Mutually repressive interactions between the gap genes giant and Krüppel define middle body regions of the Drosophila embryo

A central problem in the segmentation field is how relatively few, crudely distributed gap proteins generate organized sets of pair-rule and homeotic stripes in the early Drosophila embryo. The cascade of gene interac-tions that constitute the segmentation hierarchy result in a progressive refinement in the patterns of gene expression (reviewed by Ingham, 1988). This process transduces the broad gradients of maternal morphogens (reviewed by Niisslein-Volhard et al. 1987) into the highly refined patterns of the segment polarity genes, which are expressed within the limits of just single cells (reviewed by DiNardo and Heemskerk, 1990). Interac-tions between gap genes and pair-rule genes lead to the single most dramatic increase in the spatial complexity of gene expression that occurs during the course of early development.

Recent efforts to understand how gap genes specify stripes have centered on the pair-rule genes even-skipped (eve) (Harding et al. 1986; Macdonald et al. 1986; Frasch et al. 1987) and hairy (h) (Rushlow et al. 1989). Promoter fusion studies suggest that separate cis elements are responsible for the initiation of individual eve and hairy stripes (Goto et al. 1989; Harding et al. 1989; Hooper et al. 1989; Howard and Struhl, 1990; Pankratz et al. 1990). Proteins encoded by gap genes have been shown to bind specific sequences within the eve promoter (Stanojevič et al. 1989) and there is evidence that the same is true for the h promoter as well (Pankratz et al. 1990).

Meinhardt (1986) first proposed that borders between neighboring domains of gap gene expression define the sites where pair-rule stripes are initiated. However, a potential limitation of this model is that there are not enough gap borders to account for 7-stripe patterns of eve and h expression that are out of register with one another. Despite the fact that the gap genes hunchback (hb) (Tautz, 1988) and knirps (kni) (Pankratz et al.1989) are each expressed in two discrete domains, one in anterior regions and another in posterior regions, the three best characterized gap genes, hb, Krüppel (Kr) (Gaul et al. 1987) and kni, together define only 3 or 4 gap borders. The paradox of too few gap borders for so many stripes would be resolved by the demonstration that additional gap genes participate in this process. The segmentation gene tailless (tll) was recently shown to have the properties of a gap gene (Pignoni et al.1989), and here we provide evidence that giant (gt) (Mohler et al. 1989) is also a bona fide gap gene which interacts with the others to make pair-rule stripes.

Another mechanism for how relatively few gap genes might specify many stripes is suggested by recent immunolocalization studies, which have shown that gap proteins are expressed in broad, overlapping gradients (Stanojevič et al. 1989; Gaul and Jäckie, 1989; K. Howard, personal communication). This situation is in sharp contrast to the abutting zones of uniform expression originally envisioned by Meinhardt (1986, 1988) and supported by earlier studies (Knipple et al. 1985; Jäckie et al. 1986; Gaul and Jackie, 1987a; Tautz et al. 1987). Gradients of gap proteins raise the possibility of dose-dependent effects on pair-rule expression, while overlapping patterns suggest a combinatorial mode of regulation which could generate more stripes than the sum of the gap borders. Evidence that gap genes exert dose-dependent effects on eve and h expression stems from genetic studies (Hooper et al. 1989; Warrior and Levine, 1990). For example, it has been shown that the progressive loss of Kr⁺ gene activity causes a gradual expansion in the limits of the eve stripe 2 border. Just a 2-fold reduction in the level of Kr activity (in Kr⁺/Kr⁻ heterozygotes) causes a significant expansion in the limits of this stripe (Warrior and Levine, 1990). Various models have been proposed to account for the role of overlapping patterns of gap gene expression in the combinatorial regulation of pair-rule stripes (i.e. Carroll, 1990). Overlapping expression of the hb and Kr proteins have been implicated in the organization of eve stripes 2 and 3 (Stanojevič et al. 1989), while the overlapping Kr and kni patterns are thought to play an important role in the organization of h stripes 5 and 6 (Pankratz et al. 1990). Although not rigorously proven, the preponderance of the evidence to date suggests that striped patterns of pair-rule and homeotic genes depend on both the exact combinations and concentrations of gap proteins (Carroll and Scott, 1986; Ingham et al. 1986; White and Lehmann, 1986; Frasch and Levine, 1987; Harding and Levine, 1988; Gaul and Jäckle, 1989; Irish et al. 19896; Reinitz and Levine, 1990). It is, therefore, of considerable interest to understand how the gap genes come to be expressed in such overlapping domains.

