We describe the regulated expression of the segmen-tation gene giant (gt) during early embryogenesis. The gt protein is expressed in two broad gradients in precellu-lar embryos, one in anterior regions and the other in posterior regions. Double immunolocalization studies show that the gt patterns overlap with protein gradients specified by the gap genes hunchback (hb) and knirps (kni). Analysis of all known gap mutants, as well as mutations that disrupt each of the maternal organizing centers, indicate that maternal factors are responsible for initiating gt expression, while gap genes participate in the subsequent refinement of the pattern. The maternal morphogen bicoid (bcd) initiates the anterior gt pattern, while nanos (nos) plays a role in the posterior pattern. Gene dosage studies indicate that different thresholds of the bcd gradient might trigger hb and gt expression, resulting in overlapping but noncoincident patterns of expression. We also present evidence that different concentrations of hb protein are instructive in defining the limits of kni and gt expression within the presumptive abdomen. These results suggest that gt is a bona fide gap gene, which acts with hb, Krüppel and kni to initiate striped patterns of gene expression in the early embryo.
Maternal morphogen gradients have long been postu-lated as the primary source of positional information in insect embryos. This idea was originally put forth by Seidel (see review by French, 1988) who introduced the notion that maternal ‘organizing centers’ exist at both poles of the egg and generate pattern in the embryo. More recently, Wolpert (1969) proposed the ‘French flag model’, whereby pattern could be generated by differential responses of cells to graded concentrations of morphogens. Early support for the presence of morphogens emanating from either pole of the insect egg involved experimental manipulations such as ligature and cytoplasmic transplantation (Sander, 1976). Recent genetic and molecular studies have shown that maternally deposited morphogens are indeed present in the Drosophila egg (Frohnhöfer and Nüsslein-Volhard, 1986; Lehmann and Nüsslein-Vol-hard, 1986; Nüsslein-Volhard et al. 1987; Sander and Lehmann, 1988). In particular, the maternal morpho-gen bicoid (bcd) has been shown to determine anterior–posterior pattern in a concentration-depen-dent way by initiating a cascade of patterning events in the early embryo (Driever and Nüsslein-Volhard, 1988a, b).
Genetic screens carried out by Nüsslein-Volhard and Wieschaus (1980) identified most of the zygotic genes in the regulatory cascade that act downstream of the maternal morphogens. The segmentation gene hier-archy transduces the crude gradients of maternal morphogens into highly refined positional identities. The first step in this process involves the activation of the first zygotically expressed genes, called the gap genes, in broad overlapping domains (Gaul and Jäckie, 1987; Gaul and Jäckie, 1989; Stanojevič et al. 1989; Struhl, 1989b).S
The best documented case of an interaction between a maternal factor and a gap gene is the regulation of hunchback (hb) by the bcd morphogen (Driever et al. 1989; Struhl et al. 1989). The initiation of hb has been shown to involve a transcriptional ‘on/off switch, whereby levels of bcd protein above a minimal threshold concentration activate hb while lower levels fail to do so. The bcd protein is distributed in a broad gradient along the anterior–posterior axis of the early embryo, with peak levels present at the anterior pole (Driever and Nüsslein-Volhard, 1988a,b). High levels of bcd in anterior regions activate hb, resulting in a relatively sharp border of hb expression. The lower levels of bcd protein in more posterior regions of the embryo fail to trigger the hb promoter. The interaction between bcd and hb is consistent with the ‘French flag’ model of cell fate specification suggested by Wolpert (1969). However, according to that model, the bcd gradient should have more instructional capacity, and define additional domains of gap gene expression.
Heterologous promoters containing 5’ hb sequences attached to the reporter gene lacZ are activated by bcd in a concentration-dependent manner. Such gene fusions are expressed in progressively more posterior regions of the embryo as the dose of bcd protein is increased by varying the bcd gene copy number from 1 to 6 (Struhl et al. 1989; Driever et al. 1989). Activation of hb–lacZ fusions was shown to depend on at least two copies of bcd binding sites. Multimerization of a small 5’ hb DNA fragment containing two bcd sites resulted in the activation of hb–lacZ fusions even by low levels of the bcd protein. These results prompted the proposal that overlapping, but non-identical, domains of gap gene expression could involve the differential response of gap gene promoters containing different numbers of bcd binding sites (Driever et al. 1989; Struhl et al. 1989). Thus, a promoter such as hb is broadly expressed in response to the bcd gradient since it contains many bcd binding sites, while the expression of a second gap gene might be confined to more anterior regions by possessing fewer bcd sites. Although this proposal for establishing distinct domains of gap gene expression is quite appealing, previous studies have failed to identify a second gap gene that is differentially regulated by the bcd morphogen. Indeed, Driever et al. (1989) refer to such a hypothetical gap gene as ‘gene X’. Here we present evidence that the segmentation gene giant (gt) probably corresponds to gene X.
