In keeping with the traditional nomenclature, Arabic numerals (1–5) will be used when referring to the embryological position of a digit in the developing autopod regardless of the digit's terminal identity. Roman numerals (I–V) will be used when referring to digit identity.
The controversy over the identity of the three digits in the avian wing originated over 150 years ago with the inception of evolutionary morphology and the first suspicions of an evolutionary relationship between birds and dinosaurs (Gegenbaur,1863; Huxley,1868). The debate has endured ever since. Like in many controversial cases of character homology, conflict over the identity of the avian wing digits stems from a discordance between well-supported lines of evidence diagnosing character identity (Roth,1988). In the case of the avian wing, it is anatomical/paleontological evidence of digit form and evolutionary origins that disagrees with developmental data indicating the embryological origins of the wing digits. Reconciling this conflicting evidence provides not only a resolution to the long-standing controversy over identity of the avian wing digits, but also advances our understanding of the evolution of homologous characters (Wagner,2005; Young and Wagner,2011), and more practically speaks to the utility of different types of evidence (e.g., those of character anatomy versus character development) for identifying character homologies (Young and Wagner,2011).
Recent advancements in our knowledge of the developmental genetics of limb formation, new embryological and paleontological findings, and progress in our understanding of character individuation in general have both resolved old conflicts and created new questions in the digit identity debate. Here, we provide an overview of the historical debate and describe several attempts to reconcile the conflicting evidence of digit identity in birds. We review paleontological, embryological, and developmental genetic findings that contribute to the identification of the three digits of the avian wing and synthesize these data to show that the majority of evidence clearly supports a frame shift in digit identity as the anterior-most digit (position 2) of the bird wing is undoubtedly digit I. Finally, we evaluate the current state of the field to highlight the major unanswered questions including the mechanistic underpinnings of the frame shift, molecular confirmation of the posterior digit identities, and evolutionary timing of the digit identity frame shift in the theropod ancestors of birds.
OVERVIEW OF THE DIGIT IDENTITY EVIDENCE: CONFLICTS AND PROPOSED SOLUTIONS
As mentioned above, the dispute over identity of the avian wing digits results from conflicting diagnoses of digit identity based on anatomical/paleontological or embryological evidence. Specifically, paleontological studies of digit loss in the ancestors of birds reveal an uncommon pattern of digit reduction (Burke and Feduccia,1997; Wagner and Gauthier,1999) with the loss of digit V followed by the loss of digit IV (Gauthier,1986). This pattern of posterior digit reduction (and inferred loss) is well documented in the fossil record and is further supported by anatomical features of the remaining digits. Phalangeal formula and shape and position of the metacarpals and phalanges of the remaining forelimb digits support a I, II, III (e.g., thumb [or alula in birds], index finger, middle finger) diagnosis of digit identity (Gauthier,1986; Sereno,1993; Wagner and Gauthier,1999). However, studies of avian embryology and evolutionary patterns of digit reduction tell a different story. Instead of I, II, III, these data indicate that the embryological origins of the bird wing digits are identical to that of digits II, III, IV (e.g., the index, middle, and ring finger) in most other tetrapods (Hinchliffe and Hecht,1984; Müller and Alberch,1990; Burke and Feduccia,1997). In particular, broad examinations of limb formation across taxa lead to the discovery of a conserved pattern of digit development in amniotes (Shubin and Alberch,1986). These studies revealed that the first digit to appear develops spatially in line with the posterior skeletal elements of the zeugopod (i.e., the ulna and ulnare in the forelimb and the fibula and calcaneum in the hindlimb) and typically develops in position 4 (Burke and Alberch,1985; Shubin and Alberch,1986). In the bird wing, this digit, called the primary axis, develops in the posterior-most, fully formed digit position suggesting that it is digit IV, thus making the two anterior digits II and III (Müller and Alberch,1990; Burke and Feduccia,1997). Further supporting this diagnosis of digit identity, maintenance of digits II, III, and IV conforms to the more typical pattern of bilateral digit loss, with the loss of the last developing digits I and V preceding the loss of the earlier developing middle digits (Morse's Law; Morse,1872; Sewertzoff,1931; Greer,1990,1991; Galis et al.,2001; Shapiro,2002; Shapiro et al.,2003).
Attempts to resolve this conflicting evidence of digit identity have come in two forms. The first type accepts both the anatomical/paleontological assignment of digits I, II, III in theropods as well as embryological assignment of digits II, III, IV in birds and uses the incompatibility as an argument against the well-established theropod-bird evolutionary transition (Hinchliffe,1985; Feduccia,1996; Burke and Feduccia,1997). The second type—including the Pyramid Reduction Hypothesis (Kundrát et al.,2002; Galis et al.,2003; Kundrát,2009), the Axis Shift Hypothesis (Shubin,1994; Chatterjee,1998; Garner and Thomas,1998), and the Frame Shift Hypothesis (Wagner and Gauthier,1999)—argues, implicitly or explicitly, for precedence of either the anatomical/paleontological or embryological evidence for defining character identity and then provides a mechanistic accommodation of the conflicting pattern. Here, we provide a brief description of each proposed resolution (Fig. 1). Evidence for or against each hypothesis will be presented in the subsequent sections.
