Delimiting the conserved features of hunchback function for the trunk organization of insects

As one of the major coordinators of the Drosophila segmentation gene cascade, hunchback (Dm'hb) is required for the proper expression of patterning genes involved in both metamerization and segment identity specification (Hülskamp and Tautz, 1991). Maternally expressed Dm'hb polarizes the Drosophila embryo by forming an anterior to posterior morphogenetic gradient in the posterior half of the syncytial blastoderm, which is required for positioning the central and posterior gap gene domains(Tautz, 1988; Hülskamp et al., 1990; Struhl et al., 1992). The anterior zygotic expression of Dm'hb overlaps spatially and functionally with the maternal expression and appears to act as a canonical gap gene being required for the formation of an adjacent set of segments(Lehmann and Nüsslein-Volhard,1987).

In addition to the role during segmentation, Dm'hb also controls the expression of Hox genes. The anterior domain of Dm'hb limits the anterior borders of the Dm'Ubx and Dm'Antp expression domains (Irish et al., 1989; Lehmann and Nüsslein-Volhard,1987; Qian et al.,1991; White and Lehmann,1986; Zhang and Bienz,1992). However, the effects caused by this ectopic expression of Hox genes in Dm'hb mutants are often concealed by the segmentation defects, as the Hox genes are ectopically expressed in the segments that are deleted in the larvae (Lehmann and Nüsslein-Volhard, 1987).

The role of hunchback in segmentation has also been functionally studied in other insects with the aim to elucidate the transition from short germ to long germ embryogenesis (He et al., 2006; Liu and Kaufman,2004; Mito et al.,2005; Schröder,2003). Although the expression patterns of hunchback are well comparable, different functional roles have been ascribed to hunchback in the different insects. More or less canonical gap phenotypes were reported for Tribolium(Schröder, 2003) and Nasonia (Pultz et al.,2005). Different phenotypes, including transformations and loss of trunk segmentation were found in Oncopeltus(Liu and Kaufman, 2004), Gryllus (Mito et al.,2005) and Locusta (He et al., 2006). In addition, it is clear that the primary regulator of zygotic hunchback expression in Drosophila, bicoid, is a late evolutionary acquisition that emerged only in the higher Diptera(Stauber et al., 2002; Schröder, 2003). Thus, it appears that hunchback regulation and function has been subject to major evolutionary changes even within insects.

Even in Drosophila, the role of hunchback is more complex than it is often portrayed. Some alleles of Dm'hb produce directly a combination of homeotic transformations and trunk segmentation defects(Lehmann and Nüsslein-Volhard,1987). One allele has originally been identified as Regulator of postbithorax and hence as a homeotic gene(Bender et al., 1988). In addition, given that phenotypic effects are always a combination of loss of the gene itself and changes in downstream genes, it is often not easy to recognize possible conserved features.

Here, we re-investigate the hunchback phenotype in Tribolium and assess its role in regulating the trunk gap genes and Hox genes. We focus the study on the anterior expression region, as this reflects the key function for hunchback in organizing the Drosophila segmentation gene cascade(Hülskamp and Tautz,1991). We find that one can identify some core components of hunchback function that appear to be conserved in all insects. This includes the role in specifying anterior borders of Hox gene expression and interactions with other gap genes. However, the major function in segmentation that leads to the typical gap phenotype in Drosophila appears to be limited to long germ embryos. Hence, most similarities in hunchbackfunction appear to exist among insects that represent the ancestral type of embryogenesis, while more-derived types have developed additional features.

Tribolium castaneum strain San Bernardino beetles were reared on white flour supplemented with brewer's yeast at 30°C. The pupae for injections were obtained by collecting eggs within a time span of 9 hours of development and leaving them for about 25 days at 30°C to develop.

Parental RNA interference essays were performed as described by Bucher et al. (Bucher et al., 2002). Double-stranded RNA was injected into pupae at a concentration of 2μg/μl. We found this concentration ideal to obtain maximum penetrance for most genes. Eclosed females were mated with wild type males and reared under standard conditions (see above). Knockdown embryos were collected every second day and one collection per week was kept at 30°C to monitor RNAi penetrance at the cuticular level. The collections were performed until the phenotypic effect had decreased significantly. Embryos for in situ hybridization were taken from females showing the highest penetrance, as judged by the parallel analysis of cuticle phenotypes.

