Tetraploidy

Major Transitions in Animal Evolution: A Developmental Genetic Perspectivel

Peter W.H. Holland

SYNOPSIS. Several phases of animal evolution have undergone radical change in developmental mechanisms. I refer to these as major transitions in animal evolution. The six most important transitions in the lineage leading to humans are proposed to be: the origin of multicellularity, the origin of two-germ layers and radial symmetry, the origin of three-germ layers and bilateral symmetry, dorsoventral axis inversion, the origin of vertebrates, the origin of gnathostomes. Here I discuss the genetic changes that may have underlain these transitions. The last two transitions were accompanied by, and possibly facilitated by, large increases in gene number. This probably occurred by tetraploidy, with some of the duplicate genes being subsequently lost. The origin of three germ-layers, bilateral symmetry and a through gut also probably involved gene duplication; in this case, duplication of an ancestral ProtoHox gene cluster to yield two paralogous homeobox gene clusters, Hox and ParaHox, with roles in patterning different germ layers along the anteroposterior body axis. This event may provide a partial genetic explanation for the Cambrian explosion.

INTRODUCTION The range of body forms found across the animal kingdom is strikingly discontinuous. This discontinuity is reflected in taxonomy. For example, the taxonomic rank of phylum is used to divide the animal kingdom into sets of species that share a unique body plan comprising a particular spatial arrangement of morphological characters. This does not mean that transitions from one body plan to another are impossible in evolution, but they are certainly infrequent. Indeed, notwithstanding recent debates about the nature of Cambrian faunas (Gould, 1989; Conway Morris, 1994), a remarkably small number of animal phyla have ever evolved: probably fewer than forty. To further understand the nature of body plan evolution, including the constraints to change in body plan, it is helpful to take a developmental perspective. Body form is the end-product of a series of cellular processes, occurring during embryonic development, that convert genetic and epigenetic information into tissues, organs, relative positions, repetition and shape. From the developmental viewpoint, it is clear that all phylum-level (or subphylum-level) differences are not equal. Some phyla, whilst very different as adults, are likely to employ very similar developmental processes; their differences can be explained (in theory at least) by rather small-scale tweaking of developmental controls. Examples could include annelids compared to molluscs or pogonophorans, or onychophorans compared to arthropods. In contrast, some phases of animal evolution must have witnessed great change in developmental control mechanisms, verging on revolution. I refer to these developmental revolutions as major transitions in animal evolution.

I will not attempt to catalogue all major transition that have occurred through diversification of the animal kingdom. Instead, I concentrate on our own lineage, from the origin of animals to the evolution of humans. As explained below, I suggest there were at least six distinct phases along this lineage during which very substantial change in developmental mechanisms occurred. Each phase may have included several sequential transitions, although these are not readily divisible using data from extant taxa. Consideration of another lineage, such as that leading to flies, or that to bivalves, would share some of these major transitions, but include other transitions. Other authors have suggested very different approaches to defining major transitions in evolution. For example, MaynardSmith and Szathmary (1995) concentrate on methods of information transfer rather than developmental control. Consequently, they highlight very different events from those highlighted in this paper. Within the animal kingdom, for example, Maynard-Smith and Szathmary include only the origin of multicellularity, the origin of social groups and the origin of language in their list of major transitions; only the first coincides one of the major transitions as defined here. ANIMAL PHYLOGENY In order to identify times in evolution when developmental mechanisms were radically altered it is necessary to have a sound phylogeny of the animal kingdom. The past ten years have witnessed major steps forward in this direction, particularly through the use of ribosomal gene sequencing, pioneered for this application by Field et aL (1988). It is not possible here to review a decade of work in this area, suffice to say that ribosomal gene sequences have now been obtained from representatives of most animal phyla; analytical methods have also advanced to the stage where phylogenetic signal in a set of DNA sequences can generally be distinguished from noise. Consequently, there is now general consensus on the interrelationships of the major phyla of animals, perhaps for the first time. There are still disputes about many phylum-level relationships, of course, and a few have not been analyzed at all, but the general pattern of animal phylogeny is clear (for example, Philippe et al., 1994; Halanych et al., 1995; Aguinaldo et al., 1997). The phylogenetic hypothesis used in this paper is depicted in Figure 1.

Some aspects of Figure 1 are almost universally accepted; other aspects are far more contentious and require some justification. Figure 1 depicts a basal split between the Parazoa (sponges) and the rest of Metazoa; this is generally accepted. It is also commonly assumed that the common ancestor at this node had a body organisation of the sponge grade of organisation (Brusca and Brusca, 1990). Looking just at the lineage that will eventually lead to humans, the next side branch to diverge leads to the Cnidaria (sea anemones, jellyfish, hydroids). Again, this rather basal position for Cnidaria is almost universally accepted, as is the contention that the common ancestor at this node had a body organisation of the diploblast (two germ layer) grade of organisation (Brusca and Brusca, 1990). The subsequent phase of evolution along our lineage is more difficult to piece together. Traditional zoology textbooks generally place the so-called 'acoelomate' and 'pseudocoelomate' phyla as the next animals to diverge (for example, Barnes, 1980). However, the existence of such intermediate grades of organisation is not clearly confirmed by molecular phylogenetic studies. Ribosomal DNA sequencing suggests that the three best known and speciose groups within these categories, the Nemertea (ribbon worms), Platyhelminthes (flatworms, tapeworms, flukes) and Nematodes, should all be elevated to a more crownward position in the phylogeny, within a clade containing the classic coelomate protostomes (arthropods, annelids, molluscs and others; Turbeville et al., 1992; Balavoine, 1997; Aguinaldo et al., 1997). The elevated position for nematodes is not found in all studies, hence the dotted line on Figure 1. The removal of Nemertea and (at least some) Platyhelminthes from a basal position in the tree is more consistent between studies and is probably very sound. This means that the next phase of evolution along our lineage, after divergence of Cnidaria, was a radiation into many lineages of triploblastic (three germ layer) animals. Resolution of the precise branching order has proved problematic, and may reflect rapid radiation of several lineages (Philippe et al., 1994).

