HoxD expression in the fin-fold compartment of basal gnathostomes and implications for paired appendage evolution

The role of Homeobox transcription factors during fin and limb development have been the focus of recent work investigating the evolutionary origin of limb-specific morphologies. Here we characterize the expression of HoxD genes, as well as the cluster-associated genes Evx2 and LNP, in the paddlefish Polyodon spathula, a basal ray-finned fish. Our results demonstrate a collinear pattern of nesting in early fin buds that includes HoxD14, a gene previously thought to be isolated from global Hox regulation. We also show that in both Polyodon and the catshark Scyliorhinus canicula (a representative chondrichthyan) late phase HoxD transcripts are present in cells of the fin-fold and co-localize with And1, a component of the dermal skeleton. These new data support an ancestral role for HoxD genes in patterning the fin-folds of jawed vertebrates, and fuel new hypotheses about the evolution of cluster regulation and the potential downstream differentiation outcomes of distinct HoxD-regulated compartments.

of HoxD10-13 during bud outgrowth 33 . This first phase of collinear expression (i.e., HoxD13 is most posteriorly restricted) patterns the proximal limb (stylopod and zeugopod), and is predominantly regulated by enhancer elements positioned telomeric (3′ ) to the cluster 33,34 . In contrast, during the late phase of expression, HoxD9-13 are activated in the autopod, forming an inverse collinear arrangement (i.e., HoxD13 is most broadly expressed) that patterns the digits 33 . The enhancer elements that drive this late phase are interspersed across a region centromeric (5′ ) to the cluster and have been shown to regulate the expression of Evx2 and LNP, two non-Hox genes within the 5′ genomic neighborhood [35][36][37] . Together, these observations from mice underscore a topological switch between domains of chromatin interaction to pattern the zeugopod and autopod, respectively. Interestingly, one current model for limb development derived from these data posits that this regulatory transition creates a zone of low Hox dosage that manifests anatomically as the small, nodular bones of the wrist and ankle 34,38 .
Among non-tetrapod gnathostomes, the distribution of HoxD transcripts during paired fin development has been partially characterized in teleosts [zebrafish: refs 21, 24 and 42]. These in situ studies reveal a conserved, collinear pattern of expression in the proximal region of early fin buds, similar to the first phase of HoxD expression in tetrapods 18 . In later stages of paired fin development, HoxD boundaries shift, suggesting a change in regulatory control. The dynamics of these shifts, however, vary with lineage. In zebrafish, a late phase of HoxD11a-13a expression appears to extend into the mesenchyme of the dermal fin-fold, creating a patterning compartment in which the anterior and posterior boundaries of each paralogue are in register 24 . In contrast, current data sets for paddlefish and catshark indicate that 5′ HoxD transcripts form an inverse collinear profile 22,23,41 . In paddlefish, this late phase overlaps early expression at the site of distal radial formation in the mid-fin, and has been interpreted as evidence of a deep developmental homology (sensu refs 43 and 44) between the distal radials and digits 22 (but see ref. 38). In catshark, late phase HoxD transcripts localize to the distal margin of the fin 23,41 , and unlike actinopterygians are spatially separated from the early phase of expression 45 . Overall, these data from non-tetrapods suggest an evolutionary scenario in which distinct proximal and distal HoxD patterning compartments were already in place in the paired fins of ancestral gnathostomes 18,23,45 .
Recent studies have focused on testing whether the regulatory landscapes that control Hox expression in fins and limbs are functionally conserved. Significantly, this work has shown that several of the enhancer elements that drive digit expression in mice are represented in non-tetrapods 17,25,29,35,37,46  Additionally, there is evidence that trans-acting factors are at least partially conserved between fins and limbs, as a tetrapod specific element (mouse CsC) has been shown to drive distal fin expression when inserted into zebrafish 26 . Together, these results provide evidence that the origin of limb-specific morphologies involved modification of an ancient, conserved regulatory architecture for HoxD activation already in place in fins.
