Appendages and Gene Regulatory Networks: Lessons from the Limbless

Summary

Among squamate reptiles, dozens of lineages have independently evolved complete or partial limb reduction. This remarkable convergence of limbless and limb-reduced phenotypes provides multiple natural replicates of different ages to explore the evolution and development of the vertebrate limb and the gene regulatory network that controls its formation. The most successful and best known of the limb-reduced squamates are snakes, which evolved a limb-reduced body form more than 100 million years ago. Recent studies have revealed the unexpected finding that many ancient limb enhancers are conserved in the genomes of snakes. Analyses in limbed animals show that many of these limb enhancers are also active during development of the phallus, suggesting that these enhancers may have been retained in snakes due their importance in regulating transcription in the external genitalia. This hypothesis is substantiated by functional tests of snake enhancers, which demonstrate that snake enhancer elements have lost limb function while retaining genital enhancer function. The large degree of overlap in the gene regulatory networks deployed during limb and phallus development may act to constrain the divergence of shared gene network components and the evolution of appendage morphology. Future studies will reveal whether limb regulatory elements have undergone similar functional changes in other lineages of limb-reduced squamates.

1. Introduction

The vertebrate limb is a complex structure that requires the coordinated action of numerous genes to properly develop (Tickle, 2015). Genetic studies in humans and in model organisms have demonstrated that disruption of many different genes can lead to marked alterations in limb size and shape and can even lead to the complete failure of the limbs to develop. In natural populations, limb loss or dramatic reductions in limb size have independently evolved in many different tetrapod lineages, including amphibians, reptiles, birds, and mammals. In most limb-reduced lineages the key genetic changes that led to the evolution limb reduction remain unknown. In the case of threespine sticklebacks, however, the genetic basis of pelvic fin loss has been studied in great detail. Most stickleback populations exhibit prominent pelvic spines, which are modified hindfins that are homologous to the hindlimbs of tetrapods. Certain isolated populations have lost or reduced most structural components of the hindfin, including the spines and aspects of the pelvic girdle. Genetic mapping experiments in threespine sticklebacks have revealed that alterations at a small number of loci account for most of the pelvic-reduced phenotype in populations that exhibit pelvic fin loss, with cis-regulatory mutations at the Pitx1 gene explaining the majority of the phenotype (Chan et al., 2010; Shapiro et al., 2004). These regulatory mutations result in the loss of Pitx1 expression specifically in the pelvic region. Parallel studies in a different species, the ninespine stickleback, have demonstrated that pelvic-reduced phenotypes also map to Pitx1 in some pelvic-reduced ninespine populations (Shapiro et al., 2006). Thus, stickleback fish demonstrate that the genetic basis of appendage loss can have a relatively simple origin and certain loci, such as Pitx1, may be repeatedly used to convergently evolve similar phenotypes.

While fin or limb loss can be accomplished through a small number of mutations, the evolution of appendage loss may have significant secondary consequences for the gene regulatory network responsible for growth and patterning of the paired appendages. Components of the limb regulatory network that also perform functions in other tissues are expected to be retained after a limbless phenotype evolves, due to selective forces acting to preserve their functions in non-limb tissues. In contrast, components of the limb regulatory network that are truly limb-specific are expected to diverge and eventually be lost in the absence of selection. Across relatively short time-scales, however, even limb-specific elements are generally expected to retain function. For instance, the threespine stickleback populations that have evolved pelvic-reduction are relatively young and are believed to have lost Pitx1 expression and evolved a pelvic-reduced condition sometime within the past 10,000 to 20,000 years (McPhail and Lindsey, 1970). Over this very brief time span, one would anticipate that sequences important for the regulatory control of pelvic fin development would remain largely intact, since there has been limited opportunity for these elements to be altered through mutation. In agreement with this expectation, restoring Pitx1 expression to the pelvic region of developing sticklebacks is sufficient to fully rescue pelvic fin development in individuals from pelvic-reduced populations (Chan et al., 2010). Thus, the divergence of limb-specific elements is predicted to be most apparent in lineages with very ancient limb loss that have had significant time to accumulate mutations. The study of these ancient limbless lineages allows us to investigate how elements of the limb regulatory network evolve when selection for limb function is absent. Squamate reptiles, the group that encompasses lizards and snakes, contain multiple lineages that have evolved limb reduction and provide a powerful system for studies of gene regulation and morphological evolution of the limb.