Maternal factors have been shown to be responsible for the initiation of the gap genes (Driever et al. 1989; Struhl et al. 1989; Hülskamp et al. 1990), while the refinement of these patterns depends on cross-regulat-ory interactions among the gap genes (Jäckie et al. 1986; Pankratz et al. 1989). The maternal morphogen bicoid (bed) is critical for the activation of gap genes in anterior regions of the early embryo, while localized expression in posterior regions depends on nanos, which modulates the activity of maternal hb products (Hülskamp et al. 1989; Irish et al. 1989a; Struhl, 1989). The restriction of gap gene expression to central regions involves repression by torso, a maternal gene that is active at both poles (Gaul and Jäckie, 1987b; Tautz, 1988).

The initial gap patterns generated by these maternal factors are considerably broader than their limits of expression at the time when they are responsible for effecting pair-rule and homeotic stripes. For example, Kr transcripts are initially detectable in a region encompassing approx. 27 % of the total length of the egg in nuclear cycle 12 embryos (Gaul and Jäckie, 1989), but by cycle 14 these transcripts are restricted to just 20–22 % of the total egg length (Gaul et al. 1987; R. Kraut, unpublished observations), gt is expressed in two discrete domains along the A–P axis. The posterior pattern initially extends all the way to the posterior pole, from 30% to 0% egg length. However, ex-pression is rapidly repressed in the posterior-most regions and comes to be restricted to a narrower band, between 30 % to 17 % egg length (Mohler et al. 1989; see accompanying report by Kraut and Levine, 1991). There is evidence that the refinement of the Kr and gt patterns involves regulatory interactions with other gap genes (Jäckie et al. 1986; Mohler et al. 1989). However, in general, there is sporadic and even contradictory information regarding the role of cross-regulation in the segmentation process.

The first evidence for cross-regulation involved pairs of gap genes expressed in neighboring domains, such as hb–Kr and Kr–kni (Jäckie et al. 1986). For example, mutations in hb cause a slight expansion in the limits of the Kr pattern and vice versa. It is unclear whether these relatively weak interactions account for the refinement in the patterns of gap gene expression that occur during the course of normal development. Here we re-examine the question of cross-regulation by analyzing the effects of ectopically expressed gap products on the expression of other gap genes. We show that ectopic expression of the segmentation gene gt results in the rapid and selective repression of Kr. This result lends strong support to the view that gt is a bona fide gap gene which interacts with hb, Kr, and kni to specify stripes. Moreover, interactions between gt and Kr suggest that the strongest regulatory interactions occur among gap genes that are expressed in nonadja-cent domains. Further support for this proposal is provided by the analysis of gap gene expression in embryos that ectopically express hb and tll proteins.

Embryo collections, fixations and staining procedures were done exactly as described by Frasch et al. (1987). The polyclonal anti-gt, anti-Kr, anti-Kni, and anti-eve antibodies are all the same as those used in the accompanying report (Kraut and Levine, 1991). The histochemical stainings shown in Figs 7 and 8 were done using biotinylated secondary antibodies and DAB staining with horseradish peroxidase, as described in the ‘Elite’ ABC kit purchased from Vector laboratories (Burlingame, CA).

P-mediated germline transformation was done as originally described by Rubin and Spradling (1982). A full-length gt cDNA (see accompanying report for the details of its isolation) was placed downstream of a 315 bp DNA fragment spanning the hsp70 promoter (Kuziora and McGinnis, 1988). The cDNA includes 50 bp from the Xenopus beta-globin gene leader sequence (derived from the pNB40 cDNA library; see Brown and Kafatos, 1988) and 40 bp of the gt untranslated leader. The hsp70–gt fusion gene was cloned into the P-transformation vector, C20-X, which has a wild-type copy of the rosy gene as a selectable marker. The C20-X vector is a derivative of the Carnegie 20 vector described by Rubin and Spradling (1982). Injections were done with approximately 400 ng of P-transposon DNA along with approx. 100 ng of the pII25.7 helper P-element μl⁻¹ of injection buffer. Five independent transformed lines were obtained, and two of these were homozygosed.