From this rapid radiation, it is clear that the deuterostomes emerged as a defined branch, containing the common ancestor of echinoderms, hemichordates and chordates. The phylogeny of deuterostomes is rather clearly resolved by molecular phylogenetics (Fig. 1) and includes a surprising sister group relationship for echinoderms and hemichordates, and a chordate phylogeny involving divergence of the urochordate lineage, then the cephalochordates, and finally emergence of the vertebrate (craniate) lineage (Wada and Satoh, 1994). The nature of most ancestral forms within the deuterostomes is unclear, except that it is generally accepted that the common ancestor of cephalochordates and vertebrates had the cephalochordate grade and layout of organisation (Holland, 1996). Figure 1 resurrects the term agnatha (including a monophyletic cyclostome group of extant hagfish and lampreys) for the basal clade of vertebrates; this is contentious, since it is consistent with most molecular evidence, but not most morphological and palaeontological analyses (Stock and Whitt, 1992; Lanfranchi et al., 1994; Forey and Janvier, 1993). Whether extant agnatha are a monophyletic clade or a paraphyletic grade does not affect the present discussion. SIX PHASES OF DEVELOPMENTAL REVOLUTION From the above phylogeny, plus inferred pattern of character change, six internode phases can be identified as times when developmental control mechanisms must have undergone radical alteration (numbered 1 to 6 on Fig. 1). Changes in developmental control must ultimately be based in changes at the genetic level. Three general categories of genetic change might be associated with any phenotypic change: the causative, the permissive and the inconsequential. In practise, distinguishing these will be difficult. Simply looking for genetic differences between higher level taxa will not suffice, since over the time scales involved most of the genetic differences are likely to be inconsequential to the transition itself. Instead, it is necessary to take into account the nature of the developmental alterations, hypothesize the sorts of genes that may have been important, and then focus attention on these. An important property to examine for each character is its phylogenetic distribution, as discussed by Erwin (1993). Transition 1 marks the origin of multicellularity, and must have involved the origin of numerous cellular and developmental mechanisms. These include cell layers, cell adhesion, and spatially controlled patterns of differentiation. Genes necessary for these cellular functions include cadherins, extracellular matrix molecules, collagen, integrins and transcription factors involved in the specification of distinct cell types. In may be predicted, therefore, that many of these genes will be present in sponges (and all other animals), but not in non-animal outgroups. Representatives of several of these gene classes have been cloned from sponges, including a collagen gene, an Stype lectin, and an integrin gene (see Pancer et al., 1997). Erwin (1993) argues that the origin of collagen may have been a particularly important step permitting the origin of multicellular animals. PCR screens have also identified a number of homeobox genes in sponges (Seimeya et al., 1994), the majority of these, perhaps all, belong to classes that in higher animals are associated primarily with cell differentiation roles, as opposed to spatial patterning roles. These data are consistent with the hypothesis that many genes necessary for multicellularity evolved at transition 1, permitting the origin of animals. One major problem, however, is that insufficient molecular data have been obtained from close outgroups, such as choanoflagellates (as of October 1998, the GenBank database includes no protein-coding sequences from any choanoflagellate). Transition 2 involved the origin of defined axes of symmetry, more precisely controlled species-specific body shape, structural repetition, origin of neurones and the invention of defined inner and outer epithelial germ layers. The origin of neurones is predicted to have required genes for neurotransmitters, their receptors and ion channels; in fact, many of these genes predate the Metazoa (Erwin, 1993). It is difficult to predict which genes would be necessary for control of precise shape, repetition and the origin of germ layers, since the developmental control of these characters has been studied primarily in bilateral triploblasts rather than radially symmetrical diploblasts. Research into cnidarian development in several laboratories is addressing these questions, and a few candidate genes have emerged (Schummer et al., 1992; Naito et al., 1993; Miller and Miles, 1993; Kuhn et al., 1996). Most intriguingly, cnidarians of all three classes possess homeobox genes closely related in deduced protein sequence to the Hox genes of chordates, arthropods and nematodes. Comparison between species of cnidarian suggests there are at least five of these 'Cnox' genes, denoted Cnox1 to Cnox-5. There is also some evidence, particularly from regeneration studies in Hydra, that Cnox genes play roles in axial specification (a vital component of shape generation) and/or germ layer distinction (Schummer et al., 1992).

These sequence comparisons and expression studies have tended to highlight the similarities between Cnox and Hox genes; indeed, the former are often referred to simply as cnidarian Hox genes. Consequently, the origin of Hox genes is often considered to predate the common ancestor of Cnidaria ploblasts. This influenced the zootype hypothesis of Slack et al. (1993). In that paper, we reasoned that the origin of a Hox gene cluster (together with other genes) may have been influential in permitting the origin of animals with defined axes and germ layers. Recent studies have uncovered two principal problems with this hypothesis. One problem emerged from our studies of Hox-related genes in higher triploblasts, and will be discussed later. The other emerged from molecular phylogenetic analyses of Cnox and Hox sequences. Kuhn et al. (1996) undertook maximum parsimony and UPGMA analyses using (cnidarian) Cnox and (triploblast) Hox homeobox DNA sequences, and concluded that Cnox1 to Cnox-5 are not directly equivalent to particular Hox genes. Instead, the common ancestor of Cnidaria and triploblasts possessed a single Cnox/Hox precursor gene that diversified by independent gene duplications in the two lineages. The methodology of this analysis may be challenged in that DNA sequences are less appropriate than protein sequences for inter-phylum analysis (third codon positions could be under differing directional selection due to codon usage bias) and that neither parsimony nor UPGMA are ideal methodologies for deducing gene histories when substitution rates may differ greatly between lineages. Nonetheless, our own analyses using homeodomain protein sequences and neighbor-joining distance methods (corrected for multiple substitutions) support the essential conclusions of Kuhn et al. (N.M. Brooke and PWH.H., unpublished). Our analyses suggest that the common ancestor of cnidarians and triploblasts may have had just one, or more likely two, Cnox/Hox genes; these duplicated independently in the two lineages.

