ABSTRACT
Comparisons between Hox genes in different arthropods suggest that the diversity of Antennapedia-class homeotic genes present in modern insects had already arisen before the divergence of insects and crustaceans, probably during the Cambrian. Hox gene duplications are therefore unlikely to have occurred concomitantly with trunk segment diversification in the lineage leading to insects. Available data suggest that domains of homeotic gene expression are also generally conserved among insects, but changes in Hox gene regulation may have played a significant role in segment diversification. Differences that have been documented alter specific aspects of Hox gene regulation within segments and correlate with alterations in segment morphology rather than overt homeotic transformations.
The Drosophila Hox cluster contains several homeobox genes that are not homeotic genes -bicoid, fushi-tarazu and zen. The role of these genes during early development has been studied in some detail. It appears to be without parallel among the vertebrate Hox genes. No well conserved homologues of these genes have been found in other taxa, suggesting that they are evolving faster than the homeotic genes. Relatively divergent Antp-class genes isolated from other insects are probably homologues of fushi-tarazu, but these are almost unrecognisable outside of their homeodomains, and have accumulated approximately 10 times as many changes in their homeodomains as have homeotic genes in the same comparisons. They show conserved patterns of expression in the nervous system, but not during early development.
INTRODUCTION
In Drosophila, the differences between the trunk segments are controlled by the homeotic genes of the Antennapedia and Bithorax gene complexes (Lewis, 1978; Sánchez-Herrero et al., 1985; Kaufman et al., 1990). This paper is concerned with the extent to which changes affecting these Hox genes may have contributed to the evolution of body form and developmental mechanism in the arthropods, and in particular the insects. We begin by considering the timing of gene duplications that gave rise to the Insect Hox cluster*, in relation to the origin of invertebrate phyla and arthropod classes. Then we consider how changes in the regulation of Hox genes may have contributed to the diversification of the insects. Finally, we present evidence that the selective constraints on some genes in the Hox clusters appear to have been relaxed in some insect lineages, allowing them to diverge relatively rapidly, both in terms of sequence and role in development.
HOX GENE DUPLICATIONS AND THE ORIGIN OF ARTHROPOD BODY PLANS
An explicit hypothesis relating Hox genes to evolutionary change was articulated by E. B. Lewis in the opening paragraph of his classic paper (Lewis, 1978):
“During the evolution of the fly, two major groups of genes must have evolved: Teg suppressing’ genes which removed legs from abdominal segments of millipede like ancestors, followed by ‘haltere-promoting’ genes which suppressed the second pair of wings of four-winged ancestors…
During evolution a tandem array of redundant genes presumably diversified by mutation to produce this complex.” (the Bithorax Complex).
In essence, Lewis proposed that the evolutionary history of the homeotic gene clusters would be related to the evolution of morphology in the arthropod lineage, with increasing complexity of the gene cluster underlying increasing complexity of the body plan. No doubt he had in mind a scheme proposed by Snodgrass (1935), where the archetypal insect is derived from a superficially myriapod-like ancestor in which all the segments between the mouthparts and the anal segment were similar. In the parlance of arthropod morphologists, such an animal is homonomous; all the segments of the trunk form a single tagma, or body region. This contrasts with the regional specialisation of segment form and function that constitutes tagmosis, and characterises most arthropod classes.
Snodgrass’ scheme assumes that in the proposed insect/myriapod lineage, a homonomous trunk is the primitive state, and tagmosis a derived character. At first sight this conflicts with the observation that most representatives of the other major group of mandibulate arthropods -the Crustacea -also display clear trunk tagmosis, and typically have distinct thorax and abdomen (Schram, 1986). However, conventional phylogenies argue that these analogous states have been acquired independently in the two lineages (Brusca and Brusca, 1990): homonomous Crustacea exist (the remipedes), and are assumed to reflect the condition ancestral to this whole class (Fig. 1).