Current models for the fin to limb transition that integrate these data focus on the role of HoxD genes in patterning the endoskeleton, though with alternative perspectives on the relationship between cluster regulation and anatomical homology 17,18,28,29,38,45 . In these models, the fin-fold is considered a Hox-free zone, and its formation is thought to interrupt the ectodermal/mesenchymal signaling interactions that determine the relative size of a Hox-patterned endoskeletal mesenchyme 17,23,24 . Key to further informing models of appendage evolution is a better understanding of the phylogenetic distribution of HoxD expression relative to the proximo-distal fin compartments in basal gnathostomes. Herein, we characterize the expression of HoxD genes, as well as the cluster-related genes Evx2 and LNP, in a basal actinopterygian, the American paddlefish Polyodon spathula. Our results demonstrate a posteriorly restricted, collinear pattern of nesting in the early fin buds that includes HoxD14Beta, a gene Scientific RepoRts | 6:22720 | DOI: 10.1038/srep22720 previously thought to be insulated from canonical Hox regulation in the mesoderm 14,47 . Additionally, we show that in later stages of fin development, the collinear expression profile established in early fin buds is maintained along the site of distal radial formation, with no "autopodial-like" anterior expansion of HoxD13 (contra a previous report for Polyodon 22 ), Evx2, or LNP. Most significantly, we observed a proximo-distal dynamic to HoxD expression in which late phase transcripts of both 3′ and 5′ paralogues extended into the fin-fold mesenchyme, co-localizing with And1, the elastoidin component of the actinotrichia 48 . In order to determine the phylogenetic distribution of this condition, we compared the expression domains of And1 and HoxD12 in a representative chondrichthyan, the lesser-spotted catshark Scyliorhinus canicula. These in situs revealed an overlap between distal HoxD and And1 expression, much like that of paddlefish, suggesting HoxD-positive cells contribute to fin-fold specific tissues in both taxa (the ceratotrichia and actinotrichia of chondrichthyans and actinopterygians, respectively). Our results support an ancestral role for HoxD genes in patterning the fin-fold, bringing new information to current models for fin/limb evolution and fueling novel hypotheses about the relationship between cluster regulation and appendage morphology.

Results
Expression of HoxD cluster genes in paddlefish paired fins. A whole genome duplication event occurred in the paddlefish lineage approximately 42 million years ago, resulting in the formation of unique Alpha and Beta HoxD clusters 13 (Fig. 2a). In order to determine if both clusters are transcriptionally active during development, we surveyed our Polyodon transcriptome assembly using published genomic BAC sequences for specific HoxD Alpha and Beta paralogues 13 . These blast searches identified both Alpha and Beta sequences of HoxD8, HoxD12, and HoxD13, and Beta sequence of HoxD11. Unexpectedly, we also identified transcripts of HoxD14, a Beta cluster gene previously hypothesized to be inactive during appendage patterning in vertebrates 14,47 .
In order to characterize the spatiotemporal expression dynamics of HoxD cluster genes during paired fin development, we performed a series of in situ hybridizations using probes targeted against either Alpha or Beta paralogues. Because the distribution of labeling appeared similar between genes from duplicated clusters (compare Fig. 2b and Supplemental Fig. 1), we limit the descriptions presented here to the HoxD Beta cluster only. It is worth noting, however, that we cannot rule out the possibility of probe cross reactivity due to high sequence similarity between a given set of paralogues (e.g., HoxD11Alpha and HoxD11Beta), and that positive staining may reflect their combined distribution of transcripts. Additionally, we could not assign an Alpha/Beta identity to the HoxD4 clone used in our analyses in the absence of available genomic sequence from the 3′ end of each duplicated cluster 13 .