The implication for interpreting the evolutionary transitions in Figure 1 is that the first homeobox gene with a Hox-like sequence originated around transition 2, predating the emergence of axes and germ layers. We describe this as a ProtoHox gene, rather than a true Hox gene, since it was the precursor of Cnox genes, Hox genes and (as shown later) several other genes. It is equally important to note that an elaborate Hox gene cluster did not necessarily originate at this transition point; it is possibly a later evolutionary invention. Origin of the ProtoHox gene(s) may have been an important prerequisite to the evolution of the diploblast grade of body organisation, but it cannot have been the only gene necessary. It is interesting to note that homologues of several other homeobox genes, implicated in spatial patterning in triploblasts, have been cloned from cnidarians. These include Evx and Emx class homeobox genes. The former gene was cloned from the coral Acropora formosa (Miller and Miles, 1993), where it was shown to be physically linked to a Cnox gene. An Emx gene from Hydractinia was cloned by 0. Mokady (personal communication; GenBank Y11836). Evx genes in vertebrates play roles in patterning the posterior of the body axis, whilst Emx genes in chordates and arthropods play roles at the anterior. It will be important to deduce if the cnidarian homologues have similar roles along the oral-aboral axis; such a finding would strengthen the hypothesis that their origin, together with ProtoHox gene(s), permitted evolutionary transition 2.

Transition 3, more than any of the other transitions noted, is the one most likely to be a composite of a series of sequentially occurring developmental changes. Together, they resulted in the conversion of a two germ layer body plan (as seen in cnidarians) into a three germ layer body plan, the invention of bilateral symmetry (with distinct dorsal, ventral, left, right, anterior and posterior), the consolidation of the nervous system into a centralised axial nerve cord, and the origin of a through gut with distinct mouth, anus and intermediate regions. Insight into the genetic changes that occurred during this transition (or series of transitions) can be gained by comparison of cnidarian and triploblast genes (particularly those implicated in anteroposterior, dorsoventral, left-right and germ layer patterning). I will consider these genetic differences at the end of this article, since insight into this transition has emerged most recently. The differences are also best inter preted in the light of the other data discussed. Transition 4 is unusual in that it is not defined by invention of new structures. Instead, developmental evidence suggests that an inversion of dorsoventral patterning systems occurred somewhere along the deuterostome lineage. Much of the recent molecular data supporting this view are summarized by DeRobertis and Sasai (1996), and additional supporting data have accrued since their article. How such an axis inversion occurred is unclear and hotly debated. At one extreme, a simple rotation of the body would transpose dorsal and ventral, and simultaneously left and right. If this was the mechanism, it would hardly qualify as a major developmental change; indeed, it would be very trivial. There are many examples of animals that swim or crawl upside-down in comparison to their close relatives (water boatmen, upside-down catfish and tree sloths, for example). This simple mechanism for axis inversion is unlikely, however, since other aspects of embryology seem to differ in concert with the axis inversion. This suggests that axis inversion actually reflects a reorganisation of gastrulation, possibly concomitant with the evolution of a new oral opening in deuterostomes (Lacalli, 1996). More research is needed, particularly on axis formation and axis homologies in echinoderms and hemichordates, to clarify the picture. Until then, any genetic correlates of axis inversion are likely to remain completely unknown. Transition 5 marks the origin of vertebrates and includes invention of a new strategy for deploying cells in development (the use of pluripotential migratory neural crest cells), origin of many new cell types (such as osteoblasts and odontoblasts) and elaboration of the mesodermal germ layer by mediolateral subdivision. Transition 6 occurred within the vertebrates, after divergence of the extant agnathans, and involved invention of migratory lateral mesodermal cells, origin of two sets of patterned paired appendages, and anteroposterior diversification of the cranial visceral arches. Although much additional morphological evolution occurred between the origin of the jawed vertebrates and the emergence of humans, I suggest this did not involve radical alteration to developmental patterning, at least not on the scale of transitions 1 to 6. Insight into genetic correlates of transitions 5 and 6 have come from comparison of developmentally expressed genes between ascidians, amphioxus, hagfish, lampreys and jawed vertebrates; these are outlined in the next section.

NEW GENES IN VERTEBRATE EVOLUTION The idea that extensive gene duplications occurred during the emergence and early evolution of vertebrates is not new (Ohno, 1970). Recent molecular work has led to confirmation of this idea, but has substantially revised it. A large body of evidence now exists to demonstrate that extensive gene duplication occurred at evolutionary transition 5, and again at transition 6, on the phylogeny in Figure 1. I will not go through this evidence in detail, since most of the data have been thoroughly discussed in previous publications (for reviews see Holland et al., 1994; Holland, 1996; Holland and Garcia-Fernandez, 1996; Sharman and Holland, 1996; Sidow, 1996). The data are rather fragmentary, but internally consistent. In brief, it was found that the Hox gene clusters duplicated during early vertebrate evolution; a long list of additional gene families are now known to have duplicated between the time of divergence of cephalochordates and the radiation of jawed vertebrates. Note, however, that this period encompasses both transition 5 and transition 6. Of these, just a few have so far been shown to have duplicated before the divergence of lampreys and hagfish, and a few others later than this time. It is currently thought that the second of the gene duplication phases (transition 6), and possibly also the first (transition 5), occurred by tetraploidy of the genome. This mechanism is supported by the large number genes that duplicated and by aspects of mammalian genome organisation (Lundin, 1993).