‘Conventional’ and ‘alternative’ hypotheses concerning the origin of trunk tagmosis in relation to arthropod phylogeny. (A) Conventional view depicting the common ancestor of insects and crustaceans with a homonomous trunk. Thorax/abdomen tagmosis is assumed to have arisen independently in the two lineages. Myriapods, assumed to constitute a sister group to the insects, and remipedes within the Crustacea, retain a homonomous trunk reflecting the primitive condition for this whole lineage (B) An alternative scenario in which insects and Crustacea derive from a common ancestor that already displays well differentiated trunk segments (perhaps controlled by the differential expression of multiple Antp-class Hox genes, which we infer to have been present in this ancestor). Those myriapods (some or all) that constitute a sister group to the insects, and remipedes among the Crustacea, would then show a secondarily simplified trunk. Note that the monophyly of the myriapods is uncertain, and the position of some myriapods within this lineage has recently been challenged: Sequence data (Turbeville et al., 1991; Ballard et al., 1992; Averof, 1994) and patterns of neural development (Whitington et al., 1991) suggest that many myriapods may form an outgroup to the whole lineage depicted here.
‘Conventional’ and ‘alternative’ hypotheses concerning the origin of trunk tagmosis in relation to arthropod phylogeny. (A) Conventional view depicting the common ancestor of insects and crustaceans with a homonomous trunk. Thorax/abdomen tagmosis is assumed to have arisen independently in the two lineages. Myriapods, assumed to constitute a sister group to the insects, and remipedes within the Crustacea, retain a homonomous trunk reflecting the primitive condition for this whole lineage (B) An alternative scenario in which insects and Crustacea derive from a common ancestor that already displays well differentiated trunk segments (perhaps controlled by the differential expression of multiple Antp-class Hox genes, which we infer to have been present in this ancestor). Those myriapods (some or all) that constitute a sister group to the insects, and remipedes among the Crustacea, would then show a secondarily simplified trunk. Note that the monophyly of the myriapods is uncertain, and the position of some myriapods within this lineage has recently been challenged: Sequence data (Turbeville et al., 1991; Ballard et al., 1992; Averof, 1994) and patterns of neural development (Whitington et al., 1991) suggest that many myriapods may form an outgroup to the whole lineage depicted here.
It is not obvious why a homonomous, myriapod-like animal would need the array of homeotic genes that specify trunk segment diversity in Drosophila, most particularly the set of distinct Antp-class Hox genes that specify the thoracic and abdominal segments to be different from one another (Anten-napedia (Antp), Ultrabithorax (Ubx) and abdominal-A (abdA)). Indeed, we previously argued that the duplication leading to these particular genes may have occurred relatively late, in association with the origin of the thorax/abdomen distinction in insects (Akam et al., 1988).
One test of this model is to compare the structure and expression of Antp-class Hox genes in crustaceans and insects. If the mechanism to specify thorax and abdomen arose independently in these two lineages, we should see signs of this when we examine the molecular machinery. The first step in this comparison has been achieved (Averof and Akam, 1993). In a study of Hox gene sequences in the branchiopod crustacean Artemia, we have shown that specific homologues of Antp, Ubx and abd-A, as well as Sex Combs Reduced (Ser) and Deformed (Dfd) are present in Crustacea. Other, more divergent Hox gene classes (labial, proboscipedia, AbdominaL B) are known to be of even more ancient origin (Fig. 2; Schubert et al., 1993). We conclude that the existing diversity of homeotic genes in insects (or at least Drosophila) derives from gene duplications that occurred prior to the insect/crustacean split.
