Regulation of the Tribolium homologues of caudal and hunchback in Drosophila: evidence for maternal gradient systems in a short germ embryo

Pattern formation in the early Drosophila embryo is initiated by maternally provided gradient systems (St Johnston and Nüsslein-Volhard, 1992). One essential component of anterior-posterior axis specification is the interaction between bicoid (bcd) and caudal (cad). bcd RNA becomes localized during oogenesis at the anterior pole of the embryo. Protein translated from this localized RNA diffuses and forms an anterior-posterior gradient in the early embryo. The protein interacts with the homogeneously distributed cad RNA and represses its translation (Chan and Struhl, 1997; Driever and Nüsslein-Volhard, 1988; Dubnau and Struhl, 1996; Macdonald and Struhl, 1986; Mlodzik and Gehring, 1987; Rivera-Pomar et al., 1996). This translational regulation results in an opposing gradient for cad protein which has an essential function in activating the abdominal gap genes (Rivera-Pomar et al., 1995; Schulz and Tautz, 1995). The major role of bcd is, however, in the transcriptional activation of hunchback (hb) and other gap genes in a concentration-dependent manner. This ensures proper positioning of several bcd-dependent gap expression domains within the anterior half of the embryo (Driever and Nüsslein-Volhard, 1989; Struhl et al., 1989).

While the bcd gradient has served as a model system in understanding pattern formation in Drosophila, it is suspected not to have this role in more ancestral insects. The long-germ mode of development as found in Drosophila is probably an adaptation to its particularly rapid embryogenesis. The ancestral type of embryogenesis in insects and arthropods is the short germ type (Sander, 1983). In these embryos, the germ rudiment forms at the posterior ventral side of the egg. In extreme cases like the grasshopper, it may be restricted to only a few percent of the total egg length - which makes it difficult to imagine how an anteriorly localized bcd RNA could determine pattern formation at the posterior end of the egg. Moreover, classical experiments have only yielded evidence for a posteriorly localized organizing activity (Sander, 1976). Therefore, bcd could be considered a late addition during insect evolution and its pivotal function during embryogenesis could be restricted to higher dipterans.

We are interested in early pattern formation of the flour beetle Tribolium castaneum. Tribolium is a typical example for short germ embryogenesis, representing the ancestral type of embryogenesis in insects (Tautz et al., 1994), albeit not in its extreme form like the grasshopper. In contrast to Drosophila, only cephalic and thoracic segments, but not abdominal segments, are determined during the blastoderm stage. Furthermore, the most anterior 20% of the Tribolium blastoderm cells form an extra-embryonic membrane, the serosa. This structure is not found in this form in higher Dipterans like Drosophila, but is again an ancestral feature of insect embryogenesis. Prior to gastrulation, most blastoderm cells move from anterior and dorsal positions towards the posterior ventral region where they form the embryo proper. This germ rudiment then continues to grow from its posterior end to form a germ band which eventually encompasses all abdominal segments.

So far, it has been impossible to recover a credible bcd homologue from Tribolium. Although bcd codes for a homeobox protein which would be expected to be conserved, the comparison of bcd homologues in higher dipterans already showed that the amino-acid replacement rate in this particular gene is exceptionally high (Schröder and Sander, 1993). As the evolutionary distance between beetles and Drosophila is more than twice the maximal distance among known bcd homologues within the diptera, there might well be too many changes in the homeobox to allow recovery of a bcd homologue by low stringency screens or redundant PCR primers. Therefore, the failure to identify bcd in Tribolium does not exclude the existence of a functional bcd homologue. To approach this problem from a different angle, we decided to look at known direct target genes of bcd, i.e. cad and hb, in this species (Tc’cad and Tc’hb). These two genes are much more conserved and could indeed be recovered from Tribolium. We used assays in Drosophila to show that most likely there is a functional bcd homologue in Tribolium after all. Moreover, our results suggest a molecular mechanism for the shift in position of the hb gap domain that occurred during evolution from short germ to long germ insects. Such repositioning of gap domains is responsible for one key difference between short and long germ embryos, i.e. their distinct blastoderm fate maps.