Some very recent data have helped to refine the above picture. Sharman et al. (1997) report the cloning of an HMG-1/2 type gene from a lamprey, and demonstrate that this gene family duplicated after the divergence of lampreys but before the divergence of ray-finned fish and tetrapods. The implication is that this gene family did not duplicate at transition 5, but only at transition 6. Does this imply transition 6 was a more significant phase of gene duplication than phase 5? Not necessarily, since some genes certainly did duplicate at phase 5. Pendleton et al. (1993) estimated the number of Hox gene clusters in the sea lamprey to be three, clearly more than the single Hox gene cluster proven for amphioxus (Garcia-Fernandez and Holland,1994). Our recent PCR screens for Hox genes in the river lamprey found a remarkably similar set of genes (Sharman and Holland, 1998). Since there is a large phylogenetic distance between sea and river lampreys, this suggests that neither PCR screen suffered greatly from amplification bias; further more, we can be confident in the authenticity of the genes. Hence, the Hox gene cluster duplicated close to transition 5 and probably again at transition 6. If Hox genes show evidence of duplication at transition 5, but HMG-1/2 genes do not, is this inconsistent with tetraploidy being the mechanism of gene duplication at transition 5? If tetraploidy occurred, we must conclude that a duplicate copy of HMG-1/2 was lost before transition 5. How likely is this? A clue has come from our studies on the aromatic amino acid hydroxylase (AAAH) genes (Patton et al., 1998). Cloning of an amphioxus AAAH gene, and molecular phylogenetic analysis, reveals that these genes did duplicate during vertebrate evolution, but all of the duplicate genes were subsequently lost, leaving the same gene complement in vertebrates and in triploblast invertebrates. Lack of duplication as an explanation can be ruled out by consideration of chromosomal location of AAAH genes in mammals; this confirms duplication then loss of all duplicates. The implication is that some genes (such as AAAH genes) are prone to complete loss of every duplicate, whereas others (such as homeobox genes, Pax genes, Wnt genes and other developmentally important genes) have very low rates of loss. To understand why, we should ask what the probability is that duplicate genes acquire new roles relative to the probability of pseudogene formation. We suggest that genes with widespread or ubiquitous expression, particularly those encoding enzymes (such as AAAH genes) require rare advantageous mutations in coding regions-for example, active sites-to acquire new roles. The ubiquitously expressed HMG-1/2 genes may be close to this model. In contrast, genes with restricted expression patterns (homeobox genes, Pax genes, Wnt genes etc) may acquire new roles through occurrence of much more probable mutations in regulatory regions altering spatiotemporal expression (Patton et al., 1998).

The AAAH example also implies that tetraploidy does not necessarily result in a net increase in gene number between invertebrates and vertebrates for every gene family. Examples of gene families that do not show a net increase are not evidence against the tetraploidy hypothesis, since gene loss can occur (Fig. 2). Most gene families studied, however, do show evidence of gene duplication between invertebrates and jawed vertebrates. Our laboratory has recently added a few more gene families to the growing list: the HNF3 family of forkhead genes (Shimeld, 1997), the Pax-3/7 family (Wada et al., 1996, 1997), the Pax-2/5/8 family (Wada et al., 1998) and the Otx class of homeobox genes (Williams and Holland, 1998). In the case of HNF3, an independent duplication also occurred in the amphioxus lineage; the Otx example is interesting in that a tandem duplication of a domain within the portion was used to firmly back up molecular phylogenetic evidence for gene duplication in the vertebrate lineage. The Pax-3/7 example involved comparison of ascidian and vertebrate genes, and revealed that after gene duplication, vertebrate Pax-3/7 genes acquired new roles specifically in patterning of mesoderm. Co-option of genes from ectoderm to mesoderm following gene duplication is also a feature of Hox gene evolution (Holland and Garcia-Fernandez, 1996).

NEW GERM LAYERS, NEW GUT, NEW GENES? In the section outlining the six phases of developmental revolution, I postponed discussion of transition 3. This truly major transition involved the conversion of a twogerm layer, radially symmetrical grade of organisation to a three-germ layer, bilaterally symmetrical body plan with a centralised axial nerve cord and a specialised through gut. Very many developmental processes must have been modified and elaborated during these dramatic changes, and it may be considered a hopeless task to decipher the suite of genetic changes at their root. However, a similar gloomy prediction could have been made a few years ago for attempts to understand the genetic basis of early vertebrate evolution, yet useful progress has been made. As shown above, extensive gene duplication occurred during early vertebrate evolution and duplicate copies of genes were retained (particularly of developmentally important genes). The implication is that gene duplication produced a suite of new developmental genes that could be co-opted for new developmental roles. In this way, gene duplication permitted the major developmental changes in early vertebrate evolution. Our recent work on Hox gene evolution suggests that something analogous may have happened around transition 3, possibly permitting the rise of the triploblasts.

In attempts to understand the diversification of the homeobox gene superfamily, several studies have found that the Hox genes do not form a monophyletic gene clade in molecular phylogenetic analyses (Burglin, 1993). Five or six other classes of homeobox gene are as closely related to Hox genes as the latter genes are to other. These include the Evx/eve, Mox, Cdx/cad, Xlox and Gsx genes (we use Gsx to describe a gene class to which the mammalian Gsh-1 and Gsh-2 genes belong). The first two sets of genes are physically linked to Hox gene clusters in vertebrates (Evx closely, Mox distantly), suggesting they could have originated from one end of a Hox gene cluster during the tandem duplications that formed the gene cluster. The other three types of gene, however, are dispersed around the genome and their origin cannot be explained in this way. The generally accepted explanation has been that these genes were originally Hox genes, but they have 'escaped' from the cluster, by transposition mutations during evolution. Our recent work on the genomic organisation of these genes suggests this cannot be correct.