The minimal inferred complexity of Hox gene clusters in the lineage leading to the insects. Panels on the right summarise the diversity of Hox genes described from insects, Crustacea, annelids (Class Hirudinea. leeches) and chordales (Amphioxus, and several vertebrates; a single ‘complete’ chordate cluster is illustrated, based on Amphioxus data; Garcia-Fernandez and Holland, 1994). Genes characterised by at least the full homeobox sequence are shown as coloured boxes. Open boxes in the annelid and crustacean panels are used for genes that are assumed to exist in these laxa, but have no! yet been characterised adequately. Identical colours are used for genes that can be assigned unique homologues (orthologues) within the clusters of different taxa. Similar but non-identical colours are used for genes where orthology relationships are unclear. Boxes are joined only where linkage has been established in some member of the taxon. The assumed phylogenetic relationships of these four taxa are not disputed. Panels on the left show the minimal complexity of the Hox cluster in each of the stem groups that is implied by considering this phylogeny together with the assumed orthology relationships. A single Antp-class ancestor is shown for the basal stem group (though Strand the genes of vertebrate paralogy group 5 may derive from a second ancestral Antp-class sequence present al the base of this lineage. The origin of these genes is not clear; Bürglin. 1994). Distinct Amp-like and Ubx/abd-A-Vtke genes are shared by arthropods and annelids. All homeotic gene classes are shared by insects and Crustacea. The inclusion of a group 3 homologue in the arthropod lineages is based on unpublished data (see text), and on the isolation of a homeobox PCR fragment that has residues diagnostic for class 3 Hox genes from Limulus, a chelicerate arthropod (Cartwright et al., 1993). Gene symbols have been abbreviated as follows; Lx, Lox (leech Hox) genes; Ant, Amp; abd, Ubx/abd-A related; ab-A, abd-A; Ab-B. Abd-B.(References:Chordates; Bürglin. 1994; Garcia-Fernandez and Holland, 1994. Annelids; Loxl, Aisemberg and Macagno, 1994); Lox2,Wysocka-Diller et al., 1989; Nardelli-Haefliger and Shankland, 1992; Lox 5,6,7, Shank land et al., 1991. Crustaceans: Averof and Akam, 1993; see also legend to Fig. 4. Insects: For a summary of Drosophila gene sequences, see Bürglin. 1994; for other insects, see legend to Fig. 4.)
The minimal inferred complexity of Hox gene clusters in the lineage leading to the insects. Panels on the right summarise the diversity of Hox genes described from insects, Crustacea, annelids (Class Hirudinea. leeches) and chordales (Amphioxus, and several vertebrates; a single ‘complete’ chordate cluster is illustrated, based on Amphioxus data; Garcia-Fernandez and Holland, 1994). Genes characterised by at least the full homeobox sequence are shown as coloured boxes. Open boxes in the annelid and crustacean panels are used for genes that are assumed to exist in these laxa, but have no! yet been characterised adequately. Identical colours are used for genes that can be assigned unique homologues (orthologues) within the clusters of different taxa. Similar but non-identical colours are used for genes where orthology relationships are unclear. Boxes are joined only where linkage has been established in some member of the taxon. The assumed phylogenetic relationships of these four taxa are not disputed. Panels on the left show the minimal complexity of the Hox cluster in each of the stem groups that is implied by considering this phylogeny together with the assumed orthology relationships. A single Antp-class ancestor is shown for the basal stem group (though Strand the genes of vertebrate paralogy group 5 may derive from a second ancestral Antp-class sequence present al the base of this lineage. The origin of these genes is not clear; Bürglin. 1994). Distinct Amp-like and Ubx/abd-A-Vtke genes are shared by arthropods and annelids. All homeotic gene classes are shared by insects and Crustacea. The inclusion of a group 3 homologue in the arthropod lineages is based on unpublished data (see text), and on the isolation of a homeobox PCR fragment that has residues diagnostic for class 3 Hox genes from Limulus, a chelicerate arthropod (Cartwright et al., 1993). Gene symbols have been abbreviated as follows; Lx, Lox (leech Hox) genes; Ant, Amp; abd, Ubx/abd-A related; ab-A, abd-A; Ab-B. Abd-B.(References:Chordates; Bürglin. 1994; Garcia-Fernandez and Holland, 1994. Annelids; Loxl, Aisemberg and Macagno, 1994); Lox2,Wysocka-Diller et al., 1989; Nardelli-Haefliger and Shankland, 1992; Lox 5,6,7, Shank land et al., 1991. Crustaceans: Averof and Akam, 1993; see also legend to Fig. 4. Insects: For a summary of Drosophila gene sequences, see Bürglin. 1994; for other insects, see legend to Fig. 4.)
Arthropods recognisable as crustaceans are well documented among Cambrian fossils (Robison and Kaesler, 1987), including probable Branchiopods from the lower Cambrian, 550 Myr before present (Butterfield, 1994). In contrast, the earliest insect fossils appear significantly later, during the Devonian, 400 Myr before present (Jeram et al., 1990). These insects are likely to derive from an early crustacean-mandibulate clade, in which case the final diversification of the Antp-class Hox genes may have occurred within this arthropodan lineage. However, at least one of the relevant gene duplication events occurred earlier, before the separation of annelid and arthropod lineages: leeches possess a Ubx/abd-A type gene (Lox2), distinct from other Antp class sequences (Lox I, Lox5;Wysocka-Diller et al., 1989; Nardelli-Haefliger and Shankland, 1992).