Pyramid Reduction Hypothesis
Giving precedence to developmental data indicating the embryological origins of the digits and stressing the evolutionarily conserved bilateral pattern of digit loss, the Pyramid Reduction Hypothesis (PRH) posits that the forelimb digits of birds and their theropod ancestors have identities II, III, and IV (Fig. 1A; Kundrát et al.,2002; Welten et al.,2005; Kundrát,2009). Anatomical similarities of these digits to digits I, II, and III are thought to result from convergent evolution (Kundrát et al.,2002; Galis et al.,2003; Kundrát,2009; Xu et al.,2009).
Axis Shift Hypothesis
Favoring paleontological/anatomical evidence, the Axis Shift Hypothesis (ASH) argues that the digits of the avian wing and theropod forelimb are I, II, and III in both identity and embryological origins (Fig. 1B; Shubin,1994; Chatterjee,1998; Garner and Thomas,1998). Proponents of the ASH hypothesize that a modification in developmental timing of digit formation leads to a shift in the location of the first forming digit from its typical position 4 to position 3 (i.e., a precocious/heterochronic development of the third digit before the fourth digit; Shubin,1994; Chatterjee,1998; Garner and Thomas,1998).
Frame Shift Hypothesis
The Frame Shift Hypothesis (FSH) argues that there is no rigid connection between embryological location and the execution of a character-specific developmental program, as shown in homeotic transformations, and thus digit identity and position can be dissociated (Wagner and Gauthier,1999). As a result, the FSH identifies the digits of the avian wing as I, II, and III regardless of their developmental positions, and hypothesizes that the genetic programs that regulate development of digit identities I, II, and III have shifted and are now expressed in embryological positions 2, 3, and 4 (Fig. 1C; Wagner and Gauthier,1999; Wagner,2005).
The evaluation of evolutionary patterns is, of course, a tree-dependent exercise, and the evolutionary origin of birds from within theropod dinosaurs is one of the strongest supported phylogenetic hypotheses derived from the vertebrate fossil record (Gauthier,1986; Sereno,1999; Turner et al.,2001; Clark et al.,2002; Prum,2002). As expected from a well-supported tree, the morphological features that distinguish birds from the other major reptile crown clades have deep phylogenetic histories whose origins are distributed along the nested hierarchy of dinosaur clades that comprise the avian stem (Gauthier and de Queiroz,2001). Feathers, which are now recognized in a wide variety of stem birds (Norell and Xu,2005), are the most famous example. Filamentous integumentary structures were even discovered recently in an ornithischian (Zheng et al.,2009) suggesting that at least the early homologs of feathers were already present in the most recent common ancestor of all dinosaurs and thus predate the origin of the avian crown by at least 140 million years. It is also noteworthy that the phylogenetic inference of a dinosaur origin of birds does not require any characters from the hand, as eliminating hand characters does not change the conclusion of the phylogenetic analysis (Wagner and Gauthier,1999). Overall, this compelling evidence means that the theropod ancestry of birds is no longer under serious scrutiny.
Similarly to feathers, the characteristic tridactyl hand of crown birds originates deep in dinosaur phylogeny. The phylogenetically and stratigraphically earliest theropods of the Late Triassic retained five fingers (Sereno,1994; Nesbit et al.,2009; Martinez et al.,2011), but digits 4 and 5 were already highly reduced; digit 4 being characterized by a short and slender metacarpal and one or two clawless phalanges and digit 5 represented by a vestige of the metacarpal (Fig. 2). Digit 5 is last observed in the early Jurassic Dilophosaurus wetherilli (see Xu et al.,2009), whereas digit 4 is retained in the more crown-ward Ceratosauria (sensu stricto; see Carrano and Sampson,2008) before being reduced in the mid Jurassic to a vestigial metacarpal that either was lost independently in several lineages (minimally allosauroids, tyrannosauroids, and non-tyrannosaur coelurosaurs) or was lost near the base of Tetanurae and secondarily ossified in a small number of Jurassic species (Bever et al., in press). This general pattern of postaxial digit reduction (albeit without digit loss) also characterizes Sauropodomorpha (the sister taxon of theropods within Saurischia) and most lineages of ornithischians (see Weishampel et al.,2004) and therefore was likely established before the origin of Dinosauria. For example, Plateosaurus engelhardti, a taxon that contains perhaps the most primitive hand among sauropodomorphs (Galton and Upchurch,2004), still expresses a diminished postaxial phalangeal formula relative to Alligator mississippiensis (2-3-4-3-2 vs. 2-3-4-5-3) with ungual phalanges (or the distal-most, clawed phalanges) of digits 4 and 5 reduced to vestiges of bone. Thus, the reduction and loss of digit identities IV and V, which characterizes early theropod evolution and arguably produced the tridactyl hand expressed in crown birds, reflects a trend of posterior digit reduction that likely predates the origin of Dinosauria (Sereno,1997).
