Developmental evolution as mechanistic science: The inference from developmental mechanisms to evolutionary processes

Developmental evolution as mechanistic science: The inference from developmental mechanisms to evolutionary processes

Wagner, Gunter P

Developmental Evolution as a Mechanistic Science: The Inference from Developmental Mechanisms to Evolutionary Processes1

SYNOPSIS. Developmental Evolution (DE) contributes to various research programs in biology, such as the assessment of homology and the determination of the genetic architecture underlying species differences. The most distinctive contribution offered by DE to evolutionary biology, however, is the elucidation of the role of developmental mechanisms in the origin of evolutionary innovations. To date, explanations of evolutionary innovations have remained beyond the reach of classical evolutionary genetics, because such explanations require detailed information on the function of genes and the emergent developmental dynamics of their interactions with other genetic factors. We argue that this area has the potential to become the core of DE's disciplinary identity. The main challenge in developing a research program for DE along these lines, however, is to provide a methodological framework that accounts for the fact that developmental mechanisms continue to evolve after a character has originated. Developmental mechanisms elucidated in a derived species may therefore not provide insights into the evolutionary origin of the character in question. To meet this challenge, we propose a set of questions that may guide us in our search for valid inferences on the role of developmental mechanisms in the explanation of evolutionary innovations.

INTRODUCTION

In an influential paper from 1968, Gunter Stent characterized three clearly recognizable phases in the history of molecular biology: the so-called romantic, dogmatic and academic phases (Stent, 1968). We propose that developmental evolution (DE) is going through the same three phases, except that we prefer to call the second phase "enthusiastic" rather than dogmatic3. Today workers are still enthralled by their enthusiasm for a new enterprise; hence, DE has not yet fully entered the most mature academic phase of its history. The romantic phase of DE culminated in the Field Museum conference on Macroevolution in 1980 and the Dahlem Conference on Development and Evolution in 1981 (Bonner, 1982). At that time DE was focused on staking out its conceptual territory and defending it against established disciplines. Formative for the romantic phase were Stephen Gould's Ontogeny and Phylogeny (could, 1977), Rupert Riedl's Order in Living Organisms (Riedl, 1978) (first German edition) and Rudolf Raff and Thomas Kaufman's Embryos, Genes and Evolution (Raff and Kaufman, 1983). There was a sense of great promise shared among the workers, but the actual accumulation of knowledge was relatively slow. This picture changed radically with the breakthroughs in Drosophila developmental genetics and the discovery of homologous developmental genes in other animals. We think that this led to the enthusiastic phase of DE. More and more startling discoveries were and continue to be made. While many discoveries were overinterpreted (we decline to cite examples) in the first wave of enthusiasm, general progress in DE, however, became irreversible. Examples for this progress are the discovery of the genetic basis for the evolution of butterfly wing patterns (Brakefield et al., 1996; Keys et aL, 1999; Nijhout, 1991) and for the evolution of arthropod body regions (Averof and Akam, 1993, 1995; Carroll, 1995; Damen and Tautz, 1999). These and other cases are a clear indication that DE is maturing into a productive research enterprise. But there are still some features missing in DE that one would like to see in a mature science. One of them is an agreement about the goals of DE. We will argue in the next section that DE contributes to a set of heterogeneous research programs, but that it also makes a unique contribution, which has not been anticipated by any established research program. Furthermore we will argue that it is important for a mature DE to establish rigorous standards of evidence. It is, for example, not clear what kinds of data justify the conclusion that a particular gene is instrumental in the origin of a character, such as insect wings. Recent advances in developmental biology clearly validate the expectation that we will be able to answer such questions in the near future. But it is also clear that there are many methodological pitfalls along the way. Towards the end of this paper we will propose a set of methodological criteria (a sort of checklist or a list of "commandments," depending on one's temperament) for establishing a causal link between molecular developmental evolution and phenotypic evolution.

WHAT IS THE AGENDA OF DEVELOPMENTAL EVOLUTION?

Our survey of the work published on developmental evolution indicates that there are at least five partially overlapping research goals (Table 1).

1) The most obvious can be characterized as research in the Evolution of Development. Increasingly detailed knowledge about the mechanisms of development as a newly recognized level of biological organization invites the comparative study of these mechanisms. Perhaps the most important conclusion reached by this type of inquiry is the insight that developmental cascades constitute a level of organization and have their own evolutionary history, which is to some degree independent of the evolutionary history of the characters for which they are responsible (Abouheif, 1999; Gerhart and Kirschner, 1997; Shubin et al., 1997). Increased knowledge of the phylogenetic history of developmental mechanisms also raises the question of what evolutionary forces shape development. What causes the evolution of direct and indirect modes of development? Why are imaginal discs present in some insects and not in others? These are just a few examples of important biological questions that fit into the growing set of evolutionary research programs such as molecular, phenotypic, and behavioral evolution that focus on a specific character type.

