The genome sequences of two non-bilaterian animals, the cnidarians Nematostella vectensis and Hydra magnipapillata, have been recently completed. These new data lead to the fascinating result that the complement of Hox genes in the cnidarian ancestor is considerably lower than that in the bilaterians, although the complexity of their genome is otherwise similar (Technau et al. 2005). Thus, there is a correlation between the radiation of the Bilateria and the expansion of the Hox complex.

In the first part of this chapter, we shall present and discuss these data. In the second part, we shall present a novel hypothesis accounting for this phenomenon. In short, we surmise that the expansion of the Hox complex at the base of the Bilateria was due to a series of transposition events. Indeed, we hypothesise that the Hox genes themselves originate from transposons. The main support for this hypothesis is provided by the similarity between the homeodomain and the DNA-binding domain of bacterial integrases and eukaryotic transposases. We also examine some very precise rearrangements of the Hox complex in the Drosophilidae lineage. In the third part, we propose a scenario for the evolution of the Hox complex from the basic complement of Hox genes in the common ancestor of cnidarian and bilaterian animals. This scenario, based on our transposition hypothesis, accounts for several properties of the extant Hox genes.

TO SET THE SCENE: THE HOX EXPLOSION

The homeobox is a conserved motif found in a huge variety of eukaryotic genes, encoding a DNA-binding domain.

Several alternative hypotheses have been suggested that support various phylogenetic groupings of the individual euarthropod taxa, the chelicerates, myriapods, crustaceans and insects. The Tetraconata hypothesis suggests a sister-group relationship of insects and crustaceans in contrast to the traditional monophyletic grouping of myriapods and insects, the Tracheata or Atelocerata (see references in Stollewerk and Chipman 2006). The Mandibulata hypothesis suggests a clade consisting of insects, crustaceans and myriapods (see references in Harzsch et al. 2005). However, the relationships within this clade are being debated since this hypothesis excludes neither the Pancrustacea nor the Tracheata concept. The latest hypothesis suggests a sister-group relationship of chelicerates and myriapods. Although this theory was initially based on molecular phylogenetic analysis (Friedrich and Tautz 1995, Hwang et al. 2001, Kusche and Burmester 2001, Nardi et al. 2003, Mallatt et al. 2004, Pisani et al. 2004), recent morphological and molecular data on neurogenesis in these groups potentially support a close relationship (Stollewerk et al. 2001, 2003, Dove and Stollewerk 2003, Kadner and Stollewerk 2004, Stollewerk and Simpson 2005, Chipman and Stollewerk 2006, Stollewerk and Chipman 2006). On top of the various ideas on the relationships within the euarthropods, none of the groups except the insects is generally accepted as monophyletic.

New insights into the evolutionary relationships between the different taxa have been gained by comparing morphological features and expression patterns of genes involved in developmental processes.

Evolutionary developmental biology or evo-devo will make crucial contributions over the coming years to exploring the occupancy of morphospace. Why do species show the patterns of diversity and disparity they do, and to what extent do such patterns reflect the ways in which phenotypic variation is generated as well as the processes of natural selection? Evo-devo in appropriate study systems is providing the means to explore fully how phenotypic variation is generated by the processes intrinsic to individuals, especially those of development. Substantial progress will be made in understanding the occupancy of morphospace if the results of this type of evo-devo can then be combined with analyses of how this same variation is influenced by natural selection and other extrinsic processes to result in the patterns of evolution. Use can also be made here of recent advances in developing appropriate null models for testing adaptive versus neutral or random-walk explanations of evolution (Pie and Weitz 2005).

In combination, this type of broad evo-devo and evolutionary biology can begin to unravel how evolvability, the capacity to evolve at the genetical and developmental levels, contributes to shaping the evolution of patterns of diversity and disparity in phenotypic space (Brakefield 2006). We will then be able to examine the extent to which such phenomena as genetic channelling, developmental bias and developmental drive are reflected in patterns of evolution, whether involving change or stasis (e.g. Maynard Smith et al. 1985, Schluter 1996, Wagner and Altenberg 1996, Arthur 2001, Blows and Hoffman 2005).

