Limb modifications are fundamental to tetrapod evolution. Adaptive digit loss enables specialized functions such as running or flight, and has repeatedly evolved in parallel. But the developmental mechanisms underlying deviation from the five-digit ground state are unclear, partly owing to the hurdles involved in analysing embryos from animals that are not typically studied in the laboratory. Two papers published in this issue, from Cooper et al.1 (page 41) and Lopez-Rios et al.2 (page 46), illuminate developmental mechanisms that result in digit reduction in hoofed mammals and in rodents, and reveal surprising plasticity in the pathways that lead to evolutionary convergence. Together, these studies show that similar developmental changes may be shared by distantly related groups, but also that fundamentally different mechanisms can be used in closely related taxa to yield similar digit-loss outcomes.
During embryonic development, interacting and cross-regulating signalling centres in the early limb bud control and coordinate digit patterning (type and number) and growth3. Among these, the apical ectodermal ridge (AER) region secretes fibroblast growth factors (FGFs) that direct outgrowth, and the posterior limb bud secretes Sonic hedgehog (SHH), which regulates digit patterning dose-dependently through its posteriorly polarized distribution (Fig. 1a). Digit formation fails without SHH, but precocious SHH activation also causes digit loss by disrupting polarity and signalling-centre formation4. After digit precursors form, FGFs in the AER stimulate digit-tip elongation5, and bone morphogenetic proteins (BMPs) reshape limb borders and free individual digits through apoptotic cell death6, and modulate growth by inducing AER regression3.
Cooper et al. compared the expression of key patterning genes and remodelling activity in limbs of embryonic ungulates (hoofed mammals with 1–4 toes) and rodents (particularly jerboas, hopping desert rodents with 3 or 5 toes) — two groups that diverged around 60 million years ago (Fig. 1b). No early patterning changes distinguished a 3-toed (Dipus sagitta) from a 5-toed (Allactaga elater) jerboa species, or from the 5-toed mouse and 5-fingered forelimbs in all three rodents. Instead, in 3-toed jerboas, expanded apoptotic regions encompassed the digit I and V precursors and distal proliferative tips. Examination of the pro-apoptotic BMP pathway implicated increased expression of Msx2, which encodes a target transcription factor of the BMP pathway, as the triggering event. Secondary premature AER regression over digit tips also contributed to digit loss. Proof that MSX2 is indeed the primary instigator of adaptive digit loss in jerboas awaits genome comparisons and experimental interrogation of gene-regulatory regions in mice.
Extending their analysis to two different ungulate groups — the even-toed artiodactyls (including 2-toed camels and 4-toed pigs) and the odd-toed (1-toed) horses — Cooper et al. found that camels and horses also use cell-death mechanisms to reduce digit number. However, although apoptosis is the major factor in promoting digit loss in camels, the authors' findings suggest that MSX2 is not the instigator. In horses, although digit II and IV precursors are lost through cell death, digit I and V precursors never form, suggesting an additional patterning change that warrants further analysis. Nevertheless, these results point to cell death and growth arrest as major drivers of convergent digit loss in mammalian evolution. It will be interesting to see if progressive digit loss in other vertebrates, such as birds7, is similarly driven.
Unexpectedly, Cooper et al. found that the cell-death machinery is unaltered in some ungulates. In pigs, expression of limb-bud patterning genes loses polarity, with reduced expression of Ptch1, which encodes the SHH receptor. When unbound, PTCH1 suppresses SHH-pathway activity; SHH binding derepresses the pathway and upregulates Ptch1 expression, which sequesters SHH, forming a negative feedback loop3. Lopez-Rios et al. also found reduced Ptch1 expression in cows — the focus of their work — that was associated with non-polarized expression of other SHH targets and broader spread of SHH protein across the limb bud. Furthermore, they showed that the limb buds of mouse embryos with limb deletion of Ptch1 lose polarity and develop artiodactyl-like digits, implicating reduced Ptch1 as the primary basis for digit loss in cows.
Lopez-Rios and colleagues then used cutting-edge analysis of dynamic chromatin topology to assess interactions between gene-regulatory and promoter regions. This identified a mouse Ptch1 limb-regulatory module (LRM) that drove expression of a reporter gene mimicking Ptch1 in the limb. By contrast, the cow Ptch1 LRM was unable to drive reporter expression appropriately. Intriguingly, the LRM contains artiodactyl-specific DNA insertions that disrupt a highly evolutionarily conserved region. However, this region by itself failed to regulate reporter-gene expression appropriately, and the disrupting insertions are present in camels, which express Ptch1 normally. The DNA changes that restore or ablate the function of bovine and mouse LRMs, respectively, remain to be identified — this will be necessary to test whether the LRM change suffices for loss of bovine Ptch1 activation.
These papers show how independent molecular mechanisms can drive evolutionary convergence. It is likely that multiple genetic hits involving Ptch1 regulation, changes in BMP–MSX2-pathway regulation and other unknown genetic changes, separately and in combination, yielded digit loss in ungulates. Although cows and pigs share reduced Ptch1 expression, their extent of digit reduction differs. In addition, a shift in weight-bearing from middle-toe-centred to symmetric sharing between digits III and IV, the hallmark of even-toed artiodactyls, was already present in their pentadactyl ancestors and persists in camels — without obvious Ptch1 changes. Consequently, these artiodactyl features are either unrelated to Ptch1 expression or reflect progressive, incremental changes in Ptch1 regulation. Dissecting the diverse mechanisms converging on similar structures will require characterization of the tissue-specific regulation for each candidate gene. Mutations to regulatory regions are suspected to be major drivers of evolution, but few vertebrate examples have been characterized8,9. Lopez-Rios and colleagues' study is a benchmark for future work in this area.
It is not surprising that the SHH pathway, which is so central to digit specification, is a prime target for digit adaptation. In different Australian skink species, which have between two and five digits, digit loss correlates with the time span of Shh expression at late limb-bud stages10, a finding that has been functionally corroborated by genetic studies in mice11. But digit loss due to late SHH loss does not alter morphologies and is more related to growth arrest than patterning. Lopez-Rios and colleagues' work shows that early, precocious SHH activity can also be fine-tuned to create a symmetrical, but functional cow foot. Intriguingly, both patterning and remodelling mechanisms also lead to premature AER regression and FGF loss, which points to FGF modulation as another major driver of evolutionary limb adaptation. These mechanisms are linked by a common set of interacting signals, whose early and late developmental roles may often collaborate to achieve a final adaptive morphology.
The high incidence of convergent adaptive evolution of the limb derives from its intricate and robust — but plastic — regulatory network. Developmental mechanisms have long been considered the major constraint on evolutionary change. However, these studies invite speculation that, in response to selective pressures to adapt, the 'robustness' of the regulatory network guiding limb development evolved to be evolvable.
About this article