muscles) and those attached to the pelvis (i.e., gluteal and thigh muscles) in configurations in adults or through ontogeny (Figure 8; Diogo and Tanaka, 2014; Diogo and Ziermann, 2014; Diogo et al., 2014a,b). In the zeugopodial and autopodial muscles, however, there are numerous topological correspondences (e.g., 19 in most salamanders) between FLs and HLs in embryonic development of many tetrapods. Some are lost later in development, whereas others persist in adults (Figure 8; see numbers in Figure 10).

The FLs and HLs of humans and other tetrapods reveal further heterogeneity from early development to the adult stage (Bardeen, 1906; Lewis, 1910; Čihák, 1972). There are marked differences between the configuration of the arteries of the human FL and HL in the stylopodial (arm vs. tight) and girdle (pectoral vs. pelvic) regions, not only in adults but also in early development (Figure 9). By contrast, remarkable similarities exist between the hands and feet in humans and some other tetrapods (e.g., salamanders) in the patterns of bones, muscles, nerves, and blood vessels (Diogo et al., 2013). An analysis of development regulation of the pectoral and pelvic appendages in tetrapods and some fish helps explain these morphological patterns (Sears et al., 2015). Sears et al. provided the first detailed review of the developmental programs and gene networks of the pectoral and pelvic girdles and of the HL and FL of tetrapods. This analysis revealed that embryological origins are more similar between the zeugo- and autopodia of the FLs and HLs than between the pectoral and pelvic girdles. Known genes and regulatory gene interactions differ greatly between the pectoral and pelvic girdles, with few genes regulating patterning of both girdles, whereas there are many similar genes involved in patterning the FL and HL zeugo- and autopodia. Importantly, only some of the genes with known roles in pectoral girdle development have roles in development of the pelvic girdle, and vice versa. In these few mutually expressed genes, their functions, interacting partners, and tissue domains of expression may be different in each of the girdles. Overall, the networks that regulate patterning of the pectoral and pelvic girdles differ to a large degree, when the number of genes that are present in both networks is used as the metric. Specifically, they identified 10 genes with a role in pectoral girdle development that could be placed in the network, and 10 genes for the pelvic girdle; fewer than half (4 of 10) of the 10 genes in the pectoral girdle network are also present in the pelvic girdle network. In contrast, while the gene network regulating initial specification and outgrowth of the FL and HL fields differ in many respects, the networks regulating later stages of FL and HL outgrowth and patterning are more similar to each other. Namely, the networks regulating initial specification and outgrowth of the FL and HL share about half or slightly more than half of their genes (3 out of 5 for the FL and 3 out of 6 for the HL), whereas the networks regulating later stages of FL and HL outgrowth and patterning share almost all of their genes (27 out of 28 for the FL and 27 out of 29 for the HL). Sear et al. thus concluded that patterning of the regions of the FL and HL that represent major tetrapod innovations (i.e., the zeugo- and autopodia) occurred through the sequential (homoplastic) deployment of a similar developmental program.

Other authors have also previously proposed that similar genes were co-opted (“gene piracy”) during the transition from fish to tetrapods (e.g., Tbx4 and Tbx5 for the development of the HL and FL, respectively), rather than being ancestrally present in earlier vertebrates with pectoral and pelvic appendages (Roth, 1994; Willmer, 2003; Diogo et al., 2013). Specifically, Roth (1988, 1994) proposed that because of this “gene piracy,” the hindlimb and forelimb became anatomically more similar to each other than were the pectoral and pelvic appendages of non-tetrapod gnathostomes. This idea is defended by Diogo et al. (2013): the striking similarity between the forearm/hand and leg/foot of tetrapods is due to derived (homoplastic) events that occurred during the fin-limb transitions, and not to ancestral serial similarity between the pectoral and pelvic appendages. As such, prediction 4 is contradicted.

To test prediction 3, Diogo and Ziermann (2015) compared the musculoskeletal structures of the paired appendages of actinopterygians and chondrichthyans. They suggest that many of the strikingly similar FL-HL muscles of extant tetrapods evolved independently in each appendage because the ancestors of crown gnathostomes are predicted to have possessed only an adductor and an abductor in each fin. This is inconsistent with the idea that at least some muscles present in the tetrapod FLs and HLs were already present in the first vertebrates with pectoral and pelvic appendages. They propose that the origin of the pectoral girdle was instead likely related to head evolution, as illustrated by the cucullaris of gnathostomes such as chondrichthyans inserting onto both the branchial arches and pectoral girdle. Only later in evolution the cucullaris became differentiated into the levatores arcuum branchialium and protractor pectoralis, which gave rise to the amniote neck muscles trapezius and sternocleidomastoideus.