In contrast to the posterior digits, the identities of the preaxial fingers (digits I–III) along the avian stem are generally conserved. There is no evidence in dinosaurs, or any archosaur, of a prepollex—a rudimentary extra digit that occurs anterior to position 1 (Larsson et al.,2010)—that characterizes many frog and mammal groups and that can be present as a polymorphic condition in salamanders and possibly some lizards (Wagner et al.,2000; Fabrezi,2001; Galis et al.,2001). The anterior-most (preaxial-most) digit of theropods is characterized by a robust metacarpal whose distal articulation surface is formed by a pair of offset condyles that orient the two phalanges anteriorly. This digit I identity is actually present in the earliest ornithischians, sauropodomorphs, and theropods, all of which are bipedal, and forms the morphological basis of a grasping hand that is ancestral to Dinosauria but still expressed in the distal portions of the avian stem (e.g., Archaeopteryx lithographica; Wagner and Gauthier,1999). This grasping digit I works in combination with elongate second and third fingers (digit II surpasses digit III to become the longest finger in the hand early in theropod history) that retain the widely conserved amniote phalangeal formula of 3–4 but undergo transformations that apparently increase the grasping ability of the hand. The most notable example is the extreme elongation and twisting of phalanx III–3 in tetanurans that result in an inward-facing digit III ungual (Gauthier,1986; Wagner and Gauthier,1999).
The presumed advantages that such a hand would provide a bipedal stem bird makes the FSH a functionally elegant hypothesis, as it allows the theropod ancestors of birds to maintain the grasping capabilities associated with digits I–III despite losing the potential to develop a fully formed digit identity at position 1—a loss that is visible in the embryonic wing bud of crown birds (see a more detailed description below) and therefore evolved at some point along the avian stem. The FSH predicts that digit identities I–III were conserved continuously in the direct evolutionary line uniting the ancestral five-fingered theropod with the most recent common tridactyl ancestor of the avian crown. The most recent review of the FSH from a paleontological perspective (Bever et al., in press) recognizes that this prediction has at least two important consequences. One, the frame shift itself should be skeletally seamless; complicating efforts to infer its phylogenetic position on the theropod tree (see below). Two, the predicted conservation of digit identities I–III provides a means to test the FSH with paleontological data. If the direct ancestral lineage of crown birds can be established as having lost digit identity I, then the parsimony advantage inherent in the FSH's prediction of identity conservation is weakened. It is important to make the distinction that while the FSH predicts that digit identity I was conserved continuously at some position in the hand, it also predicts the same identity was lost at position 1 (Bever et al., in press).
The predicted conservation of digit identities I–III was recently questioned with the description of an exciting new theropod that has obviously lost the grasping digit identity I (Xu et al.,2009). Named Limusaurus inextricabilis, its four-fingered hand expresses a vestigial metacarpal at position 1 but largely retains the diagnostic morphologies of digit identities II–IV at positions 2–4 (Fig. 2). This combination does not support a frame shift of digit identities in Limusaurus. A phylogenetic analysis recovered Limusaurus as a phylogenetically early divergence within Ceratosauria, which prompted Xu et al. (2009) to conclude that its loss of digit identity I reflects the ancestral condition for the subsequent length of the avian stem. If correct, this conclusion would falsify the FSH's predicted conservation of digit identity I in bird-line theropods. Other, previously described, ceratosaurs also exhibit the loss of digit identity I (e.g., Coria et al.,2002), but because these taxa are recovered in relatively derived positions within Ceratosauria, they have never been considered especially significant for understanding digit evolution in the direct ancestors of birds. Outside of Ceratosauria, however, there are no known theropod taxa that lose digit identity I, which means that its loss in ceratosaurs is unique and thus most parsimoniously inferred as a derived feature of those taxa with no necessary implications for the ancestral lineage of crown birds (Vargas et al.,2009; Bever et al., in press). Moreover, the argument of Xu and colleagues (2009) that the frame shift took place over an extended period of time and involved a series of step-wise transformations is problematic for additional reasons (Bever et al., in press). A model in which digit identities II–IV converge on the morphology of digit identities I–III is not a frame shift but rather a form of the PRH (Kundrát et al.,2002; Galis et al.,2003; Kundrát,2009). Clarifying this point is important because the PRH requires that digit identities I–III, and possibly IV, including all the individual features that together diagnose these identities, were lost independently at their primitive positions of 1–4 and re-acquired independently at their derived positions of 2–5. Although some previously unrecognized characters were described in support of their model, the parsimony costs of such large-scale convergence were not considered in the quantitative analysis of Xu et al. (2009).