2) Another exciting implication of DE is that in some cases gene expression patterns may help to resolve longstanding questions in assessing homologies. Initially there was a tendency to assume that expression data can overrule any other kind of data relevant to homology assessment, but recently a more balanced view has emerged (Bolker and Raff, 1996; Dickinson, 1995; Galis, 1999; Miller and Wagner, 1996; Wray, 1999). Nevertheless, detailed developmental studies are an invaluable source of additional information when assessing the homology of problematic characters. One can therefore expect that DE will continue to contribute to the agenda of comparative anatomy, originally set out by Owen, Gegenbauer and others for the identification of corresponding body parts in divergent animal body plans.

3) DE also promises to fill a gap in the current account of the adaptationist program in illuminating the structure of the genotype-phenotype map (Mayr, 1983). While it is clear that adaptations result from natural selection on spontaneous heritable variation, the genetic and developmental architecture of this adaptive variation is largely unknown. To understand the dynamics of adaptation, it is necessary to know how complex an adaptive change is at the genetic level. Recent progress in evolutionary genetics demonstrates that the genetic basis of species differences is now within the reach of experimental research. For example, it has been shown that the genetic basis of bristle pattern differences on the third leg of Drosophila can be traced to variation in the cis-regulatory sequences of Ubx (Stern, 1998). Similarly, variation in the regions of the axial skeleton of anmiotes seems to be caused by changes in Hox gene expression domains (Belting et aL, 1998; Burke et aL, 1995).

4) From the beginning a goal of DE research has been to determine if developmental constraints influence patterns of evolutionary diversification (Alberch, 1983; Gould, 1977; Riedl, 1978; Wake and Larson, 1987). The locus classicus of this research are the papers by Alberch and Gale (Alberch and Gale, 1983, 1985) that demonstrated a causal link between patterns of digit reduction and the mode of digit development.

More recently, developmental and cell biological mechanisms have also been invoked to explain the evolvability of complex biological traits (Gerhart and Kirschner, 1997). Here, developmental regulation may prevent unconditionally deleterious effects of mutations (Gerhart and Kirschner, 1997), which, in turn, can facilitate the evolvability of complex characters (Wagner and Altenberg, 1996). An example is the compensatory capacity of signal transduction networks characterized by weak interactions. Genes that play a central role in signal transduction, such as ras and fos have been identified, but surprisingly, in spite of the central function of these genes in the transduction cascade, deletions of these genes have only mild effects. This is because other genes can compensate for their function. This type of compensatory interaction thus provides physiological robustness and the potential for variation, i.e., evolvability (Gerhart and Kirschner, 1997).

5) Finally, DE may lead to a mechanistic explanation of the origin of evolutionary innovations and the origin of body plans (MOller and Wagner, 1991). Evolutionary innovations and the evolution of body plans are hard to understand in population genetic terms since they involve radical changes in the genetic/developmental architecture of the phenotype. A good example of the power of DE in this context is the origin of butterfly eye spots (Keys et al., 1999). Since this innovation has produced a qualitatively new phenotypic state, a quantitative genetic account is inherently uninformative. Indeed, knowledge on the functional interactions among the participating genes and gene products is necessary for understanding the processes that led to these new characters.

Our list is certainly not comprehensive; for example it does not include the influence of developmental evolution on the study of molecular evolution (Purugganan, 1998; Sidow, 1992; Zardoya et al., 1996). We believe, however, that our list does capture the main contributions of research in DE. This list can be divided into two categories. On the one hand are those cases in which DE contributes new facts to existing research programs. These are the contributions of DE to homology assessment, the evolution of development (which can be seen as another set of biological characters that evolve) or to the explanation of the developmental architecture of adaptations. In each case the agenda is set by an established research program and DE makes ancillary, though important, contributions. The only two dimensions of DE, which are outside the scope of existing research programs, are the idea of developmental constraints and the explanation of innovations. The concept of developmental constraints expanded the scope of the neo-Darwinian theory of evolution. It was also the first context in which developmental mechanisms acquired an explanatory role in evolutionary biology (Sterelny, 2000). Similarly, evolutionary innovations are outside the scope of any current research program. Through its contribution to the solution of that question, DE genuinely expands the explanatory range of evolutionary theory. We think that this is the one area where DE will have its most lasting impact on evolutionary theory and biology in general. This assessment does not mean that we downplay the contributions of DE to other research programs, but it means that we see in the problem of innovation and the evolution of body plans a unique opportunity for DE to develop its own independent identity as a research program. In the rest of the paper we will focus on the methodological challenges of explaining evolutionary innovations through developmental mechanisms. Many of the things we say will in principle apply to any attempt to link genetic events to phenotypic evolution. However, the difficulties are more pronounced in the case of innovations because many of these events happened a long time ago.

PROBLEMS IN ESTABLISHING A CAUSAL LINK BETWEEN GENETIC CHANGE AND A PHENOTYPIC INNOVATION

The question of identifying what exactly is an innovation, is beyond the scope of this paper (Miiller and Wagner, 1991). Examples of innovations are the origin of new body parts, or any significant change in the identity of a character, such as the transformation of nodular bones into long-bones (Blanco et al., 1998). A precise explication of innovation thus requires a theory of biological characters, which is still an unresolved issue (Wagner, 2000). Furthermore, any change in the level of selection also qualifies as an innovation, such as the origin of multi-cellular organisms or the origin of eukaryotic cells (Buss, 1987; Maynard-- Smith and Szathmary, 1995). Here we will primarily discuss the origin of new body parts and changes in the identity of body parts.

