Australian lungfish fins have the typical sarcopterygian Bauplan characterized by a central metapterygial axis at the base of which homologs of the humerus, radius, and ulna can be distinguished (Fig. 1) (5, 8). More distally, in the region of distal radials and rays, the homology with the limb remains controversially discussed (Fig. 1). In tetrapods, formation of the hand domain is governed by hoxa13, and its expression marks the boundary between the zeugopod (radius/tibia and ulna/fibula) and the autopod (hand/foot) (1, 21). In ray-finned fish, expression occurs primarily in the developing dermal skeleton (19, 22, 23), tentatively suggesting that digits might derive from fin rays. However, the possibility that an endoskeletal autopodial domain was already in place in the tetrapod ancestor remains untested due to a lack of data from sarcopterygian species. Using whole-mount in situ hybridization, we detect first expression of hoxa13 posteriodistally in lungfish pectoral fins at stages 42 to 44 (Fig. 2 and fig. S1), which at stage 45 has developed into a domain covering most of the fin except for the fin base (Fig. 2). At stage 47, a strong expression is observed in the condensing elements of the metapterygial axis (fig. S2), and at stage 48/49 also appears in the lateral domains where distal radials and rays will form (Fig. 2). To unambiguously distinguish between the endochondral and dermal fin domains, we used multicolor labeling for actinodin1/2 to visualize the fin fold [which foreshadows the ray domain (17, 22)] in combination with collagen2a1 and an anti-sarcomere antibody to label the endochondral skeleton with its musculature. This further confirms that the hoxa13 expression delimits the endochondral part of the skeleton distal from the radius and ulna (Fig. 2 and fig. S2); that is, at the same position along the proximodistal axis as in developing tetrapod limbs and its expression thus demarcates an autopod-like domain in the distal fin endoskeleton.

To evaluate these findings within the context of limb evolution, we note that tetrapod hands (as well as feet) are compound structures composed of the nodular bones of the mesopodium and the long bones of the digits and metacarpals. Together, these provide the distal unit of the highly constrained tetrapod “cross-articular” Bauplan that allows the hand palm to articulate with the radius and ulna via the wrist (Fig. 1) (1, 2, 9). Whereas, the expression of hoxa13 suggests that the “hand” domain has a pre-tetrapod origin, such cross-articulating arrangement of long bones interspersed by nodular bones is not found in any fin skeleton, including that of sarcopterygians. This morphological dissonance thus implies that the fish distal fin domain underwent substantial changes along its proximal-distal axis at the evolutionary transition from fins to limbs during the evolution of the tetrapod autopod. In this regard, it is relevant that the patterning processes for the proximal-distal and anterior-posterior axes of the tetrapod hand are deeply interwoven (24) and that key genes driving the development of the distal limb have a posteriorly originating expression domain. Hoxa13, hoxd13, and hand2 initiate their expression posteriorly in the early limb bud, and it is only during subsequent outgrowth that their domains become anteriorly expanded to occupy the distal limb margin (21, 25–27). At the same time, anterior markers, such as alx4 and pax9, do not extend distally and become excluded from the hand (28–30). Therefore, a phase of “progressive posteriorization” confers a posteriorly originating identity onto the most distal aspect of our limbs. To better understand their proximal-distal patterning, we analyzed the occurrence of this process in lungfish fins. Hoxd13 and hand2 become activated in a posterior domain at stage 42/43 (Fig. 2 and fig. S1) similar to their expression in mouse limb buds. At stages 44 to 46, during elongation of the fin bud, these domains become distally extended within the posterior fin (Fig. 2 and fig. S1), and at stages 48 to 50, they extend throughout the posterior fin in parallel with the metapterygial axis (Fig. 2 and fig. S2). During this process, hand2 expression becomes slightly expanded anteriorly, whereas hoxd13 maintains its approximate anterior-posterior expression boundary. Both genes occupy a posterior territory parallel to the metapterygial axis, resulting in an anterior domain in which they are not expressed [note that we do not detect an anterior to posterior progression of hoxd13 expression, as was suggested previously (8); also see Supplementary Materials and Methods]. On this opposite side of the metapterygial axis, the anterior marker alx4 (31) shows a spatiotemporal progression similar to the posterior genes, thereby creating a near mirror-image domain (Fig. 2). Expression of the anterior gene pax9 (30) is detected only around stage 45 and is expressed similarly to alx4 in an anterior territory that extends along the proximal-distal aspect of the metapterygial axis (Fig. 2).