For cuticle preparations, embryos were collected at room temperature for ⁠, aged for another and examined under Voltalef oil. Gastrulating or older embryos were discarded, while others were heat shocked on apple juice-agar collection plates on a pre-heated metal block in a 36° –37°C water incubator. Embryos were incubated at this temperature for 20 to 45 min, and then aged for at least 24 h at room temperature. We found that heat shocks of a total time of 30 min or longer, with a slow increase in temperature were the most effective in inducing severe phenotypes. Cuticle preparations were done as described by Wieschaus et al. (1984a,b). Embryos were rinsed with heptane to remove the Voltalef oil prior to dechorionating.

Embryos used for antibody staining were treated as described above, except that they were not monitored with Voltalef oil prior to heat shocking. The embryos were heat shocked for 30 min and allowed to recover for 30 to 45 min at room temperature prior to fixation.

The following fly stocks were used for these studies. rosy⁻ for P-transformation: rosy⁵⁰⁶; spliced (dominant gain-of-function torso mutation): torso^RL3 (Klingler et al. 1988; Strecker et al. 1989); Kr⁻: Kr⁹, a null point mutation (Wieschaus et al. 1984b). Embryos from hsp70–hb P-transformants, provided by Dr Gary Struhl, were collected from a homozygous line on the third chromosome (details described in Struhl, 1989). hb⁻, gt⁻ double mutants: hb^14F21, a null mutation (Lehmann and Nüsslein-Volhard, 1987) and gt^YA82, a strong ems-induced mutation (Wieschaus et al. 1984a); hb⁻, gt⁻ double mutants were identified among mixed progeny by staining with a mixture of anti-gt and anti-Kr antibodies. Hemizygotes for the gt^YA82 allele can be identified since they express much lower levels of the gt protein than wild-type. hb⁻/hb⁻ embryos can be identified by the anterior shift in Kr expression (not seen in gt⁻ alone). Double mutants were identified by both the lack of strong gt staining and the shift in the Kr pattern. Fluorescent-stained embryos were grouped into three classes: hb⁻ homozygotes; hb⁻, gt⁻double mutants; and trans hetero-zygotes or single hb⁻ heterozygotes (which are phenotypically wild-type). These were photographed, the numbers in each class were counted and the extent of Kr expression was measured as percentage egg length for each embryo. Figures for each group were averaged and the differences in Kr expression between groups was found to be significant according to the calculated standard deviation.

Polyclonal antibodies raised against a full-length gt protein made in E. coli were used for immunolocaliz-ation studies (for details, see accompanying report by Kraut and Levine, 1991). The gt protein is expressed in both anterior and posterior regions of early embryos (Fig. 1A, red stain). These sites of expression are consistent with previous RNA localization studies (Mohler et al. 1989) and correspond to the regions of the embryonic fate map that are disrupted in gt⁻ embryos. Such mutants lack pattern elements in the fifth through seventh abdominal segments and show deletions of head structures (Gergen and Wieschaus, 1986; Petschek et al. 1987; Mohler et al. 1989). Double immunolocalization studies with mixtures of anti-gt and anti-Kr antibodies show that the two proteins are expressed in complementary sets of cells, such that the two gt domains bracket the central Kr pattern (Fig. 1A). Although the gt and Kr expression patterns appear to directly straddle one another, no cells contain detectable levels of both proteins.

Genetic studies have shown that the initiation of gt expression depends on maternal factors, particularly bed and maternal hb proteins (see accompanying report by Kraut and Levine). However, the maintenance of the normal limits of gt expression depends on cross-regulatory interactions with other gap genes. For example, there is premature loss of gt expression in kni⁻ embryos. The most dramatic cross-regulatory interaction involves Kr. Both the anterior and posterior domains of gt expression expand into central regions of Kr⁻ embryos (Fig. IB), indicating that Kr exerts a negative effect on gt expression. This repression is important for inhibiting gt expression in central regions and restricting the two gt domains near the poles.