The eggs were washed for 1 minute in 50% bleach solution and for 2 minutes in running water to remove the chorion. The fixation was performed in a scintillation vial with 3 ml PBS, 6 ml heptane and 4% formaldehyde for 30 minutes. The eggs were then devitellinized by replacing the PBS with 8 ml of methanol and by shaking thoroughly for 30 seconds. The eggs that lose the vitelline membrane become hydrophilic and move from the interphase to the hydrophilic phase. After several washes with methanol they were transferred to Eppendorf vials. The remaining eggs were passed through a 0.9 mm needle until the vitelline membrane was removed.

The gene expression profile was obtained by whole-mount in situ hybridization as previously described(Lehmann and Tautz, 1994; Tautz and Pfeifle, 1989). Digoxigenin- or fluorescein-labelled probes were detected using alkaline phosphatase-coupled antibodies and INT/BCIP (red) or NBT/BCIP (blue)substrates.

First-instar larvae were digested overnight in a 1:1 lactic acid/Hoyer's medium solution at 70°C and mounted on microscope slides. The cuticular autofluorescence in a range of 520-660 nm was detected on a Leica Confocal microscope, and maximum projection images were created from z stacks composed of 50 layers scanned four times each.

Parental RNA interference (pRNAi) was used to generate females lacking both, maternal and zygotic Tribolium hunchback (Tc'hb)function. As the mother transfers the RNAi effect to the eggs, the phenotype decays in relation to the age of the injected female, resulting in a phenotypic series (Fig. 1).

Schröder (Schröder,2003) has previously shown that loss of Tc'hb function(Tc'hb^pRNAi) does not affect the pre-gnathal segments labrum, antenna and mandible. In the strong phenotypes, the remaining segments bear no appendages and appear to have abdominal identity. As a consequence,this could be interpreted as a gap phenotype in which the maxillary, labial and thoracic segments are missing(Schröder, 2003).

However, a reanalysis of the phenotypic series suggests that the phenotype observed is a combination of transformation and loss of segments. In weak phenotypes, all segments appear to be present, but the maxillary and the labial segments are transformed into abdominal identity(Fig. 1B). In addition, the thoracic segments look also more like abdominal segments, although T2 and T3 retain some appendage stumps, which may even look like developed legs in T3(arrow in Fig. 1B). In more severe phenotypes, all segments beyond the mandibular one look like abdominal segments (Fig. 1C). Yet there are often more than the normal eight abdominal segments, suggesting that this cannot be interpreted as a simple loss of the remaining gnathal and thoracic segments. Instead, one sees a disruption of more posterior segments (arrowhead in Fig. 1C,D) and an increasing loss of posterior structures (Fig. 1D). Therefore, the Tc'hb^pRNAi phenotype can be described as a transformation of gnathal and thoracic segments into abdominal identity combined with a loss of abdominal segments in the more severe phenotypes. In the most extreme phenotypes, only the antennae and mandibles are left, followed by four segmental structures of abdominal identity (Fig. 1D).

The transformation of anterior segment identity into posterior segment identity suggests an ectopic expression of posteriorly acting Hox genes in Tc'hb^pRNAi embryos. In order to test this, we have compared the wild-type expression of the gnathal Hox genes, Tc'Dfdand Tc'Scr, as well as the trunk Hox genes Tc'Antp, Tc'Ubxand Tc'AbdA with their expression in Tc'hb^pRNAiembryos (Fig. 2). To assess the segmental register in these embryos, we have used the segment polarity gene gooseberry (Tc'gsb) as segmental marker. Tc'gsb was chosen instead of engrailed because its expression precedes the expression of other segment polarity genes(Savard et al., 2006a).