In tetrapod limbs, a further manifestation of its posteriorization is the reverse-collinear expression of hoxd genes in the long bones of the hand. In the early limb bud, these genes are activated collinearly, whereby hoxd11 is expressed more anteriorly than hoxd13, a relationship that subsequently becomes reversed in the distal domain during a second phase of regulation (1, 21, 26). Analysis of hoxd11 expression (Fig. 2) shows that, at stages 42 to 48, hoxd13 is not expressed more anteriorly than hoxd11, indicating the absence of reverse-collinear hoxd expression during lungfish fin development.

In tetrapod limbs, the phase of progressive posteriorization is established through the activity of sonic hedgehog (shh) secreted from the zone of polarizing activity (ZPA), which antagonizes the repressive action of the transcription factor gli3 (24, 32–34). This shh signal is required for the anterior expansion of hoxa13, hoxd13, hoxd11, and hand2 (33–35) and for the concomitant reduction of the anterior alx4 and pax9 domains (30, 31). In the fins of the Australian lungfish, shh expression is first detected at very low levels throughout the posterior fin around stage 43 and develops into a posterior ZPA that is present from stages 44 and 45 (Fig. 3), after which shh expression becomes undetectable. Expression of gli3, which, in turn, determines anterior limb identity (30, 31, 33–36), is detected at low intensity from stages 42 to 44 and increases strongly throughout the fin afterward at stage 45 (Fig. 3). At stage 42, gli3 is expressed with the same anterior bias as in the early mouse limb (32).

The maintenance of the ZPA during limb development depends on a feedback loop whereby shh activates the secreted BMP inhibitor gremlin1, which is required for fgf expression in the apical ectodermal ridge (AER) (37, 38). Sustained fibroblast growth factor (FGF) signaling by the posterior AER, in turn, provides a signal for ZPA survival. Gremlin1 expression in Australian lungfish fins was detected in a central domain similar to that observed in tetrapod limbs, present on both anterior and posterior sides of the metapterygial axis originating posteriorly, consistent with its activation by shh (Fig. 3 and fig. S1). Fgf8 expression in the posterior AER of Australian lungfish fins has been reported from early fin budding until the transition of the AER into an apical ectodermal fold (AEF) around stage 45 (39), coinciding with the disappearance of the ZPA and shh expression around this stage. Therefore, the components of the shh-gli3 axis and the shh-gremlin1-fgf8 feedback loop are expressed in a manner consistent with a role in anterior-posterior patterning similar to that in tetrapod limbs, although perhaps with different dynamics that could explain the different posteriorization of lungfish fins (see below).

Lungfish fins show strong resemblance to tetrapod limbs with respect to the proximal-distal patterning of their endoskeleton into an arm and hand domain by the posterior hoxa genes, whereby hoxa11 marks the stylopod-zeugopod transition (5) and hoxa13 marks the domain distal to the zeugopod. Therefore, an autopod-like domain that pre-dates the evolution of tetrapods is present in lungfish fins. At the same time, departures from limb-like proximal-distal gene expression exist, and these result from a different progression of the patterning of the anterior-posterior axis, whereby the lack of a complete phase of posteriorization translates into disparate gene expression profiles along the proximal-distal axis of limbs versus fins (Fig. 4). A diversity of morphological fin “archetypes” exists in fish. These exhibit, for instance, a reduction in the number of basal fin radials or, in teleosts, the loss of the metapterygium (10, 12), whereby each of such transitions is likely related to alterations in the patterning of the anterior-posterior fin axis (10, 40, 41). Despite these differences, a generalized expression signature can be identified that consistently distinguishes fins from limbs. In fins, hoxa13 occupies a distal domain along the entire anterior-posterior axis, but hoxd13 and hand2 remain confined to a more posterior domain and alx4 and pax9 are not distally limited (Fig. 4 and fig. S3) (10, 12, 22, 23, 40–43). The extension of a phase of progressive posteriorization in tetrapods therefore likely represents a derived heterochronic developmental mechanism that distinguishes limbs from fins.