In Stage 42 embryos, transcripts of HoxD4 and HoxD8Beta were detected along much of the length of the pectoral fin buds (Fig. 2b). In contrast, transcripts of HoxD11Beta-14Beta were more restricted in distribution, forming a collinear, posteriorly nested pattern of expression that mirrored the 3′ -5′ arrangement of paralogues within the cluster (Fig. 2a,b). In more advanced embryos (Stages 44 and 46), the expression domains of HoxD4, HoxD8Beta, and HoxD11Beta-HoxD13Beta persisted in the middle of the fin following outgrowth of the distal fin and fin-fold (see arrowheads in . We also examined the distribution of LNP and Evx2 to infer if these non-Hox genes are coordinately regulated with 5′ HoxD paralogues, as in mice where Evx2 and LNP co-localize with HoxD13 35 . In paddlefish, Evx2 expression was restricted to the posterior most region of the fin and appeared similar in distribution to HoxD14Beta (Fig. 2b). LNP, in contrast, was detected throughout the pectoral fins, with no distinct regionalization (Fig. 2b). Pelvic fins mirrored the 5′ HoxD and Evx2 expression patterns observed in pectoral fins (Fig. 2c). Transcripts of the more 3′ -ward paralogues HoxD4 and HoxD8Beta, however, were difficult to detect in whole mount.
HoxD expression in the fin-fold compartment of paddlefish. During limb development, HoxD genes are expressed in well-characterized proximal and distal zones that correspond to the stylopod/zeugopod and autopod, respectively. Given that fins are highly regionalized along the proximo-distal axis into endoskeletal radials (proximally) and a fin-fold/dermal skeleton (distally) (Fig. 1), we sought to characterize the distribution of HoxD transcripts relative to the early patterning zones that give rise to these morphologically distinct compartments. Actinodin genes encode non-collagenous (elastoidin) components of the actinotrichia and provide an early molecular marker for cells contributing to the fin-fold 48 . We identified a homologue of Actinodin1 (And1) in paddlefish, which we used for both single and double colorimetric in situs along with the pre-chondrogenic marker Sox9 to visualize the early endoskeletal radials (Fig. 3). In Stage 41 embryos, And1 transcripts were detected along the apical margin of the pectoral fin buds, forming a labeling boundary just distal to the Sox9 expression domain (Fig. 3a). In later stage fins (Stage 42-45), the relative position of the And1/Sox9 boundary was maintained despite considerable growth in both compartments. Cross sections through developing fins at these stages revealed And1 transcripts in both the fin-fold mesenchyme and the ectoderm adjacent to the basement membrane (Fig. 3b). Notably, at stage 45 the proximal margin of And1 labeling extended beyond the distal, lateral margins of Sox9 labeling, presaging the ultimate arrangement of the dermal fin supports and cartilaginous fin radials described for mature fins (see Fig. 3 in refs 49 and 50). Our And1/Sox9 data make visible the proximal boundary of the fin-fold patterning compartment at various stages of pectoral fin development, and show that prechondrogenic cells are excluded from the population of mesenchyme distal to this boundary. We next performed a series of double colorimetric in situs to determine if HoxD paralogues co-localize with And1 in the fin-fold mesenchyme. In whole mount pectoral fins at Stage 45, transcripts of HoxD11Beta-HoxD13Beta formed the same posteriorly-restricted pattern of nesting along the mid-fin (Fig. 4a) as in single in situs (Fig. 2b). Significantly, the expression of these 5′ HoxD paralogues extended distally to overlap with the And1 expression domain. Cross sections revealed differences in the profile of HoxD and And1 boundaries, depending on their position along the anterior-posterior axis (Fig. 4c). Whereas no 5′ HoxD transcripts were detected in the anterior portion of the fin, in more posterior sections HoxD11Beta-HoxD13Beta were expressed throughout the fin-fold mesenchyme, with a proximal boundary for strongest labeling that roughly aligned with that of And1 in the ectoderm. In sections near the caudal margin of the fin, no such alignment was observed, as HoxD11Beta-Hox13Beta labeling was continuous between the fin-fold and proximal fin mesenchyme.
In mice, early phase HoxD patterning of the proximal limb is primarily driven by the 3′ telomeric landscape and involves the upregulation of paralogues along the length of the cluster (i.e., HoxD1-13) 33,34 . Regulatory control over HoxD expression then transitions to the 5′ centromeric landscape to pattern the autopod, where only HoxD9-13 (i.e., those paralogues neighboring the 5′ end of the cluster) are active 33,37 . In paddlefish, our in situ data demonstrate that 5′ HoxD paralogues are expressed in the fin-fold (Fig. 4a), which like the autopod in mice represents the distal-most compartment of the appendage. In order to determine if HoxD expression in the fin-fold is restricted to genes at the 5′ end of the cluster (similar to late phase autopod patterning), we compared the expression of And1 and the more 3′ -ward HoxD paralogues HoxD4 and HoxD8Beta. In whole mount fins, strong And1 labeling in the ectoderm made it difficult to discern 3′ HoxD labeling within the fin-fold. In section, however, transcripts of both HoxD4 and HoxD8Beta were detected in the fin-fold mesenchyme (Fig. 4b), and like the 5′ paralogues examined (Fig. 4a), aligned proximally with the And1 labeling boundary in the ectoderm.