The arguments for the 2, 3, 4 embryological origins of the digits in the avian wing come from a highly conserved evolutionary pattern of digit reduction and developmental pattern of digit formation across taxa. Among tetrapods, with the exception of urodeles, loss of digits occurs in a predictable pattern (Morse's Law), that is digit I > V > II > III > IV (Morse,1872; Sewertzoff,1931; Greer,1990,1991; Shapiro,2002). This pattern of digit reduction is thought to reflect a developmental constraint imposed by the pattern and sequence of digit formation (Holder,1983) such that loss of digits occurs in the reverse order of appearance during development (i.e., the “first in, last out” rule; Alberch and Gale,1985; Shapiro,2002). The primary axis, digit IV, is the first to appear during limb formation. Following the primary axis, development proceeds anteriorly resulting in formation of the digital arch and the development of digits III, II, and I in that order (Burke and Alberch,1985; Shubin and Alberch,1986; Müller and Alberch,1990); digit V develops independently of the primary axis and digital arch (Burke and Alberch,1985; Shubin and Alberch,1986; Müller and Alberch,1990).
Investigation of distal limb formation in chickens reveals that birds are no exception to this general pattern and sequence of digit formation and reduction. Examinations of digit development reveal that the posterior-most, fully formed digit in the avian wing is located in the position of the primary axis indicating that it is in position 4, and thus the two anterior digits are inferred to be in positions 2 and 3. This finding has been confirmed by studies staining the early digit condensations. Early investigations of digit condensation using radioactive sulfate (Hinchliffe and Griffiths,1983) and Alcian blue (Müller and Alberch,1990; Burke and Feduccia,1997) revealed only four digits condensations; however, because these techniques require extracellular matrix resulting from chondrification, these techniques neglect transient digit anlagen that appear early in development before chondrification. More recently, techniques that visualize the initial condensation of pre-chondrogenic cells reveal five digit anlagen in the developing avian wing (reviewed in Wagner,2005). Several independent research groups using distinct techniques—including documentation of capillary degeneration associated with cell condensation (Kundrát et al.,2002; Kundrát,2009), staining with the condensing cell affinitive peanut agglutinin (PNA affinity; Larsson and Wagner,2002), and in situ hybridization of a transcription factor (Sox9) expressed in condensing skeletogenic cells (Welten et al.,2005)—have confirmed this finding. Importantly, in all cases, three condensations were found anterior to the primary axis (i.e., in positions 3, 2, and 1) and one condensation posterior to the primary axis (i.e., position 5) providing strong evidence that the digits of the avian wing develop in positions 2, 3, and 4.
An alternative interpretation of the staining of digit condensations is that the anterior-most condensation is not a digit, but rather a prepollex (Vargas and Fallon,2005a; Wagner,2005; Welten et al.,2005) similar to that found in several distantly related taxa including opossums and frogs (Fabrezi,2001; Galis et al.,2001). In this case, consistent with the ASH (Fig. 1B), the stained pre-chondrogenic cell condensations would consist of a prepollex and digits positions/identities I, II, III, and IV. Because the primary axis is positioned in the posterior-most, fully formed digit of the avian wing, these positional/identity digit assignments require a shift of the primary axis from position 4 to position 3/digit III. However, a recent thorough examination of pre-chondrogenic cell condensations in Alligator mississipiensis, the closest extant relative of birds, found no evidence of a prepollex using PNA affinity histochemistry, one of the techniques which revealed the pre-chondrogenic anterior digit anlage in the chick wing (Larsson et al.,2010). Moreover, there is no evidence of this type of rudimentary anterior digit in the fossil record of the theropod ancestors of birds (reviewed above). Combined, the lack of a prepollex in the ancestors and closest extant relatives of birds provides additional support for the conclusion that the digits of the avian wing develop in positions 2, 3, and 4.
What is clear from the embryological evidence is that the digits of the avian wing develop in positions 2, 3, and 4, leaving only the PRH and FSH as viable hypotheses. However, the stepwise transformation of digit identities required by the PRH is unlikely given lack of evidence of intermediate morphologies in the fossil record and the parsimony costs of this type of morphological transition (reviewed above); thus, leaving the FSH as the most likely solution to conflicting evidence of digit identity in birds. The question then becomes, does the embryological data of digit formation provide further support for the FSH? Embyrologically, the FSH has two major requirements. First, the digits must develop in positions 2, 3, and 4. This fact has been well established through work from several independent research groups and is arguably no longer in question. Second, in order for digit identities I, II, and III to develop in digit positions 2, 3, and 4, there must be a temporal and mechanistic (discussed in the subsequent section) separation of the appearance of condensations and their individuation into digits. More specifically, the processes of condensation of the digit anlagen must occur before individuation of the digits. In this case, the embryological evidence that reliably labels the digit condensation positions as 2, 3, and 4 is not relevant for digit identity determined later in development thus allowing positions 2, 3, and 4 to express the developmental genetic programs of and ultimately individuate into digits I, II, and III. Consistent with the FSH, recent experimental results provide evidence of a temporal separation of digit condensation and individuation. Using the chicken hindlimb, Fallon and colleagues have shown that transplantation of the posterior interdigit tissue—the tissue between the developing digit condensations thought to regulate digit individuation—can lead to digit identity transformations, specifically resulting in development of a digit with the typical identity of the digit anterior to the inserted interdigit (Dahn and Fallon,2000; Suzuki et al.,2008). More importantly for the temporal separation of digit condensation and individuation, transplantations of the interdigit tissue at developmental time points after formation of the digit condensations are able to transform digit identity (Suzuki et al.,2008). Later in autopod development, presumably after digit individuation, interdigit tissue transplantation no longer transforms digit identity (Suzuki et al.,2008). These findings indicate that there is a finite period between digit condensation and individuation when digit identity is flexible relative to position. Termination of this flexibility seems to occur upon digit individuation which differs among the digits with digit IV being the earliest to fix its identity followed by III, I, and finally II in the chicken hindlimb (Suzuki et al.,2008).