Molecular research into the origin of a specific character is usually triggered by the discovery of certain genes that are essential for the development of that character in some model species. The discovery of sufficient and/or necessary genetic factors for a phenotypic character naturally raises the question of whether the same genes were also involved in the evolutionary origin of that character. Usually this leads to the hypothesis that the developmental causes of a character may also be part of the evolutionary cause for the origin of the character. One could call this the

Hypothesis of congruence between developmental and evolutionary causes: a gene that is sufficient for the development of a derived character state may also have been the cause of the evolutionary origin of the character state.

This is the null-hypothesis for DE research. There are at least four specific methodical and biological problems that make the assessment of this hypothesis non-trivial. Below we will discuss these problems and will propose recommendations of how to deal with them.

Of course the hypothesis of congruence between developmental and evolutionary causes requires a few qualifications. The causes of evolutionary processes, such as selection and drift are, in part, the population genetic processes that lead to the fixation of genetic variation. In that sense there cannot be any congruence between developmental mechanisms and evolutionary processes, since developmental mechanisms act at the level of individual organisms while evolutionary mechanisms act at the level of populations. However, in the case of evolutionary innovations, the specific developmental functions of the genes involved are an important part of the explanatory narrative. To state that a genetic mutation led to a favored character, which, in turn, was selected is utterly uninformative in explaining innovation. Such an account is only sufficient for quantitative changes, identified and characterized by means of quantitative genetic techniques, where it can be shown that contributions of individual genes are not that important to the phenotypic result. In contrast, the emergence of morphological innovations depends to a large extent on the epigenetic dynamics of the involved developmental pathways (Newman and Miller, 2000). Therefore in explaining the origin of a new character, the developmental function of genes is of greater importance than in quantitative genetics. To speak of a congruence between developmental mechanism and evolutionary process means that the developmental function of a gene in the derived species was also the developmental function for which the gene was selected when the new character arose.

Character definition

To be successful, a research program of elucidating the genetic basis of phenotypic innovations demands a rigorous definition of the derived phenotypic character (-state). In other words, it is difficult to reach a consensus view of the genetic basis of a morphological innovation, if the precise definition of the new derived character remains unclear. An instructive example of this problem was reported in an important paper on the evolution of insect wings (Warren et al., 1994). It is well known that a loss of function mutation of the Ubx gene in Drosophila transforms a haltere into a wing, which suggests that Ubx controls "wing number" in Drosophila. To further test this possibility, Carroll and collaborators (Warren et al., 1994) hypothesized that Ubx may be responsible for the difference between four- and two-winged insects. In testing this, however, they found that in four-- winged insects the expression of Ubx is identical to that of Drosophila, i.e., to a two-winged insect. This finding was surprising in light of the supposed role of Ubx in establishing wing number. It turns out, as shown by Carroll and collaborators, that the differences between four and two-winged insects are caused by genetic factors downstream of Ubx. Is there is a systematic reason for their findings? A possible answer to this question can be found when one reconsiders the character definition: instead of assuming that wing number is a unitary character, Carroll and collaborators suggested that the haltere is a character state of the hind-wing (see Warren et al., 1994). In other words, fore- and hind-wings are two characters whereby the haltere is a character state of the hind-wing. Diptera, including Drosophila, have four "wings," but the function of its hind-wing is not to create uplift. One can see the haltere as a character state of the hind wing, rather than something completely different. Ubx determines hind wing identity, regardless of whether the adult character is a functional wing or a sense organ (a haltere). Therefore the differences between the Diptera and the four-- winged insect are caused by genes downstream of Ubx.

Another example of how the definition of a character can influence a research program involves investigations into the genetic basis of the fin-limb transition. There is agreement that the main difference between the fins of the sarcopterygians (lobefinned fishes) and the limbs in the earliest tetrapods is the distal most part, the autopodium (i.e., hand and foot) (Ahlberg and Milner, 1994; Coates, 1991; Shubin, 1995; Sordino and Duboule, 1996; Laurin et al., 2000). But what precisely is an autopodium? There are many morphological and developmental differences between the autopodium and the proximal parts of the limb. Some authors define the autopodium through the presence of digits (Coates, 1994; Daeschler and Shubin, 1998). So defined, the fin-limb transition has to be explained in terms of the evolution of the digital arch and the derivation of digits from radials in fins (Shubin and Alberch, 1986). Alternatively the autopodium can be defined as consisting of a segment of small nodular elements (carpals and tarsals), the so called mesopodium and distal to it a set of digits, i.e., metapodials and phalanges, the so called acropodium (Fig. 1). So defined, the question is how was the developmental boundary between the lower limb and the mesopodium established in evolution? In this paper we do not discuss evidence in favor of the one or the other hypotheses (for a more extensive discussion see Wagner and Chiu, 2000), but use it as an illustration that further demonstrates how the definition of a character strongly influences molecular studies aimed at elucidating its origin.