Pretreatment and fixation of embryos, as well as whole-mount in situ hybridization were done as described by Klingler and Gergen (1993); Tautz and Pfeifle (1989); Wolff et al. (1995). For visualization of the rather weakly expressed Tc’hb-lacZ transcripts, riboprobes yielding especially low background were used. For this end, in vitro transcription reactions (digoxigenin or fluorescein RNA labeling kit; Boehringer) were treated with DNAse I (Boehringer) and purified with Ultrafree-MC columns (100 000 NMWL, Millipore), without further precipitation and without hydrolyzation.

Simultaneous colorimetric detection of two different transcripts was done as published (Hauptmann and Gerster, 1994; Wolff et al., 1995). In addition, a novel double-fluorescence in situ hybridization procedure was developed involving HNPP/fast red (Boehringer) and tyramide signal amplification (NEN Life Science Products). Details of this protocol will be presented elsewhere but are available from the authors on request.

Tc’CAD protein was detected with an antibody described by Schulz et al. (1998). This antibody is specific for Tribolium and does not cross-react with the Dm’CAD protein in whole-mount stainings.

As wild-type strains, Drosophila melanogaster Oregon R and Tribolium castaneum San Bernardino were used. For genetic analysis, crossings were done with the Drosophila maternal-effect mutants bcd^GB, bcd^E1, bcd^2-13, bcd⁰⁸⁵ (Frohnhöfer and Nüsslein-Volhard, 1986), osk⁵⁴, osk¹⁵⁰ (Lehmann and Nüsslein-Volhard, 1986) and tor⁴ (Schüpbach and Wieschaus, 1986).

Embryos lacking maternal Dm’cad function were obtained by generating cad² germ-line clones with the FLP-DFS technique (Chou et al., 1993; Rivera-Pomar et al., 1995). Maternally and zygotically mutant embryos were generated by additional use of a cad³/CyO hb-lacZ line (all lines kindly provided by R. Rivera-Pomar).

P-element mediated transformation was done as described by Rubin and Spradling (1982). Recipient strains were either ry⁵⁰⁶ for pCarnegie20 and pHZ50 based constructs or y w^67c23 for pCasper plasmids, respectively. For all transgenic lines generated, a minimum of two (but for most constructs, five) independent insertions were examined.

To test translational regulation of the Tc’cad gene in Drosophila, a 0.9 kb XbaI/XhoI fragment from an almost full-length Tc’cad cDNA clone (Schulz et al., 1998) was cloned into a pBluescript plasmid (Stratagene) containing the maternal promoter of Dm’hb (Schulz and Tautz, 1994). Promoter plus Tc’cad gene were then transferred as a1.9 kb NotI/XhoI fragment into NotI and SalI sites of a pCarnegie20 derivative (Rubin and Spradling, 1983), a gift from S. Cohen.

For the largest Tc’hb-lacZ construct (Fig. 4B, top), a genomic 5.2 kb XhoI/SacII fragment (pBS-Tc’hb-XS5.2) was cloned into pBluescript using EcoRV and XhoI restriction sites (the genomic SacII site was blunted). The whole region could then be transferred as a 5.2 kb BamHI fragment into the pCasper-β-Gal vector (Thummel et al., 1988), resulting in an in-frame fusion with the lacZ gene within the first coding exon of Tc’hb (see also Fig. 4A,B). A similar construct was generated using pHZ50 (Hiromi and Gehring, 1987) as vector for transformation (the basal hsp70 promoter of pHZ50 was removed).

To generate deletions within the Tc’hb regulatory region (2nd to 4th construct in Fig. 4B), a 5.2 kb XhoI/XbaI fragment from pBS-Tc’hb-XS5.2 was subcloned into pGem7Zf (Promega). Nested deletions within this fragment were then generated by restriction with ClaI, XhoI/SalI or HindIII and subsequent religation. The remaining truncated Tc’hb upstream sequences were then transferred as 3.2 kb, 2.6 kb and 0.3 kb BamHI fragments, respectively, into pCasper-β-Gal.