Molecular evidence of digit identity further solidifies the conclusion from paleontological and embryological data that the FSH is the most viable hypothesis to resolve the controversy over digit identity in the bird wing. As mentioned above, a requirement of the FSH is that digit condensation and individuation must be mechanistically separable (Wagner,2005). In fact, a great deal of empirical and experimental evidence indicates that digit identity can be separated from digit position. First, there are several mutant phenotypes that exhibit digit identity transformations. For example, in one form of the rare human genetic disorder nonopposable tri-phalangeal thumb (Qazi and Kassner,1988; MIM#190600), digit I is transformed into a digit II including not only the addition of a phalanx, but also modification of metacarpal shape and digit articulation and position (reviewed in Wagner,2005). Second, experimental modifications in expression of several genes—including, but not limited to: Gli3 (e.g., Litingtung et al.,2002; Ros et al.,2003), Shh (e.g., Chiang et al.,2001; Towers et al.,2008; Vargas and Wagner,2009), 5′HoxD genes (e.g., Morgan et al.,1992), and BMPs and their antagonist Gremlin (e.g., Dahn and Fallon,2000; Benazet et al.,2009)—during limb development can decouple digit position from identity.
In addition to flexibility in the relationship between digit position and identity, patterns of gene expression in the developing chicken wing further support the FSH. First, an evolutionarily broad examination of 5′HoxD gene expression during distal limb formation has revealed a highly conserved expression pattern in the developing digits (Fig. 3A; Johnson and Tabin,1997; Vargas and Fallon,2005a; Montavon et al.,2008; Reno et al.,2008; Vargas et al.,2008; Young et al.,2009). Specifically, Hoxd13 is expressed in all of the developing digits; however expression of Hoxd12-Hoxd8 is restricted to the more posterior digit identities II–V (Fig. 3A). Importantly, this pattern is maintained regardless of the digit position. For example, in the Silkie chicken mutant an ectoptic digit II often develops anterior to digit I in the hindlimb. Despite being located anterior to the normal digit I (with no Hoxd12 or Hoxd11 expression), the ectopic anterior digit II expresses Hoxd12 and Hoxd11 as expected for its identity (Vargas and Fallon,2005a). In addition, when digit identity transformations are experimentally induced by manipulation of gene expression during limb development, the expression patterns of the 5′HoxD genes are likewise modified (e.g., Morgan et al.,1992; Knezevic et al.,1997; Chiang et al.,2001; Litingtung et al.,2002; Ros et al.,2003; Vargas and Fallon,2005a; Vargas and Wagner,2009). In birds, 5′HoxD expression patterns confirm the FSH digit identity assignments (Fig. 3B). In the developing wing, while Hoxd13 is expressed in all three digits, despite developing in digit positions 2, 3, and 4, expression of Hoxd12 and Hoxd11 is restricted to the two posterior digits (digit identities II and III in embryological positions 3 and 4, Fig. 3B; Vargas and Fallon,2005a). Beyond HoxD expression, recently, two additional genes (Mkp3 and Sef) with expression patterns supporting the frame shift in digit identity in birds were discovered (Uejima et al.,2010). Uejima and colleagues (2010) found that, unlike in early stages of development, in later stages of chicken wing limb formation (i.e., stages 25–26), expression of these two genes is anteriorly restricted, such that expression is limited to the anterior-most digit forming region position 2. In two five-digited taxa (mouse and gecko) and in the chicken hindlimb, Mkp3 and Sef exhibited a similar anterior restriction in expression late in limb development with the expression domain limited to position 1 during embryonic stages corresponding to stages 25–26 in chicken suggesting a conserved pattern of expression of these two genes in developing digit I across amniotes (Uejima et al.,2010).
From the molecular data it is clear that in the developing bird wing position 2 has the developmental genetic signature of digit identity I confirming that a frame shift in identity has occurred. In addition, these results indicate a clear association between gene expression patterns of the 5′HoxD genes (and likely Mkp3 and Sef) and digit identity; however, neither the mechanistic basis underlying the modification in HoxD expression nor a causal relationship between changes in gene expression and the digit identity frame shift have been established (see further discussion below).