"Recency bias" in cladistic character reconstruction

Another methodological challenge to DE comes from the fact that DE data have to be obtained from extant species. Under favorable circumstances and based on a wellestablished phylogeny, the ancestral developmental and morphological phenotype can be reconstructed (Maddison and Maddison, 1992). The problem for understanding evolutionary innovation is that the reconstructed character states may not reflect the events at the origin of the character. There is a "recency bias" built into cladistic character reconstruction methods. By the very nature of the cladistic method applied to recent species, the deepest node that in principle can be reconstructed is the most recent common ancestor of the extant clade. For instance, following the phylogenetic hypotheses of Ahlberg and Milner (1994) and Laurin (1998) the most recent common ancestor of extant tetrapods is most likely from the Carboniferous age (ca. 340 MA) (Fig. 2) (but see [Coates, 1994] for a dissenting view). The fin-limb transition, however, happened about 30 million years earlier in the Devonian. The limb morphology of the Devonian forms is also quite different from that of recent forms and all the carboniferous forms (Coates and Clack, 1990; Coates, 1991, 1996). Hence the most recent common ancestor of extant tetrapods is not the direct product of the fin-limb transition, but rather the product of the transformation of the archaic limbs into modern ones.

Temporal dissociation between origin and canalization of a character

Closely related to the methodological problem outlined in the last paragraph is a biological problem. The "recency bias" of cladistic character reconstruction would not be a problem for the reconstruction of the developmental and genetic bases of innovations, if the developmental architecture of the character remains unchanged after its origin. There can, however, be a temporal dissociation between the origin and the canalization of a character4. Again the difference between Devonian and recent tetrapod limbs is an excellent example.

As mentioned above, the typical limb of a recent tetrapod is pentadactyl. This limb type is fundamentally different from that of Devonian forms (Fig. 3). The Devonian tetrapods have up to eight digits and the digits can be grouped into at least two different morphological classes that differ in size and cross section (Coates, 1991; Coates and Clack, 1990). Furthermore the morphology of the mesopodium is not comparable to that of modern forms. These morphological facts show that the pentadactyl limb arose after the fin-limb transition and is the product of the canalization of the phenotype of the archaic limbs. Many developmental features that are universal (or nearly so) among recent tetrapods were not necessarily acquired during the fin-limb transition but rather by the Devonian and early Carboniferous tetrapods which evolved the pentadactyl limb. For instance almost all tetrapods examined have a digital arch from which most of the digits develop (Shubin and Alberch 1986)1. There are, however, other modes of digit development described for extant species (Fig. 4). Hence a possible explanation of the difference between archaic and modern tetrapod limbs is that archaic limbs may have used more than one mode of digit development, leading to different kinds of digits. In contrast, digit development in modern tetrapods is almost completely monopolized by the digital arch. This switch from two or three parallel modes of digit development to one (the digital arch) may explain the canalization of the modern limb. We suggest that the mode of digit development in modern tetrapods (the digital arch) may be the product of the secondary stabilization of the limb phenotype and may not have been acquired during the fin-limb transition. Indeed, only those developmental characters that explain the morphological features common to all tetrapods, including the archaic forms, are candidates for accounting for the fin limb transition. We argue elsewhere that the feature common to all tetrapods, archaic and modern, is the existence of a mesopodialacropodial configuration (Fig. 1) and that the genetic mechanism for the development of this feature is more likely involved in the fin-limb transition than mechanisms involved in digit development (Wagner and Chiu, 2000). This example also shows that paleontological data are an essential component of DE research.

Recruitment of genes into existing developmental pathways

Another fact that makes it problematic to draw inferences of evolutionary processes from developmental mechanisms is the finding that the developmental function of a gene can fundamentally change without affecting the morphology of phenotypic character(s). One of the most striking examples is the fact that even-skipped, a pair rule gene essential for the development of postcephalic segments in Drosophila, and most likely other higher insects, does not have this function in segment development in grasshoppers (Patel et aL, 1992). This shows that development and its genetic regulation is a feature with its own evolutionary history that is partially autonomous of the evolutionary history of the characters they "code for." The genetic machinery active in the development of a derived species may not be the same as the one that was active in the ancestor that originally acquired the character. Consequently the genes that have been recruited into the development of a character after its origin have nothing to do with the genetic mechanisms responsible for the origin of a character.

This and the previous argument about the possible secondary canalization of a character reflect the fact that the development of a character continues to evolve after the evolutionary origin of that character. The difference between the two cases is that in the first, canalization, the character itself shows obvious signs of morphological evolution from the archaic to modern types. In the second example the character, the insect segment, does not show an evolutionary trend in its morphological character identity. The main difference here is the mode of development (long germ versus short germ development) without any effect on the phenotypic character. This is the only reason for distinguishing between canalization and recruitment. For the elucidation of the developmental basis of innovations these two cases have the same effect, namely that one has to examine the evolution of the developmental mechanisms in addition to the evolution of the character itself.

These examples certainly do not exhaust all the possible methodological pitfalls that one is likely to encounter in research into the origin of characters. They may, however, suffice to show that the inference from developmental mechanisms to evolutionary processes needs methodological constraints in addition to those imposed by developmental genetics. In the next section we propose a small set of "questions" that may guide the formulation of a valid argument that links a genetic to a phenotypic event in evolution.

