A phylogenetic interpretation of the patterns of gene expression in Drosophila embryos

The early development of all animal embryos involves the generation of an asymmetric mass of cells oriented with respect to some basic coordinate system, the growth of this mass and the specification within it of regional identities e.g. dorsal, ventral, head, thorax, etc. Shortly afterwards, the main morphological characteristics of a given phylum are visible and ‘the body plan’ is thus defined. Examination of developing embryos suggests that multiple strategies exist to carry out these processes. A detailed understanding of these strategies together with the elucidation of common principles underlying the establishment of different body plans are import ant aims of developmental biology.

The search for these principles has traditionally relied on comparative morphology and embryology but recently genetics and molecular biology are making important contributions. Thus a fertile interaction between these two disciplines has provided some insights into the strategies used by the embryo of the fruit fly Drosophila melanogaster to generate and specify its body plan and we have learnt much about the genes controlling those processes, the molecules they encode and the pathways they define (Nüsslein Volhard et al. 1987; reviewed in Akam, 1987). Perhaps surprisingly, the primary structure of many of these molecules is highly conserved between species and across phyla and, as a consequence, connections have been established between transcriptional regulators and growth factors from many organisms and genes involved in Drosophila embryogenesis (see for example: Laughon & Scott, 1984; Rosenberg et al. 1986; Padgett et al. 1987; Rijsewski et al. 1987). These conservations are, at present, merely structural and raise the question of whether a similar conservation exists at the functional level. If this were so, it would be possible to find homologies in the mechanisms establishing body plans. Unfortunately, the answer is usually obscured by the difficulty in comparing the embryogenesis of such diverse animals as insects and vertebrates and, equally important, by our ignorance of the biological function of these molecules in Drosophila. Notwithstanding, these structural homologies provide a method to obtain molecular markers for developmental processes (Sharpe et al. 1987).

In the hope of extending our knowledge from Drosophila to other animals, we are using these homologies to generate probes to study the embryogenesis of the orthopteran insect Schistocerca gre-garia at the molecular level. This insect has a different, but comparable, mode of embryogenesis to that of Drosophila (see below) and thus we expect to rationalize at the functional level the structural homologies that we may find. In this context, we would like to discuss here segmentation in Schistocerca from the perspective provided by the ongoing work on Drosophila. An outcome of this discussion is the suggestion that many patterns of gene expression in Drosophila reflect ancestral situations or patterns maybe still present in less-evolved insects.

The body plan of the insect embryo is characterized by a mass of cells elongated and metamerized in the anteroposterior axis known as ‘the germ band’. It usually comprises a head, a telson and an overtly metamerized region which normally includes seventeen segments: three in the mouth parts, three in the thorax and a basic set of eleven in the abdomen. Despite the many classes of eggs, blastoderms, gastrulas and final morphologies embodied in the insect class, the germ band represents a constrained and obligatory developmental stage (Seidel, 1960; Sander, 1976, 1983). Because of this and since an enormous amount of the available information on Drosophila relates to the construction of its germ band, it is important to assess the relationship of Drosophila to other insects.

The embryos of Diptera like Drosophila represent an extreme form from a phylogenetic and an embryological point of view. Phylogenetically, they are the most highly evolved animals of a lineage that can be traced back to a common ancestor of present day annelids and arthropods. Embryologically, the Drosophila blastoderm contains a rather precise projection of the animal such that its cells are already determined to give rise to defined regions of the larva (Lohs Schardin et al. 1979; Technau & Campos Ortega, 1985) and the adult (Chan & Gehring, 1971; Illmensee, 1978). This highly determined blastoderm contrasts with the comparable stage of the other extreme of the annelid-arthropod lineage, the extant annelid embryo, in which the only regions specified are the head and a group of stem cells, the teloblasts, which will generate the main body region (Fernandez, 1980 and Fig. 1A,B); this mode of development is probably similar to that of the common ancestor. The embryogenesis of arthropod embryos offers a wide spectrum of variations between these two extremes (Dawydoff, 1928; Johanssen & Butt, 1941) exemplified by the subdivision of insects into short, intermediate and long germ according to the degree of representation of the animal in the primordial germ anlage (see Sander, 1976). Drosophila is a long germ insect and the degree of determination of its blastoderm is different from that of short germ insects, like the locust Schistocerca gregaria, in which the germ anlage is very small comprising only the head and a segment-building zone (Fig. 1C-F), an organization strongly reminiscent of that of the annelid embryo. Intermediate germ band insects have intermediate degrees of representation and determination.