These findings argue against the strong version of the Lewis hypothesis -that the acquisition of trunk tagmosis in insects was directly correlated with the origin of new Hox genes. However, they do not rule out the possibility that significant changes in the regulation of pre-existing Hox genes may have played a crucial part in this transition. Further work will be required to establish if and how the Antp class Hox genes are used to define trunk segment identities in Crustacea, and in homonomous arthropods (e.g. myriapods).
Even before these data are available, it is perhaps worth questioning the validity of one assumption on which the Lewis model is based, for the argument illuminates one contribution that developmental data may make to evolutionary discussions. This is the assumption that homonomy is always primitive. As developmental geneticists, we would emphasise that there is an enormous asymmetry between the invention of a complex feature (e.g. the transition from homonomy to heteronomy) and its loss (heteronomy→homonomy). An animal that has no developmental mechanism to make segments different must invent both the mechanism to specify the differences, and also the machinery to elaborate distinct, adaptive segment types. Loss of heteronomy can be achieved by what seem to us to be much simpler processes -by turning off the mechanism that specifies segments to be different, or simply by eliminating some tagma from the adult body plan altogether.
We do not seriously doubt the hypothesis that the first metamerically segmented ancestors of the arthropods were homonomous -though there is at present no way of knowing where to place that ancestor on the phylogenetic tree. We do question whether homonomy is generally a primitive trait among extant arthropods. In an environment where it is useful for all segments to have legs, this adaptive change may be relatively easy to achieve. A single Bx-C mutation will make a ‘myriapod’ of a fly (albeit not a viable one; Lewis, 1978), but we cannot conceive of the reverse transition occurring in a single step.
An analogous situation pertains to the origin and loss of insect wings. It takes many genes to make a wing (Williams and Carroll, 1993) yet, in Drosophila, we know of several single gene mutations that will suppress wing formation in an otherwise viable fly (wingless, apterous, vestigial;Lindsley and Zimm, 1992). Among the insects, where the phylogenetic context is much less ambiguous, wings are believed to have evolved only once; we do not hesitate to invoke convergence repeatedly to explain the origin of wingless taxa at many phylogenetic levels among the pterygotes, from whole orders (e.g. fleas, lice) to individual species (Kristensen, 1991).
If heteronomy is hard to achieve de novo, but useful and easy to modify once invented, we should expect to see two very different modes of evolution: clades that generate a diversity of taxa which retain essentially homonomous body plans, and clades that exploit heteronomy to fill many niches. Trilobites may exemplify the first mode; despite their abundance and taxonomic diversity throughout the Palaeozoic, virtually all trilobite families retain largely homonomous trunk segments (Eldridge, 1977). The Crustacea present a very different picture, with a Palaeozoic radiation giving rise to a huge diversity of body plans (Cisne, 1974; Schram, 1986; Briggs et al., 1992).
HOMEOTIC GENE REGULATION AND EVOLUTIONARY CHANGE WITHIN INSECTS
Hox gene duplications may have been a necessary precondition for segment diversification within the insects, but the results summarised above suggest that they were not an immediate driving force for its appearance. There is some evidence, however, that changes in the regulation of the Hox genes have occurred within insects, and may have contributed to morphological evolution. These changes are to be seen against a broadly conserved role for many Hox genes, at least in so far as can be inferred from the phenotypes of mutations in the Hox clusters of beetles and moths (Beeman et al., 1993; Tazima, 1964; Booker and Truman, 1989; Ueno et al., 1992), and from the patterns of expression of Hox genes in these and other insects (see below).
Most data are available for abd-A. In each of four species where the expression of abd-A has been examined (Drosophila, Karch et al., 1990; Macias et al., 1990; Manduca, Nagy et al., 1991; Tribolium, Stuart et al., 1993; Schistocerca, Tear et al., 1990), the gene is expressed throughout much of the abdomen, with no evidence for expression in the head or thorax. Moreover, in each case, the gene shows an abrupt anterior boundary of expression within the first abdominal segment. Where tested, this lies at the parasegmental boundary defined by engrailed expression, suggesting that, at least in this case, the precise co-ordination of segmentation and homeotic gene expression is maintained in insects as diverse as Orthoptera and Diptera (Tear et al., 1990).