A CHAIN OF QUESTIONS LINKING GENETIC TO PHENOTYPIC EVOLUTION

The natural starting point of any DE project on the origin of a novel character is the discovery of a genetic mechanism for the development of a derived character, say the limb or the insect wing. Therefore the answer to the first question we propose below will be the starting point rather than the result of a DE project. Nevertheless, the results from developmental genetics have to be scrutinized as to whether they, in fact, imply a meaningful hypothesis about the origin of a novel character.

A) What is the developmental mechanism that accounts for the derived character (-state) ?

This question actually stands for a complex of related questions that have to do with defining the phenotypic innovation and with establishing the developmental mechanism that specifically accounts for the innovation. What exactly is the phenotypic difference between the derived and the ancestral character state? What is the taxonomic group for which this character state is synapomorphic? What are the outgroup taxa and do they represent the ancestral character state? Does the identified developmental mechanism account specifically for the derived character state or is it responsible for more fundamental and more ancient (plesiomorphic) character states within the lineage?

An example to illustrate the last question is the function of Shh in limb development. Clearly Shh, which is released in the ZPA (zone of polarizing activity), is necessary for proper digit development (Riddle et al., 1993). Its role in digit development, however, is derivative of a more general and more ancient function. The ZPA determines the anterior-posterior polarity of paired appendages in vertebrates. Fins also have an anterior-posterior polarity and it seems that Shh already has this function in fins, since it is expressed in a similar location in zebrafish fin buds as it is in tetrapod limb buds (Krauss et al., 1993). A gene with an essential function in the development of a character thus may have this function because of a more ancient role than the one related to the development of the specific character.

B) Does the developmental mechanism for the derived character (-state) map to the same node on the phylogeny as the derived character (-state)?

This question is the first test of the causal efficacy of the developmental mechanism in the evolutionary process. If the developmental changes are a cause for the morphological difference then they have to be coincidental. The evolutionary change in development can neither be older nor younger than the origin of the character state it is supposed to explain. In addition, a positive answer to this question eliminates one of the problems discussed above, namely that the development of a character can continue to evolve after the character originated. Only those developmental mechanisms that were active in the ancestor can account for the evolutionary origin of the character. Our ability to answer this question is limited by the "recency bias" of character reconstruction as discussed above. There are situations where the reconstructed nodes are not a close approximation of the event that led to the origin of a character. But there are other kinds of data that could be employed to resolve the issue.

For instance one would expect that a mechanism, call it D, that depends on another mechanism, say A, will have evolved after the mechanism A. This is not an absolute rule, because upstream mechanisms can evolve after downstream mechanisms, but in general it may be used as an indication of the possible sequence of evolutionary events. For example, there are two groups of Hox genes involved in the development of the autopodium. These are the AbdB homologues on Hox gene clusters A and D. Mutational analyses have shown that the developmental function of the HoxD genes is downstream of the HoxA genes (Zakany et al., 1997). It also happens to be the case that the HoxA genes are involved in setting up the boundary between the anlage of the hand and the lower arm, while the HoxD genes regulate the number and morphology of the digits. From these data Duboule and collaborators have concluded that the function of HoxA genes may be more ancient than that of the HoxD genes. Note that the lack of direct descendants from the Devonian tetrapods prevents us from reconstructing the events that occurred between the origin of limbs and the origin of the penta-dactyl limb.

C) What are the developmental changes that occurred at the origin of the derived character (-state)?

The answer to this question requires the reconstruction of the developmental mechanism in the ancestor of the derived clade. This will only be possible if appropriate out-group species are available for investigation. This information is also necessary to constrain the possible candidates for the allelic substitutions (or other genetic changes) that caused the morphological difference. An example of such an hypothesis is the suggestion that the origin of the autopodium (hand and foot) may have been caused by the evolution of a non-overlapping expression domain of the genes Hoxa11 and Hoxa-13 (Fig. 5). In tetrapods the expression domains for these two genes are non-overlapping when the autopodium develops (Davis et al., 1995; Haack and Gruss, 1993; Yokouchi et al., 1991, 1993), while they are overlapping in zebrafish (Sordino et al., 1995). This scenario predicts that the origin of the autopodium is caused by changes in the regulation of these two genes, either by upstream regulators, mutations in cis-regulatory sequences or in the activator and repressor domains of the protein. Preliminary results show that the evolution of the derived expression pattern is coincidental with the acquisition of new putative repressor domains in the N-terminal region of the Hoxa-11 encoded protein (Chiu et al., 2000). In general the study of the sequence evolution of developmental regulatory genes is a powerful tool for detecting candidate mutations that may be responsible for developmental and phenotypic transformations (Purugganan, 1998).

D) Are the genetic differences sufficient to cause the derived character (-state)?