These differences in the representation of the different regions when projected onto the germ anlage suggest that, although the germ band is a homologous structure between insects, the mechanisms underlying its generation and specification might not be (see Sander, 1983 for a thorough discussion of this problem). It also suggests that the establishment of the insect body plan relies, to different degrees, on instructions elaborated during the early blastoderm stage: in long germ embryos, it must depend heavily upon instructions generated during the syncytial phase (Schübiger & Wood, 1977; Vogel, 1977; Fröhnhoffer et al. 1986) whilst in short germ insects this process only yields a partial body plan and must have an additional phase of specification linked to the growth of the blastoderm anlage (Krause, 1938; Krause & Sander, 1962; Sander, 1976; Mee & French, 1986a,b).

In Drosophila, it is possible to relate developmental processes to the activity of specific gene products, thus the generation of the body plan depends on the correct deployment of the segmentation genes (Nüsslein Volhard & Wieschaus, 1980) and the specification of different regions on the homeotic genes (Lewis, 1963, 1978; Garcia Bellido, 1975; Kaufman & Abbot, 1984). The products of many of these genes have been shown to share a protein-coding domain, the homeobox, which is most tightly conserved between homeotic genes and is also present with a lesser degree of homology in some segmentation genes (McGinnis et al. 1984; Scott & Weiner, 1984). Interestingly, one segmentation gene, fushi tarazu (ftz), contains a homeobox of the homeotic class and its product has been shown to be required for the spatial registration of homeotic gene activity and the metamerism of the animal (Ingham & Martinez Arias, 1986; Duncan, 1986). This observation can be used as an example to argue that these sequences and others (Bopp et al. 1986) might help define and connect functional gene networks (Frigerio et al. 1986). In the case of ftz, for example, it is its ‘homeotic class’ homeobox that allows the linkage between the segmentation and the homeotic gene networks.

The homeobox domain is conserved across phyla and, within the arthropods, it is present in orders other than Diptera (McGinnis, 1985). Although this alone does not prove the existence of homologous genes in organisms other than Drosophila, the classical work of Bateson (1894), many homeotic mutations in insects (reviewed in Garcia Bellido, 1977), and the discovery of homeotic gene complexes in the silk worm Bombix (Tazima, 1964) and the flour beetle Tribolium (Beeman, 1987) support this hypothesis. These conservations are implicit in the initial suggestion of Lewis (1963) that these genes have long evolutionary histories and relatednesses and can be used to argue that their products are good markers, both for the specification of the body plan within a phylum and as a measure of the changes concomitant with the transitions between species.

On these assumptions, we are using homeoboxes from Drosophila to obtain several homologues of segmentation genes from Schistocerca. An important demonstration from the current work on Drosophila is that, as hinted from experimental embryology, the molecular processes that define metameric units and their identities precede their visible differentiation. Consequently, although experimental embryology provides a valuable tool, the understanding of early events during embryogenesis demands the acquisition of molecular markers for the processes of interest.

It is commonly accepted that the embryonic development of any animal contains information about its phylogenetic history. Debates, however, have often arisen about the precise meaning and nature of this information (Russel, 1916; Gould, 1977; Raff & Kaufman, 1983). For example, during the embryogenesis of Schistocerca and other orthopteran embryos, all abdominal segments develop small buds which might correspond to leg primordia and, in the first abdominal segment, they often begin to grow to form pleuropodia, appendages that fulfil some function during embryogenesis (Dawydoff, 1928; Schwalm, 1988). The development of these ‘prolegs’ is common among the embryos of most insects and in some cases has been interpreted as a reflection of their myriapod ancestry (Berlese, 1913; Imms, 1956). Another example of an ancestral developmental feature can be found in the progressive generation of abdominal segments characteristic of Schistocerca and other short and intermediate germ insects; it is likely that this mode of development is related to the teloblastic growth of their annelid-like ancestor, crustaceans and other more primitive arthropods. In the Drosophila embryo, morphological connections with its ancestry can be found in the process of gastrulation, the paired origin of the gnathal appendages (Turner & Mahowald, 1977), the conservation of the neuroblast map (Thomas et al. 1984), the extension of the germ band and, of course, in the general organization of the germ band. However, there is no sign of prolegs or of a major input of growth in the development of the abdominal primordium in comparison to the thorax.