In wild-type embryos, Tc'Dfd is expressed in the mandibular and maxillary segment (Fig. 2A),followed by the expression of Tc'Scr in the labium(Fig. 2C). In Tc'hb^pRNAi embryos, we see a narrower expression domain of Tc'Dfd (Fig. 2B),while Tc'Scr is absent (Fig. 2D). This is in line with the observed cuticle phenotypes. Tc'Dfd is required for mandible specification in Triboliumand also partly for the maxilla (Brown et al., 2000). Given that the mandible is not affected in Tc'hb^pRNAi embryos, Tc'Dfd expression was expected to be retained in this segment. Tc'Scr is required for the labial segment (DeCamillis et al.,2001), but given that this is transformed into an abdominal segment, its absence is in line with the phenotype.

The three trunk Hox genes Tc'Antp, Tc'Ubx and Tc'AbdA are all expanded towards anterior in Tc'hb^pRNAi embryos,starting to be expressed from the mandibular segment onwards(Fig. 2E-J). Each of these genes has specific functions in specifying thoracic and anterior abdominal segments, in agreement with their specific anterior expression borders(Lewis et al., 2000). However,in the abdominal segments, they are all co-expressed in wild-type embryos(Fig. 2E,G,I) and they are likely to have a joint function in specifying abdominal segment identity. Accordingly, the fact that all three are co-expressed from the mandibular segment onwards in Tc'hb^pRNAi embryos is in line with the observation of the transformations into segments of abdominal identity(Fig. 1).

We can conclude from these observations that Tc'hb is required for the regulation of at least four Hox genes along the anteroposterior axis,although some of these regulatory effects may be indirect (see Discussion).

To understand the basis of the segment deletions observed in Tc'hb^pRNAi embryos, we have analysed Tc'Kr and Tc'gt as possible target genes. In Drosophila, the hunchback gradient is required to regulate other gap genes, in particular Dm'Kr, Dm'kni and Dm'gt(Hülskamp et al., 1990; Struhl et al., 1992). The homologues of Krüppel and giant have been functionally studied in Tribolium (Bucher and Klingler, 2004; Cerny et al.,2005) and we have therefore focused on these in the following.

Tc'Kr expression starts already at the blastoderm stage with a broad domain at the posterior end (Fig. 3A) (Sommer and Tautz,1993), which covers the three thoracic segment primordia in the early germband (Fig. 3C)(Cerny et al., 2005). In Tc'hb^pRNAi embryos, this domain is strongly reduced or even absent (Fig. 3B,D),indicating that Tc'hb is required for its activation. There is also a segmental expression of Tc'Kr, which appears during segment differentiation (Fig. 3E)(Cerny et al., 2005). This expression is not affected in Tc'hb^pRNAi embryos, although fewer segmental stripes are generated (Fig. 3F), in line with the loss of segments in such embryos. Thus, Tc'hb is required for the activation of the early Tc'Krdomain. This is also an essential function of hunchback in Drosophila (Hülskamp et al.,1990; Struhl et al.,1992; Schulz and Tautz,1994), i.e. this regulatory interaction appears to be conserved.

Tc'gt is initially expressed in a broad domain during blastoderm stage covering the future head and gnathal segments, but excluding the labium(Bucher and Klingler, 2004). The trunk expression appears during germband elongation(Fig. 4A) and converges to two stripes over the third thoracic and second abdominal segments(Fig. 4B)(Bucher and Klingler, 2004). In Tc'hb^pRNAi embryos, the anterior domain is not visibly affected (Fig. 4C). The expression of the trunk stripes, however, is lost (compare Fig. 4B,D). With further development, it becomes apparent that the segments that should have expressed Tc'gt are partially fused, as monitored by the Tc'gsbexpression (Fig. 4D). No further segments are produced beyond this point, at least in strong phenotypes. These results suggest that Tc'hb acts formally as an activator of Tc'gt, which would be different from its role in Drosophila, where it acts as a repressor(Struhl et al., 1992).

As cross-regulatory interactions among gap genes are known in Drosophila, we have also analyzed the effects of Tc'Kr and Tc'gt on the expression of Tc'hb.