gt was originally identified in the genetic screens of Wieschaus and Nüsslein-Volhard (1984a) as a gap gene on the basis of its effects on segmentation of the larval cuticle, which includes deletion of denticle belts A5–A7 and some head defects. However, these defects are less extensive than those seen for the gap genes hb, Kr and kni. For example, Kr mutants show deletions of the entire thorax and anterior abdomen (Wieschaus et al. 19846). Thus, gt was originally thought not to play an instrumental role in initiating segmentation (Wieschaus et al. 1984a; Gergen and Wieschaus, 1986; Petschek et al. 1987). Here and in the accompanying paper, we present evidence that gt belongs to the gap class of genes and is essential for the initiation of segmentation, since it appears to be a primary target of morphogen gradients and interacts with the other identified gap genes.
We describe additional details of the gt expression pattern during development (see Mohler et al. 1989), and show that it is a nuclear protein that is distributed in a relatively simple pattern at the blastoderm stage, characteristic of gap genes (i.e. Lehmann and Nüsslein-Volhard, 1987; Gaul et al. 1987; Nauber et al. 1988). The gt pattern consists of two broad domains, which are under the separate control of different maternal morphogenetic centers. We have analyzed the distri-bution of the gt protein in all known gap mutants and in mutants that disrupt each representative maternal organizing group, namely bcd, nanos (nos) and torso (tor) (reviewed in Nüsslein-Volhard et al. 1987). None of the gap mutants cause a failure to initiate gt expression, suggesting that activation is directly regu-lated by maternal factors. We present evidence that the anterior portion of the gt expression pattern is controlled by bcd, while the posterior gt domain seems to be regulated by the posterior pattern organizer, nos, which controls levels of the maternal hb product in posterior regions (Struhl, 1989a; Irish et al. 1989; Hülskamp et al. 1989). We present evidence that the overlapping, but non-coincident patterns of gt and hb expression involve their differential response to the bcd concentration gradient.
Materials and methods
Cloning a full-length gt cDNA
A cDNA library in the vector pNB40 (Brown and Kafatos, 1988) was screened with a genomic DNA fragment from the gt coding region (Mohler et al. 1989). Several clones, all about 1.7 kb in length, were isolated, and those with identical restriction patterns were assumed to represent the major gt transcription product. Previous studies have shown that the predominant gt mRNA has a length of 1.9 kb (Mohler et al. 1988). One of the isolated cDNAs was used to verify that it corresponds to gt by making a digoxygenin–labelled probe and hybridizing to whole-mount preparations of embryos (Tautz and Pfeifle, 1989). The pattern was the same as that observed for gt by Mohler et al. (1989), based on hybridization to tissue sections. This cDNA was then subcloned as a HindIII–Eco-RI fragment into the Bluescript SK+ vector (Stratagene), and partially sequenced from the 5’ and 3’ ends. It was found to contain 40 bp of leader sequence and the consensus initiation sequence CACCATG of Kozak (1984), followed by an extended open reading frame. The trailer is 318 bp long and contains a potential polyadenylation signal 300bp downstream from the stop codon.
Localization of gt transcripts
gt transcripts were localized in wild-type embryos according to the protocol of Tautz and Pfeifle (1989). The full-length cDNA was gel-isolated and approximately 200 ng of this fragment was labelled by random priming and digoxygenin– labelled UTP incorporation. About l/50th of this reaction was used per hybridization reaction, which was done in a total volume of 50 μl in an Eppendorf tube.
Preparation of gt protein and antisera
A full-length gt protein was produced in E. coli using the T7 expression system developed by Studier and Moffatt (1986). Oligonucleotide-directed mutagenesis was performed using the Muta-Gene Phagemid in vitro Mutagenesis kit (Bio-Rad). An NdeI restriction site was created at the initiating ATG of the cDNA in the Bio-Rad vector pTZ-19 by using the oligonucleotide GCATTAGCATATGGTTCGGTG. The pAR-gt expression plasmid was constructed by subcloning a 1.7 kb NdeI–EcoRI fragment containing the entire gt coding region into the unique NdeI and EcoRI sites of the vector. The protein was extracted from induced cultures of the bacterial strain BL21(DE3), as describcd previously (Stano-jevic et al. 1989), and electroeluted from an SDS-polyacrylamide gel and lyophilized. Alternatively, protein was prepared according to the Rubin Lab Methods Book (1986), protocol of D. Rio, p.19.2, plan B and lyophilized. Protein extracts prepared according to both methods were electrophoresed on SDS–polyacrylamide gels; two major bands at ∼50 and ∼60×103Mr were observed upon induction of the bacterial cultures. When these two bands were eluted separately and re-run on acrylamide gels, the 60×103Mr band appeared to degrade so that it generated the 50×103Mr species. Therefore, the 60×103Mr. species was assumed to be the full-length protein, and the 50×103Mr species a break-down product. Lyophilized protein prepared both by electro-elution from acrylamide gels and by the Rio method was injected into rats and guinea pigs (Pocono Farms, Canadensis, PA).