The construct containing the region distal of the P2 promoter but lacking P3 (5th construct in Fig. 4B), was generated by PCR using the High Fidelity Expand system (Boehringer). A pBluescript-specific upper primer (5′ AATACGACTCACTATAG) and a P2-5′-UTR-specific lower primer (5′ CCGACTGTTCGAAATTTTTGCAC) were used with the pBluescript subclone of Tc’hb (see above) as template. PCR products were digested with XhoI and cloned into XhoI/EcoRV sites of pBluescript. This region was then ligated as 3.0 kb BamHI fragment into the respective site of pCasper-AUG-β-Gal (Thummel et al., 1988). The last construct which contained both promoters but lacked most DNA between P2 and P3 was generated by SalI/HindIII digest of the pBluescript subclone of Tc’hb genomic region, and subsequent blunting and religation reactions. The resulting truncated region was again transferred as a 3.6 kb BamHI fragment into pCasper-β-Gal.

For in vitro binding assays, recombinant Dm’CAD protein was overexpressed in E. coli BL21 (Stratagene) and crude extracts were prepared as described by Rivera-Pomar et al. (1995). Purification of the 6xHis-tagged protein was done under denaturing conditions with TALON resin (Clontech) according to the manufacturer’s instructions. Purified protein was renatured through dialysis against Kadonaga buffer B (Kadonaga et al., 1987) with progressive reduction of urea concentrations.

DNaseI-footprinting and Maxam-Gilbert reactions were performed according to Dynan and Tijan (1983) with slight modifications (Rivera-Pomar et al., 1995), using components of the SureTrack footprinting kit (Pharmacia).

The earliest target gene of bcd during Drosophila embryogenesis is cad. bcd regulates the translation of cad by directly binding to the 3′-end of its mRNA (Chan and Struhl, 1997; Dubnau and Struhl, 1996; Rivera-Pomar et al., 1996). We have cloned a cad homologue from Tribolium and studied its expression pattern (Schulz et al., 1998). In the early blastoderm, Tc’cad mRNA is homogeneously distributed throughout the embryo. Prior to gastrulation, it becomes restricted to the posterior half of the egg, and as the germ band forms, cad mRNA gets confined to the growth zone. To detect a possible translational regulation of maternal cad mRNA, we prepared an antibody against the Tc’CAD protein. This antibody detects an initial homogeneous staining in all early blastoderm nuclei (Fig. 1A), as was observed for the maternal Tc’hb expression (Wolff et al., 1995). Subsequently, Tc’CAD protein withdraws from the anterior half of the blastoderm embryo, which could indicate translational regulation by an anteriorly localized factor (Fig. 1B). To determine whether this relates to a bcd-like activity, we transformed a construct into Drosophila bearing the Tc’cad gene under the control of a maternal promoter (see Methods). The transgene mRNA is equally distributed throughout the embryo (not shown). With the antibody specific for Tc’CAD, however, we find that Tc’CAD is excluded from the anterior quarter of the Drosophila blastoderm embryo (Fig. 1C). Therefore, Tc’cad in Drosophila is translationally repressed at the anterior pole like Dm’cad. To see if this regulation is dependent on bcd, we tested the Tc’cad construct in a Drosophila mutant background devoid of this maternal factor. Indeed we find that Tc’CAD protein is homogeneously distributed in Drosophila embryos lacking bcd activity (Fig. 1D). In principle, this effect of bcd on cad translation could be indirect, i.e. bcd might activate transcription of a factor that would regulate translation of Tc’cad. To exclude the possibility of zygotically expressed intermediate(s), we looked at Tc’CAD in unfertilized eggs (in a wild-type background). Also in this situation, a Tc’CAD protein gradient is formed (Fig. 1E). These findings strongly suggests that bcd regulates Tc’cad mRNA in the same way as it does Dm’cad RNA. Since Drosophila and Tribolium are separated by about 500 Myr of independent evolutionary history, the ability of Tc’cad mRNA to interact with BCD protein clearly demonstrates that the evolution of the Tc’cad sequence is functionally constrained and that a similar regulatory interaction must also occur in Tribolium. Therefore, this result strongly suggests that a bcd-like activity exists in the short germ embryo of Tribolium.