NONAVIAN DIGIT IDENTITY CONFLICTS
Because limb and digit reduction is common among tetrapods, it is not surprising that other cases of digit identity conflict appear in the literature. In the anuran forelimb, despite anatomical descriptions of the digits referring to them as I, II, III, and IV (Romer and Parsons,1970), the position of the primary axis in the forelimb (Oster et al.,1988) combined with expression of Hoxd11 in the anterior-most digit (in Xenopus; Satoh et al.,2006) indicates the presence of II, III, IV, and V digit identities. Another case of digit identity conflict, remarkably similar to that in birds, occurs in Chalcides chalcides, the Italian three-toed skink. In C. chalcides, the three digits of both the fore- and hindlimb seem to have undergone a frame shift in identity. As in birds, in C. chalcides, the position of the primary axis in the developing autopod indicates that the digits develop in positions 2, 3, and 4 (Steiner and Anders,1946; Young et al.,2009); however, shape of the anterior-most metacarpal/tarsal, fusion of anterior-most distal carpal and metacarpal (thought to be an anatomical characteristic of digit I; Steiner and Anders,1946; Rieppel,1992; Leal et al.,2010), and the phalangeal formula of the two anterior digits (Caputo et al.,1995) are indicative of digits I, II, and possibly III identities (Fürbringer,1870; Steiner and Anders,1946; Renous-Lecuru,1973; Young et al.,2009). Recently the suspected frame shift has been further supported using molecular data. As in the avian wing digits, expression of the 5′HoxD genes (as indicated by Hoxd11 expression) have shifted such that the expression of Hoxd11 is restricted to digit positions 3 and 4 confirming that position 2 develops into digit I (Young et al.,2009).
Evidence of an additional, independent digit identity frame shift in C. chalcides suggests that this mode of evolution may confer some benefit. For example, the digit identity frame shift may provide an escape from the hypothesized developmental constraint on digit reduction by allowing loss of embryological positions 1 and 5, but maintenance of digit I morphology and function. In fact, in both well-supported cases of digit identity frame shifts (the theropod ancestors of birds and C. chalcides), loss of digits is thought to be adaptive rather than a result of rudimentation (Steiner and Anders,1946; Wagner and Gauthier,1999; Wagner,2005). In the ancestors of birds, loss of digits is the hypothesized result of adaptive modifications for prey capture (Sereno,1999; Wagner and Gauthier,1999). In C. chalcides, the reduction of limb length and associated loss of digits are thought to improve locomotor performance in this grass-swimming species. At the same time, the reduced, but fully formed limbs and digits are thought to maintain important functions to stabilize the animal while at rest and to push away vegetation when moving at slow speeds (Orsini and Cheylan,1981).
OUT WITH THE OLD AND IN WITH THE NEW: PROBLEMS OF AVIAN DIGIT IDENTITY
The direct disagreement between paleontological/anatomical and embryological evidence has fueled the lasting debate and motivated continued investigation into the identity of the digits in the avian wing (e.g., see Hinchliffe and Hecht,1984; Burke and Feduccia,1997; Feduccia,1999; Wagner and Gauthier,1999; Galis et al.,2003,2005; Vargas and Fallon,2005b; Kundrát,2009; Vargas and Wagner,2009; Xu et al.,2009; Young et al.,2009; Larsson et al.,2010). This work has resolved several issues of the avian digit identity controversy. First, discoveries such as feathered dinosaurs have placed the origin of birds firmly within the theropod clade such that the theropod-bird transition is no longer seriously in doubt. Second, the revelation of five digit anlagen in chicken and ostrich embryos combined with a lack of evidence of a prepollex in either extinct or extant relatives of birds clearly establishes that the avian wing digits develop in positions 2, 3, and 4. Finally, highly conserved expression of several genes, most notably the 5′HoxD genes, clearly illustrates that the anterior-most digit of the avian wing expresses the molecular signature of digit identity I indicating that a frame shift, at least of digit I, has occurred. While resolving these issues has been a major step forward for the field, it has also turned the spotlight onto several other unanswered questions and in addition has created new controversies in the study of avian wing digit identity. Below we outline some of the important areas for future research.
Reinvigorating Paleontological Investigations of the Frame Shift, a New Model With New Predictions
The most obvious unanswered question regarding the frame shift and the fossil record is: can we determine with any degree of resolution where along the avian stem the frame shift occurred? This is an important question because it is basically asking whether the fossil record is likely to contribute further to frame shift research. It is also a difficult question considering that, as described above, digit morphologies should be skeletally seamless across the frame shift and therefore uninformative for circumscribing the frame shift's phylogenetic position. However, the FSH also makes some explicit predictions regarding digit loss and these, it turns out, are informative (Bever et al., in press).