There are numerous evolutionary and functional reasons for the deep spatial relation between the skull and pectoral girdle in early gnathostomes: the girdle forms the rear wall of the internal branchial chamber—a shield for the pericardial cavity and a secure insertion for the pectoral fins (Coates and Cohn, 1998; Matsuoka et al., 2005). The pectoral appendage is deeply associated to the head developmentally through the use of strikingly similar developmental mechanisms, which include a Shh-dependence in both this appendage and the PAs in chondrichthyans, a commonality also seen in the pelvic appendage (Gillis et al., 2009; Gillis and Hall, 2016). It is interesting that these observations are consistent with Gegenbaur's (1859) old—and now often discredited—idea that the pectoral appendage may be a derivative of the posterior PA region (see below).

The existing data about the evolution of both fossil and extant gnathostomes contradict prediction 4. The pectoral appendages appeared earlier than the pelvic appendages: the former is present in jawless stem gnathostomes (Janvier et al., 2004; Janvier, 2007; Wilson et al., 2007), whereas the latter appears to be restricted to jawed gnathostomes (Zhu et al., 2012; Brazeau and Friedman, 2014). One could argue that the absence of evidence is not evidence of absence. For a long time it was thought that antiarch placoderms lack pelvic appendages, but a dermal pelvic girdle was reported recently in the antiarch Parayunnanolepis, which has also a mainly dermal pectoral girdle that consists partly of perichondral elements (e.g., a scapulocoracoid; Zhu et al., 2012; Figure 10). Therefore, it is not unreasonable to predict co-occurrence of pelvic and pectoral appendages in the deeper stem of gnathostomes. However, any proposition for simultaneous appearance of both appendages would require multiple events of secondary losses in the lineages without pelvic appendages (Wilson et al., 2007).

Instead of secondary loss, the tendency appears to be independent acquisition of different types of appendages in several early vertebrate clades. This is borne out by complex taxonomic distribution of not only the pectoral and pelvic appendages, but also of paired reproductive organs and “anal” fins in the stem of gnathostomes (Sansom et al., 2013; Brazeau and Friedman, 2014; Trinajstic et al., 2014; Long et al., 2015). Co-options of gene expression may explain the homoplastic acquisitions of these appendages. Patterns of Hox and Shh expression are similar in body appendages as diverse as the head barbels of paddlefishes and the sexual claspers of sharks (Archambeault et al., 2014). These Hox patterns are probably part of an ancient module that provided a shared genetic program that was co-opted in the independent (homoplastic) formation of appendages such as genitalia, barbels, and the vent of some fishes (a medial structure that is analogous to a urethra). Similar Hox patterns also govern the development of the pelvic and pectoral, as well as the medial, fins of sharks (Freitas et al., 2006). Even in lampreys, the formation of medial fins involves Hox genes, suggesting that paired appendages originated when gene expression patterns for the median fin were redeployed in lateral plate mesoderm (Freitas et al., 2006). As such, similar developmental mechanisms are at play to form appendages as diverse as sexual organs, vents, barbels, and median, pectoral, and pelvic appendages. Accordingly, it is interesting that both the intromittent organs and paired appendages of placoderms consist of dermal and perichondral components (e.g., Trinajstic et al., 2014; Long et al., 2015). This is also the case for the pectoral appendage in osteostracans (Janvier et al., 2004). Furthermore, similar mechanisms are also used in the formation of the PAs and both the pectoral and pelvic appendages, and this similarity has led Gillis and colleagues to resuscitate Gegenbaur's idea that the pectoral appendage might be derived from the PAs, or at least be associated to a co-option of some of the mechanisms used in the formation of these arches (Gillis et al., 2009; Gillis and Hall, 2016).