The Kr pattern is not strongly disrupted in gt⁻ embryos (Gaul and Jäckie, 19876, and data not shown). At first glance this result suggests that gt does not participate in the regulation of Kr expression. However, Kr is subject to very complex and redundant control (Gaul and Jäckie, 1987b; Jäckie et al. 1988; Lehmann and Frohnhöfer, 1989; see Discussion). Evidence that gt does indeed regulate Kr expression was obtained by creating a ‘dominant’ gt mutation using the hsp70 heat shock promoter to drive ectopic expression of the gt protein.

An hsp70–gt fusion gene was introduced into the germ line by P-transformation (see Materials and methods for details). The fusion gene that was used is shown in Fig. 2. It includes the entire gt coding sequence, as well as untranslated leader sequences from gt and the Xenopus beta-globin gene (Brown and Kafatos, 1988). Several independent transformed lines containing the hsp70–gt fusion gene were obtained, and all displayed the same mutant phenotype, but with varying degrees of severity, upon transient misexpres-sion of gt following heat shock. Immunolocalization studies with anti-gt antibodies showed that heat shock caused ubiquitous expression of the gt protein at low levels in all embryonic nuclei (data not shown). After heat shock, P-transformed embryos were allowed to grow at room temperature until the secretion of the cuticle. Typical cuticular phenotypes are presented in Fig. 3.

The most consistent defect is the deletion of pattern elements in the thorax and abdomen, including the second (T2) and third (T3) thoracic segments as well as the second (A2) and fourth (A4) abdominal segments (Fig. 3A and B). The most severe defects were obtained in homozygous lines containing two copies of the hsp70–gt fusion gene. In these cases most of the pattern elements in the thorax and anterior abdomen were deleted, resulting in embryos containing normal denticles for A6 through A8 but lacking most of the segments between the posterior head and A6 (Fig. 3C). This phenotype is very similar to that observed in Kr⁻ embryos (Wieschaus et al. 1984b) (compare Fig. 3C with D), although even the most severely affected hsp70–gt transformants retain some pattern elements in the anterior abdomen (Fig. 3C). Further evidence for a link between the hsp70–gt and Kr phenotypes is that less severely affected hsp70–gt embryos resemble weaker (i.e. hypomorphic) Kr mutants. For example, the first disruptions seen in the least severely affected hsp70–gt embryos center around T2–T3 and A4, similar to the weakest Kr mutants, which lose T2, T3, A2 and A4 (Wieschaus et al. 1984b).

The striking similarity of the cuticular defects resulting from ectopic expression of gt and those seen in Kr⁻ embryos suggest that gt selectively represses Kr and does not cause a general disruption of segmentation gene activities. For example, the normal head skeleton and tail structures observed in even the most severely affected hsp70–gt transformants suggest that the activi-ties of the gap genes hb (Lehmann and Nüsslein-Volhard, 1987) and tailless (tll) (Strecker et al. 1986) are not altered significantly by ubiquitous expression of the gt protein. The specificity of the gt–Kr interaction was confirmed by immunolocalization studies, using anti-bodies against the gap proteins hb and kni, and the pair-rule protein even-skipped (eve) to stain hsp70–gt transformants after heat induction.

There is a rapid loss of Kr protein following heat shock (compare Fig. 4A and B), while the hb and kni patterns appear essentially normal (data not shown). Specific repression of Kr is also indicated by the altered eve staining patterns that are observed (Fig. 4C). Instead of 7 periodic stripes, the eve pattern consists of only two stripes (1 and 7) and two broad bands in place of stripes 2 through 6. This pattern is very similar to the altered eve pattern observed in Kr⁻ mutants (Frasch and Levine, 1987) (Fig. 4D). Control experiments indicate that the altered Kr and eve patterns are a specific consequence of ectopically expressed gt pro-tein. Untransformed parental stocks (rosy⁵⁰⁶) were subjected to the same conditions of heat shock, but consistently showed normal patterns of Kr and eve expression (data not shown).

The above results indicate that complementary patterns of Kr and gt expression involve mutually repressive gene interactions. The two genes are expressed in alternating, not adjacent, domains (summarized in Fig. 7). hb intervenes between gt and Kr in anterior regions, while kni is expressed between Kr and gt in posterior regions. In order to examine the possibility that strong regulatory interactions occur among gap genes expressed in nonadjacent domains, we examined interactions between gt and tll as well as hb and kni.

