A shift in anterior–posterior positional information underlies the fin-to-limb evolution

1) The idea that there is a broadening of the AP axis in the evolution of tetrapod limbs is not new. Shubin and Alberg (1986) and Oster, et al. (1988) outlined a metapterygial arch hypothesis that outlined a shift in AP axis as a key mechanism leading to evolution of the tertrapod limb. They supported this hypothesis with developmental and comparative analysis. Thus, finding molecular correlates of this pattern is not altogether unexpected, although the data in the present manuscript are the most extensive analysis of the patterning changes in chondricthians. Surprisingly, these papers are neither cited nor discussed, and must be included.

As suggested, we have added discussion about the digital arch model in the seventh paragraph of the Results and discussion section.

2) The data on “posteriorizing” of the Scyliorhinus fin through retinoic acid is not convincing. The authors present quite elegant experiments that uncover a limb-specific Gli3 regulator domain. However the cartilage staining of the treated fins does not look anything similar to the author's interpretation – and does not phenocopy an early tetrapod such as Tiktaalik.

We agree that the RA-treated skeletons are not a direct phenocopy of Tikitaalik and we have therefore removed that panel from Figure 4, and also replaced the scheme of evolutionary process into a comparison between specific species (S. canicula and mice). Nevertheless, our important observation remains the striking correlation between the anterior shift of gene expression patterns, and the fusion of basal bones into a single metapterygium. In the context of the fin-to-limb transition, this correlation is a very important scientific observation.

Even if the remnant posterior fragment can be argued to be stylopodial, the authors must explain how this occurred as they have shown the lack of a key functional Gli3 regulatory domain in Scyliorhinus that is necessary for the regulation of polarity in mouse and chick – presumably even after RA treatment.

We agree that this part may have been confusing – in particular we may not have distinguished clearly enough between regulation at the transcriptional level versus the protein activity level. On the one hand, we show that Gli3 in S. canicula is lacking a relevant cis-regulatory element to restrict its expression anteriorly. On the other hand, we also show that RA is indeed able to cause a shift in AP patterning for Ptch1, Hand2 and Pax9. However, there is no contradiction here: The knowledge that RA treatment increases Shh signalling in fin and limb buds is well-known from the literature (Riddle et al., 1993 Cell 75: 1401-1416; Hoffman et al., 2002 Int. J. Dev. Biol. 46: 949-956; Dahn et al., 2007 Nature 445: 311-314). In turn, the effect of Shh signalling on Gli3 is not an alteration in transcriptional regulation, but rather a modification of the protein, specifically inhibiting formation of the repressor form (which is responsible for repressing Hand2 in the anterior tissue; Vokes et al. 2008 Genes Dev. 22: 2651-2663). This reduced repressive activity allows Hand2 to be upregulated, causing the general AP shift. Thus we propose that the effect seen in the RA-treated embryos is a post-translational regulation of Gli3 – not transcriptional. In other words, the AP shift during evolution was partially driven by cis-regulatory changes, while by contrast in our RA experiments it was driven by a Shh-mediated effect on the Gli3 protein itself, but in both cases achieving similar phenotypic changes. We have made this much clearer in the main manuscript now (Results and discussion, fifth and sixth paragraphs).

Unlike the elaborate elements formed in the miniskate (Dahn et al. 2007), the data here make it seem like cartilage development has been delayed or inhibited. Even though dHand expansion is expanded after RA treatment the changes in domains are small and this may be due to developmental delay or independent effects of RA such as on proliferation (note the fin size is quite different than controls at the same stage), rather changes in polarity.

We agree it is important to address the potential non-specific effect of RA experiments, for example on general growth and development of the bud (even in this non-model system which is non-trivial to work with). We have therefore performed new experimental works to address this issue (Figure 4 and Results and discussion, fifth paragraph).

Firstly, we addressed whether the treated fin buds appear to have retarded patterning, by examining another molecular marker which reveals the progress of proximo-distal patterning – Hoxa13. We detected slightly weaker expression in RA treated embryos than that in control embryos. However, the expression domain of Hoxa13 was not obviously shifted (Figure 4E), showing that this dosage and timing of RA treatment is having a clear effect on AP patterning (the shifts of Ptch1, Hand2 and Pax9, Figure 4A-C), while causing no obvious effects on PD patterning or general growth.

Secondly, we re-evaluated the sizes of fin buds from the RA-treated experiments, and also repeated the experiment to increase the replicates of Hand2 analysis. The image in the original panel of this figure was slightly distorted due to the in situ hybridisation process. From further analysis of all RA-treated results, it is clear that we observed the anterior expansion of Hand2 expression in buds of a similar size to the controls (in Figure 4B).