Tc'hb is initially maternally expressed throughout the whole zygote and early embryo. At blastoderm stage, it forms an anterior cap in the extra-embryonic serosa and a domain in the prospective head region, which covers the head segments up to the labium(Wolff et al., 1995). This domain becomes weaker in the early germband(Fig. 5A). At later stages, Tc'hb forms a strong expression domain in the growth zone, remaining there until the end of segmentation (Fig. 5D). Finally, there is weak expression in segmental stripes(Fig. 5D).

In Tc'Kr^pRNAi embryos, the blastodermal Tc'hbexpression domains appear not to be strongly affected, although it is possible that the head domain is extended towards the posterior pole. Given that this domain develops very dynamically, it is not possible to show this unequivocally. A major effect is seen from early germband stages onwards. The posterior domain develops much earlier and is expressed much more strongly. Its anterior boundary is initially within the maxillary segment(Fig. 5B) and overlapping the normal head domain. This boundary recesses at later stages and the domain is broadly confined to the growth zone (Fig. 5E). Thus, Tc'Kr acts formally as a repressor on the posterior domain, a role that is not known from Drosophila.

In Tc'gt^pRNAi embryos, Tc'hb expression is not significantly changed. There are no visible effects on the anterior domains(Fig. 5C). The posterior expression domain in the growth zone is present, but appears to be activated earlier and in a smaller area (Fig. 5F). This effect could be caused by the segment deletions observed in these embryos (arrowhead in Fig. 5F). In Drosophila, giant has a role in regulating secondary anterior Dm'hb expression domains, but has no role for the posterior Dm'hb domain (Wu et al., 1998).

Gap genes in Drosophila are directly required for regulating the primary pair-rule stripes. Accordingly, pair-rule stripe formation is disrupted in the area of the expression of the respective gap gene. In Tribolium it seems possible that the pair-rule pattern is set up only via interactions among the pair-rule genes themselves, whereby runtand even-skipped have essential functions(Choe et al., 2006). We have therefore analyzed the expression of these genes in Tc'hb^pRNAi embryos(Fig. 6). In wild type, the first Tc'runt stripe appears in the maxillary segment, the second in the first thoracic segment. The border of hunchback expression is within the labial segment, i.e. between the two stripes. Thus, if hunchback did have a direct effect on Tc'runt expression,one would expect to see the first two stripes to be affected. This does not appear to be the case. Stripe 1 and the distance to the second stripe are practically unchanged in Tc'hb^pRNAi embryos(Fig. 6A,B). Only the formation of the further stripes is disturbed. They form a large domain rather than separate stripes (Fig. 6C,D). At later stages, only a broad domain remains visible in the growth zone(Fig. 6E,F). The situation is comparable for Tc'eve, with the complication that the pattern is more dynamic. The first Tc'eve stripe overlaps the mandibular/maxillary segments and the second the labial/T1 segments. These then split into segmental stripes (Fig. 6G). In Tc'hb^pRNAi embryos, these first two stripes are almost normally formed and only subsequent stripes are less well resolved(Fig. 6G,H). An additional difference concerns the stability of the stripes. In Tc'hb^pRNAi embryos they disappear much faster than in wild type (Fig. 6I-M). Interestingly, however, separate stripes are still seen in the growth zone(Fig. 6L,M), suggesting that stripe patterning is less disrupted for Tc'eve than for Tc'run.

Studying the phenotypic series of hunchback effects in Tribolium reveals that the phenotype is not a simple loss of several adjacent anterior segments. Instead, it is a combination of transformed segments and loss of posterior segments. This does not fit the canonical definition of gap genes in Drosophila, but fits well with the effects seen for other gap genes in Tribolium, namely that they display a combination of transformation and segment loss phenotypes(Bucher and Klingler, 2004; Cerny et al., 2005; Savard et al., 2006a). Thus,regulation of Hox genes and segmentation genes are linked features for the Tribolium gap genes. In the following, we want to argue that this core role of Tc'hb can be reasonably well defined and that it is in fact not so much different between the different insects.