The second well characterized target gene of bcd is hb. In this case, bcd acts as a transcriptional regulator on the hb promoter, thereby activating zygotic hb expression in the anterior half of the embryo. We have previously described hb expression in Tribolium (Wolff et al., 1995). There are two early zygotic expression domains of Tc’hb, one in the prospective serosa nuclei and one in the posterior half of the egg where the germ rudiment is about to form (see also left column in Fig. 2). From comparison of this pattern with that of pair-rule genes in Tribolium we have concluded that the posterior domain in Tribolium is the equivalent of the anterior domain in Drosophila, and that Drosophila (which has no serosa) lacks an equivalent for the anterior domain (Wolff et al., 1995). To see if the bcd-like activity described above may also serve as a transcriptional regulator of Tc’hb, we transformed a reporter gene construct with approximately 5 kb of Tc’hb upstream sequences into Drosophila (see Methods). The earliest expression of this Tc’hb-lacZ gene occurs in a posterior-to-anterior gradient, which is reminiscent of the CAD gradient in Drosophila (Fig. 2A, central column). Shortly afterwards, two separate domains become evident, one in the anterior region and one covering the posterior pole. These two domains are reminiscent of the early zygotic hb expression in Tribolium (Fig. 2B). Both domains later retract from the poles (Fig. 2C, D). At the end of the blastoderm stage, the posterior domain disappears (Fig. 2E). Later, in the extended germ band, we also find a neuronal expression pattern which represents a subset of neuroblasts that also express endogenous Dm’hb (Fig. 2F). Altogether, this tightly regulated expression pattern suggests that the regulatory regions of Tc’hb are faithfully recognized by regulatory factors in Drosophila. In particular, it appears that the anterior blastoderm domain of the reporter gene is the equivalent of the serosa domain in Tribolium and that its posterior blastoderm domain is the equivalent of the gap domain in Tribolium. Specifically, double staining for the endogenous Dm’hb mRNA and the RNA from the Tc’hb-lacZ reporter gene shows that both domains arise at the same time (Fig. 2A,B, right column), prior to the formation of the posterior domain of endogenous Dm’hb (Fig. 2C,D).

To see how these Tc’hb domains are regulated in Drosophila, we introduced the reporter construct into embryos of different maternally mutant backgrounds. In a hypomorphic bcd mutant background we find that the anterior domain disappears, while the posterior domain expands somewhat anteriorly (Fig. 3B). In amorphic bcd alleles, the posterior domain expands even more, but does not reach the anterior pole (Fig. 3C). The repression at the anterior pole in bcd⁻ is due to torso activity, because in a torso mutant background we find that both reporter gene domains are still visible, but do not recede from the poles (Fig. 3E). In a oskar mutant background (where embryos lack nanos activity), both domains remain unchanged (Fig. 3D). These results indicate that the anterior domain of Tc’hb-lacZ is activated by bcd (directly or indirectly). In contrast, the posterior domain of the reporter gene is repressed by bcd as well as by torso. Thus, this domain is regulated unlike any Dm’hb domain, i.e. it is neither activated by bcd (like the anterior domain of Dm’hb), nor is it activated by the terminal system (like the posterior domain of Dm’hb; Margolis et al., 1995).

In Drosophila, cad is known to act as a transcriptional activator of posteriorly expressed genes (Rivera-Pomar et al., 1995). To test the influence of Dm’cad on the posterior Tc’hb-lacZ domain in Tribolium, we generated embryos maternally mutant for Dm’cad (see Methods). In this genetic background, we find that the posterior domain of Tc’hb-lacZ is indeed lost, while the anterior domain is unaffected (Fig. 3F). To check whether CAD functions as a direct activator of Tc’hb, we constructed a series of reporter gene constructs with different Tc’hb upstream fragments. This allowed us to pinpoint a region around the P2 promoter (Wolff et al., 1995) as the region that mediates expression of the posterior Tc’hb-lacZ domain in Drosophila (Fig. 4). Using in vitro footprinting on this fragment, we find that Dm’CAD binds at multiple sites in this region. These sites are similar to the previously published CAD binding consensus (see legend for Fig. 4; Dearolf et al., 1989b; Rivera-Pomar et al., 1995). Thus, the high density of CAD binding sites in the fragment that mediates the CAD-dependent regulation strongly suggests that cad functions as a direct activator of Tc’hb in Drosophila. This finding can also explain why the earliest expression of the Tc’hb-lacZ construct in Drosophila occurs in a way that reflects the maternal cad gradient (Fig. 2A). Further, we note that the observed negative regulation of the posterior reporter gene domain by bcd (Fig. 3B,C) is likely to be indirect, since in bcd mutants the Dm’cad expression domain expands anteriorly, which must result in concomitant expansion of this lacZ domain.