Embryo fixation and antibody staining
Embryo collections, fixations and staining procedures were done as describcd in Frasch et al. 1987. For anti-gt sera, both 50×103Mr and 60×103Mr species of gt protein were eluted separately for injection into rats. These sera gave identical staining patterns in embryos. Also crude gt bacterial extracts from the Rio protocol were injected into guinea pigs and rats. Rat anti-gt sera were diluted 1:250 in a buffer consisting of 1×PBS, 0.5 M NaCl, 0.1 % Tween 80, and 1 % BSA. All other sera used in this study were diluted in the same buffer, and were used at the following dilutions: guinea pig anti-gt, 1:500; rabbit anti-Kr, 1:100; rabbit anti-eve 1:500; mouse anti-hb, 1:500; guinea pig anti-kni 1:200. Kr and eve antisera were the same as those used by Stanojevič et al. (1989), and Frasch and Levine (1987), respectively. The hb antiserum was made against a full-length hb protein produced in E. coli (Stanojevië et al. 1989). kni antiserum was obtained from Dr Ken Howard. Indirect immunofluorescence was done with TRITC-conjugated or FITC-conjugated secondary antibodies to rat, rabbit, guinea pig and mouse, purchased from Jackson Immunoresearch (Bethesda, Md.). Photomicroscopy was done as describcd by Stanojevič et al. (1989).
The following fly stocks were used for embryo collections. bcd−: bcdE1, a strong, ems-induced small deletion mutation (Frohnhdfer and Nüsslein-Volhard, 1986; Berleth et al. 1988); bcd− embryos were collected from a mixed population of heterozygotes and homozygotes; bcd multiple copies: the BB9+BB16 stock has two wild-type copies of bcd and two extra copies balanced over the second chromosome, making the total number of copies four or six in a balanced population of females; exu− and stau−: exuPJ42, a strong allele, and stauHL54, a hypomorph (mutations are describcd in Schüp-bach and Wieschaus, 1986; Tearle and Nüsslein-Volhard, 1986). All of the above stocks were obtained from Dr Gary Struhl, and bcd duplications are describcd in Struhl et al. (1989). Other maternal mutants used in this study were, nanos−. nanosL7, a hypomorph (G. Struhl, personal com-munication) provided by Dr Ruth Lehmann (Nüsslein-Volhard et al. 1987); torso−: torsoPM51, a strong allele (Tearle and Niisslein-Volhard, 1987; Schiipbach and Wieschaus, 1986); torso dominant gain-of-function mutants: splcRL3 was provided by Dr Trudi Schüpbach (Schüpbach and Wieschaus, 1989), and tor04021 was provided by Dr Martin Klingler (Klingler et al. 1988). The gap mutants used in this study were hb−: hb14F21 (Lehmann and Nüsslein-Volhard, 1987); Kr−: Kr9 (Preiss et al. 1985; Redemann et al. 1988); kni−: kniIID48 (Tearle and Nüsslein-Volhard, 1987), all of which are strong alleles; tll−: tll1, a hypomorph (Strecker et al. 1986,1988); and gt−: gtYA82, a strong allele (Gergen and Wieschaus, 1986; Wieschaus et al. 1984a). All of the gap mutant stocks were obtained from Dr Jym Mohler.
RNA expression pattern
gt transcripts were localized in whole-mount prep-arations of wild-type embryos using a full-length gt cDNA (see Materials and methods). The results that were obtained (Fig. 1A–D) are consistent with the previously published pattern based on hybridization to tissue sections (Mohler et al. 1989). Transcripts are first detected during nuclear cycle 12 as two broad bands encircling the embryo. The anterior band extends from about 80% to 60% egg-length (Fig. 1A), while the posterior band extends from about 30% to 0% egg-length (data not shown). The distribution of transcripts changes rapidly during the next two cleavage cycles, and the posterior band recedes from the pole, becoming narrower and now covers approximately 35 % to 25 % egg-length. The anterior domain splits into two bands at about mid-nuclear cycle 14, each of which includes approximately one and a half parasegments in width (about 71% to 63% and 85% to 77% egg-length, respectively). Simultaneously with the splitting of the anterior domains, expression is lost in the ventral-most portion of the anterior stripe (Fig. IB, C, arrow), and by late cell cycle 14 a third anterior stripe appears in a dorsal patch very close to the anterior pole of the embryo, at 97% to 91% egg-length (Fig. ID, arrow). At the end of cycle 14, a ventral patch reappears at the base of the middle head stripe (Fig. ID). As the anterior pattern unfolds, the posterior domain fades slightly. By the time gastrulation is under way, the posterior domain is almost completely gone and the head staining becomes very intense. Anterior ex-pression persists throughout germ band elongation and shortening (data not shown).