In some other cases, the domains of homeotic gene expression are broadly conserved, but the precise limits of expression are not the same as those in Drosophila. For example, the Abd-B genes in Tribolium and Schistocerca appear to be expressed in the epidermis of only the most posterior abdominal segments (from A8p back; Kelsh et al., 1993; J. He & R. Denell, personal communication), whereas in Drosophila Abd-B is also expressed in the more anterior abdominal segments, A5p-A8a. This anterior expression of the gene, which is necessary for normal development, is regulated quite independently of that in the more posterior abdomen: It derives from a separate promoter and is activated through different regulatory elements (Boulet et al., 1991). Altered regulation of the Abd-B gene may have played a role in the evolution of the modified abdomen that is characteristic of the higher Diptera.
Comparisons between Schistocerca and Drosophila reveal one clear case of ‘molecular heterochrony’. In Drosophila, abd-A is normally never expressed in the ninth or more posterior abdominal segments (parasegment 14). All of these segments fuse with A8 to form the ‘terminalia’. In Schistocerca, abd-A is expressed at high levels in A9 and A10 during early embryogenesis, but this expression is rapidly lost, at the same time that Abd-B expression extends anteriorly (Tear et al., 1990; Kelsh et al., 1994). It seems most likely that this is due to the direct repression of the abd-A gene by Abd-B protein, for this same interaction has been demonstrated in Drosophila. In Abd-B mutants of Drosophila, abd-A expression extends posteriorly into A9 (Macias et al., 1990), and this segment now develops characteristics of the more anterior abdomen, including a denticle belt (Sánchez-Herrero et al., 1985). Thus a regulatory interaction that occurs as a temporal sequence after segment formation in Schistocerca, appears to be ‘hard wired’ at an earlier stage in Drosophila development, and is only apparent in mutants.
The two cases above suggest that alterations in the timing or extent of homeotic gene expression may be used to modify segment development -even though they do not result in striking ‘homeotic’ transformations. However, there are a number of observations in the entomological literature suggestive of homeotic transformations that have been fixed in the course of evolution. One that we are unlikely to be able to study in detail concerns the Monura -an extinct group of insects that flourished in the Carboniferous (Kukalova-Peck, 1991). These insects, in common with a number of other early groups, had reduced but well-formed legs (with terminal claws) on each abdominal segment Al to A10: thorax-abdomen tagmosis was not complete. However, in contrast with all other known insects, in the Monura, these legs were also present on the terminal abdominal segment, in place of the filiform cerci that are the usual appendages of All. The presumed sister group (Paleodictyoptera) and outgroups (Thysanura, Symphyla) to the Monura had cerci, so it is possible that a homeotic transformation of cerci to leglets occurred in this particular order. Alternatively, of course, this ‘simple’ character state may represent a retained primitive feature that has been lost several times in other groups.
The case of the Strepsipteran haltere is more amenable to experimental analysis. The Strepsiptera are a small order of parasitic insects. The adult females are neotenic, retaining larval morphology. It is the males that are interesting for our purposes (Whiting and Wheeler, 1994). They have only a single pair of wings, but these are developed from the third thoracic segment, not the second, as in the Diptera. The dorsal appendages of the second thoracic segment are reduced to halteres, which in many details resemble the halteres derived from the third thoracic segment of Diptera. Until recently, this resemblance has been ascribed to convergence, for the Strepsiptera have been assumed to be only distantly related to the Diptera. Ribosomal DNA sequence data now provides strong evidence that the Strepsiptera are in fact the sister group to the Diptera. This prompted Whiting and Wheeler to suggest that the Strepsipteran haltere is indeed homologous to the haltere of Diptera, and that this developmental pathway has been adopted in T2 of the Strepsiptera, presumably by a change in the segmental regulation of Ubx. This hypothesis is open to test, using a recently described antibody that cross-reacts with Ubx protein in a wide range of insect species (Kelsh et al., 1994).