We cloned homologues of the Gsx, Xlox and Cdx homeobox genes from the cephalochordate amphioxus, finding just single copies of each gene type. A genomic walk, using both cosmid and bacteriophage libraries, revealed the surprising finding that all three genes are physically linked (Brooke, et al., 1998). The Gsx homeobox is just 25 kb from the Xlox homeobox, which in turn is just 7.5 kb from Cdx. Clearly, these genes are truly adjacent, and consequently form a novel homeobox gene cluster. We call this the ParaHox gene cluster, for reasons explained below. Clustering of Gsx, Xlox and Cdx-even in one species-disproves the notion that these genes could have escaped from a Hox gene cluster independently. They could conceivably have been transposed as a cassette of three adjacent genes, but such a model is unlikely on the basis of phylogenetic analysis of the sequences. Molecular phylogeny reveals that the Xlox homeobox genes are close relatives of the paralogy group 3 Hox genes of vertebrates (equivalent to zerknult of insects). The affinities of Gsx and Cdx are less certain, but the same phylogenetic analyses suggest they are related to Hox genes of the anterior (paralogy groups 1 and 2) and posterior (paralogy groups 9 to 13) genes respectively. Only one model is consistent with both the clustered organisation in amphioxus and the molecular phylogenetic analyses: Gsx, Xlox and Cdx must be the remnants of a homeobox gene cluster duplication. The other product of this duplication was the Hox gene cluster itself. The two gene clusters are of equal age, since they are paralogues (hence the name, ParaHox gene cluster). Figure 3 shows further details of this model. The finding that Hox and ParaHox gene clusters arose by duplication of an ancestral gene cluster (the ProtoHox gene cluster) raises two important questions for evolutionary biology. When did it happen, and what were the consequences? The first can be answered by determining the phylogenetic distribution of true Hox and ParaHox genes. Unambiguous Xlox and Cdx genes have been cloned from several higher triploblast phyla, including chordates, arthropods, annelids and nematodes; they have not been found in Cnidaria (see Brooke et al. 1998, for references). Unambiguous Hox genes have been cloned from every higher triploblast phyla in which they have been sought (see Holland and Garcia-Fernandez, 1996, and Brooke et al., 1998, for references). This list includes not only the major coelomate phyla (such as arthropods, annelids, molluscs, echinoderms, hemichordates, chordates), but also nematodes, platyhelminthes and nemerteans: triploblasts that were originally placed as acoelomates and pseudocoelomates before ribosomal gene sequencing forced a reappraisal of this concept (see earlier). In contrast, the Cnox genes of Cnidaria, although often referred to as Hox genes, may be derived from independent duplications of a ProtoHox gene or genes (see earlier). Indeed, our own analyses suggest they may be no closer phylogenetically to Hox than to ParaHox genes (N. M. Brooke and P.WH.H., unpublished). Definitive insight into the affinities of Cnox genes, however, must await analysis of their genomic organisation, further sampling from cnidarians and additional data from ParaHox genes in protostome triploblasts. The conclusion from current data is that a ProtoHox gene cluster probably duplicated on the triploblast stem lineage, close to transition 3 in Figure 1.

Did ProtoHox gene cluster duplication have any implications for the origin of either three-germ layers, bilateral symmetry, a centralised nerve cord, or a through gut? As thoroughly discussed elsewhere, Hox gene clusters play roles in spatial patterning along the main anteroposterior body axis (and some secondary axes); certainly in chordates, this role was originally confined to ectodermal derivatives, with later recruitment to mesodermal derivatives (Holland and Garcia-Fernandez, 1996) Expression of Hox genes in neurectodermal derivatives is also seen in other phyla, including arthropods, annelids and nematodes. Taken together, these lines of evidence suggest that the fundamental and original role of Hox genes was to pattern the ectoderm or neurectoderm of triploblasts. ParaHox genes have been less intensively studied, but the data so far suggest an intriguing parallel to Hox genes. The first member of the Xlox gene family to be identified was an endodermally expressed homeobox gene in Xenopus, XlHbox8 (Wright et al., 1988). This is now thought to be the orthologue of a single gene in mammals, also expressed in endoderm. This gene has been cloned by several groups and variously called Ipf-1, STF-1 or IDX-I; the International Committee on Standardized Nomenclature for Mice has now renamed this gene Pdx-i (Offield et al., 1996). Gene targeting has demonstrated that Pdx-1 is essential for specification of a particular anteroposterior region of the endoderm: that fated to become pancreas and rostral duodenum (Jonsonn et al., 1994; Offield et al., 1996). This is analogous to the role of Hox genes in specification of particular anteroposterior regions of ectoderm or mesoderm. The amphioxus orthologue, AmphiXlox, is also expressed in a band of endoderm, suggesting evolutionary conservation (Brooke et al., 1998). Xlox genes are also well studied in two species of leech (phylum Annelida). Although annelids are phylogenetically very distant from chordates, these genes are also expressed in spatially restricted regions of endoderm. In the leech Hirudo medicinalis, this expression has been elaborated further with tandem duplication of the Xlox gene; the three descendent Xlox genes (Lox3A-C) are expressed in endoderm (Wysocka-Diller et al., 1995). The similarity between coelomate protostome and chordate Xlox genes suggests evolutionary conservation. The implication is that the Xlox gene of an ancestral triploblast also had a role in patterning the central portion of a through gut. The first Cdx gene cloned was the Drosophila gene caudal. Initial expression studies suggested that the zygotic expression of this gene was distributed through posterior endoderm and into the (ectodermal) hindgut (Mlodzik and Gehring, 1987). More recent studies highlight the hindgut expression as most important (Calleja et al., 1996; G. Morata, personal communication). This tissue marks the posterior boundary of the endoderm. The Cdx gene family of vertebrates is more complex, probably comprising three genes: Cdx-1, Cdx-4 and a gene currently denoted Cdx-2 in mouse or CDX-3 in humans (Duprey et al., 1988; Gamer and Wright, 1993). The genes have rather different expression patterns, complicating inferences about the ancestral role for Cdx in chordates. One expression site shared by all vertebrate Cdx genes is the posterior endoderm, however, suggesting this was the original expression site before gene duplication in the vertebrates (Holland et al., 1992). Consistent with this, the single amphioxus homologue, AmphiCdx, is expressed in posterior gut (Brooke et al., 1998). As with the Xlox genes above, the similar expression data from both protostomes and chordates suggests evolutionary conservation. Thus, the Cdx gene of an ancestral tribloblast probably played a role in development of the posterior extremity of the gut, close to the anus. The third ParaHox gene, Gsx, has only been isolated from chordates to date. If our model of ParaHox and Hox gene cluster evolution (Fig. 3) is correct, however, it should be as ancient as Cdx or Xlox, and (unless secondarily lost) should be present in several coelomate protostomes. Vertebrates have two Gsx genes, Gsh-1 and Gsh2, compared to one gene detected in amphioxus. Both genes have complex spatiotemporal expression in the developing mouse brain (Valerius et al., 1995; Szucsik et al., 1997). The single amphioxus homologue, AmphiGsx, is also expressed in the developing brain homologue (Brooke et al., 1998). The mouse genes have been deleted in turn by gene targeting, but they result in rather different phenotypes.