Protein expression pattern
The same cDNA used for examining the gt transcript pattern was used to make full-length protein in bacterial cells, and polyclonal antisera were generated against this product (see Materials and methods). The pattern seen in embryos stained with these sera (Fig. 1E-H) corresponds closely to the RNA expression pattern. The gt protein is first detected during nuclear cycle 13, within 10 min of the first appearance of the RNA. An important feature of the gt protein distribution is that it is nuclear, which introduces the possibility that gt is a transcription factor (see accompanying report by Kraut and Levine (1991); Small et al. unpublished data).
Co-localization of gap proteins
Direct comparisons of the gap gene expression patterns with gt can help elucidate ways in which gt may be functioning to control segmentation. To achieve this, we have carried out double immunolocalization of gt in pairwise combinations with each of the other cloned gap genes for which there are antibodies.
We have found that gt overlaps extensively with the hb and kni expression patterns, but is complementary to the Krüppel (Kr) pattern. The gt pattern is consistent with previously identified patterns of gap genes, in that neighboring gap domains overlap where their ex-pression intensities decline (see Stanojevič et al. 1989; Gaul and Jäckie, 1989). Double staining for hb and gt (Fig. 2) reveals that the anterior hb pattern entirely encompasses the anterior gt domain, while the posterior hb stripe overlaps with gt’s posterior domain by a few cells (Fig. 2B).
The knirps (kni) posterior domain similarly overlaps the gt posterior pattern by a few cells on the anterior edge of the gt band, gfs posterior domain is therefore straddled by kni at its anterior and hb at its posterior margins. At the anterior of the embryo, the kni and gt domains are completely exclusive, and remain so throughout the evolution of both patterns during cell cycle 14 (Fig. 2D–F). The central Kr domain and the two gt domains are entirely non-overlapping. The significance of this finding and the regulatory interac-tions between the two genes are discussed in the accompanying paper by Kraut and Levine (1991).
gt expression in gap mutants
The gt staining pattern in all of the known gap mutants demonstrates genetic interactions between gt and other gap genes, which are suggested by their relative expression patterns. Kr (Wieschaus et al. 1984b) has a strong effect on gt expression, in that both gt domains are expanded toward the center of Kr− embryos (Fig. 3D,E). This result suggests that Kr is a strong repressor of gt expression since only very low levels of the Kr protein are detected at the margins of the gt domains. Several lines of evidence suggest that the gt–Kr interaction is direct (see Discussion and Kraut and Levine, 1991). The reduced expression of the expanded posterior gt domain in Kr− is probably due to the premature loss of kni expression seen in these mutants (Pankratz et al. 1989; see below).
The gt protein distribution is only slightly altered in hb mutants (Fig. 3A–C). The posterior domain is expanded slightly toward the posterior, into the domain which hb normally occupies. The anterior domain is slightly shifted toward the anterior of the embryo, possibly due to repression by ectopic Kr, which has been shown to expand anteriorly in hb− embryos (Jackie et al. 1986). This suggests that hb has a slight repressive effect on gt, which is (notably) only operating in the posterior of the embryo. Not surpris-ingly, hb is not a repressor of the anterior gt pattern since proteins encoded by the two gap genes overlap extensively in these regions.
The major effect of kni− (Nauber et al. 1988) is a premature reduction in the level of gt protein in the posterior domain (Fig. 4A). In the head domain, the middle stripe persists as an unbroken band covering the entire ventral side, in contrast to wild-type, where it is ventrally and ventral-laterally repressed (see Fig. 1G,H). This result establishes kni as a dual regulator of gt: in head regions it represses, and in the tail it maintains high levels of gt. This maintenance role for kni on gt in the posterior domain is reminiscent of the positive effect that Kr exerts on kni (Pankratz et al. 1989).
Finally, we tested the effect of tailless−(tll) on gt expression (Fig. 4D–F, Strecker et al. 1986; Pignoni et al. 1990). tll exerts a negative effect on gt expression, in that the posterior domain expands posteriorly, to 12 % egg length. This result indicates that tll either directly or indirectly represses gt expression at the posterior pole, and probably participates in the normal refinement of the gt pattern during early development (see Fig. 1A, B). In tll−, as in all the gap mutants, the relative positioning of the gt and eve patterns remains constant, (as shown by double stains in Figs 3,4,5,6 and 7), even though both patterns are disrupted. This implies a coordinated movement, so that, for example, in tll− the posterior gt band still encompasses stripe six of eve, and the anterior margin of eve stripe 1 still coincides with gt head stripe 2.