However, we should be cautious in making such predictions. This dramatic modification of adult segment morphology may depend entirely on the regulation of genes acting downstream of the homeotics. Even if it does involve modification of homeotic gene expression, it may involve only their regulation in a subset of tissues at a particular stage in development (e.g. in the developing imaginai discs). It is not necessary to assume that evolutionary change affecting the homeotic genes must involve major, saltatory change in the morphology of whole segments. It may much more frequently proceed by incremental changes affecting levels of expression, or by the alteration of those enhancer elements within their promoters that control activity in particular structures within a segment.
This view is supported by careful studies of the role of one homeotic gene, Ubx, in Drosophila (J. Castelli-Gair and M. Akam, unpublished data). The Ubx gene is responsible for defining the identities of two quite different (para)segments, parasegment 5 (T3) and parasegment 6 (Al). This differential function does not appear to depend on the segment-specific expression of different protein variants (Busturia et al., 1990). Rather it seems to involve the precise temporal and spatial regulation of the gene in parasegment 5 (Fig. 3). Examples of how changes in the precise pattern of within-segment regulation may have contributed to insect segment diversity are provided by the work of Carroll et al. (this volume). They show, for example, that the loss of expression of abd-A specifically within some limb buds may allow the development of larval abdominal prolegs in Lepidoptera.
Segment specification involves precise spatial regulation of homeotic genes in Drosophila. A region of the Drosophila embryo at extended germ band stage (maxillary to second abdominal segments) has been stained for protein produced by the Ubx gene (black). At this stage, homeotic genes are regulated in parasegmental domains. t/Zw plays no role in segment specification in parasegment 4 (PS4; posterior first thoracic segment/anterior second). Near ubiquitous expression of Ubx characterises parasegment 6 (PS6, T3p/Ala), and can be shown to specify PS6 by experimental manipulation (Gonzalez-Reyes et al., 1990; Mann and Hogness, 1990). Spatially regulated expression of Ubx characterises parasegmenl 5 (PS5, T2p/T3A), and is necessary for its correct specification. In this preparation, counter staining for the product of a distal-less reporter gene (Vachon et al., 1992) marks the developing leg discs of the thoracic segments (brown). Scale, 10 μm.
Segment specification involves precise spatial regulation of homeotic genes in Drosophila. A region of the Drosophila embryo at extended germ band stage (maxillary to second abdominal segments) has been stained for protein produced by the Ubx gene (black). At this stage, homeotic genes are regulated in parasegmental domains. t/Zw plays no role in segment specification in parasegment 4 (PS4; posterior first thoracic segment/anterior second). Near ubiquitous expression of Ubx characterises parasegment 6 (PS6, T3p/Ala), and can be shown to specify PS6 by experimental manipulation (Gonzalez-Reyes et al., 1990; Mann and Hogness, 1990). Spatially regulated expression of Ubx characterises parasegmenl 5 (PS5, T2p/T3A), and is necessary for its correct specification. In this preparation, counter staining for the product of a distal-less reporter gene (Vachon et al., 1992) marks the developing leg discs of the thoracic segments (brown). Scale, 10 μm.
HOX GENES THAT GOT AWAY
Up to this point, we have taken for granted that Hox genes cloned from any insect can be identified without question as the specific homologue (orthologue) of one of the Hox cluster genes defined in Drosophila. In general, this is the case. With few exceptions, the homeodomains are almost identical, and gene-specific sequences around the homeodomain and extending upstream to the hexapeptide motif (Bürglin, 1994) are sufficiently similar to make identification unambiguous.
This is true, however, only for homologues of the ‘homeotic’ genes within the Drosophila clusters (labial, proboscipedia, Dfd, Scr, Antp, Ubx, abd-A and Abd-Bf Well-conserved homologues have not been isolated for the several homeobox genes of the Antp complex that have ‘anomalous’ roles in development. There are four such genes in Drosophila melanogaster -bicoid, fushi-taruzu and the two zen genes.
Bicoid is only expressed maternally. It encodes a gradient morphogen in the early (syncytial) embryo (Driever and Nüsslein-Volhard, 1988). Fushi-taruzu (ftz) is a ‘pair-rule’ segmentation gene (Kaufman et al., 1990). It is re-used during neurogenesis to specify the identity of certain neurons in each trunk segment (Doe et al., 1988). Zenl is required very early for the specification of the dorsal pattern elements in the embryo (Ferguson and Anderson, 1991). None of these roles is analogous to that of the typical homeotic genes, or to those of the Hox genes in vertebrates, so far as we know. The existence of these genes within the insect Hox clusters has been something of an enigma, but isolation of several ‘divergent’ Hox genes from other insects throws some light on their origin.