In Drosophila, generation and specification of the basic body pattern along the anteroposterior axis take place simultaneously during blastoderm formation. These two processes rely on an early definition of asymmetries in the zygote in response to maternal information (Nüsslein Volhard et al. 1987) and on a complex network of interactions between maternal and zygotic segmentation gene products (reviewed in Akam, 1987). In the ectoderm, an important outcome of these interactions is the generation of a primary pattern of homeotic gene expression defining broad regions of the blastoderm (see for example Akam & Martinez Arias, 1985), and the activation of a ground set of cell states defined by the onset of segmentpolarity gene activity (Weir & Kornberg, 1985; Baumgartner et al. 1987; Ingham et al. 1988). After germ band extension, when cell division resumes and some morphogenetic movements occur, changes take place in the patterns of expression of homeotic (Akam & Martinez Arias, 1985; Martinez Arias et al. 1987) and segment-polarity genes; the first ones become modulated within and between metameric units, the latter also undergo refinements and, in addition, new segment-polarity genes are activated (see, for example, transcripts from the gooseberry region in Baumgartner et al. 1987); this second phase of gene expression relies on cell interactions (Martinez Arias et al. 1988; DiNardo et al. 1988).

An important feature of the above developmental profile is the dynamic patterns of gene expression. Although during blastoderm formation this plasticity is clear in the transient expression of pair-rule genes (for example: Hafen et al. 1984a; Ingham et al. 1985; McDonald et al. 1986), throughout development it is particularly clear in the patterns of expression of the homeotic and segment-polarity genes. Most of these changing patterns are difficult to relate to the pheno-type produced by the absence of the corresponding gene and, while they obviously reflect different levels of transcriptional control, it has been suggested that, in the case of the homeotics, they also reflect the phylogenetic history of the genes (Martinez Arias, 1987). In this manner, atavisms which are not morphologically visible in Drosophila because of the speed of its development, can be observed in the changing expression of the homeotic genes.

The wild-type expression of the Antennapedia gene provides a detailed example of this. In the embryo, mutations in Antp result in the transformation of T2 and T3 towards a novel nonthoracic segment (Waki-moto & Kaufman, 1981). This loss of thoracic character is reinforced by the damage observed in the Keilin’s organs, which are thought to represent leg rudiments (Keilin, 1915). During imaginal development, Antp mutations result in defects in proximal leg development and diverse transformations in the thorax (Struhl, 1981; Abbott & Kaufman, 1986). Underlying these phenotypes, there is a gene with a complex molecular structure (Garber et al. 1983; Scott et al. 1983), two independent promoters (Pl and P2) (Schneuwly et al. 1986; Laughon et al. 1987) and multiple levels of regulation (Irish et al. 1988).

From the onset of their expression at blastoderm, both promoters have very different patterns and regulations (Martinez Arias, Bermingham & Scott, unpublished data); while P2 behaves like most of the other homeotic gene promoters at blastoderm, in that it defines precise metameric domains (PS4 and PS14) (Fig. 2B,D) and is critically dependent on the product of the ftz gene, Pl defines a broad unmodulated domain (Fig. 2A,C) and is independent of ftz (Ingham & Martinez Arias, 1986). Shortly after germ band extension, the situation changes and for the rest of embryogenesis Pl behaves like the promoters of Dfd, Ser and Ubx, in its spatial and temporal patterns of expression, modulation and cross-regulation (Martinez Arias, 1987; Fig. 2E,G), while P2 displays different patterns: it is transiently expressed in the ectoderm of PS3,4,5 and, throughout the rest of development, in the nervous system from PS3 to the posteriormost metameric unit, PSIS (Martinez Arias, Bermingham & Scott, unpublished data). In the epidermis, this expression is clearly modulated but, in the nervous system, it is restricted to a subset of cells in each metamere. This broad domain of Antp expression can be seen first during gastrulation when Pl expression extends to PS12/13 and is bounded by P2 expression dorsolaterally in PS14. Given the requirement for Antp function to implement thoracic development, this pattern of expression can be interpreted as a transient stage in which the abdominal anlage has thoracic character. This transient ‘thoracic pattern’ soon disappears when the products of the BX-C, which have been deployed in the abdominal anlage, are translated and repress the expression of Antp. Indeed, in embryos lacking the BX-C, this repression does not take place and Antp expression extends into the abdomen (Hafen et al. 1984b; Carroll et al. 1986). Interestingly in these embryos, Keilin’s organs (Lewis, 1978) and leg discs (Bate & Martinez Arias, unpublished observations) develop in every abdominal segment suggesting a partial reversion to an ancestral, myriapod-like, condition. Thus, it is possible to envisage the pattern of expression of Antp as an ancestral pattern upon which the pattern of expression of other homeotic genes has elaborated, during evolution, the pattern of the Drosophila embryo; anteriorly with Ser, posteriorly with Ubx and abdA (Martinez Arias, 1987). The ancestry of Antp is still reflected in its broad domain of expression in the CNS (Fig. 2G,H), which probably reflects the early pattern and a requirement that is maintained for some yet unknown important function.

