ABSTRACT
caudal (cad) is a maternally and zygotically expressed gene in Drosophila whereby the two phases of expression can functionally replace each other. The zygotic expression forms an abdominal and a posterior domain, whereby only the posterior domain has so far been studied with respect to its regulation and function. We show here that the abdominal cad domain is regulated by the hunchback (hb) gradient through repression at high concentrations and activation at low concentrations of HB protein. To study the function of the abdominal cad domain in the absence of redundant interactions, we have utilized an experimental system in which the embryo lacks the normal bicoid (bcd) and hb expression. An artificial hb gradient is then intro-duced into such embryos, which results in an induction of an ectopic zygotic cad domain in the more anterior region. Employing this system, we show that the cad domain functions by activating the expression of the abdominal gap genes knirps (kni) and giant (gt). We conclude that cad is the so far missing region-specific activator of abdominal segmentation genes.
INTRODUCTION
Drosophila early development proceeds in a hierarchical manner in which the genes of one level of the hierarchy interact with each other and then regulate the genes of the next level in the hierarchy (St. Johnston and Nüsslein-Volhard, 1992; Pankratz and Jäckle, 1993). Several of these regulatory circuits are redundantly controlled either by different genes, or by different phases of expression of the same gene. The so far best studied example in this respect is the gap gene hb, which is expressed maternally and zygotically. The two phases of expression are under separate control of the anterior and posterior maternal systems (Tautz, 1988) and have partially redundant functions. The maternal expression of hb results in a protein gradient in the abdominal region of the embryo, which serves to regulate other gap genes in a concentration-dependent manner (Hülskamp et al., 1990; Struhl et al., 1992). However, essentially the same gradient function can be provided by the zygotic expression alone and the maternal expression is therefore dispensable under appropriate experimental conditions (Lehmann and Nüsslein-Volhard, 1987). Furthermore, some of the functions of the hb gradient are also provided by the bcd gradient (Hülskamp et al., 1990; Hoch et al., 1991; Struhl et al., 1992) and there is, in fact, a synergism between bcd and hb with respect to certain regulatory functions (Simpson-Brose et al., 1994). However, the hb gradient is also capable of regulating its target genes independently of bcd (Schulz and Tautz, 1994).
cad, as hb, has a maternal and a zygotic phase of expression. The maternal RNA is at first homogeneously distributed, but is under translational control from the anterior maternal system (Macdonald and Struhl, 1986; Mlodzik and Gehring, 1987a,b). The maternally derived CAD protein becomes thus distributed in a posterior-anterior concentration gradient. Zygotic cad expression at the blastoderm stage is often described as occurring only in a small posterior domain, which forms a 3-to 4-cell wide stripe. However, Macdonald and Struhl (1986) noticed that embryos devoid of maternal cad expression show an abdominal expression of cad, which is usually obscured by the maternal CAD protein gradient. We show here that this domain can be more easily visualized by whole-mount in situ hybridization. This allows direct study of its regulation by other genes, without the necessity to remove the maternal expression of cad.
Embryos maternally mutant for cad are viable, though they show occasional abdominal segmentation defects. Embryos zygotically mutant for cad are not viable, though this is not because of segmentation defects, but due to later functions of cad (Macdonald and Struhl, 1986). Only embryos lacking both maternal and zygotic expression of cad show severe segmen-tation defects (Macdonald and Struhl, 1986). This indicates that, as with hb, the maternal and zygotic expression of cad are redundant at least with respect to the segmentation function. Nonetheless, it is evident that cad does play an important role in the segmentation of the embryo, though its place in the genetic hierarchy was so far unclear.
We show here that the abdominal cad expression domain is under the control of the hb gradient. It is activated at low concentrations of hb and repressed at high concentrations. We show also that the abdominal cad domain itself is then required to activate the gap genes kni and gt.
MATERIALS AND METHODS
The construction and the expression pattern of the hb-bcd3′UTR construct are described in Schulz and Tautz (1994). In short, the hb-bcd3′UTR construct carries the maternal hb promoter, the full hb coding sequence and part of the bcd 3′-end. The 3′-end was chosen such that it causes the anterior localization of the hb RNA, but does not include the nanos response elements (NRE; Wharton and Struhl, 1991). It is therefore not under the regulatory control of nos. HB protein is expressed as an anterior-posterior gradient from this construct which has a shape similar to that described for bcd (Driever and Nüsslein-Volhard, 1988). The nos-bcd3′UTR construct is described in Gavis and Lehmann (1992) and utilizes also the bcd 3′-end for anterior localization. The hb-bcd3′UTR construct has no phe-notypic effects on wild-type flies, while the nos-bcd3′UTR construct acts as a dominant maternal sterile. However, it can be propagated paternally and can thus be used in certain mating schemes.