The best studied of these are putative homologues of fushi-tarazu, isolated from Tribolium (Brown et al., 1994) and Schis-tocerca (Dawes et al., 1994). Neither of these genes shows extensive similarity to ftz (or any other Hox gene) outside of the homeobox, and even within the homeobox they are much more divergent than is typical for the other Hox genes (Fig. 4). However, both of these genes are closely linked to the Hox cluster, and both show a pattern of expression in the nervous system very similar to that of the fushi-tarazu gene in Drosophila.
Relative divergence of homeotic and other Hox cluster homeodomains within Arthropods. Species for which Hox gene sequences are available have been grouped according to approximate phylogenetic distance from Drosophila melanogaster. Rough estimates of divergence times are given below, based on immunological similarity (within Diptera; Beverley and Wilson, 1984)) or on the first appearance of other taxonomic groups in the fossil record (Kukalova-Peck, 1991). Other Drosophila subgenera, (40-60 My): D. hydeiftz (Jost and D. Maier, personal communication; EMBL accession X79494); other Dipteran family/suborder, (100-200My): Musca bicoid (n.b. 6 changes in 44 residues of homeodomain sequenced (Sommer and Tautz, 1991); Aedes abd-A (Eggleston, 1992); other Endopterygote insects (250My): Bombyx (Lepidoptera) Ubx, abd-A (Ueno et al., 1992); Manduca (Lepidoptera) abd-A (Nagy et al., 1991); Apis (Hymenoptera) Antp, Scr, Dfd (Fleig et al., 1988; Walldorf et al., 1989); Tribolium (Coleóptera) abd-A (Stuart et al., 1993), ftz (Brown et al., 1994). Schistocerca (Orthoptera, approx. 300My) abd-A (Tear et al., 1990), Abd-B (Kelsh et al., 1993), Scr (Akam et al., 1988), Dax (ftz-related gene (Dawes et al., 1994); crustacean: Artemia (Anostraca, 550 My; Averof and Akam, 1993). The assignment of the divergent Artemia gene AfHxl to the ftz homology group is tentative, but is consistent with the fast divergence of these genes. For comparison, the divergence between some annelid or vertebrate Hox sequences and their closest Drosophila match are shown on the right.
Relative divergence of homeotic and other Hox cluster homeodomains within Arthropods. Species for which Hox gene sequences are available have been grouped according to approximate phylogenetic distance from Drosophila melanogaster. Rough estimates of divergence times are given below, based on immunological similarity (within Diptera; Beverley and Wilson, 1984)) or on the first appearance of other taxonomic groups in the fossil record (Kukalova-Peck, 1991). Other Drosophila subgenera, (40-60 My): D. hydeiftz (Jost and D. Maier, personal communication; EMBL accession X79494); other Dipteran family/suborder, (100-200My): Musca bicoid (n.b. 6 changes in 44 residues of homeodomain sequenced (Sommer and Tautz, 1991); Aedes abd-A (Eggleston, 1992); other Endopterygote insects (250My): Bombyx (Lepidoptera) Ubx, abd-A (Ueno et al., 1992); Manduca (Lepidoptera) abd-A (Nagy et al., 1991); Apis (Hymenoptera) Antp, Scr, Dfd (Fleig et al., 1988; Walldorf et al., 1989); Tribolium (Coleóptera) abd-A (Stuart et al., 1993), ftz (Brown et al., 1994). Schistocerca (Orthoptera, approx. 300My) abd-A (Tear et al., 1990), Abd-B (Kelsh et al., 1993), Scr (Akam et al., 1988), Dax (ftz-related gene (Dawes et al., 1994); crustacean: Artemia (Anostraca, 550 My; Averof and Akam, 1993). The assignment of the divergent Artemia gene AfHxl to the ftz homology group is tentative, but is consistent with the fast divergence of these genes. For comparison, the divergence between some annelid or vertebrate Hox sequences and their closest Drosophila match are shown on the right.
