Perhaps the most widely acknowledged character of snakes is the loss of limbs. However, only in derived macrostomatans is the loss of limbs total. Scolecophideans, pythons, and boas all retain vestiges of hind limbs or girdle elements. With the demonstration that Pachyrhachis and Haasiophis are macrostomatans (Tchernov et al. 2000), the occurrence of hind limbs with a complete bony complement attains added significance. Although Tchernov et al. (2000); see also Zaher and Rieppel 1999b) entertained the possibility (on the basis of parsimony analysis) that legs may have re-evolved in macrostomatans, that possibility now seems remote. In and of itself, the loss and re-attainment of a complex element or character within a specific monophyletic group is always less parsimonious than a single loss. The interpretation of limb reacquisition as more parsimonious is possible only in the context of tree topology based on other characters not related to the one in question, and on the acceptance of reduced limbs as a basal synapomorphy within Serpentes.
Tree topology is accepted on the basis of parsimony. Character interpretation as homoplasy, synapomorphy, or symplesiomorphy is guided by the test of congruence. But neither parsimony analysis nor interpretations of characters based on tests of congruence are explicitly accompanied by an understanding of genetic, developmental, or other biological processes involved with the loss or gain of a character. Therefore, analysis and interpretation of character distribution should, when possible, be performed in the context of an understanding of biological processes rooted in empirical evidence and experimentation.
Independent of limb development, the morphological interpretation of the skull presented here is consistent with a phylogenetic placement of Pachyrhachis as a basal macrostomatan. Retention of limbs in a relatively advanced snake has implications for the developmental model of limblessness and axial patterning as presented by Cohn and Tickle (1999) who studied the genetic control and developmental process of limb loss in snakes. Their model accepted Pachyrhachis as the sister taxon to Serpentes and thus favored loss of distal limb elements in basal Serpentes, and total loss in advanced snakes. In that study, fibroblast growth factors were grafted to Python limb buds and successfully stimulated outgrowth. They were unable to stimulate apical ridge development, but did demonstrate polarizing potential by grafting the Python limb buds to the wing area of mutant wingless chicken embryos. The development of limbs in treated wingless chicken embryos demonstrated that Python hind limb mesenchyme conserves coding for limb development and polarization, and that suppression of limb development in Python is due to absence of apical ridge development.
Hind limbs occasionally occur in whales as in the humpback whale specimen documented by Andrews (1921) that possessed a cartilaginous femur, osseous tibia, cartilaginous tarsus, and osseous metatarsal. Reduction of hind limbs in whales is the result of arrested limb bud development (Bejder and Hall 2002). In humpback whales limb bud development persists longer than in odontocetes, partially explaining the more frequent expression in humpbacks. Presumably the loss of hind limbs in whales is accomplished by similar developmental and genetic processes as in other limbless vertebrates. However, the occasional occurrence of hind limbs in whales is a variant within a species and is quantitatively different from the fixed presence of hind limbs implicit in Pachyrhachis and Haasiophis. The variable presence of hind limbs in whales speaks to the process of limb loss, and does not necessarily speak to the genetic or developmental processes of reacquisition unless the assumption is made that the processes and controls elucidated by Cohn and Tickle (1999) are easily reversed. There is no direct supporting evidence to suggest that reacquisition of limbs after loss has ever occurred.
In insects, the reacquisition of wings has recently been hypothesized (Whiting et al. 2003). That study, based on molecular phylogenetic analysis, suggested the reacquisition (versus multiple loss of wings) based on parsimony analysis, similar to the situation we face with limb presence in basal macrostomatans. However, the conclusion that wings re-evolved after loss is again based on the parsimony analysis and tests of congruence of character distribution, a distribution that may have nothing to do with the genetic control, development, or other fundamental biological attributes of wings. In that case, one could reasonably predict that with further acquisition and analysis of character data, the hypothesis of re-evolution of insect wings will be falsified.
Clearly, expansion of the thoracic identity in the axial skeleton, most likely controlled by Hox gene expression domains, is an early development in snake evolution. Additionally, Hox gene expression related to apical ridge development in hind limb buds may control further limb loss. The recognition of the derived nature of Pachyrhachis demonstrates that leg loss under such a model is more variable in snakes than recognized previously because the hind limb skeleton of Pachyrhachis is better developed than in either boas and pythons or worm snakes (scolecophideans). However, this variation merely implies that hind limb loss in snakes is more complicated than a progressive decrease in development through time, and that hind limbs were reduced separately in more than one clade. This also appears to be the case in a recent study of limblessness in anguid lizards (Wiens ands Slingluff 2001). Given the hypothesis that snakes are evolved from within Squamata, retention of hind limbs, a plesiomorphic structure in tetrapods, is not surprising nor is it informative in a cladistic sense.
The interpretation of limb retention in Pachyrhachis and related forms implies multiple limb reductions in scolecophidians, alethinophidians, and within macrostomatans. This scenario appears likely given our current understanding of developmental mechanisms controlling limb expression, does not require an additional ad hoc hypothesis of biological processes controlling redevelopment, and is therefore a more parsimonious conclusion compared to reacquisition. The total elimination of hind limb specification and its skeletal remnants is characteristic of advanced macrostomatan snakes only.
Various interpretations of observed morphology placed Pachyrhachis in a more primitive and basal position among snakes, or in a derived phylogenetic position as a macrostomatan. Computed tomography allows a more thorough examination of Pachyrhachis problematicus than previously available and therefore provides for revised reconstruction and modeling of the skull, especially with respect to the morphology and position of the quadrates, the identity of the circumorbital bones, and the identity and position of the ectopterygoid. There is no compelling evidence for retention of a squamosal, a jugal, vertical orientation of the quadrate, or retention of multiple mental foramina. This analysis confirms the presence of a posterior free-ending projection of the supratemporals with distal expansion providing sole support of the quadrates, quadrate with no suprastpedial process but with a stylohyal process. The contact of the exoccipitals above the foramen magnum is ambiguous. Our analysis favors the phylogentic hypothesis that Pachyrhachis is a basal macrostomatan, consistent with the conclusions of Zaher (1998; see also Zaher and Rieppel 1999b; Tchernov et al. 2000).