Evolutionary theory is the philosophical backbone of biology. Interestingly, contemporary research in evolutionary biology involves several parallel lines of investigations that build on different philosophies and aim for different kinds of explanations and mechanisms. At its extreme, at least three independent research activities are actively promoted in evolutionary biology: neo-Darwinism with a population genetics research agenda analyses the evolution of populations by natural selection (Amundson 2005). Molecular phylogeny tries to reconstruct historical patterns and the phylogenetic relationship of organisms using cladistic approaches. And finally, comparative morphology, and more recent ‘evo-devo’ research, build on the evolution of ontogeny and try to show how modifications of development (ontogeny) result in evolutionary novelties (Valentine 2004, Kirschner and Gerhard 2005).

All of these agenda are actively propagated and they all consider themselves to follow the Darwinian logic. Surprisingly, however, there is hardly any cross-talk between these disciplines and even worse, these research fields ignore each other to a certain extent. Several authors have emphasised the different research strategies and philosophies in contemporary evolutionary biology, i.e. neo-Darwinism and evolutionary developmental biology (Wilkins 2002, Amundson 2005). Despite these obvious problems and lack of interactions, we are in need of a true synthesis of evolutionary thought. And such a synthesis must include both population genetic and developmental thinking. In this context, homology could be an important concept.

If evo-devo is a discipline in its own right, is there a distinctive set of biological systems and methods of investigation through which it is currently advancing? Although evo-devo probably does not rely upon specific tools of analysis unknown in other fields of biological research, because of its particular relationships to both evolutionary and developmental biology evo-devo exhibits a specific combination of model systems and research tools. In other words, to use a fashionable term in developmental genetics, it has its own toolkit. However, what is most distinctive about evo-devo materials and methods is that, precisely because tools devised in other fields are here used at the borders of their original range of application, investigations in this interdisciplinary territory periodically need a critical evaluation of the sharpness, precision and adequacy of these tools. A survey of this important work is offered in this section.

The model organism approach has become the lingua franca in modern biology. However, a good model for medicine, where one searches for conserved features shared with humans, is not necessarily a good model for understanding evolutionary change. Athanasia Tzika and Michel Milinkovitch (Chapter 7) tackle the problem of model organism choice in evo-devo studies. The authors propose a pragmatic optimisation approach that incorporates criteria suggested by evolutionary history such as the phylogenetic position of candidate model species and the presence of ancestral/derived character states, along with practical attributes such as the feasibility of handling, housing and breeding.

According to Darwinism, evolution occurs because, in populations, there is heritable phenotypic variation, and then ecological factors differentially affect the contribution of each of those variants to the next generation. Thus, to understand the way in which phenotypes in a population change over generations (this is the direction of evolutionary change) two questions need to be addressed: (1) which phenotypic variants arise in each generation, and (2) which of these variants are filtered out by ecological factors in each generation. In each generation, and assuming no dramatic genomic rearrangements, developmental dynamics determine which morphological variation arises from genetic and environmental variation. Developmental dynamics are currently not very well understood and thus the question of which phenotypic variants arise in each generation is not well understood either. A different emphasis is given to each of these two questions by different approaches or schools of thought in evolutionary biology.

For many evolutionary biologists, especially those close to the core of neo-Darwinism, this lack of understanding about development has not always been perceived as a limit on progress in understanding morphological evolution (Haldane 1932, Mayr 1982). For some authors (Haldane 1932) this does not derive from lack of understanding development. Instead, natural selection is seen as the main or unique force determining how the phenotype changes (Fisher 1930, Charlesworth et al. 1982). The question of which phenotypic variation arises and what its role is in determining how phenotypic distributions change over time is either not addressed or assumed to be unanswerable.

INTRODUCTION

Our green and living world is a continuum in space and time. This view is well expressed in the ‘continuum model’ proposed by botanists and biophilosophers such as Arber (1950) and Sattler (1996). As an opposite view we may accept the green world around us as consisting of discrete units on several hierarchical levels. This view is called here the ‘discontinuum model’ or the ‘classical model’ because it has been the predominant view in biological textbooks for decades. Branching and repetition of developmental units (e.g. cells, meristems, modules, leaves, phytomers) are omnipresent as developmental processes in multicellular plants. These processes resemble the process of segmentation in various metazoan phyla, also occasionally leading to fuzzy borderlines between consecutive developmental units (Minelli and Fusco 2004, Prusinkiewicz 2004, Rutishauser and Moline 2005). Perspectivists studying plants accept structural and developmental categories such as cells, meristems, modules, leaves and phytomers as mind-born, simplified concepts reflecting certain aspects of the structural diversity (Sattler and Rutishauser 1990, Hay and Mabberley 1994).