2. Limb loss and the conservation of limb enhancers in snake genomes

Among tetrapod lineages that evolved limb loss, snakes are one the most successful with approximately 3400 described species (Pincheira-Donoso et al., 2013). The common ancestor of all extant snake species lived more than 110 million years ago, but it is likely that limb-reduction evolved even earlier (Hsiang et al., 2015). Hence, a limb-reduced phenotype is an ancient feature of this group. The ancient origins of limb loss in snakes are not only apparent from the fossil record, but are also evident from sequence analyses of limb-specific genes. A recent investigation of claw-specific keratin genes in snake genomes revealed a series of inactivating mutations and gene deletions (Emerling, 2017). Most of the claw keratin gene HA1 was found to be completely deleted from the genomes of six different snake species. In addition, though the first exon of HA1 is still present in all six species examined, the exon contains multiple inactivating mutations. These inactivating mutations include an 8 bp insertion shared by all six snake species, indicating that the functional inactivation of HA1 in snakes is a very old event. Further analyses of a second claw keratin gene, HB1, is consistent with the complete loss of this gene from snake genomes. In contrast, both HA1 and HB1 are intact in the genomes of limbed reptiles, including the lizard species Anolis carolinensis and Gekko japonicus. The divergence and loss of claw-specific keratin genes in snakes aligns well with our expectation that limb-specific genetic elements will degenerate in lineages that have evolved a limbless phenotype.

We recently sought to investigate the functional constraints on limb enhancers by determining the fate of ancient tetrapod limb cis-regulatory elements in snakes (Infante et al., 2015). Over the past 25 years, dozens of different mammalian limb enhancers have been identified through regulatory analyses of various limb patterning genes. When we examined the conservation patterns of these enhancers, we found that many of these known limb enhancers are conserved in the genomes of limbed reptiles and in the genomes of snakes (Fig. 1). Overall, the pattern of conservation reflects the shared evolutionary history of squamates, with some enhancers being diverged relative to mouse, but others highly conserved. However, there is at least one clear exception to the general pattern of limb enhancer conservation we observe in snakes. The ZRS enhancer, which is essential for driving Shh expression in the posterior limb bud (Lettice et al., 2002; Sagai et al., 2005), exhibits extremely high levels of sequence divergence in advanced snakes (Fig. 1; Kvon et al., 2016; Leal and Cohn, 2016). In addition, functional tests of python and cobra ZRS in mice revealed a complete or virtually complete loss of enhancer activity. Among published limb enhancers, the ZRS is notable for its extreme limb specificity. Thus, the ZRS displays high levels of sequence and functional divergence in snakes, just as we would expect to see for extremely limb-specific components of the limb regulatory network that are no longer under functional selection. Many of the other published limb enhancers are known to be active in non-limb tissues, providing a plausible explanation for their high degree of conservation in snake genomes.

Figure 1.

Sequence conservation of known limb enhancers in reptiles. Mouse enhancer DNA sequence was used to query turtle (gray square), lizard (gray triangle), python (green circle), boa (blue circle), and cobra (orange circle) genomes for similar sequences with the LASTZ aligner. The conservation metric (Y-axis) is the alignment score scaled by enhancer size. Known enhancers (X-axis) are arranged based on average conservation score. The black arrow indicates conservation score of the cobra ZRS enhancer sequence.

In order to examine limb enhancer conservation in snakes in a more systematic fashion, we performed comparative sequence analyses on a series of 65 limb-specific enhancers retrieved from the VISTA enhancer database. The enhancers in this database were identified as part of a large-scale effort to discover mammalian enhancers through the use of reporter assays in transgenic mouse embryos (Visel et al., 2007). Approximately two-thirds of the limb-specific VISTA enhancers are conserved in reptiles (Infante et al., 2015). Unlike the ZRS, however, we found that these enhancers showed relatively modest levels of sequence divergence in snakes relative to limbed reptiles. The tissue specificity of the VISTA enhancers was originally assessed based on the ability of the orthologous mouse or human sequences to drive reporter gene expression at a single stage of mouse development. This leaves open the possibility that the limb-specific VISTA enhancers are active in additional tissues during other stages of development. Moreover, transgenic reporter assays test whether an enhancer is sufficient to drive reporter gene expression from a heterologous promoter, but an enhancer may regulate transcription in additional tissues or cell types in its native context. Finally, the trans-acting environment that the snake enhancers operate in is clearly not identical to the mouse, and even highly conserved snake enhancers contain sequence differences from their mammalian counterparts. Therefore, the conservation of the VISTA limb enhancers in snakes suggests that they perform some function in these limbless animals, but the conservation of these elements does not tell us what those functions are.