This is experimentally the most difficult question to answer. It requires the identification of genetic differences that may account for the developmental differences and the ability to introduce genetic elements into in-group as well as out-group species. Advances in transgenic technology, however, make the thought of performing these experiments less crazy than they were a few years ago. For instance there is increasing evidence that specific changes in the cis-- regulatory region of Hox genes may account for differences in the size of body regions of vertebrates (Belting et al., 1998; Burke et al., 1995). Ideally one would like to cause a transformation towards the derived state in an outgroup species by introducing a genetic element from a derived species and an atavistic character state by introducing a genetic element from an outgroup species into a derived species. We are not aware of a case where both experiments have been performed.

To summarize, if all the answers to these questions support a hypothesis about the developmental mechanism for the origin of a novel phenotypic character, then the conclusion is all but unavoidable that this mechanism in fact was instrumental in causing the origin of the derived character (-state).

CONCLUSIONS

Developmental Evolution makes important contributions to the agenda of established research programs, such as the assessment of homology and the evolution of adaptations. In addition, DE opens up new areas of research that have not been part of any of the established research programs, especially in elucidating the genetic factors that are responsible for the origin of evolutionary innovations, and, in particular, the origin of new characters. We think that it is this area in which DE will eventually make its most distinctive contribution to evolutionary biology and where it may find its own disciplinary identity. The main methodological challenge of this research program is to link developmental mechanisms to evolutionary processes. We have exemplified the problems with a number of examples and suggested a "check list" for constructing a valid chain of inferences between developmental mechanisms and evolutionary innovation. From this list we conclude that a successful DE project needs essential input from at least five biological disciplines: 1) developmental biology, which provides insights into the proximate mechanisms of character development; 2) evolutionary genetics, which provides insights into the forces acting on the developmental genes; 3) systematics, which provides us with the comparative methods for testing the evolutionary assertions; 4) comparative anatomy, which helps us to coherently define the characters and the character states we seek to understand; and 5) paleontology, which provides us with information about the sequence of phenotypic transformations that led to the character states found in extant forms. All of these contributions are essential for DE.

AcKNOWLEDGMENTS

The authors are grateful to Frank Ruddle, Terri Williams and the members of the Ruddle and Wagner labs for stimulating discussions on the subject of this papers. The financial support by NSF grant 9905403 is gratefully acknowledged. CHC is supported by a NSF-Sloan postdoctoral fellowship.

From the symposium Evolutionary Developmental Biology: Paradigms, Problems, and Prospects presented at the Annual Meeting of the Society for Integrative and Comparative Biology, 4-8 January 2000, at Atlanta, Georgia.

3 For a discussion of the prehistory of DE please refer to the papers by Hall (2000), Gilbert (2000), Laubichler (2000) and Dietrich (2000).

4 For a definition and discussion of the canalization concept see Gibson and Wagner (2000).

5 In addition the most posterior digit, DV, develops from an independent condensation. The only exception to the Shubin-Alberch model of digital arch development are the urodeles.

REFERENCES

Abouheif, E. 1999. Establishing homology criteria for regulatory gene networks: Prospects and challenges. In G. R. Bock and G. Cardew (eds.), Homology, pp. 207-221. J Wiley & Sons, Chichester, England.

Ahlberg, P E. and A. R. Milner. 1994. The origin and early diversification of tetrapods. Nature 368:507514.

Alberch, P. 1983. Morphological variation in the neotropical salamander genus Bolitoglossa. Evolution 37:906-919.

Alberch, P and E. A. Gale. 1983. Size dependence during the development of the amphibian foot. Colchicine-induced digital loss and reduction. J. Embryol. Exp. Morph. 76:177-197.

Alberch, P and E. A. Gale. 1985. A developmental analysis of evolutionary trend: Digital reduction in Amphibians. Evolution 39:8-23.

Averof, M. and M. Akam. 1993. HOM/Hox genes of Anemia: Implications for the origin of insect and crustacean body plans. Curr. Biol. 3:73-78.

Averof, M. and M. Akam. 1995. Hox genes and the diversification of insect and crustacean body plans. Nature 376:420-423.

Belting, H. G., C. S. Shashikant, and F H. Ruddle. 1998. Modification of expression and cis-regulation of Hoxc8 in the evolution of divergent axial morphologies. Proc. Nat. Acad. Sci. U.S.A. 95: 2355-2360.

Blanco, M. J. and P. Alberch. 1992. Caenogenesis, developmental variability, and evolution in the carpus and tarsus of the marbled newt Triturus marmoratus. Evolution 46:677-687.

Blanco, M. J., B. Y. Misof, and G. P. Wagner. 1998. Heterochronic differences of Hoxa- 11 expression in Xenopus fore- and hind limb development: evidence for a lower limb identity of the anuran ankle bones. Dev. Genes Evol. 208:175-187.

Bolker, J. A. and R. A. Raff. 1996. Developmental genetics and traditional homology. BioEssays 18: 489-494.

Bonner, J. T. (ed.) 1982. Evolution and development. Springer-Verlag, Berlin, Heidelberg, New York. Brakefield, P M., J. Gates, D. Keys, E Kesbeke, P.

Wijngaarden et al., 1996. Development, plasticity and evolution of butterfly eyespot patterns. Nature 384:236-242.

Burke, A. C., C. E. Nelson, B. A. Morgan, and C.

