Segmental polarity and identity in the abdomen of Drosophila is controlled by the relative position of gap gene expression

The anterior-posterior pattern in the Drosophila egg is initially evident both in specialized structures of the egg cell covers and within the cytoplasm. Characteristic specializations of the chorion include the dorsal appendages and micropyle at the anterior end and the aeropyle at the posterior end of the egg. Within the egg cytoplasm at the posterior pole are the polar granules which are determinants of the germ line fate (Mahowald, 1962). As development proceeds, additional pattern is established as segments of the embryo become arranged into functional units such as the head, thorax and abdomen. Each segment subsequently acquires a polarized pattern as rows of denticles appear at the anterior margin of each segment followed by a region of naked cuticle. The anterior and posterior ends of the embryo do not develop segmental characteristics, but rather specialized structures referred to as the acron and telson at each end respectively.

The establishment and maintenance of the polar pattern in the embryo is directed by a genetic pathway whose principles can be summarized as follows (for review see Akam, 1987; Ingham, 1988; Nüsslein-Volhard et al. 1987): the basic information for the embryonic pattern is provided by maternal genes. Mutations in these genes affect the development of specific regions of the embryo. By and large, these mutations interfere only with the embryonic pattern without disturbing the pattern of the extra-embryonic egg covers. On the basis of their phenotype the maternal genes can be grouped into three sets: the anterior group that affects the development of the head and thorax; the posterior group that affects the development of the abdomen; and the terminal group that affects the most anterior (acron) and most posterior (telson) structures. Maternal gene products provided by the mother and stored in the egg cell are required for the spatial regulation of transcription of the first genes expressed by the embryo, the gap genes.

The gap genes are expressed in nonrepetitive domains and the products of neighboring gap genes are thought to overlap at the borders (U. Gaul et al. in preparation). Several different types of experiments indicate that gap gene expression is controlled both by the products of the maternal genes described above and by the products of the gap genes themselves. First, the phenotype of mutations in gap genes resembles the phenotype of mutations in the maternal genes (for review see Lehmann, 1988). Second, mutations in one gap gene affect expression of other gap genes (Jäckle et al. 1986). Accordingly, the maternal anterior system controls the zygotic expression of the gap gene hunchback (hb) in the anterior half of the embryo (Tautz, 1988), the maternal posterior system affects expression of knirps (kni) in the prospective abdominal region (Nauber et al. 1988), and the maternal terminal system regulates the activity of the gap gene tailless (til) at the termini (Klingler et al. 1988). Later in development, the polar arrangement of individual segments is established by the patterned expression of pair-rule and segment polarity genes (for review see Ingham, 1988; Ingham and Gergen, 1988).

What are the mechanisms that establish the polarity described above? Do the maternal gene products directly control the expression of gap genes, pair-rule genes and segment polarity genes, or are the maternal gene products at the top of a hierarchy in which the only genes they control directly are the gap genes? Presently, most of our information comes from studies of the gene, bicoid (bed), involved in development of the anterior pattern. Bicoid encodes an RNA that is highly concentrated at the anterior end of the egg (Berleth et al. 1988). The product of this RNA is a homeo-domain containing protein that is distributed in an anterior-posterior gradient (Driever and Nüsslein-Volhard, 1988). The bed protein activates transcription by binding to the promoter of the gap gene hb (Driever and Nüsslein-Volhard, 1989). Activation of hb is dependent on the concentration of the bicoid protein and consequently the hb gene is activated only in the anterior half of wild-type embryos (Struhl et al. 1989; Driever et al.1989). These experiments show that the anterior system controls gap gene expression directly by transcriptional activation. However, these experiments do not resolve whether the anterior system operates as a hierarchical system or whether the maternal genes also directly control the expression of pair-rule and segment polarity genes.

A key gene within the posterior group of genes is nanos (nos) (Nüsslein-Volhard et al. 1987; Nüsslein-Volhard and Lehmann in preparation). Genetic experiments argue that transcriptional activation of the gap gene kni by nos involves hb (Hülskamp et al. 1989; Irish et al. 1989; Struhl, 1989). In addition to the zygotic hb expression which is under the transcriptional control of bed, hb is also expressed maternally. The distribution of the maternal hb product is controlled by the posterior system, nos does not affect the transcription of the maternal product but seems to interfere with the stability and/or translation of the maternal hb product such that hb is absent from the posterior region of the embryo. Loss of nos function results in uniform, rather than anterior, distribution of maternal hb RNA and protein (Tautz, 1988). The abnormally high hb concentration within the posterior half of the embryo is correlated with the suppression of transcription of kni (Nauber et al. 1988). Finally embryos that are derived from a germ line deficient for the nos and hb products can develop a normal segmental pattern suggesting a normal pattern of kni expression. This leaves us with the question of how, in the absence of posterior maternal information, a normal abdominal segmentation pattern develops. One hypothesis is that the distribution of gap gene expression plays a critical role in the establishment of the segmental pattern independently of maternal information. According to this hypothesis, maternal genes would act in a strict hierarchical manner and maternal information would become dispensable, once the gap gene expression pattern is established. This hypothesis was tested by transplanting posterior pole plasm into embryos of different genetic backgrounds. We show that the polarity and identity of the posterior segments can be determined by the autonomous action of gap genes.