The fossil record is clear that the initial digit loss in theropods occurred at position 5. The next step has always been assumed to be the loss of digit IV followed at some unknown point by the frame shift, which leaves position 1 without a digit identity. This inferred sequence (posterior digit loss [PDL] model; Fig. 2) places the frame shift of digits I–III at some point after the loss of digit IV in the Jurassic but before the origin of the avian crown in the late Cretaceous. Considering that theropods empirically continue the pattern of posterior digit reduction ancestral for Dinosauria, this frame shift model appears reasonable. However, a pattern of digit reduction does not necessarily reflect the positional order of digit loss, and because the digits at positions 1 and 5 tend to be the last to develop ontogenetically, they tend to be the first lost phylogenetically (reviewed above). Therefore, if digits IV and V are lost before the loss of digit I at position 1, then theropods express a derived pattern of digit loss. If theropods conserve Morse's Law, which would be the more parsimonious solution, then the frame shift is pulled back in time to a point between the loss of digits IV and V (bilateral digit loss [BDL] model; Fig. 2). This frame shift model involves a much more narrowly delimited area of the theropod tree that includes the early history of Tetanurae and the stem of the slightly more-inclusive clade Averostra (Ceratosauria + Tetanura). It also means that digits I, II, III, and IV all were frame shifted and that the tridactyl hand expressed in crown birds was achieved when digit identity IV was lost from position 5. Deciding at which of these two phylogenetic positions the frame shift occurred (i.e., basal Tetanurae or Averostra) should depend on whether or not the ceratosaur hand is frame shifted. As previously discussed, there is no evidence of a frame shift in ceratosaurs, which supports the frame shift in the more derived position within Tetanurae. The only issue with this position is that it requires the reduced developmental potential at position 1 in crown birds and the loss of digit identity I at position 1 in Limusaurus to be the product of independent evolutionary events. This suggests the possibility of an even more parsimonious model of the frame shift.
The Developmental Variability Model of the frame shift (Bever et al., in press) predicts the loss of digit V directly facilitated the frame shift and that the frame shift was introduced as a polymorphic character state. In other words, after the loss of digit V, an area in time and tree space was established in which the probability that digits I–IV could develop from either positions 1–4 or 2–5 was relatively high. Because both the shifted and unshifted hands were expressing digits I–IV, this zone of developmental variability would remain phenotypically cryptic until the subsequent reduction of developmental potential at position 1. This reduction, which is visible in crown birds, would leave the shifted individuals unaffected (their developmental pathways had already moved away from position 1). In contrast, the unshifted individuals lost digit identity I but retained identities II–IV in their ancestral positions 2–4. This is the combination of identities expressed in the hand of Limusaurus, supporting the conclusion that Limusaurus is derived from the unshifted portion of the zone of developmental variability while the shifted portion continued to express digit identities I–III at positions 2–4 and eventually gave rise to crown birds.
Testing directly for a zone of developmental variability in early theropods is, of course, a challenge, but the Exaptation model does raise several secondary questions and predictions that can be pursued with extant species. For example, would a zone of ancestral developmental polymorphism leave some fingerprint in the developmental data themselves, and if so, what would that fingerprint look like? What is the relationship between the relative length of the forelimb and variability in the hand? Can an extreme shortening of the forelimb, like that found in ceratosaurs, decrease the selection pressure on the morphology and development of the hand to such an extent that digit identities decouple, resulting in the aberrant morphologies and homoplasy observed in ceratosaurs?
Developmental Genetics of Limb Formation: Finding the Frame Shift Mechanism
Growing evidence in support of a digit identity frame shift in the avian stem (reviewed above) promises an exciting phase of discovery in this field as it is motivating work focusing on the mechanistic underpinnings of the transformation in digit identity in the avian wing. While an association between digit identity and expression patterns of the 5′HoxD genes has been clearly established, whether the modification in HoxD expression resulted in the digit identity frame shift in birds is not clear. However, modifications in expression of the 5′HoxD genes and subsequent digit identity transformations have been shown to accompany experimentally induced changes in expression of genes upstream in the limb development pathway. Knowledge of the developmental genetic underpinnings of digit determination highlights two potential mechanisms. The first is a modification in Shh signaling early in limb formation and the second is a modification of 5′HoxD expression itself as both play established roles digit identity determination (Shh: Riddle et al.,1993; Wang et al.,2000; Yang et al.,2003; 5′HoxD: reviewed in Vargas and Fallon,2005a).