In summary, the results presented here suggest that regulatory interactions with the gap genes modulate gt expression, but are not required for its initiation. Below we describe the role of maternal factors in the initiation of gt expression.
gt expression in maternal mutants
gt is among the earliest zygotic genes to be activated during embryogenesis. Given its very early expression, it seems likely that maternal morphogens could directly determine gt’s expression pattern in a way that may be analogous to their action on hb and kni (Gaul and Jäckie, 1989). In order to assess the effects of maternal products on gt expression, one representative mutant from each of the three maternal organizing groups was selected and stained with gt antibody.
bcd has been established as the anterior maternal morphogen (Frohnhöfer and Nüsslein-Volhard, 1986; Driever and Nüsslein-Volhard, 1988a,b) responsible for determining head and thoracic structures. One mode of action is its direct activation of hb (Driever and Nüsslein-Volhard, 1989; Struhl et al. 1989). But because hb− exhibits only a subset of the defects seen in the bcd mutant, hb clearly cannot be the only product that responds to bcd (Frohnhöfer and Nüsslein-Volhard, 1986; Driever et al. 1989). Therefore, we suspected that gt, which is expressed in the head but in a smaller region than hb, may also be a target for the bcd morphogen. Embryos from bcd homozygotes fail to initiate the anterior gt domain. Fig. 5 shows changes in the gt pattern observed in bcd mutants. Although the anterior domain of expression is absent, the posterior gt domain is correctly initiated. Co-localization studies with anti-eve and anti-gt antibodies indicate that there is a coordinated shift of the remaining part of the gt pattern with the remaining eve stripes, as is seen for all other mutants tested.
We believe that in addition to being a target of bcd, gt may also respond to the posterior morphogen nos (Nüsslein-Volhard et al. 1987; Lehmann, 1988). In nos mutant embryos, the gt posterior domain is usually absent, but is occasionally seen as a very weak and transient stripe (Fig. 6). nos+ gene activity is known to prevent maternal hb expression in the posterior of the embryo, and thereby allow the initiation of the abdomen-determining gene kni (Hülskamp et al. 1989; Irish et al. 1989a; Struhl, 1989a). Again, the apparent positive effect that nos exerts on gt is likely to be indirect. It has been proposed that nos represses maternal hb expression post-transcriptionally by either degrading hb transcripts and/or blocking their trans-lation. The effect of nos on gt may be mediated solely through maternal hb. In the absence of nos, maternal hb protein is expressed throughout the egg, and may repress both gt and kni expression.
The gt pattern is also affected in embryos lacking the maternal product tor, which is responsible for specify-ing terminal structures (acron and telson) (Klingler et al. 1988; Schüpbach and Wieschaus, 1989). In these mutants, the abdominal gt stripe extends all the way to the posterior pole. The anterior pole is also affected in that the anterior-most head patch of gt disappears, while the remainder of the anterior pattern is expanded and shifted (Fig. 7). However, expression does not extend all the way to the tip, indicating independent repression by the bcd system (see Discussion).
tor− exceeds tll− in its effect on gt expression, implying that tll is not the only repressor of gt at the poles. Examination of tor− embryos by double staining with both eve and gt reveals a coordinated movement of the gt pattern and the eve stripes, similar to the effect seen in tll− mutants. This lends further support to the notion that particular gap gene combinations establish the individual pair-rule stripes (Gaul and Jäckie, 1989; Stanojevič et al. 1989).
Coordinate regulation of gt and hb by the bcd concentration gradient
According to the ‘French Flag’ model for localized patterns of gap gene expression, the overlapping, but non-coincident hb and gt patterns in anterior regions of the early embryo could involve the differential response of these two genes to distinct thresholds of the bcd morphogen. Perhaps the hb domain extends beyond the limits of the gt pattern because it is activated by even low levels of bcd protein, gt might be restricted to more anterior regions due to a less sensitive response to bcd. As a first step towards testing this idea, we examined the relative patterns of hb and gt expression in embryos containing varying doses of the bcd gene. The two patterns show a coordinated shift with increasing doses of bcd (Fig. 8), from one copy of the bcd gene (bcd+/bcd− [Fig. 8A]) where gt and hb are shifted slightly toward the anterior, to six copies (Fig. 8C), where both hb and gt show a marked expansion into posterior regions.