All these structures provide an illustrative example of deep homology (Shubin et al., 2009), which overlaps with the developmental criterion of serial homology (sensu Wagner, 1994) and assumes at least some degree of parallelism (homoplasy) under the evolutionary criterion of serial homology (Shubin et al., 1997; Hall, 2003; Wake et al., 2011). These iterative correspondences are broken by historical (evolutionary) discontinuity of phenotype. That is, there might be continuity of some genes/developmental mechanisms that account for the similarity between the pectoral and pelvic appendages. However, there is no historical continuity in phenotype in the sense that ancestrally these appendages were not similar to each other. There likely was no true morphological duplication of the whole appendages (Diogo et al., 2013). Within the context of the idea propagated by Gillis and colleagues, the evolutionarily later acquisition of the pelvic appendage could mainly be seen in two different ways. Firstly, it could be related to a co-option of some of the mechanisms shared by both the pectoral appendage and the PAs. In this regard, it is interesting that the pelvic fin converts into a large gill-like organ during breeding season in males of the lungfish Lepidosiren paradoxa (Foxon, 1933). A non-mutually exclusive alternative is a co-option of some of the mechanisms shared by other appendages, such as the medial (unpaired) appendages. In this sense, it is worthy to note that various authors have pointed out that both the embryologic development and adult configuration of the pelvic appendage in various fishes (e.g., Acipenser, Polyodon) are much more similar to those seen in the dorsal and anal fins than in the pectoral appendages (Danforth, 1913; Mabee and Noordsy, 2004). A different, third hypothesis was proposed by Tabin and Laufer (1993), but this hypothesis does not match the temporal sequence according to the fossil record available so far. Namely, after recognizing that the similarity between the structures of the fore- and hindlimbs of tetrapods is very likely the result of parallel (homoplasic) evolution, they suggested that hox genes were originally expressed only in the pelvic appendage. Then, later in the evolutionary history of gnathostomes, the subsequent ectopic activation of Hox genes in the developing pectoral appendage resulted in a homeotic transformation, imposing the pattern of the pelvic appendage on the pectoral appendage by initiating signaling centers originally specific to the hindlimb, such as the zone of polarizing activity.

Aside from this latter idea, fossil evidence also rejects prediction 3. In the antiarch Parayunnanolepis—the earliest and plesiomorphic stem gnathostome with both pectoral and pelvic appendages—these appendages are different anatomically (Figure 10; Zhu et al., 2012). This goes in line with a detailed, long-term work on the paired appendages of early vertebrates and their evolution in gnathostomes, which show that the pectoral and pelvic appendages are primitively distinct from each other in the stem of gnathostomes and only became more similar later, in osteichthyans and particularly in tetrapods (Coates and Cohn, 1998, 1999). Specifically, Coates and Cohn (1999, p. 678) stated that “pelvic fins therefore neither originate as simple copies, nor as identical serial homologs of the pectorals […] Patterns of primitive fin phylogeny therefore provide little evidence of parallel (or concerted) evolution between pectoral and pelvic appendages […] close similarity between pectoral and pelvic fins is therefore a specialized feature which is developed most clearly within sarcopterygian (lobe-finned) osteichthyans; but even in this taxon, which includes tetrapods and their closest fish-like relatives, pectoral and pelvic fins primitively differ.” Janvier et al. (2004) also pointed out that while the pectoral appendages of osteostracans are likely homologous with those of gnathostomes, the pelvic fins of the gnathostomes likely had a “rather different history”; in the gnathostome stem, the pelvic endoskeleton does not show the same process of radial formation that the pectoral appendages do (Janvier et al., 2004).

In summary, the four predictions of the serial homology model (1–4) are rejected on the basis of developmental, comparative, and fossil evidence. We support the “serial parallelism and bottleneck” model as an alternative (Figure 10; Diogo et al., 2013; Diogo and Molnar, 2014). The pectoral appendages appeared earlier than the pelvic ones, and were anatomically and functionally part of a cohesive structural unit that may have included the head. This link was likely developmental as well, as indicated by the Shh dependence (Gillis et al., 2009) and by the cucullaris—a muscle connecting the head and pectoral girdle in placoderms and chondrichthyans (Trinajstic et al., 2013)—as part of the cardiopharyngeal field (Diogo et al., 2015). The pectoral and pelvic girdles and the muscles attached to them remained markedly different across gnathostomes. In all tetrapod clades listed in Figure 10, no pelvic-thigh muscle corresponds topologically to a pectoral-arm muscle.

The FL-HL similarities in zeugopodial and autopodial bones and muscles were therefore acquired secondarily across the fish-tetrapod transitions (Figure 10). The zeugopodia and autopodia of tetrapods have ossified skeletons, and the developmental patterns are in general quite different from those in distal elements of the pectoral and pelvic fins in fishes (Don et al., 2013). The zeugopodia- and autopodia-specific patterns include cooption of paralogues (e.g., Tbx4 and Tbx5) as a result of “gene piracy” (Roth, 1994), and these cooptions explain similarity between the distal regions of these limbs in tetrapods (a “similarity bottleneck”: Figure 10). In salamanders such as axolotls, 19 muscles/muscle groups of the leg-foot clearly seem to “correspond” topologically to forearm-hand structures (Figure 4).