Shh is expressed from a discrete region of the posterior limb bud called the zone of polarizing activity (ZPA). Expression and diffusion of Shh from ZPA cells organizes the anterior–posterior axis and influences development of the distal limb. Shh is only expressed by cells in this region; however, diffusion and signaling activity of the morphogen extends the role of Shh anteriorly (Wang et al.,2000; te Welscher et al.,2002). Across the anterior–posterior axis differences in both the type of Shh exposure (e.g., cells can be exposed due to expression of Shh, diffusion of Shh expressed from surrounding cells, or may not be exposed to Shh signaling at all) and duration of Shh exposure establish distinct morphogenetic fields that are thought to make up the future digits (Tickle et al.,1975; Harfe et al.,2004; McGlinn and Tabin,2006). In fact, naturally occurring and experimentally induced modifications of the Shh signaling pattern can lead to changes in late phase 5′HoxD expression and subsequent loss and transformation of digit identities (Wada et al., 1999; Chiang et al.,2001; Litingtung et al.,2002; Ros et al.,2003; Vargas and Wagner,2009). For example, evolved digit loss in some species of Australian lizard Hemiergis results from a shortened duration of Shh expression resulting in decreased cell proliferation and digit loss (Shapiro et al.,2003). In addition, experimental manipulation of Shh signaling in the developing chicken embryo results in an additional digit identity frame shift in the wing. When developing chicken embryos are treated Cyclopamine, a Smoothened inhibitor that interferes with Shh signaling (Incardona et al.,1998; Scherz et al.,2007), an additional frame shift in digit identity is induced resulting in the development of a bidactyl wing with digits I and II in positions 3 and 4 (Vargas and Wagner,2009). Moreover, this digit identity frame shift is associated with a shift in spatial expression patterns of the 5′HoxD genes—Hoxd11 and Hoxd12 expression is restricted to position 4, digit 2 (Vargas and Wagner,2009) suggesting that a modification of Shh in the developing bird wing may underlie the frame shift in digit identity. However, there is no known modification in Shh expression or activity in the developing avian wing.
Alternatively, the modification in HoxD expression in the developing bird wing may result from a regulatory change in the HoxD genes themselves. Recently, two regulatory elements responsible for the conserved pattern of 5′HoxD expression in the developing digits were discovered. Two long-range cis-regulatory elements, the global control region (GCR) and Prox (Spitz et al.,2003; Gonzalez et al.,2007) interact to form a complex which regulates transcription of all the 5′HoxD genes (Montavon et al.,2008). Differences in transcriptional activity the 5′HoxD genes depends on proximity of the gene to the cis-regulatory element with Hoxd13 exhibiting the highest relative expression and Hoxd8 the lowest (Montavon et al.,2008). In addition, differences in affinity of gene-specific promoters to the regulatory complex also influence levels of gene expression (Montavon et al.,2008). In the developing autopod, this collinear regulation of 5′HoxD expression results in a correlation between transcriptional activity of the 5′HoxD gene and the spatial distribution of gene expression across the developing digits (Montavon et al.,2008), suggesting that modifications in the spatial patterns of 5′HoxD expression, like that associated with the digit identity frame shift in birds, may result from a corresponding adjustment in transcriptional activity of the 5′HoxD genes.
Potentially initiating a new controversy, recent evidence in support of a mechanistic role for both regulatory change in the 5′HoxD genes and a modification in Shh (more specifically, ZPA cell activity) was presented. A forelimb-specific down-regulation of the HoxD genes was shown to occur late in development of the chicken wing (stages 28–30; Young and Wagner,2010). This change in 5′HoxD expression occurs independent of a regulatory change in Shh, as there was no evidence of a forelimb-specific down-regulation in Shh expression (Young and Wagner,2010). However, in a different study, Tamura and colleagues (2011) documented an anterior displacement of ZPA cells in the chicken forelimb early in digit development (stage 20). These cells, originally situated along the anterior–posterior axis in a position consistent with that of digit IV in the hindlimb, were displaced from the ZPA potentially freeing them to receive developmental signals responsible for the designation of digit identity III (Tamura et al.,2011). Both studies show that the developmental pathway of chicken wing digits is derived compared with those in the hindlimb or to the mouse. These facts are strong evidence for the FSH, because they confirm that the avian hand developmental pathway has been altered in evolution.
While both of these studies support the FSH, they have distinct implications for the mechanisms of digit identity frame shift itself. The down-regulation of HoxD expression occurs late in development (stages 28–30), potentially after individuation of digit IV (in the hindlimb, stage 26, Suzuki et al.,2008) suggesting that the frame shift may be incomplete (i.e., does not include all three digits of the avian wing). Alternatively, it is possible that only two digit identities exist: digit I and the posterior digits. Digit I is the only digit to develop in the absence of Shh signaling and without expression of Hoxd12–8. In addition to sharing developmental genetic features, with the exception of phalangeal number and length, the posterior digits are often anatomically quite similar (e.g., in the human hand), suggesting that differences among these digits may simply be morphological variants (i.e., character states, Wagner,2007) of the same character identity. However, this is unlikely in the bird wing, as three digits have very distinct morphologies. Moreover, the anterior displacement of ZPA cells occurs before individuation of any of the digits (stage 20). As a potential mechanism underlying the frame shift of digit III, this finding suggests that all three digits have unique identities and have undergone a frame shift. Further investigations of digit identity determination in birds are necessary to tease apart the details of the frame shift and its underlying mechanisms.
The authors thank John Fallon for the invitation to contribute a manuscript to the special issue on limb development. We thank H. Larsson and two anonymous reviewers for comments that improved this manuscript.