The setting of Hox gene expression domains appears to be particularly sensitive to hunchback function and may be the key feature for understanding its role. Changes in Hox gene expression are also one of the hallmarks of the allelic series of hunchback phenotypes in Drosophila. In hypomorphic class III alleles(Lehmann and Nusslein-Volhard,1987), Dm'Ubx is ectopically expressed in the thoracic segments (White and Lehmann,1986). In the region where four metameres should have formed(corresponding to two thoracic and two abdominal segments), only two enlarged metameres spanning this entire region appear. Owing to a resizing process,which involves cell death, these two enlarged metameres approach wild type width later in development (White and Lehmann, 1986). Because of ectopic Dm'Ubx expression,they are specified as abdominal segments. Hence, the phenotype is characterized as a loss of T2 and T3 in the larvae, although the remaining resized metameres are composed of primordial cells of thoracic and abdominal segments.

Other hunchback alleles in Drosophila are directly characterized by homeotic transformations of anterior segments into abdominal identity (Lehmann and Nüsslein-Volhard, 1987; Hülskamp et al., 1994) or act as dominant regulators of Hox genes(Bender et al., 1988). Some Dm'Ubx enhancers have been shown to bind HB protein, i.e. the regulatory interaction appears to be direct(Qian et al., 1991). Similar results were also obtained for Dm'AbdA regulation and enhancers(Casares and Sánchez-Herrero,1995; Irish et al.,1989; Shimell et al.,2000) and there is evidence that Dm'Scr, Dm'Antp and Dm'AbdB are also regulated by Dm'hb(Casares and Sánchez-Herrero,1995; Riley et al.,1987; Wu et al.,2001).

Our results suggest that Tc'hb acts formally as a repressor on Tc'Antp, Tc'Ubx and Tc'AbdA, as the expression domain of all three expands towards anterior in Tc'hb^pRNAi embryos, and formally as an activator of Tc'Scr, as its expression is lost in Tc'hb^pRNAi embryos. It seems clear, however, that some of these effects are indirect (Fig. 7). The expansion of the Tc'Antp domain in Tc'hb^pRNAi embryos may be the reason for the repression of Tc'Scr, as Dm'Antp is known to have an epistatic effect(posterior prevalence) on Dm'Scr in Drosophila(Carroll et al., 1988; Pelaz et al., 1993).

In Drosophila, hunchback does not act as a repressor on Dm'Antp. Instead, the secondary blastoderm expression of Dm'hb, the PS4 stripe expression, acts as an activator of Dm'Antp in this domain (Wu et al., 2001). An equivalent of the PS4 stripe expression is missing in Tribolium (Wolff et al.,1995) and a conserved regulatory interaction cannot be expected for this aspect. Thus, the repression effect of Tc'hb on Tc'Antp is not a conserved feature, at least not in Drosophila.

A possible direct repressor function of Tc'hb on Tc'Ubxis not obvious, as Tc'hb expression does not visibly reach to the anterior border of the Tc'Ubx trunk expression domain. However, the regulatory effect might be mediated via epigenetic regulation. It was proposed that Dm'hb initiates the formation of a silencing complex and that another protein, apparently dMi-2 (Kehle et al., 1998), takes over the role of Dm'HB protein when Dm'HB levels start to decline. In Tribolium, this mechanism would imply that only a few cells in the growth zone, which show no HB protein expression(Wolff et al., 1995), might retain the capacity to express Tc'Ubx, even though the actual transcription may start later.

The second consistent feature of Tc'hb function is the interaction with other gap genes, most notably Tc'Kr(Fig. 7A). In Drosophila,Krüppel is regulated by many other genes(Gaul et al., 1987), but the only activators that were identified are Dm'bcd and Dm'hb(Hoch et al., 1992; Hülskamp et al., 1990; Struhl et al., 1992). Dm'bcd is a late addition in higher Dipterans(Stauber et al., 2002), so only Dm'hb is a candidate for a conserved positive regulator. Moreover, it has been shown that Dm'hb alone is capable of establishing a functional Dm'Kr expression domain(Schulz and Tautz, 1994). Hence, the finding that the Tc'Kr domain is dependent on Tc'hb is in line with the activation role of hunchback on Krüppel observed in Drosophila. Given that Tc'Kr expression starts already at blastoderm stage at the posterior pole, in the region where Tc'hb forms a short gradient, it would seem likely that this effect is direct, i.e. this may be another conserved feature of hunchback function.