Since the reviewer considered that the change in Hand2 expression was small, we have also chosen to boost the result by analysis of another AP marker – Pax9 which is expressed in the anterior tissue (Figure 4C). As with Hand2 and Ptch1, this marker also displays a “posteriorisation” (or anterior shift of positional information) as it’s graded levels are reduced after RA treatment. Also important to note is that again the size of the fin bud is not reduced.

3) The expression patterns in the fins are quite late in comparison with comparable sized limbs such that these signaling events may pattern more distal structures. Changes in patterning of proximal bones such as the humerus would be expected to occur earlier, such as HH stage 20 in the chick. Could the authors comment on how such late expression changes would be expected to alter early developmental patterning events in the shark?

Indeed, we agree that the timing of these patterning processes is important point to address, and so we have added new gene expression data to clarify the time point at which the proximal skeletal elements begin to be patterned. Figure 1–figure supplement 2E now shows that Sox9 expression starts in the proximal elements between stage 28 and late stage 29. This is the same period for which we show the more posteriorly-biased gene expression patterns in S. canicula (Figure 1–figure supplement A-D) and also the RA-treatment experiments (Figure 4).

4) Correlation of positional value shift (expression boundaries) in fin/limb bud to resultant adult morphology isn't very tight. The one example that we're shown of an experimentally RA-dosed cat shark pectoral fin is juxtaposed with the pectoral fin of Tiktaalic. What we're not shown is the adult fin morphology of Callorhinchus, which is also dominated by the metapterygium (and its various, more distal, radials). Callorhinchus pectoral fin skeletons are morphologically like those of Tiktaalik and its various finned relatives.

Although the fin skeleton of Callorhinchus has a strong metapterygium, it also clearly has a large propterygium, which is in contrast to Tiktaalik, which only has a single proximal element (metapterygium/ stylopod). To address this comment we have therefore followed the reviewers’ suggestion, and generated a new cartilage staining of Callorhinchus and included it in Figure 2B.

5) ‘Partial posteriorisation’ (the authors' term) is a repeated phenomenon in the broad spectrum of paired fin evolution. It looks like this results in a ‘necessary but insufficient’ item of the fin-to-limb agenda, and likewise it might or might not be part of the history of chimaeroid fins. For example, how does Callorhinchus have a metapterygium-dominated pectoral fin, but seems not to have employed/evolved the positional value shift at the heart of this study?

We agree that the loss of anterior elements is a repeated phenomenon. However, we do not believe this has happened in the Callorhinchus milii pectoral fin skeleton (which we now show in Figure 2B), because its skeleton still has a large proximal propterygium (annotated as “pro”). We therefore see no clear evidence against our hypothesis that an AP shift was involved in the fin-to-limb transition.

6) The connection between the part of the paper that paper that documents the changes in Gli3 expression and function are not clearly linked to the final part of the paper that uses RA treatment to shift AP patterning. The authors need to make a clearer statement about exactly how they link together these observations using the experimental manipulations they have done. It seems that the RA treatment is a very indirect way to try and link together the observations on Gli3 to the actual effects of shifting AP patterning. The link seems very indirect and RA treatment is bound to have many effects that are independent of the changes in Gli3.

We agree that RA treatment is an indirect way to perturb AP patterning, although we must point out (a) that experimental embryology in a non-model species like this is challenging, and also (b) that precisely this method for boosting Shh signalling has been used in important previous papers about fin patterning in the mini-skate (Dahn et al., 2007 Nature 445: 311-314) although here we are drawing novel and distinct conclusions about the shift in AP patterning (which nevertheless do not contradict this previous paper).

To further address this general point, we have now strengthened and clarified the manuscript in the following ways (Results and discussion, fifth and sixth paragraphs):

Firstly, we realised that our proposal may have been confusing (the link between the results on Gli3 regulation and the RA experiments) because in Figure 3 we show that S. caniculus lacks the cis-regulatory elements to be inhibited in the posterior tissue, and yet in the RA experiments we report that an AP shift is induced. This point is discussed above (point 2). Essentially, we did not explain the known relation between Gli3 and RA, and we did not distinguish strongly enough in the text between transcriptional regulation and post-translational regulation. We have now re-written this to make it clear: At the genetic/evolutionary level, our results suggest that acquisition of element 1586 was a way to shift Gli3 expression anteriorly. However, in the experiments extensive Shh signalling triggered by RA causes a similar AP shift, but through a different mechanism inhibiting production of Gli3 repressor form (rather than altered regulation of Gli3 transcription). The important result is that we have two cases of an anterior shift: (a) the comparison between sharks and tetrapods, (b) the RA-treated shark buds, and in both cases two important observations are correlated: (i) the anterior shifting of genes, (ii) a reduction or loss of anterior basal bones, leaving a metapterygium-dominated arrangement.