Embryos derived from females homozygous for any one of the posterior group genes lack all abdominal segmentation, while the head, thorax and telson develop normally (Boswell and Mahowald, 1985; Schiipbach and Wieschaus, 1986, 1989; Lehmann and Nüsslein-Volhard, 1986, 1987). The abdominal phenotype of posterior group mutants can be completely rescued by injection of wild-type cytoplasm into mutant embryos (Lehmann and Nüsslein-Volhard, 1986, 1987, and in preparation). Only posterior pole plasm is active and best rescue is achieved when posterior pole plasm is transplanted into the prospective abdominal region. The site of localization, the posterior pole, is separated from the prospective abdominal region by the telson, whose development is under the control of the terminal genes. After transplantation of posterior pole plasm into the middle of the embryo, a normal anteroposterior pattern of abdominal segmentation is restored (Fig. 1B,C). Since at the site of injection the normal orientation of abdominal segments is retained, we conclude that posterior activity can not control the abdominal pattern in a concentration-dependent manner. It is more likely that a certain threshold of posterior activity is required for segmentation to occur, but that the orientation of segments within the segmented region is established independently. In the case of bed, on the other hand, a direct quantitative relationship between pattern and concentration has been shown (Frohnhöfer and Nüsslein-Volhard, 1986): injection of high amounts of anterior cytoplasm into the anterior pole of embryos derived from bed females leads to the development of head structures, while lower concentrations give rise to thoracic structures.

Although normal pattern requires all three maternal systems, some pattern can be generated in the presence of only one or two of the systems. The precise type of pattern that develops, however, depends upon which maternal systems are active. This point is addressed by injection of posterior pole plasm into the middle of bed, nos embryos, which lack both anterior and posterior information, or by injection into bed, nos,torsolike (tsl) mutant embryos, which lack all three systems of maternal information (Fig. 1). tsl is a maternal gene and a member of the terminal group (Frohnhöfer, 1987).

If left uninjected, embryos derived from bed, nos females will develop two telsons in mirror image, due to the presence of the terminal system (Fig. ID, Nüsslein-Volhard et al. 1987). Embryos derived from triple mutant bed, nos, tsl females develop no anterior-posterior pattern (Fig. 1G). After injection of posterior pole plasm into the middle of the double or triple mutant embryos, abdominal segments are formed in mirror image. The polarity of the mirror image depends on the maternal genotype. In bed, nos embryos the duplicated sets of segments are formed in a posterior-anterior-posterior (P-A-P) orientation such that the posterior segments (A8) are juxtaposed to the two telsons (Fig. IE). After injection into bed, nos, tsl embryos, however, the embryos develop the reverse mirror image with abdominal segments oriented toward the middle and the more anterior abdominal segments formed at both ends (A-P-A, Fig. 1H). These findings indicate that, although posterior activity is crucial for the establishment of segmentation within the abdominal region, the anterior-posterior polarity of the segments depends upon additional positional information. Since parts of the normal abdominal pattern can be established in the absence of anterior and terminal maternal systems, it is possible that this additional positional information is encoded by gap genes.

Three gap genes, hb, Kr, and kni have been characterized on the molecular level and the distribution of their RNA and protein patterns have been described. During the early syncytial blastoderm stages, hb is found within the anterior 50 % of the embryo (for RNA: Tautz et al. 1987, for protein: Tautz, 1988) Kr is found between 40% and 55% egg length (for RNA: Knipple et al. 1985, for protein: Gaul et al. 1987), and kni is expressed between 30% and 45 % egg length (for RNA: Nauber, 1988) (0%= posterior pole). Although the distribution of the til product has not been determined, genetic evidence suggests that til is active at the anterior and posterior ends of the embryo (Strecker et al. 1986; Klingler et al. 1988). Thus, the following anterior-posterior order of gap gene expression would be predicted: tailless, hunchback, Krüppel, knirps and tailless (Fig. 3B). (All gap genes analyzed so far have additional domains of expression which have been omitted from this discussion.)