We also examined gt expression in the maternal mutants staufen (stau) and exuperantia (exu) (Schüp-bach and Wieschaus, 1986) in order to examine changes in the limits of gt and hb expression as a consequence of alterations in the slope of the bcd gradient (Fig. 8D,E). exu− alters the distribution of bcd mRNA, so that low levels of bcd protein are present relatively evenly throughout the embryo (Frohnhöfer and Nüsslein-Volhard, 1987; Driever and Nüsslein-Volhard, 1988b; Berleth et al. 1988). stau− also alters the distribution of bcd products, but does not completely eradicate the concentration gradient of bcd protein. Instead, it results in a very shallow gradient of bcd, with reduced amounts of protein in anterior regions of the embryo (Driever and Nüsslein-Volhard, 1988b). gt’s posterior domain is abolished in stau−, and the anterior domain is shifted anteriorly (Fig. 8D), similar to the shift seen in embryos heterozygous for bcd−. hb is also shifted anteriorly in concert with gt. Both genes are also affected in exu mutant embryos, but in a different way from that seen in stau−. exu− strongly reduces gt expression in both anterior and posterior regions, but the positioning of the pattern is not drastically altered (Fig. 8E). The embryo shown here represents the strongest expression of gt seen in exu− mutants. Most mutant embryos completely lack gt expression. In contrast, 66 is expressed at high levels, but is restricted to somewhat more anterior regions than in wild-type.
gt expression in nos, Kr double mutants
Maternal morphogens are thought to directly activate certain gap genes, such as hb (Driever et al. 1989; Struhl et al. 1989). In addition, it has been proposed that localized gap expression patterns involve regulatory interactions among the gap genes (Jäckie et al. 1986). In order to determine the extent to which these two mechanisms operate to regulate gt expression, we examined embryos that were doubly mutant for nos− and Kr−. The aim here was to determine whether the loss of the gt expression pattern observed in the maternal mutant nos was due to ectopic expression of maternal 66 products, or an indirect effect due to the Kr repressor. Previous studies have shown that the central Kr domain expands posteriorly in nos− mutants (Gaul and Jäckie, 1987a). Such an altered Kr pattern could repress gt, since we have shown that the gt domains expand in Kr− embryos (Fig. 3D,E). If Kr repression is responsible for the loss of gt expression in nos− embryos, then gt should ‘reappear’ in nos−,Kr−double mutants. In fact, we found that Kr− does not rescue gt expression in a nos− background, since all embryos from a nos−/nos−, Kr−/CyO mutant stock are missing the posterior domain, similar to that seen in nos− (data not shown). This result suggests that Kr does not define the limits of gt–s posterior domain, but instead the pattern is more likely regulated by maternal hb repressor. Once established, the maintenance of the gt pattern involves regulatory interactions with Kr.
The genetic studies presented here suggest that maternal factors are responsible for the initiation of gt expression in anterior and posterior regions of the early embryo. The maternal morphogen bcd (Nüsslein-Volhard et al. 1987) is important for the initiation of the anterior pattern, while posterior expression is con-trolled in part by 66 (Lehmann and Nüsslein-Volhard, 1987). The limits of the anterior gt domain are set by the levels of the bcd protein, and the non-coincident patterns of hb and gt expression involve their differen-tial response to distinct thresholds of the bcd morpho-gen. Repression by maternal hb products determines the overlapping patterns of kni and gt expression in posterior regions, which are maintained by cross-regulatory interactions among gap genes (J äckie et al. 1986). The genetic control of the gt expression pattern is summarized in Fig. 9.
Initiation of anterior gt expression
The bcd morphogen appears to regulate the initiation of the anterior gt pattern through a ‘French flag’ mechan-ism (Wolpert, 1969). The gt promoter is either active or repressed depending on the level of the bcd protein present in a given nucleus in the syncytial blastoderm. Peak levels of bcd, present in the anterior-most regions, coincide with sites where gt is not activated. Intermedi-ate levels of the morphogen activate gt expression, while low levels, present in more posterior regions, are insufficient for activation, bcd probably exerts a direct effect on gt expression since null mutations in each of the known gap genes fail to abolish its initiation, and gfs early expression makes intermediate regulatory steps unlikely. It may be that the bcd protein binds directly to gfs promoter to activate its expression. The promoter might contain multiple bcd binding sites, as was shown for the hb promoter (Driever and Niisslein-Volhard, 1989) and .the morphogen might fail to activate gt when either all of these sites are occupied or too few are occupied. According to this ‘titration’ model, activation is achieved only when some of the bcd sites are bound with morphogen. Thus, in anterior and posterior regions of the early embryo there are either too high or too low levels of bcd protein to activate gt expression. A similar model has been proposed for the regulation of the gap gene Kr by bcd (Gaul and Jäckie, 1989; Hülskamp et al. 1990).
An alternative explanation for localized gt expression is that repression in the anterior-most regions involves an unknown intermediate gene(s) which is activated by peak levels of bcd (Driever et al. 1989). This model is consistent with results of transient cotransfection assays, which have shown that bcd can function as a sequence-specific transcriptional activator. Such studies have failed to demonstrate a repressor activity for the bcd protein (Driever and Nüsslein-Volhard, 1989; Struhl et al. 1989). The initiation of any such putative intermediate target gene is also sensitive to the concentration of the bcd protein since increasing copies of the bcd gene result in a progressively broader zone at the anterior pole where gt is repressed (see Fig. 8).