Previous studies suggest that tll does indeed exert a negative effect on gt expression (Mohler et al. 1989; see accompanying report by Kraut and Levine). In tll⁻ embryos the posterior gt domain expands towards the posterior pole. The negative effect that tll exerts on gt could account for the normal refinement of the gt pattern seen in wild-type embryos. Upon its initiation the posterior gt domain normally extends to the posterior pole, but soon thereafter gt expression is rapidly lost in the posterior-most regions, tll has been cloned and characterized and recently shown to be expressed from 0 % to 15 % egg length (where 0 % =the posterior pole and 100 % =the anterior pole) (Pignoni et al. 1990). The posterior tll domain closely coincides with the region where gt expression is lost during the course of normal development.

Additional support for a strong regulatory interaction between gt and tll stems from studies on the pattern of gt expression in embryos where tll is active in ectopic regions. As discussed above, tll expression is normally restricted to the poles where it is activated by the maternal torso (tor) gene (Klingler, 1989; Strecker et al. 1989). tor encodes a transmembrane tyrosine kinase that is ubiquitously distributed throughout the embryo, but is activated only at the poles where it somehow directs the initiation of tll expression (Klingler et al. 1988; Casanova and Struhl, 1989). Several different dominant gain-of-function tor mutations have been identified, including a temperature-sensitive mutation, tor^RL3, also called spliced (Schüpbach and Wieschaus, 1989). These are thought to cause constitutive acti-vation of the tor tyrosine kinase, resulting in ubiquitous expression of tll (Klingler, 1989; Strecker et al. 1989). The consequences of such misexpression are quite severe, resulting in embryos lacking virtually all middle body segments. Previous studies have shown that this mutant phenotype results from the repression of Kr in middle body regions (Klingler, 1989) coupled with an expansion in the limits of hb expression into central regions (Warrior and Levine, 1990). Ectopic tll activity also causes a severe reduction in the posterior gt pattern, as shown in Fig. 6. This repression is either due to ectopically expressed tll products, or is an indirect consequence resulting from the expanded hb pattern (see below). Either way, this result establishes a strong regulatory interaction between gt and a second gap gene, tll that is normally expressed in a nonadjacent region.

Previous studies have shown that ectopic expression of the hb protein in early embryos results in a phenotype similar to that observed in nanos mutants (Hülskamp et al. 1989; Irish et al. 1989a; Struhl, 1989). This observation was taken as evidence that nanos organizes the posterior pattern by repressing the activity of maternal hb products, thereby restricting hb function to anterior regions of early embryos. Both the kni and gt posterior patterns are severely reduced or absent in nanos⁻ embryos (see accompanying report by Kraut and Levine), raising the possibility that hb is a strong repressor of these genes. To test this we examined the expression of kni and gt in P-transformed lines carrying an hsp70–hb fusion gene (Struhl, 1989).

Heat shock results in strong ubiquitous expression of the hb protein (Fig. 5A), as was shown previously (Strùhl, 1989). Both the posterior kni and gt patterns are virtually abolished in these embryos (Fig. 5B and C). This result provides additional evidence that the strongest cross-regulatory interactions occur among gap genes expressed in nonadjacent domains (see Fig. 7 summary). In marked contrast to the effect that ectopic hb expression has on kni and gt, Kr expression remains quite strong (Fig. 5D). Thus, hb has little or no repressive effect on Kr expression, indicating that regulatory interactions between these neighboring genes are quite weak. The severe posterior expansion of the Kr pattern seen in hsp70–hb embryos (Fig. 5D) is probably due to the loss of both kni and gt products, which normally repress Kr expression (Jäckie et al. 1986; Fig. 4B).

The ectopic expression studies and mutant analysis described above indicate that gt is a strong repressor of Kr. However, the results of these studies are paradoxi-cal in that the Kr expression pattern is not disrupted in gt⁻ embryos until after cellularization (Gaul and Jäckie, 1987b; Reinitz and Levine, unpublished observation). Such mutants show only a very mild anterior and posterior expansion of the Kr pattern during gastru-lation (data not shown). One possible explanation for this apparent inconsistency is that gt functions either redundantly or cooperatively with hb and kni to define the anterior and posterior limits of the Kr pattern, respectively (see Fig. 7 summary). If repression of Kr is achieved through cooperative interactions between hb and gt and/or gt and kni, then the expansion of the Kr pattern should be greater when both products are removed in double mutants than in single mutants. To test this possibility we examined the Kr staining pattern in hb⁻, gt⁻ double mutants (Fig. 8). Previous studies have shown that there is an anterior expansion of the Kr pattern in hb⁻ embryos (Jäckie et al. 1986; Harding and Levine, 1988). As indicated above, gt mutants have little or no effect on the Kr pattern prior to cellularization. In hb⁻, gt⁻ double mutants there is a dramatic anterior expansion of the Kr pattern in pre-cellular embryos, which is significantly greater than the broadening observed in hb⁻ or gt⁻ single mutants (Fig. 8A).