3. Connections between the limb and genital gene regulatory networks

The conservation of ancient tetrapod limb enhancers in snake genomes, prompted us to reevaluate the activity of these enhancers in other tissues. One logical tissue to examine was the embryonic genitalia. Over the past two decades, a growing body of evidence has pointed toward substantial overlap between the sets of genes that regulate formation of the limbs and the external genitalia. For instance, mutations in several different genes result in human syndromes characterized by defects of the limbs and phallus (Afzal et al., 2000; Bamshad et al., 1997; Del Campo et al., 1999; Kang et al., 1997; Mortlock and Innis, 1997; Person et al., 2010). Moreover, studies in mice have demonstrated that key components of the hedgehog, Fgf, Bmp, and Wnt signaling pathways play essential roles in the development of both the limbs and the genital tubercle (Haraguchi et al., 2001; Lin et al., 2013; Suzuki et al., 2003; Yamaguchi et al., 1999). Recent global comparisons of gene expression between the limbs and genital appendages of amniotes have also demonstrated a large degree of overlap in the sets of genes that are transcribed during the development of these different appendage types (Tschopp et al., 2014), and studies of Hox clusters have revealed similar cis-regulatory topologies in the digits and external genitalia (Lonfat et al., 2014).

Given the overlapping network of genes that function during development of the limb and genital appendages, we hypothesized there might also be a large degree of overlap between the cis-regulatory elements that are used to drive gene expression in different appendages. A dual role for many enhancers in limb and genital appendages could also explain the preservation of limb enhancers in snakes. By using enrichment of the H3K27ac histone modification as a proxy to assess cis-regulatory activity (Creyghton et al., 2010), we examined enhancer activity in the limbs and external genitalia of mouse embryos and the embryos of Anolis carolinensis, the green anole lizard. Our analyses revealed high levels of shared cis-regulatory activity between developing limbs and genitalia with many enhancers showing evidence of activity in forelimbs, hindlimbs, and genitalia in both species.

The discovery that many enhancers display shared patterns of activity in the embryonic limbs and external genitalia of limbed amniotes, led us to assess the function of limb-genital enhancers of limbed and limb-reduced reptiles. Previous work in mice demonstrated that the HLEB element, an enhancer of the Tbx4 gene, is capable of driving reporter gene expression in the hindlimbs and genital tubercle of transgenic mouse embryos (Menke et al., 2008). Our subsequent tests of Anolis lizard HLEB in transgenic mouse embryos revealed that the lizard enhancer drives patterns of hindlimb and genital expression very similar to the mouse version. This suggests that the dual hindlimb-genital enhancer activity of mouse and Anolis HLEB represents the ancestral amniote activity pattern of this cis-regulatory element (Infante et al., 2015). In contrast, transgenic mouse assays demonstrated that the HLEB elements of king cobra and Burmese python have lost hindlimb enhancer function while retaining activity in the genitalia. Guerreiro and colleagues have observed similar functional alterations in snake enhancer elements from the HoxD gene cluster (Guerreiro et al., 2016). For instance, the mouse version of the Prox enhancer element drives reporter gene expression in the embryonic limbs and genital tubercle of mouse embryos, while the orthologous corn snake element can only drive expression in the genital tubercle. A second enhancer from the HoxD cluster, the Island I element, also shows clear evidence of limb enhancer activity loss in snakes. The mouse and chicken versions of the Island I enhancer are active only in the limbs, and the Anolis ortholog displays activity in both the limbs and the genital tubercle. In contrast, corn snake Island I drives reporter gene expression in the genital tubercle but not the limbs when tested in mouse embryos. Thus, the snake orthologs of the HLEB, Prox, and Island I enhancers have all lost limb activity while retaining genital activity. Together, the current evidence suggests that the loss of limb enhancer activity in snakes is a genome-wide phenomenon.