Deletion of Gsh causes dwarfism that is traceable to defects in development of the adenohypophysis (anterior pituitary; Li et al., 1996); deletion of Gsh-2 causes defects in the forebrain and hindbrain (Szucsik et al., 1997). As with the Cdx genes, the presence of more than one gene complicates interpretation; in the case of Gsh genes, the rather different inferred roles makes inference of ancestral role even more difficult. This may become clearer once a double homozygote phenotype is described (Gsh-1 I , Gsh-2 -), since this could uncover redundant roles shared by the genes since their origin by gene duplication. It is intriguing, however, that the Gsh-1 phenotype primarily affects the adenohypophysis, since this develops from a dorsal diverticulum of the oral cavity: Rathke's pouch. This structure is ectodermal in mammals, but endodermal in the more basal hagfish (Gorbman, 1983); in either case, it is a region of gut at the extreme anterior, at the junction between germ layers.

If a role in formation of Rathke's pouch does reflect an ancestral role for Gsx genes, then the ParaHox genes present a rather straightforward picture. The genes are linked in the order Gsx, Xlox then Cdx. The first gene (Gsx) has roles in the anterior most extremity of the through gut (and brain) and is paralogous to the most anterior Hox genes (with roles in anterior ectoderm). The next ParaHox gene (Xlox) specifies fate of a precise anteroposterior region of endoderm, and is paralogous to group 3 Hox genes (with roles in fate specification in ectoderm). The final ParaHox gene (Cdx) has roles in the posterior extremity of the gut, and is paralogous to the posterior Hox genes. The two gene clusters, related by gene duplication, seem to have ancient patterning roles in different germ layers.

In the light of these functional considerations, the fact that the origin of Hox and ParaHox genes may date to transition 3 (see above) is very intriguing. This transition involved the invention of a through gut (the extremities of which may now specified by ParaHox genes) and the evolution of three germ layers (with Hox and ParaHox genes fulfilling patterning roles in different germ layers). We suggest that the ancestral ProtoHox gene cluster could not have fulfilled these distinct roles before duplication. This implies that duplication of the ancestral ProtoHox gene cluster, on the stem lineage of triploblasts, was a necessary prerequisite for the evolution of three germ layers, and also for the origin of a throughgut with mouth and anus. This gene duplication event was permissive for the origin of bilateral triploblasts with a through-gut. CAMBRIAN CONCLUSIONS Many of the evolutionary transitions discussed in this paper seem very different. All are defined as phylogenetic internodes that underwent major change to developmental patterning, but they had rather different consequences. For example, transitions 1, 2, 3 and 5 involve clear increases in complexity of the body plan, whilst this is not the case for transition 4 (and arguably 6). Every transition is likely to have involved a suite of genetic changes, but in most cases we are scratching the surface in deducing what they were. There does seem to be parallel, however, between the type of genetic event occurring at transition 3 and those occurring at transition 6. In each case, duplication of a homeobox gene cluster occurred; this was accompanied by elaboration of developmental roles and may have been permissive to the evolution of increased body plan complexity. In the case of early vertebrate evolution, Hox gene cluster duplication was not an isolated event, but occurred in concert with the duplication of many other genes. It remains to be seen whether the same is true of the ProtoHox gene cluster duplication. This will be an important question to resolve. Similarly, we have little insight into the ancestral role of the ProtoHox gene cluster before it duplicated. Its genes probably obeyed spatial colinearity (physical gene order matching position of deployment in the embryo), since both Hox and ParaHox genes share this property. Which tissues utilized this colinear expression is unclear

There may also be a parallel between triploblast origins and vertebrate origins from the perspective of evolutionary radiations. Vertebrates are the most speciose subphylum of chordates, by a very large margin (approximately 45,000 vertebrates, 25 cephalochordates, 3,000 urochordates). There is no evidence from the fossil record to suggest that cephalochordates or urochordates ever achieved great diversity. It seems, therefore, that chordates were not particular successful (judged by diversity) until the advent of the vertebrate body plan. Perhaps the genetic changes close to vertebrate origins not only enabled major developmental change, but also in turn enabled successful adaptive radiation of the taxon.

Triploblast origins may show this phenomenon even more clearly, depending on how one interprets the fossil record. The Cambrian explosion-the rapid appearance of a diversity of new body plans in the Cambrian-is well known, but subject to a variety of interpretations. Some have argued that the Cambrian explosion reflects a phase of increased fossil preservation rather than cladogenesis (Wray et al., 1996), but this conclusion has been challenged by Ayala et al. (1998). The consensus view at present seems to be that the Cambrian explosion was a real phenomenon, reflecting a time of rapid diversification of body plans. Controversy then focuses on the types of animals encompassed by the radiation, and those excluded. The conclusion reached by Conway Morris (1994), from palaeontological evidence, is that the Cambrian explosion affected animals of the coelomate triploblast grade of organisation. The same conclusion was reached by Philippe et al. (1994) from consideration of molecular phylogenies. Both lines of evidence suggest that diploblastic animals existed before the explosion, and did not radiate simultaneously. Rather, diploblast diversity may have dramatically decreased during the Cambrian, particularly if the enigmatic Ediacaran fossils are indeed related to the extant diploblasts (Conway Morris, 1994; Fig. 4).