The mutant genotypes of the embryos could usually be assessed unequivocally on the basis of the expression patterns of the genes studied. This was in some cases facilitated by double labelling (see text). Finally, where necessary, we made sure that the staining patterns that were allocated to certain genotypes did occur in the expected fre-quencies predicted from the mating schemes.
cad2 and tsl691 alleles were used in the different crosses. The regu-lation of cad was further studied by in situ hybridization in nosL7; bcdE1, hb14F and bcdE1, tslo35 alleles.
During the course of these and other experiments (Schulz, 1995), we noticed that the suppression of bcd in the embryos carrying the nos-bcd3′UTR construct is not complete. There is still a residual bcd activity at the most anterior tip of the embryos, which becomes apparent in the combinations with tsl. Since the repression effect of the terminal system on the gap genes is removed under these con-ditions, a weak activating effect of the residual bcd activity on hb becomes effective and a small anterior hb domain, accompanied by a small Kr and cad domain, is expressed (data not shown). Thus, the respective experiments with the tsl combinations were all done in a double mutant combination with bcd, fully to exclude potentially interfering effects from bcd.
Whole-mount in situ hybridizations were basically done as described in Tautz and Pfeifle (1989), but using the modifications for RNA probes described in Klingler and Gergen (1993).
RESULTS
Zygotic cad expression
Employing whole-mount in situ hybridization instead of antibody staining, it is possible to resolve the zygotic cad expression domain in the abdominal region of the embryo (Fig. 1A). Its first onset of expression appears to occur at stage 13, though this is difficult to determine exactly since the maternal cad RNA is still present in this region at earlier stages. However, given that the maternal and zygotic expression of cad can functionally replace each other (Macdonald and Struhl, 1986), it is possible that the expression appears already prior to stage 13. The domain is under the control of the anterior maternal system, since it shifts anteriorly in embryos zygoti-cally mutant for hb (Fig. 1B) and maternally mutant for bcd (Fig. 1C). In contrast to the posterior cad domain (Mlodzik and Gehring, 1987b), it is not under the control of the terminal system, as it is still present in embryos mutant for tsl (Fig. 2A) and embryos double mutant for bcd and tsl (Fig. 1D). Finally, the domain disappears in embryos maternally mutant for the posterior control gene nos (Fig. 1E), suggesting a negative reg-ulatory effect of hb (see below). These embryos retain only the posterior domain and a dorsal expression in the abdominal region, which is of as yet unknown function.
hb control of the abdominal cad domain
To understand the regulation of the abdominal cad domain further, we have utilized a construct that leads to an artificial hb gradient in the anterior third of the embryo (the hb-bcd3′UTR construct; Schulz and Tautz, 1994). This construct can be crossed into different types of mutant backgrounds and permits study of the effects of the hb gradient in the absence of other potentially interfering genetic functions. To create embryos that are devoid of such interfering functions, in particular those of the normal maternal hb expression as well as of bcd expression, we have used another artificial construct. This construct places nos at the anterior pole of the embryo (the nos-bcd3′UTR construct; Gavis and Lehmann, 1992). nos functions by inhibiting the translation of both bcd and hb RNA (Wharton and Struhl, 1991; Gavis and Lehmann, 1992). This function is mediated by short elements in the untranslated 3′-end of the mRNAs of these two genes, the nanos response elements (NRE) (Wharton and Struhl, 1991). The NREs were excluded from both the nos-bcd3′UTR construct and the hb-bcd3′UTR construct making them insensitive to the transla-tional regulation by nos.
When the nos-bcd3′UTR construct is crossed into embryos that are maternally mutant for tsl, one can create embryos that are devoid of all known regulatory activities that are required for anterior-posterior segmentation. The abdominal cad domain is indeed absent in these embryos (Fig. 2B). If one now introduces the hb-bcd3′UTR construct into such an embryo, one finds that a cad domain appears again, albeit at a much more anterior position (Fig. 2C), consistent with the fact that the hb gradient is now located more anteriorly (Schulz and Tautz, 1994). This result indicates that hb acts both as an activator and as a repressor on this cad domain. The activation effect is evident from the fact that the cad domain depends on the presence of hb in this background. The repression effect can be inferred from the fact that cad is not expressed in the most anterior portion of these embryos, where hb is present at high concentrations. Furthermore, repression of the abdominal cad domain by hb can also be inferred from the results in nos mutant background described above (Fig. 1E). Under these conditions, the maternal HB protein is present at relatively high levels in the abdominal region of the embryo and can therefore repress its target genes in this region (Tautz, 1988; Wang and Lehmann, 1991).