To see whether cad could indeed mediate the activation of hb in Tribolium, we doubly stained Tc’cad and Tc’hb in early embryos. For this we employed a newly developed procedure for fluorescent double in situ hybridization suitable for confocal imaging (see Methods). Fig. 5 clearly shows that the earliest zygotic activation of Tc’hb in the posterior half of the embryo occurs at a time when Tc’cad is restricted to the same region, indicating that this domain could well be directly activated by CAD in Tribolium embryos.

One way to investigate gene regulation in species not amenable to germline transformation is to transfer the gene in question into a model system like Drosophila to carry out functional experiments. This functional approach is especially important in cases where there is no obvious sequence conservation, as is the case for both, hb upstream DNA as well as for the cad 3′UTR. We have so far investigated the regulation of four Tribolium genes in Drosophila, and in all four cases we obtained meaningful expression patterns. As shown here, Tc’cad translation is properly regulated in Drosophila, and the Tc’hb promoter is regulated in a spatial pattern very similar to that in Tribolium. In two more attempts, we tested Tc’hairy (C. W., unpublished) and Tc’tailless (C. Eckert and R. S., unpublished) upstream DNA in Drosophila. Also in these cases, we obtained patterns reminiscent of the expression in Tribolium. Therefore, it appears that the molecular machinery in both species is sufficiently similar for Drosophila regulatory factors to properly recognize Tribolium sequences. Since it is often assumed that evolutionary changes are due to changes in gene regulation, this technique should help us to understand important aspects of the evolution of early patterning mechanisms.

The data presented in Fig. 1 show that Tc’cad is posttranscriptionally regulated in Drosophila embryos, and that this regulation depends on bcd. Since a Tc’CAD protein gradient is also formed in unfertilized eggs, no zygotic target gene of bcd can act as an intermediate, which makes it very likely that bcd itself binds Tc’cad mRNA. This ability of Tc’cad to interact with bcd can most easily be explained by the assumption that cad in Tribolium is translationally regulated by a bcd homologue. The failure by us and others to isolate this homologue may indicate that bcd has acquired sequence changes in the homeobox which render detection by PCR and library screening more difficult.

Additional support for this interpretation comes from our finding that a Tc’hb-lacZ transgene in Drosophila is regulated by bcd as well. In this case, our promoter mapping experiments could not place this regulatory input in a defined fragment.

However, we note that the upstream region surveyed shows multiple scattered bcd consensus binding sites. Such a distribution of bcd binding sites over a larger region also has been observed in Musca domestica (Bonneton et al., 1997). Moreover, expression of the anterior domain of Tc’hb-lacZ is very similar to that of artificial constructs carrying low-affinity bcd binding sites (Ronchi et al., 1993). Yet, we cannot exclude the possibility of indirect action, i.e. zygotic genes depending on bcd could be responsible for the anterior expression of Tc’hb-lacZ. One potential intermediate factor, orthodenticle, is expressed in a pattern similar to the anterior domain of Tc’hb-lacZ (Finkelstein and Perrimon, 1990). However, this domain is still present in a otd^YH mutant background (data not shown). Nevertheless, other anteriorly expressed genes (not all of which are known to date) might still be involved in this regulation. Therefore, we regard the regulation of Tc’hb-lacZ by bcd as less persuasive evidence for the existence of a bcd homologue in Tribolium than the regulation of Tc’cad.