To test the idea that neighboring gap genes may determine the orientation of the abdominal segments in the embryo, we analyzed the pattern of gap gene expression in different mutant backgrounds. These experiments allowed us to interpret changes in the segmentation as a consequence of changes in gap gene juxtaposition. As in the previous experiments, posterior pole plasm was injected into the middle of mutant embryos. We studied changes in Kr expression in embryos lacking either the posterior, the anterior and posterior, or all three systems of maternal information. We also examined the pattern of the Kr protein after injection of posterior pole plasm into embryos of these different maternal backgrounds. When embryos had reached the appropriate developmental stage, the distribution of Kr protein was detected with anti-Krilppel antibodies (Fig. 2A,B). The results of this study, and of studies by others (Gaul and Jackie, 1987), indicate that all maternal systems act negatively on the expression of Kr. In uninjected nos mutant embryos Kr protein extends further posteriorly than in wildtype (Fig. 2C). In the double mutant bed, nos embryos, the Kr domain is expanded anteriorly as well as posteriorly (Fig. 2E). Finally in the triple mutant, Kr is expressed homogeneously throughout the embryo (Fig. 2G). Injection of posterior pole plasm into mutant embryos leads to the suppression of Kr at the site of injection. After injection into a nos mutant embryo, the Kr domain narrows, and resembles the domain of Kr expression found in wildtype embryos (Fig. 2D). After injection into the bed nos double mutant, Kr is completely suppressed from the middle region in some embryos, while in other embryos weak staining remains (Fig. 2F). After injection of posterior activity into the triple mutants, Kr expression is confined to both ends of the embryo (Fig. 2H).

In our injection experiments, we monitored the distribution of Kr protein but not the expression pattern of the other gap genes. In order to reconstruct the approximate pattern of different gap genes in injected embryos, we extrapolated the pattern of the other gap genes from previous experiments (Gaul and Jackie, 1987; Tautz, 1988; Nauber et al. 1988). Since injection of posterior pole plasm into the middle of a nos mutant embryo (Fig. 3A) can restore the wildtype pattern, it seems likely that the expression pattern of all gap genes in such an embryo resembles that of the wildtype. We therefore infer that the injection of posterior pole plasm will activate kni expression at the site of injection. A normal orientation of the segmental pattern would thus require that the kni domain be bordered anteriorly by Kr and posteriorly by til (Fig. 3B). After injection into the bed nos double mutant, Kr is suppressed and kni is presumably expressed throughout the central region. Thus the kni domain must be flanked by the til domains at either end (Fig. 3D). The juxtaposition of kni and til would lead to a P-A-P pattern duplication (cf. Fig. IE). Finally, after injection into a triple mutant embryo, Kr is suppressed in the central region and kni is presumably expressed in this region. In contrast to the previous experiment, in these mutant embryos kni must be flanked by Kr whose expression remains at the ends (Fig. 3F). This order of kni and Kr expression would lead to an A-P-A pattern (cf. Fig. 1H) in accordance with the more anterior expression of Kr in the normal pattern.

The expression of kni together with either til or Kr seems sufficient to promote the formation of some part of the abdomen. The cooperation of either pair of gap genes determines both the polarity and the identity of the abdominal segments. If til is juxtaposed to kni, a posterior-anterior mirror-image duplication of the posterior abdomen (A8− A4/5− A8, c.f. Fig. IE) is formed. Juxtaposition of kni with Kr on the other hand results in an anterior-posterior mirror-image duplication of the anterior abdomen (A1/2− A4/5− A1/2, c.f. Fig. 1H). Thus the polarity as well as the identity of each abdominal segment is controlled coordinately by inter actions among gap genes and their products. This suggests that the relative position of the domains of gap gene expression establishes the pattern of homeotic gene expression, which determines the identity of each segment (White and Lehmann, 1986; Irish et al. 1989).

To test whether changes in the relative position of the domains of gap gene products do indeed affect polarity independently of maternal information, we injected embryos of identical maternal genotype but which differed in their zygotic genotype. Females homozygous for bed and nos and heterozygous for the zygotic lethal gene til were crossed with heterozygous til males. While all of the progeny lacked anterior and posterior maternal information, a quarter of these embryos were homozygous mutant for til and lacked telson structures. Injection into embryos from this cross resulted in two phenotypes. The majority of embryos developed duplications of the posterior abdomen in P− A− P orientation. As described for bed nos embryos (Fig. IF) this indicates that kni is flanked by til and Kr expression is suppressed. One quarter of the embryos, the til embryos, developed anterior duplications of abdominal segments in the A− P− A orientation. As described for injected bed, nos, tsl embryos (Fig. II) we infer that in these embryos kni is flanked by Kr. This result indicates that abdominal polarity is established according to the spatial arrangement of gap gene expression. In the wildtype, however, the different maternal genes ensure the correct positioning of the gap genes and thus ensure normal polarity.