HoxD expression in the fin-fold compartment of chondrichthyans. To test whether Hox patterning of the fin-fold compartment is a derived condition of paddlefish or a more general feature of gnathostomes, we extended our analysis to the lesser-spotted catshark Scyliorhinus canicula as a representative chondrichthyan. Chondrichthyan dermal fin skeletons are composed of ceratotrichia, which are histologically and chemically similar to actinotrichia 51 . Blast searches of the catshark transcriptome assembly returned a putative Actinodin  Fig. 2b). Significantly, cross sections confirm that 5′ HoxD expression extends distally to overlap the And1 expression domain. 5′ HoxD transcripts were not detected in the anterior fin (sections labeled 1). More posterior sections reveal expression throughout the fin-fold mesenchyme, with a proximal boundary that roughly aligned with ectodermal And1 (sections labeled 2). In sections near the caudal margin of the fin, 5′ HoxD expression is continuous between the fin-fold and proximal fin mesenchyme (sections labeled 3). (b) Co-expression of the more 3′ -ward HoxD genes HoxD8Beta, and HoxD4 (purple) with And1 (red) reveals transcripts of both Hox paralogues within the fin-fold mesenchyme and aligned proximally with the And1 labeling boundary in the ectoderm (sections labeled 1, 2). Anterior is left, distal is up for all whole mounts. Ventral is left, distal is up for all sections. Scale bars = 200 nm.
homologue, which we identified as ScAnd1. Whole mount in situs revealed ScAnd1 expression in the distal pectoral, pelvic, and median fins ( Fig. 5a and Supplemental Fig. 3), consistent with the distribution of labeling in actinopterygians (this study, and ref. 48). Cross sections through Stage 30 pectoral fins confirmed that ScAnd1 expression was both mesenchymal and ectodermal and marked the boundaries of the fin-fold (Fig. 5a).
In catshark, 5′ HoxD genes are expressed in early and late waves, patterning the proximal and distal regions of the fin, respectively 23 . In order to determine if the distribution of late phase HoxD transcripts extends into the fin-fold compartment, we compared the expression of HoxD12 with our ScAnd1 results. In Stage 30 pectoral and pelvic fins, HoxD12 labeling formed distinct domains proximally and distally, consistent with previously published results 23 (Fig. 5b). Significantly, cross sections showed considerable overlap between the distal domain of HoxD12 and ScAnd1, indicating HoxD expression in the fin-fold mesenchyme of catshark (Fig. 5b).

Discussion
In this study, we have characterized the expression of HoxD and HoxD cluster-related genes during paired fin development in the American Paddlefish Polyodon spathula. These results demonstrate an early collinear arrangement of 5′ HoxD nesting, similar to that of other vertebrates [21][22][23][24]33 and consistent with a model in which the first phase of HoxD activation is part of a conserved pathway for establishing Shh in the ZPA and appendage asymmetry [52][53][54][55] . Our results also demonstrate that in more advanced fins, collinear HoxD nesting is maintained along the presumptive site of distal radial formation following outgrowth of the fin-fold. This persistent mid-fin expression resembles the later dynamics of early phase patterning in limbs, where HoxD expression in the nascent limb bud continues in the fore-limb forming territory following outgrowth of the autopod 33,38 . Notably, we did not observe an extended anterior sweep of expression of HoxD13 (contra a previous report 22 ), HoxD14, or the cluster-related genes Evx2 or LNP, in any of the stages examined. These observations provide evidence that the inverse collinear profile considered a hallmark of digit patterning in tetrapods may not be present in paddlefish (or a more inclusive clade, see ref. 24).