Outside of the homeodomain, short conserved peptides are shared uniquely between Drosophila ftz and the Tribolium ftz-like gene, and between the Tribolium and Schistocerca genes. In the latter case, the conserved region is centred on a YPWM sequence just upstream of the homeodomain. This so-called hexapeptide motif is characteristic of the homeotic Hox genes, and generally well conserved between distant species, often showing extended similarity between insect and vertebrate genes of the same class. Its presence in the Schistocerca and Tribolium genes strongly suggests that, on structural grounds, these should be regarded as ‘typical’ Hox genes. However, note that no such motif is conserved in the Drosophila ftz gene.
In contrast to the conserved expression of ftz in the nervous system, these putative ftz homologues show different patterns of expression during early development. Drosophila ftz is expressed in parasegmental stripes at two-segment intervals, and is needed for the formation of alternate segment boundaries (Lawrence and Johnston, 1989). In Tribolium, the, ftz gene is expressed in less well resolved stripes, and is apparently not needed for making the correct number of segments (Stuart et al., 1991; Brown et al., 1994). The Schistocerca gene is expressed in a posterior domain of the early embryo, but never in stripes (Dawes et al., 1994); its function in early development is unknown.
We believe that all of these /iz-related genes derive from a single Anip-class Hox gene that acquired novel functions in arthropod development. They are not involved in the maintenance of segmental identity in the mesoderm or the epidermis. They may have a role in very early patterning of the embryo in all insects, but this role appears to be different in different insects. In Schistocerca and Tribolium, whatever constraints maintain the integrity of the YPWM motif and its flanking sequences are retained, but the function of ftz in Drosophila appears no longer to impose the same sequence requirement. Because the changes in the homeobox are confined to regions other than those that contact DNA, we suggest that it may be the lower complexity of protein-protein interactions that allows ftz to diverge more rapidly than the canonical homeotic genes.
The other ‘non-homeotic’ genes of the Drosophila Ant-C are in some ways analogous to ftz. Their homeoboxes are very divergent, with respect to all other non-Dipteran homeobox genes; they lack any recognisable hexapeptide motif; comparison between Dipteran species shows that the bicoid homeobox is diverging rapidly (Sommer and Tautz, 1991; Schroder and Sander, 1993). Activity capable of rescuing bicoid mutants can be recovered from the eggs of other cyclorapphan Diptera, but was not detected in honeybee, or even lower Diptera (Schroder and Sander, 1993). Searches for bicoid genes in other insects have been unsuccessful (I. Dawson, S. Frenk and M. A., unpublished results).
We think that these genes may be derived, like/fz, from Hox genes which, in the lineage leading to Drosophila, have escaped from the conservative selection that characterises homeotic genes. We have recently isolated from Schistocerca a cDNA that lends some support to this hypothesis. It encodes a homeodomain protein that shows some similarities to the proboscipedia and Zen genes, but has no close homologue in Drosophila. It resembles most closely Hox class 3 vertebrate sequences (F. F., unpublished results). As yet we know nothing of the function or expression of this gene in Schistocerca, but we guess that it may derive from an ancestral sequence that gave rise, after considerable divergence, to the zen and/or bicoid genes of Drosophila.
ACKNOWLEDGEMENTS
Work from the authors laboratory was supported by the Wellcome Trust. We thank Nick Butterfield, Dieter Maier, Sue Brown and Rob Denell for communicating results prior to publication; Geoffrey Fryer and Hans-Georg Frohnhôfer for their advice at the early stages of our Crustacean work, Peter Holland and David Stern for valuable comments on the manuscript.
References
Here we shall use the term Hox genes to refer generically to the homeobox genes contained within the arthropod homeotic gene clusters, as well as their vertebrate homologues. The term Hom/Hox, never euphonic, is now redundant. The epithet Hox should no longer be applied to other classes of homeobox genes (Scott, 1992). We include within the ‘Hox’ designation the non-homeotic genes of the Drosophila Hox clusters (bcd,ftz, zen). The final section of this manuscript provides some justification for this usage. We use the term Antennapedia-class (Antp-class) Hox genes more restrictively, to refer to that subset of the Hox genes that have a homeobox conforming to the Antp-class consensus defined by Scott et al (1989). This includes all the central genes of the Hox clusters, but excludes the labial, proboscipedia, Deformed and Abdominal-B classes (vertebrate paralogy groups 1-4, 9-13).