A key focus of evolutionary developmental biology (evo-devo) in recent years has been to elucidate the evolution of developmental mechanisms as a means to reconstructing the hypothetical last common ancestors of various clades. Prominent among such reconstructions have been proposals as to the nature of the mysterious Urbilateria, originally defined as the last common ancestor (LCA) of the extant Bilateria (Ecdysozoa, Lophotrochozoa and Deuterostomia) (De Robertis and Sasai 1996, Kimmel 1996). Indeed, drawings of this animal can now be found, as well as detailed information on the genetics and morphological processes that it used to construct its gut, heart, eyes, appendages, segments and body region identities (Gilbert and Singer 2006). Perhaps surprisingly, however, no explanations have yet been offered of how it might have achieved the successful reproduction that must have been necessary for it to give rise to still surviving lineages. This chapter will examine the comparative data available on the specification of bilaterian reproductive systems during development, with special emphasis on the cells containing the genetic hereditary material, the germ cells, and speculate on the possible gonad structure and reproductive strategy of Urbilateria.

Before proceeding, we should clarify our expectations as to what the study of extant species can tell us about Urbilateria. In this chapter, I wish to avoid suggesting that extant reproductive systems are simply variations on a defined metazoan reproductive ‘Bauplan’ theme; the great weakness of the current evo-devo approach stems from dilution of explanatory force with inappropriate fixations on strict, confining definitions of this kind (Scholtz 2004, 2005, Hübner 2005).

GENE CLASSIFICATION IS AN ESSENTIAL PRECURSOR TO EVO-DEVO

A sensible classification of developmental control genes and an understanding of their phylogeny are essential to any endeavour of molecular evolutionary developmental biology (evo-devo) or comparative genomics, since it is crucial that the structure, expression and function of orthologous genes are being compared between taxa. This is particularly true for the homeobox genes, for which there are confusing and conflicting names and classifications that hinder our investigation and understanding of their evolution and their role in animal evo-devo (I will restrict myself here to consideration of animal homeobox genes). Since these genes are central components of most developmental processes, are important indicators of major transitions in animal genome evolution, and are often found to be targets and/or agents of the evolution of development, then we must continue to improve and coordinate our classifications of these genes as more data become available from a greater array of taxa in this age of genomics.

CONVENTIONS

Animal homeobox genes can be divided, on the basis of their sequence similarities, into two major classes (ANTP and PRD) along with several minor classes (TALE, LIM, POU, ZF, cut, prox, HNF and SIX; Bürglin 2005, Edvardsen et al. 2005, Holland and Takahashi 2005). It is in the ANTP-class that most confusion and discrepancy exists, and so I shall concentrate on this class and attempt to resolve at least some of the confusion.

Stephen Jay Gould opens the Prospectus of his influential Ontogeny and Phylogeny (Gould 1977) with the following quotation from Van Valen (1973): ‘A plausible argument could be made that evolution is the control of development by ecology. Oddly, neither area has figured importantly in evolutionary theory since Darwin, who contributed much to each. This is being slowly repaired for ecology … but development is still neglected.’

As accurate as these comments may have been in 1977, today, 30 years later, they no longer hold true: two new fields centred on the study of organismal development have now emerged in modern biology. One of them, which has successfully married the traditional fields of embryology and genetics, is the field of developmental genetics. The other one is known as developmental evolution, evolutionary developmental biology or simply evo-devo, and is the primary subject of this book and this chapter.

The evo-devo field has set as its ultimate goal to provide a mechanistic explanation of how developmental mechanisms changed during evolution, and how these alterations are causally linked to modifications in morphological patterns (Holland 1999). These questions are most relevant, as, so far, the formal structure of the evolutionary theory has been based upon the dynamics of alleles, individuals and populations under selective pressures and genetic drift ‘assuming’ the prior existence of these entities (Fontana and Buss 1993).