The regulatory interaction with giant, however, is clearly not conserved. In Drosophila, hunchback is a strong repressor of giant, i.e. the anterior expression border of the posterior domain is set by a low concentration of the HB protein gradient at blastoderm stage(Struhl et al., 1992). By contrast, in Tribolium, we find formally an activating effect of hunchback on giant. However, at the time where Tc'gt becomes expressed in the trunk, there is no contact to the Tc'hb domain, i.e. this effect is likely to be indirect. Tc'Kr cannot be the mediator of this effect, as loss of Tc'Kr alone does not lead to a complete loss of the trunk Tc'gt stripes (Cerny et al.,2005). Instead, the effect may be caused by a combination of Tc'Kr and Tc'mlpt. Tc'mlpt expression in the trunk is strongly reduced in Tc'hb^pRNAi embryos and Tc'gtexpression is lost in Tc'mlpt^pRNAi embryos(Savard et al., 2006a). This combined loss of Tc'Kr and Tc'mlpt in Tc'hb^pRNAi embryos may account for the loss of Tc'gt expression in the trunk. Thus, we can conclude that the apparent direct interaction between hunchback and giant in Drosophila is not a conserved feature of hunchback function,but has probably been acquired in the lineage towards the higher Diptera.

Liu and Kaufman (Liu and Kaufman,2004) have studied hunchback function in the intermediate germband insect Oncopeltus (Of'hb) using also a parental RNAi approach. Their phenotypic series is almost identical to the series we have found for Tribolium. Only the most extreme phenotype is stronger in Tribolium. They find also misexpression of Hox genes and conclude that Of'hb has the same dual role that we find for Tribolium. Hence, hunchback functions in Oncopeltusand Tribolium are likely to be very similar, possibly even at the level of the interactions with the other gap genes. For example, the segmental defects seen in the abdomen of intermediate strength Of'hb^pRNAi phenotypes appear to correspond to the segments that are also affected in Tribolium and which may be related to the loss of the two stripes of trunk giant expression.

Mito et al. (Mito et al.,2005) have studied hunchback function in Gryllus(Gb'hb). Although they suggest that there are distinctly different functions for hunchback in this species, it seems that the core findings are nonetheless very much in line with our results in Tribolium and the results in Oncopeltus. Again, ectopic expression of Hox genes is the first effect seen in weak Gb'hb^pRNAi phenotypes, accompanied with signs of transformation of thoracic segments. Stronger Gb'hb^pRNAiphenotypes show a progressive loss of abdominal segments. The most severe phenotypes described by these authors are not as strong as those found for Tribolium or Oncopeltus, but it is naturally difficult to ensure that the parental RNAi effect is fully penetrant. These authors have also studied expression of Gb'Scr and Gb'Kr and find,similar to our results in Tribolium, that expression is severely reduced or even absent in Gb'hb^pRNAi embryos.

He et al. (He et al., 2006)have shown parental RNAi phenotypes for hunchback in Locusta(Lm'hb^pRNAi) and conclude that some of these appear to be different from those found for Tribolium, Oncopeltus and Gryllus. In the weakest Lm'hb^pRNAi phenotype,they find only abdominal effects, but no anterior transformation effects,suggesting that the Hox gene misregulation effect is not as sensitive as in the other species. However, we interpret their most frequent Lm'hb^pRNAi phenotype (class II) as embryos where the head and thoracic segments are transformed into abdominal segments and where segmentation stops after this. This would be comparable to the strongest phenotypes in Tribolium, Oncopeltus and Gryllus, although even fewer segments appear to be produced in Locusta, possibly because the germ anlage is so extremely short in this species. The even stronger Lm'hb^pRNAi phenotypes observed by these authors(class III) appear to be related to a separate function of hunchbackin the extra-embryonic membrane (He et al., 2006).