In addition to early patterning of the proximal fin bud, our paddlefish in situ results reveal a late phase of HoxD expression that co-localizes with And1 in the mesenchyme of the fin-fold, a region absent of pre-chondrocytes. These new data, together with a report of late HoxD expression in the fin-fold of zebrafish 24 , suggest this character may be shared among actinopterygians. Chondrichthyans provide a critical outgroup for testing this hypothesis. Our data for catshark confirm the spatially distinct early and late phases of HoxD activation described by Freitas and colleagues 23 , and demonstrate overlap between the late distal phase of HoxD12 labeling and that of And1. In tetrapods, most studies of Hox function during limb development have focused on endoskeletal patterning and the effects of gene abrogation on the formation of the stylopod, zeugopod, and autopod [56][57][58][59][60][61][62] . Recent work, however, has shown that Hox genes have a broader role in limbs, and are required for normal muscle and tendon formation independent of skeletal phenotype 62 . While the anatomical outcome of HoxA/D loss in non-tetrapods remains elusive, the co-localization of HoxD and And1 labeling in a representative actinopterygian and chondrichthyan provide evidence of Hox-positive cells contributing to fin-fold specific tissues (i.e., the elastoidin component of the actinotrichia in paddlefish and ceratotrichia in catshark), and suggest an ancient role for Hox genes in integrating aspects of appendage formation beyond endoskeletal patterning.
Our unexpected find, the first report of appendicular expression of HoxD14Beta, is contra recent hypotheses that Hox14 paralogues are isolated from canonical regulation of the cluster 14,47 . Instead, our results demonstrate that transcripts of HoxD14Beta are posteriorly-restricted within the fins, and together with other cluster paralogues form a collinear arrangement of nesting, as would be expected with canonical early phase regulation. These results suggest an ancestral role for HoxD14 in appendage patterning in gnathostomes, one that has been lost in most lineages 14,47 yet retained, at least in terms of expression, in paddlefish.
Current evidence from comparative gene expression, genomics, and transgenic assays support a model in which bimodal regulation of the HoxD cluster is partially conserved between fins and limbs 25,26,28,29,46 . The apparent antiquity of this regulatory strategy 28 , together with the data presented here, raise the intriguing and underexplored possibility that HoxD patterning of the proximal fin and fin-fold may reflect separate controls from the 3′ and 5′ landscapes, respectively. In tetrapods, late phase HoxD-related expression in the autopod is limited to HoxD9-13 (those paralogues nearest the 5′ end of the cluster) and the 5′ flanking genes Evx2 and LNP, both of which are co-regulated with HoxD13 33,35 . We predicted a similar restriction in cluster activity in the fin-folds of paddlefish, but found transcripts of the 3′ -ward members HoxD4 and HoxD8Beta throughout the fin-fold mesenchyme. This difference, combined with a lack of anteriorized HoxD13, Evx2 and LNP, may reflect lineage specific variation in landscape/cluster interactions between tetrapods and paddlefish, or more generally actinopterygians 24 . Additional comparative data characterizing the expression of 3′ HoxD paralogues during fin development in a chondrichthyan will inform inferences about ancestral patterns of cluster activation, and fuel new hypotheses as to the regulatory dynamics of the 3′ and 5′ landscapes during appendage formation in non-tetrapods. Interestingly, current models for limb development posit that chromatin re-organization between early (proximal) and late (distal) HoxD expression creates an intermediate zone of low Hox dosage that manifests anatomically as the small bones of the wrist 32,34,38 . In fins, a series of small, endochondral bones (distal radials) often separate the proximal radials and fin-fold (e.g., Fig. 1). An ancestral role for HoxD genes in patterning the distal fin compartment makes it tempting to speculate that a similar mechanism may explain appendage morphologies in non-tetrapods. However, not all patterns meet the predictions of this model (e.g., strong HoxD labeling in the presumptive site of distal radial