If, at the level of specification of different body regions, it is possible to obtain some phylogenetic information from the patterns of expression of homeotic genes, it is also possible to obtain similar information about the generation of the body plan from the patterns of expression of some segment-polarity genes. In this respect, an important difference between the short and long germ insects is that while the latter set up the basic positional information at blastoderm (Howard & Ingham, 1986; Ingham et al. 1988) and then intercalate new values by changing the patterns of genes already active and by activating new genes (Martinez Arias et al. 1988; DiNardo et al. 1988), in the former, for most of the thoracic and all of the abdominal metameres, there is not enough positional information in the blastoderm anlage. Thus it is likely that those mechanisms that after blastoderm play an important role in refining and elaborating patterns of segment-polarity gene expression through cell interactions, are those generating positional values in short germ insects. Therefore, the activation of engrailed and other segment-polarity genes by the pair-rule gene products in Drosophila, probably represents an evolutionary adaptation (or exaptation) to the quick generation of the body plan which takes place during blastoderm. This mechanism can be shown to be independent of another which, under certain experimental circumstances, can activate engrailed in every metamere after blastoderm (DiNardo et al. 1988) and which requires the activity of segment-polarity gene products (DiNardo et al. 1988; Martinez Arias et al. 1988). This second mechanism is likely to be responsible for the generation of positional values in short germ insects.

Experimental and descriptive embryology indicate that a fundamental difference exists between short and long germ insects in the generation and specification of the abdominal region during embryogenesis. In Drosophila, the generation of the abdomen relies on the operation during the syncytial blastoderm of a group of loci known as the grandchildless-knirps group (Nüsslein Volhard et al. 1987; Schüpbach & Wieschaus, 1986). All of these loci, with few exceptions, are maternal and their absence results in embryos lacking most or all of the abdominal segments. For example, embryos mutant for oskar represent an extreme form of this phenotypic series (Lehmann & Nüsslein Volhard, 1986); they differentiate a normal head, mouth parts, first three thoracic segments and telson, but lack all abdominal segments. This defect is foreshadowed at blastoderm by the abnormal expression of gap gene products (Gaul & Jackie, 1987; Tautz, 1987) and by a manifest defect of ftz expression in the abdominal primordium (Carroll et al. 1986).

In the wild type, the different regions of the embryo are specified at blastoderm through the activation of the homeotic genes in restricted and defined spatial domains. The abdominal primordium between PS6 and PS13 is specified at this stage by the activation of the elements of the BX-C, Ubx and adbA. Of these, the first one to be transcribed is Ubx in a region between 10 and 50 % EL. At the same time, the thoracic primordium is specified by the activation of Antp, P2 in PS4 and Pl in a broad domain spanning PS4, 5 and 6. Using these patterns of expression as guides to regional specification, we can infer a fate map for ask mutant blastoderms. In these embryos, initially there is no Ubx expression, Antp Pl is almost normal and Antp P2 has a broad domain from 10 to 40 % EL (Irish et al. 1988; Fig. 3C). This pattern can be interpreted as a shift in the fate map of the prospective abdominal anlage from abdominal to thoracic. The number of segments in this primordium is also altered; while, in the wild type, engrailed is expressed in 14 evenly spaced stripes along the anteroposterior axis (Weir & Kornberg, 1986; Fig. 3A), in ask mutant blastoderms, the first five stripes together with the last one are normal, but stripes 6 through 13 are fused in a single broad stripe (Fig. 3B).

After gastrulation, Ubx is activated in these embryos in an unusual pattern within the abdominal anlage (Irish et al. 1988); this activation is independent of the maternal information and most likely reflects zygotic functions, probably pair-rule and segment-polarity gene products which, in the wild type, serve to modulate the early expression. In this manner, the abdominal anlage of an ask mutant embryo is not only reduced in size and initially thoracic in character, but later starts acquiring some character through the effect of segmentation functions on homeotic genes. As for the case of positional values, we believe that in the embryos of short germ insects, homeotic genes are activated through the action of segment-polarity gene products during the growth of the primordia.