The distal aspect of the limbs becomes posteriorized under the influence of shh secreted from the ZPA, and shh activates the same posterior genetic program in fish (12, 18, 40). However, the response in fins appears to be different from that in tetrapod limbs, as the expression of hoxa13 and gremlin1 becomes anteriorly displaced, whereas that of hoxd13 and hand2 remains posterior, and the expression of anterior markers alx4 and pax9 persists distally. Different temporal requirements for shh signaling have been demonstrated in limbs (34), whereby gremlin1 is fully activated by brief stimulation, whereas the anteriorization of hoxd13 depends on prolonged exposure. To our knowledge, hoxa13, hand2, alx4, and pax9 have not been analyzed in this context yet, but on the basis of the available data, the patterns observed in fins are best interpreted as a developmentally truncated response to shh, resulting in an incomplete posteriorization of the distal domain as compared to limbs. The time window during which a clear ZPA is observed in lungfish fins is relatively short, and alternative mechanisms could be responsible for its termination in fins and limbs. In limbs, persistence of the ZPA depends on sustained FGF signaling from the AER, which is dependent on the bone morphogenetic protein (BMP) inhibition by gremlin1 (37, 38). In the Australian lungfish, both fgf8 expression (39) and the ZPA disappear around the same time when the AER transforms into AEF upon the formation of the fin fold, and there is evidence that in fins, these events are related. In zebrafish, for example, the experimental perturbation of fin fold formation results in persisting fgf8 expression (16) and an extended ZPA (17) as well as anteriorized hoxd13 expression (17). This therefore suggests that the formation of the fin fold may be instrumental in the termination of the ZPA in fish fins, for instance, by acting as a physical barrier between epithelial FGF signaling and the mesenchymal ZPA. The evolutionary loss of the fin fold in tetrapods thus might contribute to a longer-lived ZPA during limb outgrowth, resulting in an increased posteriorization of distal limbs as compared to fish fins. Alternatively, the evolution of additional hoxd cluster regulation in tetrapods (1, 14) could have contributed to an enhanced response to the posteriorizing shh signal. Hoxd genes in digits are regulated by a complex regulatory region containing numerous partially redundant enhancers (44). In addition to a quantitative decrease in expression levels, a posteriorly restricted hoxd expression pattern is observed in several deletion mutants affecting this region (44). A combination of enhancer dosage effects via the acquisition of novel hoxd enhancers, in combination with sustained shh signaling by the ZPA, might therefore cause the anteriorized hoxd gene expression in the distal tetrapod limb. Such a scenario concurs with the proposed role of quantitative and heterochronic shifts in the shh/gli3 axis as a driver of the morphological divergence in paired appendages (10, 40, 41).

Although an autopod-like domain forms during the proximal-distal differentiation of lungfish fins, this domain is morphologically different from that in tetrapod limbs and fails to complete the progressive posteriorization, as is typically associated with the formation of digits. Whereas the function of hoxa13 is linked to the entire hand domain and might specifically contribute to the nodular character of the mesopodium, the distal expression of hoxd13 is intrinsically related to the long-bone territory of the hand comprising the metacarpals/metatarsals and digits (1). The different expression dynamics of these genes in lungfish fins therefore indicate separated evolutionary and developmental trajectories for hand and digit domains, as is also suggested by paleontological data on the morphological sequence of distal fin evolution in sarcopterygians (11). In this scheme, an ancestral autopod-like field defined by distal hoxa13 pre-dates the evolution of digits and was coopted in tetrapods to form the hands and feet. The concomitant emergence of digits as neomorphic structures (1, 2, 13, 45) likely required additional changes in the genetic modules downstream of the posteriodistal genes analyzed here in the Neoceratodus fins. Further comparative analysis of digit-specific gene regulatory networks (46, 47) in fins will therefore yield insights into the molecular and developmental mechanisms that shaped the fin-to-limb transition.