Tabin. 1995. Hox genes and the evolution of vertebrate axial morphology. Development 121:333346.

Buss, L. W. 1987. The evolution of individuality. Columbia University Press, New York.

Carroll, S. B. 1995. Homeotic genes and the evolution of arthropods and chordates. Nature 376:479-485. Chiu, C.-H., D. Nonaka, L. Xue, C. T. Amemiya, and

G. P Wagner. 2000. Evolution of Hoxa- 11 in lineages phylogenetically positioned along the finlimb transition. Mol. Phyl. Evol. (In press)

Coates, M. 1991. New paleontological contributions to limb ontogeny and phylogeny. In R. J. Hinchliffe (ed.), Developmental patterning of the vertebrate limb, pp. 325-337 Plenum Press, New York.

Coates, M. I. 1994. The origin of vertebrate limbs. Development 1994 Suppl:169-180.

Coates, M. 1. 1996. The Devonian tetrapod Acanthostega gunnari Jarvik: Postcranial anatomy, basal tetrapod interrelationships and patterns of skeletal evolution. Trans. Roy. Soc. Edinburgh: Earth Sci. 87:363-427.

Coates, M. I. and J. A. Clack. 1990. Polydactyly in the earliest known tetrapod limbs. Nature 347:6669.

Daeschler, E. B. and N. Shubin. 1998. Fish with fingers? Nature 391:133.

Damen, W. G. M. and D. Tautz. 1999. Abdominal-B expression in a spider suggests a general role of Abdominal-B in specifying the genital structure. J. Exp. Zoo. (Mol. Dev. Evol.) 285:85-91.

Davis, A. P, D. P Witte, H. M. Hsieh-Li, S. S. Potter, and M. R. Capecchi. 1995. Absence of radius and ulna in mice lacking hoxa-11 and hoxd- 11. Nature 375:791-795.

Dickinson, W. J. 1995. Molecules and morphology: Where is the homology? Trends in Genetics 11: 119-121.

Dietrich, M. 2000. From hopeful monsters to homeotic effects: Richard Goldschmidt's integration of development, evolution, and genetics. Amer. Zool. 40:738-747.

Galis, F 1999. On the homology of structures and Hox genes: The vertebral column. In G. R. Bock and G. Cardew (eds.), Homology, pp. 80-90 by J Wiley, Chichester, England.

Gerhart, J. and M. Kirschner. 1997. Cells, Embryos and Evolution. Blackwell Science, Malden, MA. Gibson, G. and G. P. Wagner. 2000. Canalization in

evolutionary genetics: A stabilizing theory? BioEssays 22:372-380.

Gilbert, S. E 2000. Diachronic biology melts evodevo: C. H. Waddington's approach to evolutionary developmental biology. Amer. Zool. 40:729737.

Gould, S. J. 1977. Ontogeny and phylogeny. The Belknap Press of Harvard University Press, Cambridge, MA.

Haack, H. and P Gruss. 1993. The establishment of murine Hox-1 expression domains during patterning of the limb. Dev. Bio. 157:410-422.

Hall, B. K. 2000. Balfour, Garstang, and deBeer: The first century of evolutionary biology. Amer. Zool. 40:718-728.

Hinchliffe, J. R. and E. I. Vorobyeva. 1999. Developmental basis of limb homology in urodeles: heterochronic evidence from the primitive hynobiid family. In G. R. Bock and G. Cardew (eds.), Homology, pp. 95-105. J. Wiley, Chichester, England.

Keys, D. N., D. L. Lewis, J. E. Selegue, B. J. Pearson, L. V. Goodrich, R. L. Johnson, J. Gates, M. P Scott, and S. B. Carroll. 1999. Recruitment of a hedgehog regulatory circuit in butterfly eyespot evolution. Science 283:532-534.

Krauss, S., J. P Concordet, and P W. Ingham. 1993. A functionally conserved homolog of the Drosophila segment polarity gene hh is expressed in tissues with polarizing activity in zebrafish embryos. Cell 75:1431-1444.

Laubichler, M. D. 2000. Homology in development and the development of the homology concept. Amer. Zool. 40:777-788.

Laurin, M. 1998. A reevaluation of the origin of pentadactyly. Evolution 52:1476-1482.

Laurin, M., M. Girondot, and A. de Ricqles. 2000. Early tetrapod evolution. TREE 5:118-123. Maddison, W. P and D. R. Maddison, 1992. MacClade

Version 3: Analysis of phylogeny and character evolution. Sinauer Ass., Sunderland, MA. Maynard-Smith, J. and E. Szathmary. 1995. The major

transitions in evolution. W. H. Freeman, Oxford, New York, Heidelberg.

Mayr, E. 1983. How to carry out the adaptationist program? Am. Nat. 121:324-334.

MUller, G. B. and G. P. Wagner. 1991. Novelty in evolution: Restructuring the concept. Ann. Rev. Ecol. Syst. 22:229-256.

Miller, G. B. and G. P. Wagner. 1996. Homology, Hox genes and developmental integration. Amer. Zool. 36:4-13.

Newman, S. A. and G. B. Miller. 2000. Epigenetic mechanisms of character origination. In G. P Wagner. (ed.), The character concept in evolutionary biology, pp. 559-579 Academic Press, San Diego.