In light of the similar gene regulatory networks that are active in embryonic limbs and genitalia, it seems likely that appendage enhancers that are active in the limbs and the phallus are controlled by overlapping sets of transcriptional regulators in limb and genital appendages. Nevertheless, the selective loss of limb enhancer function of snake limb-genital enhancers demonstrates that the limb and genital activities of these enhancers are separable to some extent. While the sequence changes responsible for altering the function of snake limb-genital enhancers remain unknown, in principle, the loss of limb enhancer activity could be caused by mutations that remove binding sites for limb-specific factors that activate transcription. Alternatively, limb enhancer activity could be repressed through the gain of binding sites for transcriptional repressors. A recent study in Drosophila sechellia demonstrated that both of these mechanisms contributed to the evolutionary loss of enhancer activity at the shavenbaby locus (Preger-Ben Noon et al., 2016). Hence, there is precedent for the gain and loss of transcription factor binding sites as means for enhancer function to be lost. Functional studies of the HLEB, Prox, Island I, and other limb-genital enhancers are required to determine whether similar mechanisms have led to the selective loss of limb enhancer function in snakes.

Since experimental data has shown that the snake orthologs of multiple limb-genital enhancers have lost limb function while retaining genital activity, can we conclude that limb enhancers are conserved in snakes primarily due to their roles in regulating gene expression during genital formation? Considering the diverse activity patterns exhibited by enhancers (Visel et al., 2007), it seems likely that some limb enhancers are maintained in snakes due to functions that lie outside of the external genitalia. Supporting evidence for this comes from our functional studies of the HLEB element in mice. Deletion of HLEB from the Tbx4 mouse locus results in reduced Tbx4 transcription in the embryonic hindlimbs and genital tubercle and associated morphological alterations in these different appendage types (Infante et al., 2015). However, we also found a variety kidney defects and a high incidence of vaginal defects in HLEB knockout mice. Though enhancer deletion studies are not currently feasible in snakes, our mouse results suggest that the importance of HLEB function is not limited to the hindlimb and phallus. Therefore, while we believe that genital enhancer function is one important factor that has contributed the conservation of limb enhancers, this is unlikely to be the only reason limb enhancers have been maintained in snake genomes.

4. Pleiotropy and functional constraints on appendage evolution

The discovery that the posterior Hox genes play essential roles in development of the digits and external genitalia was among the first molecular evidence that similar developmental mechanisms contribute the formation of limbs and genitalia. This discovery also led to the hypothesis that the amniote phallus may have evolved through co-option of limb or digit regulatory cassettes (Kondo et al., 1997). The large number of enhancer elements that are shared between the limb and genital appendages is consistent with the co-option hypothesis, although it remains possible that the evolution of both the amniote phallus and vertebrate fins/limbs involved the convergent co-option of a suite of core appendage enhancers that predates the evolution of either the phallus or fins/limbs (Shubin et al., 1997; Lonfat et al., 2014). Regardless of the order in which shared enhancers were recruited during evolution to form different appendages, the existence of large numbers of limb-genital enhancers may have significant implications for the evolution of enhancer function and appendage morphology.

Pleiotropy is a force that acts to constrain evolution. Considering the large degree of overlap in the gene regulatory networks deployed during limb and phallus development, is there evidence that the evolution of these distinct appendage types is constrained due to their shared regulatory interactions? On initial examination, it might appear that constraints on limb development are not playing a major role in restricting phallus evolution. Among squamate reptiles, the morphology of hemipenes (the paired intromittent organs of males) is tremendously variable and is frequently used as a taxonomic character to distinguish related species (Arnold, 1986). Moreover, a quantitative analysis of trait evolution in 25 species in the lizard genus Anolis determined that genital morphology evolves six times more rapidly than non-genital traits (Klaczko et al., 2015). Yet despite the rapid evolution of external genitalia in squamates, a recent study by Klaczko and colleagues found that hindlimb size correlates with hemipenis shape in Anolis lizards (Klaczko et al., 2017). This same study presented compelling qualitative data showing that snakes display substantially greater diversity in hemipenial shape than lizards. This suggests that the evolution of genital morphology is accelerated in snakes relative to limbed lizards. This led Klaczko et al. to propose that hemipenis evolution may be less constrained in limbless species. Future investigations can test this hypothesis by quantifying the pace of morphological diversification in the genitalia of additional limbed and limbless lineages.