‘Distinctive features’, ‘key apomorphies’, ‘novelties’, ‘innovations’. All these terms point to the fact that some characters in an organism are ‘special’ in some respect, and these have attracted the most interest from biologists and challenged the explanatory capacity of evolutionary theories. These are characters that emerged as new features during the evolution of a lineage, or that are believed to be the principal cause of a successful phyletic radiation. ‘Evolution is tinkering’ and in any such feature there are both new and conserved components, even if it is not always obvious what is new and what is not. By taking development into the picture, can evo-devo provide deeper insight into the origin, evolution and diversification of such characters? The answer seems to be yes.

An initial example is offered by Cassandra Extavour (Chapter 17), who addresses the evolution of bilaterian reproductive systems by examining comparative data on somatic gonad and germ cell specification during development. Surprisingly enough, although reproduction is possibly the most important quality for a living being, the evolution of bilaterian reproductive systems and strategies has received comparably little attention. Was the last common ancestor of bilaterians (the so-called Urbilateria) hermaphroditic? Did it possess a sequestered germ line, or distinctive gonads? The data presented here can tell us what kinds of general features, or basic pattern, Urbilateria's reproductive system was likely to have had, thus accounting for the systems found in extant bilaterian lineages.

Comparison is fundamental to any evolutionary developmental analysis (e.g. Alberch 1985, Rieppel 1988, Dohle 1989, Minelli 2003, Scholtz 2005, Deutsch 2006, Jenner 2006, Breidbach and Ghiselin 2007). However, evo-devo as a discipline evolved from a mix of experimental and descriptive approaches to development. Accordingly, different weight is put on the method of studying development in an evolutionary framework depending on a researcher's scientific background. Here I want to evaluate the different approaches and their contribution to addressing evolutionary questions. I stress that only the comparative approach offers a direct method of studying development with respect to evolutionary changes. Descriptive and comparative approaches are often interpreted as being less ‘exact’ than experimental studies because they deal with untestable scenarios. Here I want to show that comparative approaches are a direct means to study evolution if the latter is accepted as the general framework for reasoning about causality and changes and the link between the two.

Since its inception in the early 1980s, evo-devo has evolved into a mature discipline. This is manifest in the naming of research groups, scientific journals and books, professional meetings and societies. Despite such formal attributes of a scientific discipline it is often unclear what constitutes its conceptual distinctiveness. Does evo-devo have its own set of specific questions and research methods? Does it solve biological problems that cannot be solved by other approaches? And does it represent a significant change in the theoretical understanding of development and evolution? That is, in which way do the goals, the empirical programs and the theories of evo-devo research differ from those of neighbouring disciplines such as developmental biology or evolutionary biology? The present chapter provides a concise overview of the current status of evo-devo as a discipline. This requires a short reflection on its history.

CONCEPTUAL FOUNDATIONS

The parallels between embryonic stages and the ‘scale of beings’ had already been contemplated in pre-Darwinian times, and the foundation of a scientific theory of evolution was significantly influenced by embryological arguments. Darwin called embryology ‘by far the strongest single class of facts in favour of a change of form’, and his first sketches of a phylogenetic tree seem to have been inspired by tree-like renderings of embryological differences between species (Richards 1992). Much of the early work in evolutionary biology focused on the uses of embryonic characters for taxonomical purposes.

One major classical justification of using a model metazoan species for experimentation has been that discoveries of biological phenomena in that species could be extrapolated to other multicellular species. Because the chances that this extrapolation is valid in humans depend on the phylogenetic distance between humans and the model species, many researchers have somewhat sacrificed the major benefits of small size, short generation time and ease of manipulation that characterise some invertebrates in order to use species that humans can more readily relate to, such as the laboratory mouse (Mus musculus). However, the community of biologists has continued to use additional model species because each of the selected taxa have specific features that make experimental manipulation easier (e.g. easy-to-score morphological variation and giant polytene chromosomes in Drosophila melanogaster, or accurate description of the largely invariant complete cell lineage and full neural connectivity in the roundworm Caenorhabditis elegans).

Ever since the molecular genetic revolution, a constant concern has been the possibility of manipulating the genome of model species. For example, generations of Drosophila scientists have developed and applied ingenious approaches that allow, in principle, screening for any phenotype at any stage of development (reviewed in St Johnston 2002). Even for the mouse model, multiple techniques, such as homologous recombination, tissue-specific activation/inactivation techniques, cloning and RNA interference (RNAi), have been developed for performing genotype- or phenotype-driven experiments. Furthermore, recent access to full genome sequences makes genome engineering of some model species easier.