Pultz et al. (Pultz et al.,2005) found that the headless mutant in Nasonia is an apparent null allele of hunchback (Nv'hb). They find also misregulation of Hox genes in Nv'hb mutant embryos but the phenotype is not easily comparable with the ones found for Tribolium, Gryllusor Oncopeltus. Instead, the Nv'hb mutant phenotype mimics the Drosophila phenotype in showing a large deletion of anterior segments, as well as loss of posterior abdominal segments(Pultz et al., 1999). However,as hypomorphic Nv'hb alleles are not available, it is difficult to assess whether these would show a homeotic transformation, as we see it in Drosophila. The more extensive loss of head segments in Nv'hb mutant embryos may be explained by the fact that bicoid is partially redundant with hunchback function in Drosophila (Hülskamp et al.,1990), i.e. might rescue some of the phenotypic effects. As there is no bicoid in Nasonia, this effect would not occur.

Bucher and Klingler (Bucher and Klingler, 2004), Cerny et al.(Cerny et al., 2005) and Choe et al. (Choe et al., 2006) have suggested that the action of the pair-rule genes may not be as strongly coupled to the gap genes in Tribolium as it is known in Drosophila. Choe et al. (Choe et al., 2006) have even suggested that the segmentation process may largely be controlled by interactions among the pair-rule genes themselves. This inference is also supported by the analysis of Tc'hb function. If Tc'hb were directly involved in setting segmental boundaries, one would expect that the major phenotypic effect would occur in or around the domain where it is expressed. However, the first two pair-rule stripes of Tc'runt appear to form more or less normally in Tc'hb^pRNAi embryos and disruption of the patterning is seen only for the subsequent stripes. This is in line with the observation that four segments are still formed in the most extreme Tc'hb^pRNAi phenotypes(Fig. 1D), although they are transformed into abdominal identity. A similar pattern is seen for Tc'eve, although this is more complex owing to the fast splitting of the primary stripes and the fast disappearance of the anterior stripes in Tc'hb^pRNAi embryos. Thus, there is no indication that Tc'hb is directly involved in regulating the anterior pair-rule stripes. However, it is evident that the regulation of the Hox genes and the setting of segmental boundaries have to be coupled by some mechanism, but it is still not known how this is achieved.

Short (or intermediate) germ embryogenesis is the ancestral form of embryogenesis in insects (Tautz et al.,1994). The hunchback function found in these types of embryos should be taken as a reference when considering conserved and diverged features. Interestingly, most details of hunchback function are fully comparable between Tribolium, Oncopeltus and Gryllus,although these insects belong to different orders that have a longer evolutionary separation time than, for example, beetles and flies(Savard et al., 2006b).

The two key features of hunchback function are clearly the regulation of Ultrabithorax, as well as the activation of Krüppel (Fig. 7A). These features are well documented in Drosophila and it seems now clear that they are ancestral. By contrast, the effect on Antp, Scr and giant appear to be partially indirect and partially not conserved, at least not with respect to the exact type of interaction.

Most intriguingly, the name-giving `gap' function does not belong to the conserved core elements of hunchback function, but appears to have evolved independently in Drosophila and Nasonia (e.g. long germ embryos). The term `gap gene' is therefore not appropriate for the hunchback function in most insects and appears also not appropriate for Krüppel and giant in Tribolium(Bucher and Klingler, 2004; Cerny et al., 2005). Thus, one could consider to revive another name that has been used to describe the gap genes, namely `cardinal genes'. Meinhardt(Meinhardt, 1986) has proposed this name in the context of his segmentation model for Drosophila. He proposed the existence of `cardinal regions' that would be set up by maternal gradients and the genes expressed in these regions would be required for regulating pair-rule expression. Interestingly, he concluded that this mechanism would only be required for long germ embryos, because the sequential segment formation in short germ embryos could be achieved by pair-rule gene interactions alone. However, Akam (Akam,1987) has then pointed out that gap genes regulate both segmentation genes and Hox genes in broader domains and that the term`cardinal genes' should reflect both of these aspects. Given that the Hox gene regulation appears to be the more conserved function of gap genes, it would indeed seem appropriate to adopt the term `cardinal genes' for this gene class, at least for other insects.

H.M. was a fellow of the International Graduate School of Genetics and Functional Genomics in Cologne. In addition, this work was supported by the SFB 572 and SFB 680.