Links between the limb and genital gene regulatory networks may also have consequences for the ability of species to re-evolve appendages. For instance, there is evidence indicating that limb elements have reappeared in squamate species with limb-reduced ancestors (Kohlsdorf and Wagner, 2006; Siler and Brown, 2011). The continued requirement for limb enhancers in the genitalia might help to preserve regulatory interactions required for limb development long after a limbless phenotype evolves and, thus, allow an easier path for limbless species to re-evolve limbs. Applying similar logic, limbed species that have evolved phallus reduction might retain key regulatory elements for genital growth due to the role of these elements in the limb. Herrera and colleagues have shown that the extremely reduced phallus of chickens evolved through the up-regulation of Bmp signaling, which induces cell death (Herrera et al., 2013). Remarkably, inhibition of Bmp signaling is sufficient to prevent regression and promote outgrowth of the chicken genital tubercle. The preservation of the genital tubercle outgrowth program in chickens could be explained by selective retention of core outgrowth cis-regulatory elements that are also important for limb development, but a much deeper investigation of the regulatory interactions that occur in the appendages of avians with different phallus morphology is needed to substantiate this speculation.

5. Convergent limb loss in squamate reptiles

Although snakes are the most well-known and successful group of limbless squamate reptile, there are at least 24 additional independent instances of limb loss in the group (Sites et al., 2011; Wiens et al., 2006). Time-calibrated phylogenies indicate that these limbless lineages evolved at very different times across squamate evolutionary history (Brandley et al., 2008; Wiens et al., 2006; Zheng and Wiens, 2016). Some of these limbless lineages are more ancient than snakes, originating approximately 200 million years ago at the base of the squamate tree. Others are more recent, evolving perhaps 3.6 million years ago (Skinner et al., 2008). Additionally, partial limb reduction involving the loss of limb elements from the forelimb, hindlimbs, or both, has evolved more than 50 times independently (Wiens et al., 2006). These intermediate limb-reduced forms evolved as long as 60 million years ago and persist today, indicating that they are not simply transitional forms to a limbless state. Instead they may represent responses to their own particular locomotor or other selective pressures (Brandley et al., 2008). There is also evidence supporting the reappearance of limb elements in species with limb-reduced ancestors (Kohlsdorf and Wagner, 2006; Siler and Brown, 2011). This variation in timing and degree of limb reduction provides multiple dimensions to compare the effects of limb loss on the limb regulatory network.

Both the frequency and the broad spectrum of limb reduction severity is unique among tetrapod vertebrates, and the utility of squamates for studies of limb morphology and evolution has long been recognized by comparative anatomists (Lande, 1978; Greer, 1991). The extraordinary convergence of limbless and limb-reduced phenotypes provides multiple natural replicates of different ages to explore a number of interesting questions concerning the evolution and development of the vertebrate limb. In species that appear to have re-evolved limb elements, are the ancestral regulatory and developmental pathways resurrected, or are new ones recruited? To answer these questions will require a combination of developmental studies, comparative genomics, and functional genomics.

6. Prospects for comparative and functional studies in reptiles

Although there are many model organisms that have been advanced for developmental studies of vertebrates, relatively little developmental work has been performed on lizards and snakes. Changes in Hox gene expression domains have been implicated in the evolution of the elongate body form and forelimb loss in snakes (Cohn and Tickle, 1999) and decreased Shh expression has been linked to limb reduction in Hemiergis skinks (Shapiro et al., 2003). However, the sequencing of the genome of the anole lizard Anolis carolinensis, the first reptile to have its genome sequenced, created a renewed interest in reptile development (Alföldi et al., 2011; Guerreiro and Duboule, 2014; Park et al., 2014; Sanger, 2012). The availability of a high-quality lizard genome assembly provided a key evolutionary node for comparative genomics analyses (Feiner, 2016). Since the publication of the A. carolinensis genome, several additional squamate genomes have become available, creating new opportunities for comparative genomic studies within this group (Castoe et al., 2013; Georges et al., 2015; Song et al., 2015; Vonk et al., 2013; Xiong et al., 2016).