However, gene piracy alone is not sufficient to explain the striking similarity between the leg-foot and forearm-hand muscles of some derived tetrapods such as horses or humans (Figure 10). In humans, 19 muscles/muscle groups of the leg-foot have clear counterparts in the forearm-hand structures. However, the number is only 16 in quadrupedal mammals (e.g., rats) and 14 in reptiles (e.g., lizards), respectively (Figure 10). Importantly, many of the human FL muscles/muscle groups with clear counterparts in the HL were acquired independently within primates (Diogo et al., 2013). For example, the adductor hallucis and adductor pollicis are similar because they have well-differentiated transverse and oblique heads. However, the heads of the adductor pollicis only became well-differentiated in the node leading to catarrhines (Old World monkeys + hominoids), whereas those of the adductor hallucis are already well-differentiated in primitive primates such as lemurs.

The 'similarity bottlenecks' leading to derived taxa such as humans are also influenced by topological and functional factors. For example, the only way to have a functional abductor hallucis or pollicis brevis (both muscles were acquired as homoplasies during tetrapod evolution) is to have a muscle lying on the radial or tibial side and then having it insert onto the radial or tibial side of the first digit of the hand or foot. Another crucial point is that various leg-foot muscles that are strikingly similar topologically to forearm-hand muscles in derived taxa do not develop from similar anlagen. For example, the FL muscle extensor pollicis longus and the HL extensor hallucis longus of humans are remarkably similar topologically, but the former derives from the anlage of the short extensors of the hand, whereas the latter derives from the anlage of the long extensors of the leg (Diogo et al., 2013; Diogo and Wood, 2015). In summary, during the evolutionary history of the tetrapod FLs and HLs, there are cases of evolutionary divergence leading to differences, whereas substantial evolutionary parallelism occurred in subsequent “similarity bottlenecks.” The strikingly similar FL and HL muscles are therefore the result of a complex interplay between ontogenetic, topological, functional factors, and not due to serial homology between the appendages in non-tetrapod ancestors.

Serial pattern can arise via assimilation between structures that have different evolutionary histories. We present two case studies in the origin of the jaw and serial patterning of the paired appendages among vertebrates. These two structures have long been considered in the context of differential specialization of initially polyisomeric (identically or similarly patterned) units, but we suggest an alternative mechanism: serial assimilation of anisomeric (dissimilarly patterned) structures. At the level of musculoskeletal patterning, the pharyngeal apparatus are serially patterned only in jawed vertebrates. The mandibular and non-mandibular PA derivatives have separate evolutionary histories and have paralleled each other since the origin of the jaw. Spatial confinement of the PA I-derived structures likely allowed serial assimilation of the PA derivatives via acquisition of the Dlx code across the pharynx. A similar pattern is found in the evolution of the paired appendages: they appeared at different times, and were significantly different morphologically in early gnathostomes, at which stage the pectoral appendages were anatomically and functionally more closely related to the head than to the pelvic appendages. Many of the similarities between these appendages were secondarily acquired in parallel during and after the fin-limb transition (likely due, at least in part, to co-option or “gene piracy”).

These two examples reveal a break in historical continuity of serial pattern shared between corresponding parts. Although the concept of serial homology has generated a controversy over different definitions and criteria, the decoupling of historical continuity and developmental correspondence is real in the cases of the vertebrate pharyngeal apparatus and paired appendages, and a polyisomeric ancestral phenotype likely is not. Under the criterion of historical continuity, differential specializations of serial homologs are an untenable explanation for serial patterns observed among these structures. We suggest two antidotes to assuming a polyisomeric ancestral phenotype a priori. First, anisomeric patterns should be tested as an alternative hypothesis. Second, similarity can only be assessed in particular context, such as at specific developmental stages or within certain clades. It remains to be explored whether or not serial assimilation helps account for the evolution of other serial structures such as vertebrae, but polyisomeric archetypes needs not serve as a default ancestral state.

TM drafted the section on jaws, Introduction, and Concluding Remarks. RD drafted the section on paired appendages.

Howard University, National Sciences and Engineering Research Council (Canada).