Phenotypic effects
Placing of the hb-bcd3′UTR construct into embryos carrying the nos-bcd3′UTR construct leads to a full restoration of the abdominal segment pattern (Schulz and Tautz, 1994; Fig. 3). We have previously shown that this is due to proper regulation of the more posterior gap genes by the artificial hb gradient (Schulz and Tautz, 1994). Interestingly, the removal of zygotic cad from these embryos results in a severe disruption of the abdominal segmentation pattern (Fig. 3C). This indicates that the zygotic cad domain has a direct role in abdominal segment formation, at least under these conditions. We note that the maternal cad expression is still present in these embryos. However, in contrast to the normal situation (Macdonald and Struhl, 1986), it is apparently not capable of rescuing the lack of the zygotic cad expression. This incongruence might be due to the fact that bcd is also not present in the nos-bcd3′UTR construct embryos. Thus it seems possible that bcd might also contribute to the specific cad functions in the abdomen. There is indeed molecular evidence for such a synergistic interaction between cad and bcd (H. Jäckle, personal communication). However, a direct test by analysing the phenotype of embryos that are maternally mutant for bcd and zygotically mutant for cad does not support this inference, since the bcd phenotype is not enhanced in the absence of cad (results not shown). However, the two situations are not directly comparable, since the maternal hb gradient that is present in bcd mutant embryos provides a different concentration range and thus a somewhat different set of regulatory interactions to the artificial hb gradient in the embryos carrying the nos-bcd3′UTR construct. This effect was also noted previously with respect to the rescue function of the hb-bcd3′UTR construct in the bcd− background (Schulz and Tautz, 1994).
cad as activator for kni and gt
To track the molecular basis of the phenotypic defects seen in the above experiments, we have analysed the expression patterns of kni and gt in the respective mutant backgrounds. To facilitate the identification of the genotypes and to provide internal standards for quantitative comparisons, we have done simultaneous in situ hybridizations with two different probes. In the case of kni, we have used hb and, in the case of gt, we have used tll as a second probe.
Embryos carrying only the nos-bcd3′UTR construct show a duplication of the posterior hb stripe at late blastoderm stage (Fig. 4A). kni is only very weakly expressed in this situation, since it requires the presence of Kr as an additional enhancer of expression (Pankratz et al., 1989). Kr, however, is absent in these embryos (Gavis and Lehmann, 1992). Bringing the hb-bcd3′UTR construct into this background restores Kr expression (Schulz and Tautz, 1994) and concomitantly estab-lishes a kni domain (Fig. 4B). This domain becomes signifi-cantly weaker in embryos lacking zygotic cad expression (Fig. 4D), indicating that cad acts as an activator of kni.
gt is expressed in a broad central domain in the nos-bcd3′UTR construct embryos (Fig. 5A), but becomes somewhat repressed from the anterior region in the presence of the hb-bcd3′UTR construct (Fig. 5B). It was previously shown that hb acts as a concentration-dependent repressor of gt (Eldon and Pirotta, 1991; Kraut and Levine, 1991a; Struhl et al., 1992) and one would therefore have expected a much stronger effect on the gt expression, when the hb gradient is introduced. However, the fact that gt is still weakly present in the anterior region suggests that hb might have ectopically activated an activator of gt which then counteracts the repression effect of hb on gt (compare Schulz and Tautz, 1994). It becomes evident that this activator is cad, since genetic removal of zygotic cad expression removes the anterior gt expression (Fig. 5C).
Our data include some evidence that the activation effect of cad on gt may be dosage sensitive. Embryos heterozygous for zygotic cad show a weakening of the expression of gt in the anterior region (not shown). This reduction in gt expression becomes particularly evident when the embryos are stained with the kni probe, which is an indirect indicator of gt function. It was previously shown that high levels of gt act as a repressor on kni and thus set its posterior border (Eldon and Pirotta, 1991). The posterior border of the kni domain seen in Fig. 4B is controlled by this effect. Accordingly, weakening of gt expression by removing one zygotic copy of cad has an effect on this border. It shifts posteriorly and becomes fuzzy in embryos heterozygous for cad (Fig. 4C). In contrast, the anterior part of the kni domain appears to be expressed at the same level as with two copies of cad in the embryo (compare Fig. 4B). This indicates that the activation effect of cad on kni can be achieved with lower levels than the activation of gt by cad.