Our findings do have important implications for how we interpret the morphological changes that occurred at the fin-to-limb transition and how the hand and digit domains originated during evolution. In the last century, several competing transformational hypotheses have been proposed for the transition of fins into limbs. One influential model, eloquently articulated by Shubin and Alberg (48) but also presented by earlier authors in simpler form (3, 4, 45), predicts that digits arose from postaxial fin radials through their translocation to the distal domain by a bending of the metapterygial axis. This model has its limitations (49), and no such actual bending or causative cell movements have been demonstrated during tetrapod limb development. These morphological changes therefore likely reflect the transcriptional respecification of the anterior domain, thereby creating the illusion of metapterygial “bending.” We report differences in the patterning of fins and limbs that show a remarkable congruence with this inferred anterior respecification. In the lungfish, hoxd13 is restricted to the domain from which postaxial fin radials will develop, whereas hoxa13 is expressed more widely in the central elements of the metapterygial axis. Consequently, the presence and absence of a phase of distal expansion of hoxd13 correlate with bent versus straight metapterygial axes in Australian lungfish and tetrapods, respectively. Our results thus indicate that the most parsimonious identification of “digit precursors” in sarcopterygian fish lies in postaxial fin radials (Fig. 4).

Australian lungfish have the same basic anterior-posterior skeletal organization as the extinct lineage of tetrapodomorphs represented by Tiktaalik (7), whereby distal radials are present on preaxial and postaxial sides of a central axis. The distal fin radials of Elpistostege, a recently described more crownward tetrapodomorph, have been identified as digit precursors and are more tetrapod-like in their presence on the distal fin margin (50). Although, these still have a postaxial bias and do not cross-articulate with the radius as in tetrapods. This thus suggests that the posteriorization of the distal paired appendages evolved close to, or coinciding with, the origin of tetrapods and was involved in a widening and anterior expansion of the distal radial or digit domain. We propose that the derived Bauplan of tetrapod limbs, i.e., with the digits juxtaposed to the lower arm in a cross-articulating position, arose from an ancestral autopod-like domain that became modified through the extension of a phase of developmental posteriorization. In this form, limbs with coherent hands and feet provided our ancestors 400 million years ago with an adaptation to conquer the land and, remarkably, have stayed with us ever since as the most distinguishing feature of the tetrapod radiation.

In situ hybridization was performed according to Woltering et al. (51), with modifications (see Supplementary Materials and Methods).

Probes were amplified from cDNA or genomicDNA using Taq polymerase and subsequently cloned in the pGEMT vector (Promega A3600). Gene-specific primers are given in table S1. For all lungfish probes shown in Figs. 2 and 3 and fig. S2 (except hand2, shh, pax9, and gremlin1), two probes spanning different regions of the gene were cloned to increase the total span used in the in situ hybridization experiments. A separate probe set was used for the in situ hybridization experiments shown in fig. S1 (see Supplementary Materials and Methods).

Australian lungfish staging was according to Kemp (52). Larvae were obtained, as described before (39), and were collected between stages 42 (3 to 4 weeks after fertilization) and 50 (10 to 12 weeks after fertilization). All lungfish care and experimental procedures were approved by the Animal Research 347 Authority (ARA) at Macquarie University (ARA 2009/039). Astatotilapia burtoni embryos were obtained before, as described in (53) under permit Az. #T15/05TFA.

Imaging was carried out using a Leica MZ10F binocular stereomicroscope equipped with either a DMC2900 camera (color) or a DFC3000G camera (fluorescent) and LAS v4.5 software using the Z-stacking option. Fluorescent images were constructed as overlays using false colors. Images were enhanced using brightness, contrast, and sharpening settings in Adobe Photoshop software. Figures 2 and 3 and fig. S2 show left or right fins imaged with the ventral side up oriented to place their anterior/preaxial side to the left (i.e., right fins were mirrored). All probes were analyzed on a minimum of two specimens per stage shown, and no significant differences were observed between left and right fins.

A lungfish embryonic RNA sequencing transcriptome was assembled from the RNA extracted from embryonic fin tissue (see Supplementary Materials and Methods). Target transcripts were identified using Basic Local Alignment Search Tool (BLAST) with coelacanth, spotted gar, or Xenopus orthologous sequences as query. Phylogenetic analysis (see Supplementary Materials and Methods) was performed using orthologs and paralogs from human, spotted gar, elephant shark, and Australian lungfish to confirm correct orthology assignment (gene trees are shown in figs. S4 to S11).