As an alternative interpretation one might consider that some other anteriorly localized factor could regulate cad in Tribolium. Intriguingly, a cad homologue in Caenorhabditis appears to be regulated by mex-3, which binds RNA with KH domains (Draper et al., 1996; Hunter and Kenyon, 1996). In Drosophila, bcd is so far the only known regulator of cad translation, and it is almost certain that bcd alone is responsible for the spatial aspect of the regulation. However, other factors may participate in this interaction, binding to the same 3′UTR sequences. Conceivably, such a ‘cofactor’ could form an anterior gradient in Tribolium and spatially regulate Tc’cad. This could explain conservation of the interacting mRNA sequences which then might still be recognized by bcd in Drosophila. If this model was true, it would be very interesting to see to what degree this factor would mimic, in Tribolium, the role of bcd in Drosophila, i.e how it would become localized (by a mechanism similar to bcd?), and how it might regulate anteriorly positioned expression domains like the anterior domain of Tc’hb (should this factor also be able, like bcd, to regulate translation as well as transcription of target genes?). We note that only the early expression of Tc’hb (Fig. 2B, left column) is likely to be under direct control of an anterior (presumably maternal) gradient in Tribolium. The later, sharply demarcated expression in the serosa cells (Fig. 2E, left column) is likely under zygotic control, possibly through interaction with Tc’zen which is also expressed in all serosa cells (Falciani et al., 1996).

That Tribolium has a functional bcd gene appears to us the more parsimonious explanation, however. If true, probably all holometabolous insects have bcd. It is then still an open question, if its range extends to hemimetabolous insects of the ‘extreme short germ’ type (Tautz et al., 1994), like Schistocerca, where the embryo forms in an area far from the anterior pole. If bcd indeed functions in these embryos, it would either have to become localized to a different position within the embryo, or it may form a protein gradient by some other mechanism. In any case, our evidence for the existence of Tc’bcd encourages new efforts to isolate the gene from Tribolium, and this then should provide sequence information to allow isolation of the gene from additional phylogenetic groups.

While our results provide evidence for a functional bcd homologue in Tribolium, they suggest a role somewhat different from that in Drosophila. An important function of bcd in Drosophila is the activation of the hb gap domain which is located anteriorly in this species. In Tribolium, the posterior blastoderm domain of Tc’hb must exert this segmentation function. However, an anteriorly positioned bcd gradient in Tribolium is not a good candidate for activating transcription in the posterior Tc’hb domain. The reporter gene data show that Tc’hb upstream sequences are capable of mediating expression in two domains in Drosophila embryos. The similarity of this pattern to that of hb in Tribolium suggests that the anterior reporter gene domain corresponds to the serosa domain of Tc’hb, while the posterior one represents the Tc’hb domain that carries the gap gene function. The anterior reporter gene domain is dependent on Dm’bcd activity, which is consistent with our interpretation that a bcd homologue is present also in Tribolium. The posterior domain of the reporter gene, however, depends on Dm’cad. Also this is consistent with the situation in Tribolium if we assume that this reporter gene domain corresponds to the gap domain of Tc’hb, because this posterior Tc’hb domain overlaps with the Tc’cad gradient in a way that suggests that it could be activated by Tc’cad. Therefore, it appears that a qualitative switch has occurred during evolution of the hb gene, since this domain is activated by cad in one species, and by bcd in the other (see below).

An alternative interpretation, namely that the posterior Tc’hb-lacZ domain was the equivalent of the posterior Dm’hb domain, would also imply that a regulatory switch has occurred in the evolution of hb, in this case a switch from activation by the terminal system in Drosophila (Margolis et al., 1995) to activation by cad in Tribolium. However, this alternative interpretation is not likely to be true, because the equivalent of the posterior Dm’hb domain in Tribolium arises at a much later stage, in the growing germ band (Wolff et al., 1995). We find it implausible that the regulatory environment in the growing germ band of Tribolium should be similar to that of the early syncitial blastoderm in Drosophila. Moreover, we have shown that the posterior domain of the reporter gene is negatively regulated by the terminal system (Fig. 3). However, the major executor of the terminal system in Drosophila – the tailless (tll) gene – is not expressed in the Tribolium germ band. Instead, Tc’tll is expressed in the blastoderm (R. S. unpublished) where it is well suited to repress the Tc’hb gap domain at the posterior pole. Thus, the observed inhibition of the reporter gene by the terminal system in Drosophila is consistent with the first, but not with this second interpretation.