The prospects for investigating the evolution of appendage regulatory networks and enhancer function in squamates using comparative sequence analyses and functional genomic approaches, such as ChIP-seq, are excellent. As the genomes of additional limbed and limb-reduced squamate species are sequenced, global patterns of sequence evolution can be explored to investigate questions that are central to understanding the evolution of gene regulatory networks in animals with different appendage morphologies. For instance, are the same developmental pathways altered in the independent instances of limb loss? What is the fate of regulatory elements controlling limb development in other species or lineages with ancient limb loss? Can the different ages of limb-reduced and limbless species be used to reveal a characteristic progression during the evolution of limb reduction? Has there been a systematic gain or loss of binding sites for particular transcription factors in appendage enhancers of snakes or other limbless lineages?

Despite the power of comparative genomic approaches, functional studies will be required to determine the impact of specific sequence changes on morphology, and the activity of individual genes and enhancers in squamates. Thus far, functional tests of reptilian enhancers and transcription factors have come primarily through the introduction of reptilian DNA sequences into mice to produce transgenic and knockin mouse models (Guerreiro et al., 2016; 2013; Infante et al., 2015; Kvon et al., 2016; Leal and Cohn, 2016). Although the use of mouse models has yielded important functional insights into reptilian genes and enhancers, the mammalian and reptilian lineages diverged from a common ancestor more than 300 million years ago (Pyron, 2010). Clearly it would be more ideal to perform functional tests of reptilian genes and cis-regulatory elements directly within the species from which they originate or from other, closely related reptiles. Therefore, the lack of transgenic and genome editing approaches in reptiles has remained a serious barrier to functional investigations of appendage development in these species. In an effort to overcome this obstacle, Nomura and colleagues have recently adapted chick in ovo electroporation methods for use in gecko and turtle embryos (Nomura et al., 2015). Lentiviral-based transgenesis has also been successfully employed for lineage tracing in lizard embryos and RNAi in turtle embryos (Ge et al., 2017; Tschopp et al., 2014). The manipulation of reptilian embryos has also become feasible through the development of ex ovo culture systems that can support embryonic reptile development outside of the shell (Figure 2; Nomura et al., 2015). These approaches provide promising new avenues for functional investigations of the gene regulatory networks that are active in reptilian appendages.

Figure 2.

Comparison of limb morphologies between an Anolis sagrei embryo cultured ex ovo and a control embryo incubated within its egg. (A) Stage 8 embryo at start of culture. (a) Stage 8 forelimb. (B) Stage 8 embryo after 6 days in culture has developed to stage 11. (b) Forelimb of ex ovo cultured embryo. (C) Stage 11 control embryo incubated naturally. (c) Stage 11 forelimb. Embryos were incubated at 29℃, and an oxygenated roller bottle culture system was used as reported in Nomura et al., 2015.

7. Conclusions

Though we know very little concerning the initial genetic mutations that led to the evolution of limb reduction in snakes, the available data is consistent with a genome-wide loss of limb enhancer activity in extant snakes. It remains unclear whether the loss of limb enhancer activity is solely a result of neutral evolution or whether selective pressures to alter enhancer activity (e.g., to optimize enhancer activity in the genitalia or other tissues) are also driving functional divergence of these enhancers. However, we propose that selection to maintain pre-existing functions in the external genitalia has been one force that has acted to preserve many limb enhancers in snakes. As genome sequencing and assembly costs continue to drop and the genomes of additional squamate reptiles are sequenced, we will learn whether similar patterns of enhancer divergence have occurred in other limb-reduced lineages. Finally, we anticipate that the establishment of new transgenic and genome-editing approaches will enable direct functional investigations of development in reptiles with different limb phenotypes. In the future squamate reptiles will undoubtedly continue to serve as an important system for unravelling the genetic pathways controlling vertebrate limb development and evolution.