It seems that a very rapid, explosive increase in diversity followed the invention of the triploblast body plan. As with the vertebrate example, it is possible that genetic changes occurring on the triploblast stem lineage (transition 3) not only enabled major developmental change, but also successful adaptive radiation of the taxon. The origin of distinct Hox and ParaHox gene clusters may have paved the way for the evolution of a through gut (enabling burrowing and feeding to be combined), three distinct germ layers (facilitating locomotion), and ultimately the Cambrian explosion (Fig. 4). ACKNOWLEDGMENTS I thank present and past members of my laboratory, and collaborators, for generating the data discussed here and their contributions to its interpretation. In particular, I thank Jordi Garcia-Fernandez, Nina Brooke, Mari Kobayashi, Graham Luke, Simon Patton, Anna Sharman, Seb Shimeld, Hiroshi Wada and Nic Williams.

' From the symposium Evolutionary Relationships of Metazoan Phyla: Advances, Problems, and Approaches presented at the Annual Meeting of the Society for Integrative and Comparative Biology, 3-7 January 1998, at Boston, Massachusetts. 2 E-mail: p.w.h.holland@rdg.ac.uk

REFERENCES

Aguinaldo, A. M., J. M. Turbeville, L. S. Linford, M. C. Rivera, J. R. Garey, R. A. Raff, and J. A. Lake. 1997. Evidence for a clade of nematodes, arthropods and other moulting animals. Nature 387: 489-493.

Ayala, E J., A. Rzhetsky, and E J. Ayala. 1998. Origin of the metazoan phyla: Molecular clocks confirm paleontological estimates. Procl. Nat. Acad. Sci. U.S.A. 95:606611.

Balavoine, G. 1997. Early emergence of platyhelminths contradicted by rRNA and Hox gene data. C.R. Acad. Sci. Paris 320:83-94.

Barnes, R. D. 1980. Invertebrate zoology, 4th ed. HoltSaunders, Japan.

Brooke, N. M., J. Garcia-Fernandez, and P W. H. Holland. 1998. The ParaHox gene cluster is an evolutionary sister of the Hox gene cluster. Nature. 392:920-922.

Brusca, R. C. and G. J. Brusca. 1990. Invertebrates.

Sinauer Associates, Sunderland, Massachusetts. Burglin, T. R. 1993. A comprehensive classification of homeobox genes. In Guidebook to the homeobox genes D Duboule, (ed.) Oxford University Press, Oxford.

Calleja, M., E. Moreno, S. Pelaz, and G. Morata. 1996. Visualization of gene expression in living adult Drosophila. Science 274:252-255. Conway Morris, S. 1994. Why molecular biology needs palaeontology. Devel. Suppl., 1-13. DeRobertis, E. M. and Y. Sasai. 1996. A common plan for dorsoventral patterning in bilateria. Nature 380:37-40.

Duprey, P., K. Chowdhury, G. R. Dressler, R. Balling, D. Simon, J. Guenet, and P Gruss. 1988. A mouse gene homologous to the Drosophila gene caudal is expressed in epithelial cells from the embryonic intestine. Genes Dev. 2:1647-1654. Erwin, D. H. 1993. The origin of metazoan development: A palaeobiological perspective. Biol. J. Linn. Soc. 50:255-274.

Field, K. G., G. J. Olsen, D. J. Lane, S. J. Giovannoni, M. T. Ghiselin, E. C. Raff, N. R. Pace, and R. A. Raff. 1988. Molecular phylogeny of the animal kingdom. Science 239:748-753.

Forey, P and P. Janvier. 1993. Agnathans and the ongin of jawed vertebrates. Nature 361:129-134. Gamer, L. M. and C. V. E. Wright. 1993. Murine Cdx4 bears striking similarities to the Drosophila caudal gene in its homeodomain sequence and early expression pattern. Mech. Dev. 43:71-81. Garcia-Fernandez, J. and P W. H. Holland. 1994. Archetypal organization of the amphioxus Hox gene cluster. Nature 370:563-566.

Gorbman, A. 1983. Early Development of the hagfish pituitary gland: Evidence for the endodermal origin of the adenohypophysis. Amer. Zool. 23:639654.

Gould, S. J. 1989. Wonderful life. Hutchinson Radius. Halanych, K. M., J. D. Bacheller, A. M. A. Aguinaldo, S. M. Liva, D. M. Hillis, and J. A. Lake. 1995. Evidence from 18S ribosomal DNA that lophophorates are protostome animals. Science 267: 1641-1643.

Holland, P. W. H. 1996. Molecular biology of lancelets: Insights into development and evolution. Israel J. Zool. 42:S247-S272.

Holland, P. W. H. and J. Garcia-Fernandez. 1996. Hox genes and chordate evolution. Develop. Biol. 173: 382-395.

Holland, P W. H., P W. Ingham, and S. Krauss. 1992. Development and evolution: Mice and flies head to head. Nature 358:627-628.

Holland, P W. H., J. Garcia-Fernandez, N. A. Williams, and A. Sidow. 1994. Gene duplications and the origins of vertebrate development. Development Suppl., 125-133.

Jonsonn, J., L. Carlsson, T. Edlund, and H. Edlund. 1994. IPF-1 is required for pancreas formation in mice. Nature 371:606-609.

Kuhn, K., B. Streit, and B. Schierwater. 1996. Homeobox genes in the cnidarian Eleutheria dichotoma: Evolutionary implications for the origin of the Antennapedia-class (HOM/Hox) genes. Mol. Phylo. Evol. 6:30-38.

Lacalli, T. 1996. Dorsoventral inversion: A phylogenetic perspective. BioEssays 18:251-254. Lanfranchi, G., A. Pallavicini, P. Laveder, and G. Vale. 1994. Ancestral hemoglobin switching in lampreys. Devel. Biol. 164:402-408. Li, H., P S. Zeitler, M. T Valerius, K. Small, and S. S. Potter. 1996. Gsh-rs an orphan Hox gene, is required for normal pituitary development. EMBO J. 15:714-724.