The negative regulation of Tc’hb-lacZ by the terminal system in Drosophila is separable from the activating elements in the promoter mapping experiments, suggesting that it is mediated via specific regulatory elements (Fig. 4). Together with our observations about Tc’tll, this supports a blastoderm function for a terminal system in Tribolium as well. In the context of the Tribolium embryo, however, a terminal influence is only obvious for the posterior pole of the egg. Retraction of Tc’HB protein from the posterior pole is probably due to nanos activity at early stages, since maternal hb RNA at the posterior end of the embryo appears to be translationally regulated (Wolff et al., 1995). Our new data suggest that at subsequent stages the continued retraction of Tc’hb mRNA from the posterior pole (see Fig. 2) is due to transcriptional repression by a terminal activity. We have no evidence for an anterior terminal activity in Tribolium so far, as the serosa domain in that species always fully covers the anterior pole of the embryo. Thus, the repression of the serosa domain from the anterior pole in Drosophila may reflect the symmetrical localization of torso activity in Drosophila. In short germ embryos, where no embryonic structures are formed at the anterior pole, terminal activity may not be present at the anterior pole.

We have previously argued that probably there is nanos activity in Tribolium (Wolff et al., 1995), and in this paper we provide evidence for anterior (bcd) and terminal (torso) systems in this species. Thus, all three of the anterior-posterior maternal systems known from Drosophila appear to be conserved in at least this short germ insect. In Tribolium, the gap domains of hb and Krüppel (Kr) occupy more posterior positions than in Drosophila, as abdominal segments are not represented in the blastoderm stage. The gap domain of Dm’hb is located in the anterior half of the embryo (Tautz et al., 1987), and Dm’Kr is expressed in the center of the embryo (Knipple et al., 1985). In Tribolium, the corresponding domains of Tc’hb and Tc’Kr occupy the posterior half of the embryo (Wolff et al., 1995) and the posterior pole (Sommer and Tautz, 1993), respectively. Since the maternal positional information in the Tribolium embryo appears to be similar to that in Drosophila, these changes must be due to differences in how these gap genes sense the positional information. In the case of hb, our results suggest which changes in the hb upstream region were necessary to achieve its change in expression. Assuming that the Tribolium situation represents the ancestral state (Tautz et al., 1994), it appears that the evolution of long germ development was accompanied by a loss of the cad regulatory input in parallel to the acquisition of (additional) bcd activating input.

We believe that the key to understanding this qualitative switch lies in the serosa expression domain of Tc’hb. Our reporter gene data suggest that this domain may already be activated by bcd in Tribolium. To explain the switch in the regulation of the gap domain, one can envision an intermediate state, where the serosa domain and the embryonic domain have fused into a single domain (Fig. 6). To achieve this, the evolution of a few additional bcd binding sites in the hb upstream region would have been sufficient. In this intermediate stage both bcd and cad would have acted as activators on the gap domain of hb. Subsequent loss of cad regulation could then have moved the posterior boundary of this combined domain towards the anterior. We note that the Tc’hb gene has three known promoters one of which appears to be specialized for mediating the cad regulation. In Drosophila, only two promoters are present, neither of which has a known responsiveness to cad. Thus, in all likelihood, the cad dependent promoter and its associated enhancer was lost. As we have found no other enhancer activity for later expression patterns of hb in the cad dependent fragment, the loss of this region could have been a single step. Intriguingly, a combined serosa and gap domain is still evident in the lower dipteran Clogmia. In this fly, hb is expressed in a large anterior domain, from which at later stages also the serosa is recruited (Rohr et al., personal communication). This mechanism – modification of the way how gap genes sense maternal positional information while this information itself remains constant – can explain how the blastoderm fate map changed during evolution of short germ insects to insects with long germ embryos. Moreover, it represents an intriguing example for the importance of regulatory adaptation during the evolution of developmental processes.

We thank G. Büttner for technical assistance, T. Zimmermann for help with confocal imaging, A. Ephrussi, C. Nüsslein-Volhard, R. Rivera, G. Struhl, as well as the Bloomington Stock Center for Drosophila strains, R. Rivera, S. Brown, C. Thummel, S. Cohen and N. Patel for DNA constructs and other reagents, and C. David and M. Kania for comments on earlier versions of the manuscript. We also thank the anonymous referees for their suggestions. This work was supported by grants from the Deutsche Forschungsgemeinschaft to D. T. and M. K.