The expanded Kr pattern in hb⁻, gt⁻ double mutants is similar to that observed in bcd⁻ embryos (Fig. 8B, compare with A). This expansion of Kr observed in bed mutants prompted the proposal that high concen-trations of the bed morphogen repress Kr, thereby restricting it to central regions (Gaul and Jäckie, 1987b; Hülskamp et al. 1990). However, our results suggest that bed may not directly repress Kr, but instead might exert a negative effect on Kr expression via the hb and gt repressors. The anterior hb and gt patterns are activated by bed, and both are absent in bcd⁻ embryos (Kraut and Levine, 1991), resulting in a situation similar to that seen in hb⁻, gt⁻ double mutants.

We have shown that the complementary patterns of gt and Kr expression depend on mutually repressive gene interactions. The demonstration that gt selectively represses Kr provides strong evidence that gt is a bona fide gap gene which participates with hb, Kr, and kni to initiate striped patterns of gene expression in the early embryo. Previous studies on the regulation of gap gene expression have focused primarily on interactions between gap genes expressed in neighboring domains (Jackie et al. 1986; Pankratz et al. 1989). An important implication of the current study is that the strongest regulatory interactions occur among gap genes ex-pressed in nonadjacent domains. In addition to interac-tions between Kr and gt, we have shown that hb strongly represses the expression of kni and gt. Finally, studies on double mutants suggest that gt functions cooperatively with hb to restrict Kr expression to central regions of the early embryo.

There has been considerable debate regarding the role of gt in the segmentation process. Hesitation to include it among the ranks of the gap genes stems from the comparatively mild cuticular disruptions seen in gt⁻ mutants (Wieschaus et al. 1984a; Petschek et al. 1987; Mohler et al. 1989). The head defects and partial fusions of abdominal segments A5–A7 observed for gt are not nearly as dramatic as the altered segmentation patterns seen in hb⁻, Kr⁻, or kni⁻ embryos (Nüsslein-Volhard and Wieschaus, 1980; Wieschaus et al. 1984b; Lehmann and Nüsslein-Volhard, 1987; Nauber et al. 1988). In the case of Kr, there is a deletion of the entire thorax and first five abdominal segments along with a mirror-image duplication of A6 (Wieschaus et al. 1984b). Further-more, gt mutants exert a relatively weak effect on the initiation of pair-rule stripes (Carroll and Scott, 1986; Frasch and Levine, 1987; K. Howard, personal com-munication). hb, Kr and kni mutations result in severe disruptions in the initiation of the 7–stripe eve pattern, with complete fusions or deletions of at least three of the stripes. In contrast, gt⁻ embryos show partial fusions of eve stripes 1 and 2, as well as 5 and 6 (Frasch and Levine, 1987).

The demonstration that ectopic expression of gt causes the rapid and specific repression of Kr provides strong evidence that gt is a bona fide gap gene, which participates with hb, Kr, and kni to generate pair-rule and homeotic stripes. The hsp70 promoter has been used quite extensively to drive ectopic expression of a variety of segmentation genes, including the gap gene hb (Struhl, 1989), the pair-rule genes ftz (Struhl, 1985), eve (G. Struhl and M. Levine, unpublished results) and hairy (Ish-Horowicz and Pinchin, 1987) and the segment polarity genes engrailed (Poole and Kornberg, 1988) and wingless (P. Lawrence, personal communi-cation). In no case does the misexpression of a member of one segmentation gene class lead to a mutant phenotype characteristic of another class. For example, misexpression of hb causes a gap mutant phenotype, while misexpression of ftz or hairy cause pair-rule phenotypes. Thus, an important criterion for the classification of gt as a gap gene is that it influences the expression of other gap genes. This is fulfilled by the demonstration that ectopic expression of gt leads to a Kr⁻ phenotype.