Garcia Bellido (1977) expanded on the idea, implicit in Lewis (1963), that homeotic mutations are atavic mutations. Following this idea, we believe that embryos from oskar mutant mothers, develop an atavic condition in which the fate map of the abdominal region is close to that of the embryo of a short germ ancestor. This fate map cannot develop as it would in a short germ insect and differentiates into a single segment with a final cuticular phenotype which is largely the result of postblastoderm regulatory events (Lehmann & Nüsslein Volhard, 1986).

The above considerations and interpretations lead to the view that the first instar larva of Drosophila is generated stepwise by modifying ancestral patterns of spatial expression of very conserved genes. Indeed, in the case of the homeotics, once a gene defines a spatial domain, it is modified within that domain by new gene products or by new interactions with preexisting ones; these changes often are brought about by new regulatory elements. This interpretation suggests that certain genes and certain patterns should be conserved in other less-specialized insect embryos and, as suggested by the patterns of Antp expression in the wild-type and ask mutant blasto-derms, we would expect the Antp gene to be the one specifying the blastoderm of short germ insects and Ubx and abdA to be under the control of zygotic, maybe segment-polarity, gene products. Also, we would expect segment-polarity genes to be deployed in restricted domains in a manner similar to that which restricts the expression of wg in ftz mutants, i.e. probably through cell interactions.

To test some of these predictions, we are isolating segmentation genes from the locust Schistocerca gre-garia. We have constructed a genomic library from testes of Schistocerca gregaria adults and screened it with mixtures of homeoboxes from Drosophila segmentation genes (Tear, Akam & Martinez Arias, in preparation). We obtained several homologous clones one of which cross hybridized with the engrailed and even skipped homeoboxes. Further analysis indicated that this clone also cross hybridized with homeoboxes from the homeotic class. We used a small homeobox-containing fragment in in situ hybridization experiments to sections of 50 % embryos of Schistocerca gregaria (see Bentley et al. 1979 for a reference on the staging) and observed a pattern of expression restricted to the abdomen and extending from the middle of the first abdominal segment to the middle of the eighth (Tear, Akam & Martinez Arias, in preparation). This domain could be parasegmental (Martinez Arias & Lawrence, 1985) and is similar to that of the abdA gene from Drosophila (McGinnis et al. 1984; A. Rowe & M. Akam, personal communication). Sequence analysis of the Schistocerca gene homoeobox proved it to be identical, at the protein level, to the Drosophila one (Tear, Akam & Martinez Arias, unpublished observations).

Metamerization in insect embryos is tightly linked to the establishment of a body plan and can be divided into two processes: the generation of metameric primordia and the specification of ‘cellular identities’ on these primordia. Here, we have discussed these processes in embryos of short and long germ insects from the perspective of Drosophila gene expression. We have suggested that the changing expression of homeotic and segment-polarity genes in Drosophila provides clues about their expression in other insects and have made some suggestions about these pat terns. As a prelude to testing these hypotheses we are cloning some homologues of Drosophila segmentation genes from the locust Schistocerca gregaria to use as markers during embryogenesis.

Our conclusions agree with already existing ideas about the phylogenetic importance of segmentation mechanisms (Sander, 1983) or homeotic genes (Lewis, 1963, 1978; Garcia Bellido, 1977) and suggest that a very important driving force in the evolution of the arthropod lineage is changes in the regulation of homeotic and segment-polarity genes. Thus, for example, Ubx and abdA might evolve from having a zygotic control, tightly linked to segment-polarity gene activity in short germ insects, to being dependent on maternal information for their initial deployment during the specification of the abdomen in long germ insects.

These ideas stress the notion that developmental systems, by the nature of their genetic hardware, are inherently plastic. This is apparent in two properties of this hardware: one, the modular nature of the control elements regulating the genes (see, for example, Hiromi et al. 1986), which allows easy addition, elimination and exchange of control sequences. The other, the combinatorial nature of the processes regulated by their products and the synergistic effects often produced by these combinations (Ingham et al. 1988; Doe et al. 1988; Irish et al. 1988) which also allows for the creation of diversity with minor modifications of preexisting elements. In consequence, it is because of the natural plasticity of developmental systems and their tendency to change that ontogeny is a very important force driving phylogeny.

We would like to thank M. Akam, I. Dawson, P. Ingham, E. Sanchez Herrero, H. Skaer and D. Sheperd for many useful and stimulating discussions. GT is supported by an MRC grant, MB by the SERC and AMA by a Wellcome Trust Senior fellowship in biomedical sciences.