formation in paddlefish), suggesting that modulation of Hox dosage and its effects on cell behavior in the context of local signaling environments differ across lineages. Our results demonstrate that further comparative work, with an increasing focus on basal gnathostomes, will be necessary to elucidate the ancestral roles of Hox genes in appendage patterning, roles that may involve a great flexibility in skeletogenic differentiation outcomes. Transcriptome. Paddlefish larvae from early pectoral fin development stages (stages [41][42][43][44] 63 were preserved in RNAlater and then dissected into pectoral fin and reference tissue (operculum) samples. After RNA library construction, Illumina TM sequencing of the transcriptome was performed by the Centre for Applied Genomics -Toronto (Toronto, Canada) producing a total of 65,682,405 pectoral fin raw reads (72 bp nonpaired). We used the Trinity program (default settings) 66  PCR products were cloned into pGEM ® -T Easy Vector (Promega), which was purified and sequenced using Genscript DNA Sequencing Services. All sequences were analyzed using Unipro UGENE and BLAST (http:// blast.ncbi.nlm.nih.gov/Blast.cgi). Orthology for the Actinodin family members was confirmed using maximum likelihood in Mega6.06. GenBank accession numbers for this work are KU744647-KU744655. For in situ hybridization experiments, the entire PCR product obtained from each primer set was used as template for in vitro transcription (see below). Plasmids containing paddlefish fragments of HoxD11Alpha, HoxD11Beta, HoxD13Alpha, and HoxD13Beta were obtained through Life Technologies Gene Art Services and were based on published genomic BAC clones 13 . All new probes generated during this study ranged from approximately 480 to 900 base pairs in length. Paddlefish probes for HoxD4, HoxD8Beta, and HoxD14Beta were complementary to regions of coding exons one and two, including the homeodomain. Probes for HoxD11Alpha and HoxD11Beta were complementary to a region of coding exon one, coding exon two, including the homeodomain, and 3′ UTR. Probes for HoxD12Alpha, HoxD12Beta, HoxD13Alpha, and HoxD13Beta were complementary to the first coding exon and excluded the homeodomain. Probes for Sox9, And1, and LNP were targeted against coding region. Probe for Evx2 was targeted against coding region and 3′ UTR. Probes HoxD13-EF527821 and ScHoxD12 were from Refs 22, and 23 respectively.

Methods
In situ hybridization. Plasmids were linearized for probe template using restriction enzymes (New England Biolabs, NEB) and any resultant 3′ overhangs were blunted using DNA Polymerase I, Large Klenow (NEB). Probe synthesis was carried out using SP6, T7, or T3 RNA polymerases (Promega) and either Digoxigenin or Fluorescein RNA labeling mixes (Roche) per manufacturer′ s instructions. Whole-mount in situ hybridizations were as described in ref. 67 at a hybridization temperature of 69-70 °C, with the following modifications for double colorimetric in situs: Both Digoxigenin (Sox9, HoxD4, HoxD8Beta, HoxD11Beta-HoxD13Beta) and Fluorescein (And1) labeled probes were included during the hybridization step. Digoxigenin-labeled probes were detected with Anti-Digoxigenin-AP, Fab fragments (1:2000 dilution; Roche) and developed with BM-Purple (Roche). Following the first color reaction, embryos were incubated in 0.1M Glycine (pH 2.2) for 30 minutes to inactivate AP, and then washed thoroughly with MABT. Fluorescein-labeled probes were detected with Anti-Fluorescein-AP, Fab fragments (1:1000 dilution; Roche) and developed using Fast Red Tablets (Roche) per manufacturer's instructions.
Histology. Embryos used for sectioning were incubated in successive 10%, 20%, and 30% sucrose in PBS solutions for a minimum of 3 hours each at room temperature, and then transferred to a 1:1 solution of TBS Tissue Freezing Medium (Fisher): 30% Sucrose in PBS for 8 hours. Embryos were then embedded in TBS Tissue Freezing Medium in disposable plastic moulds and frozen for 10 minutes using isopentane chilled with liquid nitrogen. Embryos were cryosectioned at 16 μm on a Leica CM1850 cryostat, dried for two hours at 35 °C on a slide warmer, and cover-slipped with 25% glycerol in PBS. Sections were imaged using a Zeiss AxioImager.M2 compound microscope equipped with an Axiocam 503 color camera and Zen 2012 (blue edition) software.