The comparatively mild gt phenotype may be due to the existence of one or more ‘gt-like’ genes that partially duplicate gt function. This possibility is suggested by the nature of the gt protein. It contains a b type leucine zipper (E. Eldon and V. Pirrotta, personal communi-cation) that mediates a sequence-specific DNA binding activity (T. Hoey and R. Kraut, unpublished results). Leucine zipper proteins have been shown to bind DNA as obligate dimers (reviewed by Vinson et al. 1989). In many instances such proteins function as heterodimers, as has been shown in detail for the mammalian proto-oncogenes fos and jun (Nakabeppu and Nathans, 1989; Gentz et al. 1989; Turner and Tjian, 1989). Perhaps a second leucine zipper protein active in the head partly duplicates gt activity. The removal of both gt and this putative gt-like gene might be expected to cause a far more severe phenotype than that observed in gt⁻ mutants. Support for the existence of a second giant-like gene stems from recent aneuploidy studies, which have permitted the systematic search for novel gap gene activities (Vavra and Carroll, 1989). These studies revealed a potentially novel gap gene on the left arm of the second chromosome that is similar to gt with respect to the regulatory effect it exerts on the initiation of the 7-stripe eve pattern.

It is quite possible that regulatory interactions between gt and Kr are direct and occur at the level of transcription. Both the gt and Kr proteins possess sequence-specific DNA binding activities (T. Hoey and R. Kraut, unpublished results; Stanojevič et al. 1989; Treisman and Desplans, 1989). As mentioned above, gt encodes a leucine zipper protein, while Kr contains multiple copies of the zinc finger motif (Rosenberg et al. 1986). In Kr⁻ embryos the two gt domains expand towards central regions of the embryo, which normally contain high levels of the Kr repressor. The most straight-forward interpretation of this result is that the Kr protein binds to specific sites within the gt promoter to repress its expression in central regions of wild-type embryos. Consistent with this possibility is the recent demonstration that the Kr protein functions as a transcriptional repressor in transient cotransfection assays (Licht et al. 1990).

A potential problem with this model of direct repression is the finding that the expansion of the gt pattern in Kr⁻ is relatively mild (see Fig. IB) and does not include central regions containing peak levels of the Kr protein in wild-type embryos. Why is there not a more dramatic expansion in the limits of gt expression in Kr⁻ embryos? This could be explained by spatial constraints on the distribution of the gt activators. The accompanying report by Kraut and Levine (1991) suggests that the maternal morphogen bed activates gt expression in anterior regions. The bed protein is distributed in a broad gradient along the anterior-posterior axis, with peak levels at the anterior pole (Driever and Nüsslein-Volhard, 1988). This gradient is not altered in Kr⁻ embryos, and the limited posterior expansion of the gt pattern could reflect the need for a certain minimal threshold level of bed protein to trigger gt expression. In central regions, there might not be a sufficient level of bed to activate gt even in the absence of the Kr repressor. Similar arguments suggest that the posterior gt pattern depends on a localized activator emanating from the posterior pole, but its identity is currently unknown.

The restriction of Kr products to central regions of the early embryo involves multiple tiers of repression. One set of repressors emanate from the anterior and posterior poles, under the control of maternal tor activity (Gaul and Jäckie, 19876). Dominant gain-of-function tor mutations abolish the central Kr pattern, and severely disrupt segmentation (Klingler, 1989). This effect is probably due to ectopic activation of tll throughout the embryo, since the removal of tll in a tor gain-of-function mutant background (i.e. spliced⁻,tll⁻ double mutants) rescues segmentation (Klingler, 1989; Strecker et al. 1989). While not directly tested, it is likely that the central Kr expression pattern is fully restored in such double mutants.

