Gene  expression  in  the  python  vestigial  hindlimbs

Pythons have vestigial hindlimbs and represent an excellent model system to study the molecular mechanisms that underlie limb development. Previous work has shown that python limbs lack the thickening at the distal edge of the ectoderm termed AER (apical ectodermal ridge). In addition, AER markers such as Fgf2 were not found anywhere in the python limb bud ectoderm (Cohn and Tickle, 1999). Limb growth depends on another signaling center, the zone of polarizing activity (ZPA) that expresses sonic hedgehog (Shh).

Interestingly, although Shh could not be detected in the python hindlimb mesenchyme, its expression could be activated with exposure to FGF signaling (Cohn and Tickle, 1999). Even though this study allowed the detection of important changes in gene expression, the techniques available at the time limited the approach to a small number of genes. We therefore revisited this fascinating limb loss model by making use of high-throughput technologies that are now readily available. In this context we microdissected python hindlimb buds buds of two python embryos at 0.5 dpo and analyzed their transcriptomic profiles by RNA-seq. Furthermore, we compared them with those obtained form genital buds and the tail tip of the same embryos, two structures where 5’ Hox genes also operate and whose development could contribute to other diverged aspects of the snake-like morphology.

As a first approach, we analyzed the general aspects of gene expression in these three tissues. A principal component analysis showed that the tail tip samples separate from those of the limb and the genitals over the first component. The latter two tissues, however, only show variance between them over the second component (Figure 22A). The genes whose expression varies the most among tissues mostly belonged to gene ontology classes related to embryonic development, organogenesis and morphogenesis (Figure 22B). The genes with the highest degree of variation between samples were either specific to a particular tissue or similarly expressed in both limbs and genitals (Figure 22C). Interestingly, some genes known to be involved in tetrapod limb development, such as Hand2 and Bmp2, were amongst the group of most variable genes.

Figure 22 – RNA-seq analysis of python embryo hindlimbs, genitals and tail tip A. Principal component analysis performed using the 500 more variable genes across samples. Samples coming from different embryonic tissues (genitals, limb or tail) are color-coded and duplicates differ by shape. B. 15 classes of gene ontology terms where the biggest number of identified most variable genes are classified. C. Heatmap of the 100 most variable genes across samples. The different samples are ordered and can be seen at the top of the graph. Duplicates are represented in blue and green.

Since pythons have vestigial hindlimbs, we set out to investigate how the molecular pathway underlying limb development has been modified in these snakes. Limb development has been extensively studied mostly in mouse and chicken models, which has allowed for the identification of the main factors involved and their regulatory activity. FGF10 is thought to be the mesenchymal factor that initiates the process of AER induction by activating Wnt3a/Wnt3 that subsequently initiates Fgf8 expression in the mouse limb bud ectoderm. The FGF8 at the AER then maintains Fgf10 expression in the underlying mesenchyme via a feedback loop (Fernandez-Teran and Ros, 2008). Our RNA-seq dataset shows that, similarly to what had previously been described for Fgf2 in python limb buds (Cohn and Tickle, 1999), Fgf8 is not expressed in any of the collected samples except for the tail tip that showed only negligible transcript levels (Figure 23A). Since the AER is not formed in python limb buds

(Cohn and Tickle, 1999) we analyzed the expression of genes involved in AER induction such as Fgf10 and Wnt3a. Interestingly, we found that both these genes were expressed in the python hindlimb buds (Figure 23A). These results suggest that the signaling pathways underlying Fgf8 activation and AER induction are present but fail to activate the necessary downstream processes.

BMPs have also been found to have a role in limb development and AER induction (Pizette et al., 2001). We therefore analyzed the expression of Bmp2 and found this gene to be expressed in the python hindlimb in higher levels than the other two tissues (Figure 23A). We also looked at Bmp4 and Bmp1r and found them to also be expressed in the hindlimb (data not shown). Because Shh has been described to be absent from python hindlimbs we checked for expression of Hand2, a gene that has been linked to Shh activation. Indeed, Hand2 loss of function results in the absence of Shh in the developing mouse limb bud (Galli et al., 2010).

However, we find that Hand2 is strongly expressed both in the python hindlimb and genitalia (Figure 23A).

Hoxd and Hoxa genes have also been shown to be necessary for Shh to be activated (Kmita et al., 2005). In turn, Shh expression is required to maintain “posterior” Hoxa and Hoxd gene expression by preventing the repressive form of Gli3 (Gli3R) to be formed in the posterior region of the developing limb bud. Accordingly, Shh mutants show negligible expression levels of Hoxa13 and Hoxd13 (te Welscher et al., 2002). Accordingly, Hoxd13 as well as other 5’-located Hoxd genes were expressed at low levels in the python hindlimbs (Figure 23B). In contrast, Hoxa13 and Hoxa11 were unexpectedly highly expressed (Figure 23B). This was striking because Hoxa13 expression is almost not detected in the mouse and chicken early limb primordium and is rather strongly expressed in the autopod, a region of the limb bud that patterns digits.

We made use of a transcriptome dataset for the green anole lizard limbs and genitals that has recently been released (Tschopp et al., 2014) to investigate HoxA gene expression in this species. Interestingly, we did not find the same high level of Hoxa13 expression in the lizard early limb indicating that the expression of this gene in limb buds is likely to be snake-specific (Figure 24).

In addition, python limbs and genitals showed distinct expression levels of 5’ Hoxa and Hoxd genes. While in mouse limbs Hoxa13 and 5’ Hoxd genes are highly expressed in developing genitals, the python counterpart only revealed high levels of Hoxd13 and Hoxd11 gene expression. Interestingly, in the lizard genitals, Hoxa13 also appears to be expressed at low levels (Figure 24). Hoxc13 and Hoxb13, on the other hand, are expressed as expected at high levels in the python tail tip but not in genital or limb tissue.

Recent results suggest that HoxA and HoxD clusters are necessary for the activation of Fgf expression at the AER via Gremlin1 (Grem1) and Fgf10 downregulation (Sheth et al.,

2013). We therefore analyzed expression of the python Grem1 gene, a BMP antagonist expressed in the limb mesenchyme essential for the correct expression of ectodermal Fgf genes (Zuniga et al., 1999). We found that Grem1, similarly to Fgf10 is expressed in the python limb, indicating that the few Hox genes present at the python hindlimb are sufficient to drive expression of this gene.

Altogether, these results show that genes known to be necessary for limb growth and differentiation, both for AER induction or Shh activation in the posterior limb are unexpectedly expressed in the python hindlimb. In addition, Hox gene expression at the python hindlimb is mostly restricted to the HoxA cluster and includes both mouse zeugopod-specific genes such as Hoxa11 and the mouse autopod-zeugopod-specific Hoxa13.

Figure 23 – Python embryonic tissue RNA-seq mapping over genes involved in limb development

A. RNA-seq profiles over genes that have a role in AER formation and Fgf or Shh activation. B. Mapping of RNA-seq reads over selected Hox genes. Exons estimated either by cufflinks or by blast with exons of closely related species.

Figure 24 – RNA-seq profiles of python and lizard limbs and genitals

RNA-seq mapping of python hindlimb bud and genitals (top), as well as lizard limb and genitals over the HoxA cluster (bottom). The y-axis is adjusted according to total number of aligned reads over the respective genomes between samples of the same species. Hox genes are represented by black boxes.