It should be noted that, although our experiments were designed to study the function of only the zygotic cad expression domain, it seems likely that the maternal expression of cad, which is still present in the embryos described above, has the same function. This could explain why both kni and gt are expressed even in the absence of zygotic cad.
DISCUSSION
We have established that cad acts very high up in the genetic segmentation gene hierarchy. We found that the maternal hb gradient regulates a zygotic cad domain and that one can manipulate this regulation such that the function of cad as an activator of the abdominal gap genes kni and gt becomes apparent. Previously, this function of cad was concealed by redundant regulatory effects, both of its own maternal expression, but also by partial compensation through the other maternal systems. In our experiments, we were able to remove the interference from these systems and have thus obtained a clearer picture of the functional role of cad.
Region-specific activators for the abdominal kni and gt expression domains have not been found previously. It was therefore thought that they are under the control of a general activator and that their expression domains are only delimited by repression effects (Kraut and Levine, 1991a,b; Eldon and Pirotta, 1991; Pankratz et al., 1992). The situation for kni is, however, more complicated, since Kr acts as an enhancer of kni expression (Pankratz et al., 1989). In contrast, the activator for the weak primary expression of kni could not be identified (Pankratz et al., 1992). Our results suggest now that the abdominal cad expression domain acts as a specific activator for both kni and gt and that this activation is region specific. It has previously been suggested that cad acts also as a specific activator for the pair-rule gene fushi tarazu in the abdominal region (Dearolf et al., 1989).
Our experiments deal only with the regulatory effects of the zygotic abdominal cad domain. However, it is likely that the same function is provided by the maternal expression. The acti-vation of the posterior gap genes is required fairly early during development, at a time when the zygotic cad domain may not yet be fully established. However, cad is maternally present in the respective region, prior to the activation of the zygotic domain. It is therefore likely that the maternal CAD protein functions as the first activator of the abdominal gap genes and that the zygotic domain takes over this function when the maternal protein is degraded. This is probably the reason why the zygotic domain can fully replace the maternal cad function when it is absent (Macdonald and Struhl, 1986).
The morphogenetic capacities of the hb gradient are further emphasized by our experiments. hb has previously been shown to be a concentration-dependent activator and repressor of Kr (Hülskamp et al., 1990; Struhl et al., 1992; Schulz and Tautz, 1994). The same finding can now be extended to the abdominal zygotic cad expression domain. Given that hb also acts as a concentration-dependent repressor on the anterior borders of kni and gt expression (Hülskamp et al., 1990; Struhl et al., 1992), we have to conclude that at least six different threshold concentrations are read from the hb gradient. Though we have shown that hb can autonomously achieve this regulation (Schulz and Tautz, 1994), it is also clear that the embryos are much more vulnerable to regulatory side effects when they are under the sole control of the hb gradient. It is therefore not sur-prising that most regulatory effects of the hb gradient are redundantly provided by other regulatory circuits. It appears that the embryo uses, on the one hand, the maximum of spatial information that is contained in a morphogenetic gradient, while, on the other hand, safeguards itself against a potential misinterpretation of this information by utilizing redundant pathways (Tautz, 1992). Our experimental system for studying the regulatory interactions between genes in the absence of other pattern-forming systems provides a solution to this redundancy problem. In principle, it should be possible to study the function of the other gap genes in a similar way to analyse the regulatory capacities of the morphogenetic gradients that they are supposed to provide (Hülskamp and Tautz, 1991; Pankratz and Jäckle, 1993).
ACKNOWLEDGEMENTS
We thank Gary Struhl for the cad fly stock, Walter Gehring for cad DNA, Ruth Lehmann for the flies carrying the nos-bcd3′UTR construct and Herbert Jäckle for communicating results prior to pub-lication. Furthermore we are indebted to Gerd Jürgens, Wolfgang Lukowitz, Irene Rambold and the members of the laboratory for help with various aspects of this work. We thank Charles David and Wolfgang Lukowitz for critically reading the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft, DFG, grant Ta99/5.