An implication of the current study is that Kr is subject to a second level of repression which is completely independent of tor. The second set of repressors are controlled by the maternal factors bed and nanos, bed activates gt and hb in anterior regions of the embryo (Driever et al. 1989; Struhl et al. 1989), where they help define the anterior margin of the central Kr pattern, nanos modulates the activity of maternal hb (Hülskamp et al. 1989; Irish et al. 1989a; Struhl, 1989), which participates in the localization of kni and gt expression within the presumptive abdomen (Hülskamp et al. 1990; Kraut and Levine, ac-companying report), kni and gt set the posterior limit of the central Kr pattern (see summary of gap patterns in Fig. 7). Kr expression is not significantly expanded in tor⁻ embryos due to the independent action of the bed, nanos system (Jäckie et al. 1988). In bcd⁻ or nos⁻ embryos there are only partial expansions of Kr towards the poles since tor still functions to activate the til repressor (Gaul and Jäckie, 1987b; Lehmann and Frohnhöfer, 1989). Redundant repressors are likely to ensure the accurate and reproducible localization of Kr to central regions of the embryo.

It is possible that gt functions in a cooperative manner with hb and kni to repress Kr expression. For example, the dramatic anterior expansion of the Kr pattern seen in hb⁻, gt⁻ double mutants (Fig. 8A) raises the possibility that optimal repression of Kr involves hb–gt protein-protein interactions. Consistent with this possi-bility is the observation that even the most severely affected hsp70–gt transformants retain a remnant of the Kr pattern, which corresponds to the peak of the normal Kr domain in parasegment (PS6) (Stanojevič et al. 1989) (see Fig. 4B). Previous immunolocalization studies have shown that the anterior hb domain extends through PS5, but does not include PS6 (Tautz, 1988; Stanojevič et al. 1989; see summary in Fig. 7). Thus, it would appear that ectopic gt proteins repress Kr only in those nuclei that also contain at least low levels of hb. Similarly, the kni expression pattern extends anteriorly through PS7 but does not include PS6 (K. Howard, and own observations), suggesting that gt might act in concert with kni to repress the posterior portion of the central Kr domain and leave PS6 expression intact. This idea is supported by the dramatic posterior expansion of Kr that we observe in hsp70–hb embryos, concomi-tant with the loss of both gt and kni (see Fig. 5).

Such a combinatorial mechanism for Kr repression suggests that the bell-shaped distribution profile of the Kr protein (see Fig. 7) is a direct consequence of the action of the maternal organizing centers. For example, hb and gt appear to be activated by different thresholds of the bed gradient (Kraut and Levine, 1991). Since hb is triggered by a lower threshold than gt, the hb pattern extends at least 8 to 10 cells posterior to the limit of gt. Since hb is only a weak repressor it reduces, but does not abolish, Kr expression in those cells containing little or no gt. A precipitous reduction in Kr occurs in cells containing at least low levels of both repressors, culminating with a sharp border formed by cells containing high levels of both proteins.

The specification of cell fate in the early embryo appears to depend on both the exact concentrations and combinations of gap proteins (Gaul and Jäckie, 1989; Lehmann and Frohnhöfer, 1989; Stanojevič et al. 1989; Warrior and Levine, 1990). Here we have presented evidence that strong regulatory interactions between gap genes expressed in nonadjacent domains are critical for establishing the precise patterns of their expression. The original studies on cross-regulatory interactions involved hb and Kr as well as Kr and kni (Jäckie et al. 1986), which are expressed in adjacent domains (see Fig. 7 summary). These interactions are relatively weak in that only partial expansions of the Kr pattern are observed in either hb⁻ or kni⁻ embryos and misexpres-sion of hb does not reduce Kr levels. In contrast, interactions between hb & kni and gt & Kr, which are expressed in alternating domains, are quite strong. Ectopic expression of hb in early embryos causes a failure to initiate kni and gt expression, but has virtually no effect on the neighboring Kr pattern. Similarly, misexpression of gt results in the strong repression of Kr, but has no discernible effects on the neighboring hb and kni patterns. These results suggest that the limits of gap gene expression depend most critically on regulat-ory interactions between genes expressed in nonadja-cent domains. Weak interactions between neighboring gap genes coupled with strong interactions between nonadjacent genes could account for the establishment of precisely balanced, overlapping gradients of gap gene expression.

We thank Gary Struhl for HShb stocks, Martin Klingler for tor gain of function mutant stocks, Chris Rushlow for Kr antibody, Ken Howard for kni antibody, Jym Mohler for gt^YA82 stock and help with analyzing cuticles, Ulricke Gaul for the Kr⁹ stock, and Elizabeth Eldon and Vincent Pirrotta for discussing unpublished results. This work was supported by a grant from the NIH (GM 34431).