Initiation of posterior gt expression
Plausible mechanisms have been put forth for bcd–s activation of the gap patterns in the anterior of the embryo, whereas it is still a mystery as to how posterior gap domains are initiated. So far, no abdominal activators of kni or gt are known; however, we propose that the overlapping patterns of gt and kni expression in posterior regions involve their differential repression by maternally expressed hb protein (Fig. 9). The hb transcript is uniformly distributed throughout the cytoplasm of unfertilized eggs and early embryos (Tautz et al. 1987). This transcript is not stably translated in all regions of the embryo, such that transcripts in anterior regions are more stably translated than those in posterior regions. This results in a protein concen-tration gradient by cleavage cycle 11–12 with peak levels in anterior regions, and lower levels in more posterior regions (Tautz, 1988).
Unlike the situation for bcd, which is a direct transcriptional activator, the most likely mechanism for hb–s action is that it exerts a repressive effect on kni and gt, thereby restricting their expression to posterior regions lacking maternal hb protein (Hülskamp et al. 1989; Irish et al. 1989a; Struhl, 1989a). The fact that the patterns of kni and gt are out of register could be explained if the two genes are repressed by different threshold levels of hb, with gt being more sensitive to the hb repressor. This would restrict gt expression to a region posterior to the kni pattern. We examined the gt and kni patterns in embryos where hb is overexpressed under the control of the hsp70 promoter (Struhl, 1989a). Embryos exposed to a high dose of ectopic hb at early stages completely failed to initiate both kni’s and gt–s posterior domains (Kraul and Levine, 1991).
The concentration gradient of maternal hb protein is generated by nos, which somehow interferes with hb expression in posterior regions (Tautz, 1988). It has been proposed that nos constitutes the ‘sink’ for the hb gradient, either by blocking its translation or stability (Hülskamp et al. 1989; Irish et al. 1989a; Struhl, 1989a). nos activity is thought to emanate from posterior regions of the egg (Sander and Lehmann, 1988; Nüsslein-Volhard and Roth, 1989). If this is correct, a critical level of nos expression near the posterior pole would define the site where gt can be activated due to the complete absence of hb repressor. In slightly more anterior regions, there are somewhat lower levels of nos, permitting the expression of just enough hb to repress gt. In nos− embryos, the hb protein is distributed along the entire length of the embryo (Tautz, 1988), thereby blocking the initiation of kni and gt in posterior regions.
Recent studies have shown that the only purpose of nanos is to suppress maternal hb in posterior regions of early embryos (Hulskamp et al. 1989; Irish et al. 1989a; Struhl, 1989a). Double mutants that lack both maternal hb and nanos show a normal segmentation pattern, and even survive to adulthood. Although not specifically tested, it is likely that kni and gt are correctly initiated in such mutants, suggesting that maternal hb is not an essential morphogen. However, it is conceivable that in wild-type embryos the zygotic hb gradient functions redundantly with the maternal hb gradient to trigger overlapping patterns of kni and gt expression. The maternal and zygotic hb gradients show similar slopes along the anterior-posterior axis. Zygotic hb ex-pression is correctly activated by the bcd morphogen in embryos lacking both nos and maternal hb products, giving a gradient of expression that is quite similar to the transient maternal gradient seen in wild-type. Perhaps this zygotic gradient is sufficient to generate overlapping kni and gt expression patterns in embryos lacking both maternal hb and nos products.
Since we have seen that hb is capable of repressing gt in the posterior of the embryo, the problem arises as to how gt escapes repression by the high levels of hb that exist in anterior regions. A competitive mechanism between bcd and hb might allow gt to be expressed in anterior regions, whereby the bcd activator occupies gt–s promoter more avidly, excluding hb. It is also possible that the gt promoter contains at least two discrete and autonomous elements, one responsible for expression in anterior regions and another that directs posterior expression. As discussed above, it is likely that bcd would activate a putative anterior promoter element directly via bcd binding sites. This element would be insensitive to repression by hb if it lacks hb binding sites. In contrast, a posterior promoter element might contain numerous hb binding sites, thereby making it very sensitive to repression by even low levels of the hb protein. An important implication of this model is that repression by hb does not occur over long distances within the gt promoter. In anterior regions there are high levels of hb protein, which should fill all hypothetical hb binding sites within the posterior promoter element. According to the model, the hb proteins bound to the posterior element are unable to repress over a long enough distance to interfere with the activation of the anterior element by bcd.