Secondly, we have addressed the problem of possible non-specific effects of RA (as described above for point 2). We have examined Hoxa13 expression in RA treated embryos to assess possible effects on proximo-distal (PD) patterning, which is another major role of RA in tetrapod limb buds. Although expression of Hoxa13 in RA treated embryos was slightly weaker than in control embryos, nevertheless PD patterning appears hardly affected compared to the clear AP shifts in Hand2 and Pax9 expression (as the PD boundary of Hoxa13 expression is the same as controls, Figure 4E).

Would the authors expect that the reduction of Shh signaling in a tetrapod increase the number of bones in the anterior part of the limb, or better yet what would happen if they expressed Gli3 in a more posterior region? At the very least, the authors need to make a stronger link here. Without this, the work would not be particularly compelling of broad interest.

Gli3 overexpression in a posterior limb bud may not be enough to increase the number of the proximal elements (stylopod), because the phenotype in the stylopod always appears in combination with Gli3 and other gene knockouts. Nevertheless, it is clear that in these cases the expected result (of element reduction) is indeed seen: Gli3^-/-;Plzf^-/- mice lack the femur (Barna, Pandolfi, and Niswander, 2005 Nature 436, 277-281), Gli3^-/;Alx4^-/- mice show reduction of humerus (Panman et al., 2005 Int. J. Dev. Biol. 49, 443-448). Therefore, to increase the number of stylopod elements (i.e. to get limbs back to the ancestral condition), Gli3 and additional genes are likely required. Since Alx4 and Hand2 are already expressed in the S. canicula pectoral fin bud, and Plzf is involved only in the hindlimb bud, we do not have any candidates that might be involved in the fin-to-limb evolution. Although S. canicula genome is not sequenced yet, systematic studies at whole genome level such as ChIP-seq analysis in S. canicula fin bud would provide a more complete picture of evolutionary mechanism of the loss of the anterior elements in the future (Results and discussion, eighth paragraph).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

1) The reviewers remain unconvinced of the interpretation of the RA work and identity of the resulting elements. As this is a key component of their conclusions that alteration of RA/Shh signaling can cause reduction of the pro/mesopterygium, similar to the fin-limb morphological transition, it is important that this is clear to the reader to be able to interpret their findings. For example the severe RA treated fin shown has a strongly stained element (comparable to the proximal elements of the DMSO treated fin) on the anterior side – not compatible with a reduction of a pro/meso component of the fin. Given the lightly staining of the treated fins compared with the control, it would be helpful if the authors could make a schematic of their interpretation of the resulting pattern and explain why there are patterning/staining differences.

As suggested, we have added a schematic of our interpretation to Figure 4F. In particular, since the element on the anterior side of metapterygium in the severe phenotype is not directly attached to the pectoral girdle, we interpret it as a fused radial attached to the metapterygium (indicated by ** in Figure 4F). Therefore, we believe that the skeletal pattern of the severe phenotype represents a loss of anterior proximal elements. We have added the details of our interpretation to the main text as well (Results and discussion, fifth paragraph).

The different levels of staining between controls and RA treated fins seem to reflect individual differences in exposure time to Alcian Blue, as the other control sample show a lighter staining. We have therefore replaced the control figure with the lighter stained sample (Figure 4F).

2) The title for Figure 4 should reflect that it is RA mediated signaling that was tested not Shh. Shh levels or activity were not directly tested by gene over-expression or like means.

We have changed the title of Figure 4 into “RA treatment causes ectopic activation of Shh signaling and loss of anterior skeletal elements.”

3) Results and discussion, sixth paragraph, the conclusions that “loss of the anterior proximal elements during evolution appears partially ‘driven’ by cis regulatory changes” has not been directly shown. What is shown is that changes are associated with morphological transitions and may have been a component of the changes leading to the evolution of these forms. They also may have been secondary and not directly involved.

As suggested, we have changed the conclusion into “while the loss of the anterior proximal elements during evolution was associated with cis-regulatory changes of Gli3…” (Results and discussion, sixth paragraph).