The activity of the posterior group genes is required for the normal expression pattern of zygotic segmentation genes (Gaul and Jackie, 1987; Carroll et al. 1986; Lehmann, 1988). Cytoplasmic transplantation experiments suggested that posterior activity is distributed in a gradient with its source at the posterior end (Lehmann and Nüsslein-Volhard, 1986). It might have been conceivable that at a given position along the anterior-posterior axis the concentration of posterior activity would specify the segmental pattern and its orientation. However, this study suggests that the posterior group genes do not control the polarity of the segmental pattern in a concentration-dependent manner. Indeed, we can show that, irrespective of the maternal genotype, the polarity of the abdominal segments depends on the presence or absence of the zygotic gap gene til. Our data do not rule out that the expression of other yet unknown gap genes is important for abdominal segmentation as well. We would predict that in different genetic backgrounds the pattern of expression of such genes would be altered coordinately with kni and Kr.

Our conclusions may not be limited to the posterior system: double mutants between exuperantia which affects the localization of bicoid RNA (Berleth et al. 1988) and mutants which affect the posterior system result in mirror image duplications of anterior segments (Schiipbach and Wieschaus, 1986). The polarity and identity of the pattern elements formed seem independent of the orientation of the bicoid gradient (Struhl et al. 1989) and may thus be directed by the relative position of gap gene expression. These findings may suggest that all maternal systems act in a strict hierarchical manner, such that once the domains of gap gene expression have been established, maternal information becomes dispensable.

It is not clear how the pattern of gap gene expression is translated into the repetitive transverse stripes of pair-rule and segment polarity gene expression. It is conceivable from the experiments presented here that a particular combination of two gap genes initiates the expression of a given stripe of a pair-rule gene. In a particular region, however, the relative concentration of a single gap gene product may be critical for the expression of a particular pair-rule gene stripe. Indeed, the medial abdominal segments (A3-A6) are most sensitive to reduction in kni activity. According to the fate map, the domain of kni expression (between 30% and 45% egg length, Nauber, 1988) gives rise to the primordia of segments A3-A5 (Lohs-Schardin et al. 1979). Thus a direct correlation can be established between the development of the medial abdominal segments and the domain of kni expression.

The activation of kni by the posterior group genes is indirect and mediated by the negative effect of posterior activity upon hb. In wildtype embryos, the maternal hb product is distributed in a shallow anterior-posterior gradient. This gradient depends on posterior activity (Tautz, 1988) and small changes in posterior activity are reflected in changes in the fate map (Lehmann, 1988). In the absence of posterior activity, the maternal hb product is evenly distributed in the embryo. This even distribution of maternal hb product permits an extension of the Kr domain posteriorly while kni expression is completely suppressed (Tautz, 1988; Nauber et al. 1988). The expression of the gap genes Kr and kni may thus be very sensitive to changes in the hb gradient within the abdominal region. Since the concentration of maternal hb product is quite low in comparison to the concentration of the zygotic product, low concentrations of hb may have a positive influence on Kr expression while high concentrations of til have a negative effect on Kr expression (Jfickle et al. 1986).

A direct positive influence of hb upon Kr may also account for the establishment of a normal pattern in the absence of the maternal hb product. As we have shown, the proper position of the domain of Kr expression is critical for the establishment of normal polarity. Therefore, knowledge of the mechanisms by which the positions of Kr and kni expression are controlled in these embryos is critical to understand how normal polarity is established in the absence of maternal nos and hb products. At present we can only speculate. Perhaps low levels of zygotic hb product are sufficient to activate Kr and repress kni in the anterior abdominal region and thereby set up an asymmetry in the expression of the two gap gene products. Alternatively, it is conceivable that a low concentration of bed activates Kr while a high concentration of bed inhibits Kr. Although attractive, this model seems unlikely since, in the absence of bed, Kr is expressed (Gaul and Jhckle, 1987 and this study). Analysis of this model and other models will require a more detailed understanding of the mechanisms that control kni and hb expression in the early embryo.

R.L. would like to thank Steve Burden for encouragement and help, Anne Ephrussi, Doug Barker, Francisco Pelegri and Charlotte Wang for criticism on the manuscript and Aji Kron for help with word processing. We are thankful to Ulrike Gaul and Herbert Jäckie for the anti-Krüppel antibody. R.L. acknowledges the support received while at the MRC Laboratory of Molecular Biology, Cambridge, UK.