Differential regulation of gt and hb by bcd
Overlapping patterns of hb and gt expression in anterior regions, and kni and gt in posterior regions, could arise from the differential regulation of these genes by a common set of maternal factors. As discussed above, the out-of-register patterns of kni and gt in posterior regions could involve their differential repression by distinct thresholds of the hb gradient. Similarly, the noncoincident hb and gt patterns in anterior regions could be due to their differential response to the bcd gradient. It has been proposed that there might be a gene whose domain of expression is a subset of the anterior hb pattern, designated ‘gene X’ (Driever et al. 1989). This was originally based on the observation that the phenotype of hb is not as severe as that seen in bcd− (Frohnhöfer and Nüsslein-Volhard, 1986). Recent support for this notion stems from the finding that progressively more posterior patterns of expression are obtained from fusion genes driven by synthetic pro-moters containing increasing numbers of bcd binding sites (Driever et al. 1989; Struhl et al. 1989). This result established the principle that the limits of a bcd target gene are determined by the configuration, strength and number of bcd binding sites present in its promoter. Target genes containing relatively few and/or low affinity binding sites would be activated only by peak levels of the bcd protein, and therefore be restricted to the anterior-most regions of the embryo. In contrast, a target gene such as hb that contains multiple, high affinity binding sites should be activated by even low levels of bcd, and therefore be expressed from the anterior pole into relatively posterior regions of the embryo.
We propose that the early pattern of gt expression fulfils the requirement for a gene ‘X’. The restriction of gt to a subset of the hb anterior pattern could be due to the presence of fewer high affinity bcd binding sites in the gt promoter as compared with the hb promoter (Driever and Nüsslein-Volhard, 1989). During the final stages of preparing this manuscript other ostensible targets of the bcd morphogen have been identified, including empty spiracles (Dalton et al. 1989) and orthodenticle, which are homeobox genes, and button-head (Cohen and Jürgens, 1990; Finkelstein and Perrimon, 1990). It is conceivable that all of these genes are direct targets of bcd, although it is possible that some are only indirectly regulated by bcd. Here we have shown that changes in the number of bcd copies results in a correlative alteration in the limits of gt expression at very early stages of embryogenesis (prior to cellularization; see Fig. 8). This early effect of bcd dose on gt expression suggests that bcd is likely to act directly on gt.
Mutations in the maternal genes exu and stau (Schüpbach and Wieschaus, 1986) were used to examine the effects of altering the shape and amplitude of the bcd gradient (Berleth et al. 1988; Driever and Nüsslein-Volhard, 19886) on the patterns of hb and gt expression, exu causes such a severe reduction in the overall levels of bcd protein that gt expression is nearly abolished. In contrast, the reduced levels of bcd are still sufficient to permit activation of hb, albeit in a narrower domain than normal. These results provide further evidence that hb can be activated by a lower threshold of bcd morphogen than that required to initiate gt, which is probably the basis for gfs restriction to a subset of the hb pattern in wild-type embryos. The coordi-nated shift of the gt and hb patterns seen in stau mutants is also consistent with a direct response of both hb and gt to bcd levels, since the effect of stau is to drop the overall amplitude of the bcd concentration gradient (Berleth et al. 1988; Driever and Nüsslein-Volhard, 19886).
Cross-regulation between gt and other gap genes
It has been proposed that the maintenance of discrete gap gene expression patterns involves cross-regulatory interactions among the gap genes (Jäckie et al. 1986). Weak repressive interactions have been shown to occur between hb and Kr, in that mutations in either gene causes a slight expansion in the limits of the other. There is evidence that the maintenance of the posterior kni pattern depends on positive regulation by Kr (Pankratz et al. 1989). Here we have provided further support for the notion that cross-regulatory interactions are important for refining the limits of gap gene expression.
The maintenance of the two initial domains of gt expression near the anterior and posterior poles involves repression by Kr (Fig. 3D). And, in a manner analogous to the role Kr plays in the maintenance of the kni pattern, we have found that gt is sustained by kni (Fig. 4A). Although it has been claimed that Kr is a ‘direct’ activator of kni expression (Pankratz et al. 1989), it is equally likely that it functions as an indirect anti-repressor which ‘buffers’ kni from hb. In Kr− embryos, the hb domain expands posteriorly, where it would be expected to exert a strong repressive effect on kni expression. Consistent with this model is the finding that the Kr protein functions as a strong transcriptional repressor in transient co transfection assays; thus far these studies have failed to reveal an activating activity for Kr (Licht et al. 1990). Other genetic interactions involving Kr can also be interpreted as being brought about by either repression or anti-repression (i.e. Harding and Levine, 1988; Irish et al. 19896). Similarly, the positive effect that kni exerts on the posterior gt pattern could involve an anti-repression mechanism, whereby kni restricts the Kr repressor to central regions of the embryo.
We thank Christine Rushlow for the Kr antibody, Ken Howard for the knirps antibody, Gary Struhl for fly stocks, and David Kosman for technical assistance. This work was funded by grant GM34431 from the NIH.