Nijhout, H. F. 1991. The development and evolution of butterfly wing pattern. Smithonian Institution Press, Washington and London.

Patel, N. H., E. E Ball, and C. S. Goodman. 1992. Changing role of even-skipped during the evolution of insect pattern formation. Nature 357:339342.

Purugganan, M. D. 1998. The molecular evolution of development. BioEssays 20:700-711.

Raff, R. A. and I C. Kaufman. 1983. Embryos, genes, and evolution. Macmillan Publishing Co., Inc., New York.

Riddle, R. D., R. L. Johnson, E. Laufer, and C. Tabin. 1993. Sonic hedgehog mediates the polarizing activity of the ZPA. Cell 75:1401-1416.

Riedl, R. 1978. Order in living organisms. Wiley, New York.

Schmalhausen, J. J. 1910. Die Entwicklung des Extremitatenskeletts von Salamandrella Kayserlingii. Anat. Anz. 37:431-446.

Shubin, N. 1995. The evolution of paired fins and the

origin of tetrapod limbs. Evolutionary Biology 28: 39-86.

Shubin, N., C. Tabin, and S. Carroll. 1997. Fossils, genes and the evolution of animal limbs. Nature 388:639-648.

Shubin, N. H. and P Alberch. 1986. A morphogenetic approach to the origin and basic organization of the tetrapod limb. Evol. Biol. 20:319-387.

Sidow, A. 1992. Diversification of the Wnt gene family in the ancestral lineage of vertebrates. Proc. Nat. Acad. Sci. U.S.A. 89:5098-5102.

Sordino, P and D. Duboule. 1996. A molecular approach to the evolution of vertebrate paired appendages. TREE 11:114-119.

Sordino, P., E v. d. Hoeven, and D. Duboule. 1995. Hox gene expression in teleost fins and the origin of vertebrate digits. Nature 375:678-681.

Stent, G. 1968. That was the molecular biology that was. Science 160:390-395.

Sterelny, K. 2000. Development, evolution and adaptation. Phil. Science (In press)

Stern, D. L. 1998. A role of Ultrabithorax in morphological differences between Drosophila species. Nature 396:463-466.

Vorobyeva, E. I. and J. R. Hinchliffe. 1996. Developmental pattern and morphology of Salanandrella keyserlingii limbs (Amphibia, Hynobiidae) including some evolutionary aspects. Russ. J. Herpetol. 3:68-81.

Vorobyeva, E. L, 0. P 0l'shevskaya, and J. R. Hinchliffe. 1997. Specific features of development of the paired limbs in Ranodon sibiricus Kessler (Hynobiidae, Caudata). Russ. J. Dev. Biol. 28: 150-158.

Wagner, G. P (ed.) 2000. The character concept in evolutionary biology. Academic Press, San Diego, CA.

Wagner, G. P and L. Altenberg. 1996. Complex ad

aptations and the evolution of evolvability. Evolution 50:967-976.

Wagner, G. P and C.-H. Chiu. 2000. The tetrapod limb: A definition and a hypothesis on its origin In G. B. Muller and S. Newman (eds.), Origins of organismal form. MIT Press, Cambridge, MA. (In press)

Wake, D. B. and A. Larson. 1987. Multidimensional analysis of an evolving lineage. Science 238:4248.

Warren, R. W, L. Nagy, J. Senegue, J. Gates, and S. Carroll. 1994. Evolution of homeotic gene regulation and function in flies and butterflies. Nature 372:458-461.

Wray, G. A. 1999. Evolutionary dissociations between homologous genes and homologous structures. In G. R. Bock and G. Cardew (eds.), Homology, pp. 189-202. J. Wiley, Chichester, England.

Yokouchi, Y., H. Sasaki, and A. Kuroiwa. 1991. Homeobox gene expression correlated with the bifurcation process of limb cartilage development. Nature 353:443-445.

Yokouchi, Y., M. Yamamoto, T Toyota, H. Sasaki, and A. Kuroiwa. 1993. Regulatory interaction of positional signalings on coordinate expression of homeobox genes in developing limb buds. Prog. Clin. Biol. Res. 383A:71-78.

Zakany, J., C. Fromental-Ramain, X. Warot, and D. Duboule. 1997. Regulation of number and size of digits by posterior Hox genes: A dose-dependent mechanism with potential evolutionary implications. Proc. Nat. Acad. Sci. U.S.A. 94:1369513700.

Zardoya, R., E. Abouheif, and A. Meyer. 1996. Evolutionary analyses of hedgehog and Hoxd-10 genes in fish species closely related to the zebrafish. Proc. Nat. Acad. Sci. U.S.A. 93:1303613041.

GINTER P WAGNER,2,* CHI-HUA CHIU,* AND MANFRED LAUBICHLER^

*Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut 06520-8106 ^Program in History of Science, Princeton University, Princeton, New Jersey 08544

2E-mail: gunter.wagner@yale.edu

Copyright Society for Integrative and Comparative Biology Nov 2000
Provided by ProQuest Information and Learning Company. All rights Reserved