Lundin, L. 1993. Evolution of the vertebrate genome as reflected in paralogous chromosomal regions in man and the house mouse. Genomics 16:1-19. Maynard-Smith, J. and E. Szathmary. 1995. The major transitions in evolution. W. H. Freeman, Oxford. Miller, D. J. and A. Miles. 1993. Homeobox genes and

the zootype. Nature 365:215-216. Mlodzik, M. and W. J. Gehring. 1987. Expression of the caudal gene in the germ line of Drosophila: Formation of an RNA and protein gradient during early embryogenesis. Cell 48:465-478. Naito, M., H. Ishiguro, T. Fujisawa, and Y. Kurosawa. 1993. Presence of eight distinct homeobox-containing genes in cnidarians. FEBS Lett. 333:271274.

Offield, M. E, T. L. Jetton, P Labosky, M. Ray, R. Stein, M. Magnuson, B. L. M. Hogan, and C. V. E. Wright. 1996. Pdx-l is required for the development and differentiation of the pancreas and rostral duodenum. Dev. Biol. 122:983-995. Ohno, S. 1970. Evolution by gene duplication. Springer-Verlag, Heidelberg.

Pancer, Z., M. Kruse, I. Muller, and W. E. G. Muller. 1997. On the origin of metazoan adhesion receptors: Cloning of integrin alpha subunit from the sponge Geodia cydonium. Mol. Biol. Evol. 14: 391-398.

Patton, S. J., G. N. Luke, and P W. H. Holland. 1998. Complex history of chromosomal paralogy regions: Insights from amphioxus amino acid hydroxylase and insulin gene families. Mol. Biol. Evol. (In press).

Pendleton, J. W., B. K. Nagai, M. T. Murtha, and E H. Ruddle. 1993. Expansion of the Hox gene family and the evolution of chordates. Proc. Natl. Acad. Sci. U.S.A. 90:6300-6304.

Philippe, H., A. Chenuil, and A. Adoutte. 1994. Can the Cambrian explosion be inferred through molecular phylogeny? Development Suppl., pp. IS25.

Schummer, M., I. Scheurlen, C. Schaller, and B. Galhot. 1992. HOM/Hox homeobox genes are present in hydra (Chlorohydra viridissima) and are differentially expressed during regeneration. EMBO J. 11:1815-1823.

Seimeya, M., K. Ishiguro, K. Miura, and Y. Watanabe. 1994. Homeobox-containing genes in the most primitive metazoa, the sponges. Eur. J. Biochem. 221:219-225.

Sharman, A. C. and P. W. H. Holland. 1996. Conservation, duplication and divergence of developmental genes during chordate evolution. Neth. J. Zool. 46:47-67.

Sharman, A. C. and P W. H. Holland. 1998. Estimation of Hox gene cluster number in lampreys. Int. J. Devel. Biol. 42:617-620.

Sharman, A. C., A. Hay-Schmidt, and P W. H. Holland. 1997. Cloning and analysis of an HMG gene from the lamprey, Lampetra fluviatilis: gene duplication in vertebrate evolution. Gene 184:99105.

Shimeld, S. M. 1997. Characterisation of amphioxus HNF-3 genes: Conserved expression in the notochord and floor plate. Developmental Biology 183:74-85.

Sidow, A. 1996. Gen(om)e duplications in the evolution of early vertebrates. Curr. Opinion Gen. Devel. 6:715-722.

Slack, J. M. W., P W. H. Holland, and C. E Graham. 1993. The zootype and the phylotypic stage. Nature 361:490-492.

Stock, D. W. and G. S. Whitt. 1992. Evidence from 18S ribosomal RNA sequences that lampreys and hagfish form a natural group. Science 257:787789.

Szucsik, J. C., D. P Witte, H. Li, S. K. Pixley, K. M. Small, and S. S. Potter. 1997. Altered forebrain and hindbrain development in mice mutant for the Gsh-2 homeobox gene. Devel. Biol. 191:230-242. Turbeville, J. M., K. G. Field, and R. A. Raff. 1992. Phylogenetic position of Phylum Nemertini inferred from 18S ribosomal RNA sequences-molecular data as a test of morphological character homology. Mol. Biol. Evolution 9:235-249. Valerius, M. T, H. Li, J. L. Stock, M. Weinstein, S. Kaur, G. Singh, and S. S. Potter. 1995. Gsh-l: A novel murine homeobox gene expressed in the central nervous system. Devel. Dynamics 203: 337-351.

Wada, H. and N. Satoh. 1994. Details of the evolutionary history from invertebrates, as deduced from the sequences of 18S rDNA. Proc. Natl. Acad. Sci. U.S.A. 91:1801-1804. Wada, H., H. Saiga, N. Satoh, and P. W. H. Holland. 1998. Tripartite organisation of the ancestral chordate brain and the antiquity of placodes: Insights from ascidian Pax-2/5/8, Hox and Otx genes. Development 125:1113-1122.

Wada, H., P W. H. Holland, and N. Satoh. 1996. Origin of patterning in neural tubes. Nature 384:123. Wada, H., P W. H. Holland, S. Sato, H. Yamamoto, and N. Satoh. 1997. Dorsal tube is partially dorsalised by overexpression of HrPax-3/7: the ascidian homologue of Pax-3 and Pax-7. Devel. Biol. 187:240-252.

Williams, N. A., and P W. H. Holland. (1998) Gene and domain duplication in the chordate Otx gene family: Insights from amphioxus Otx. Mol. Biol. Evol. 15:600-607.

Wray, G. A., J. S. Levinton, and L. H. Shapiro. 1996. Molecular evidence for deep Precambrian divergences among metazoan phyla. Science 274:568573.

Wright, C. V. E., P. Schnegelsberg, and E. M. DeRobertis. 1988. XIHbox 8-a novel Xenopus homeo protein restricted to a narrow-band of endoderm. Development 104:787-794. Wysocka-Diller J., G. O. Aisemberg, and E. R. Macagmo. 1995. A novel homeobox cluster expressed in repeated structures of the midgut. Dev. Biol. 171:439-447.

Corresponding Editor: Douglas H. Erwin

Copyright Society for Integrative and Comparative Biology Dec 1998
Provided by ProQuest Information and Learning Company. All rights Reserved