bicoid and the terminal system activate tailless expression in the early Drosophila embryo

Four maternal systems of genetic information are involved in setting up the pattern of the Drosophila embryo. Three establish positions along the anterior-posterior axis and one establishes positions along the dorsal-ventral axis (see review by Nusslein-Volhard, 1991). The activity of these systems results in the transcriptional activation of zygotic genes in an ordered array that subdivides the embryo into different regions. In particular, the anterior system controls the gnathal and thoracic regions, while the terminal system controls the two termini, i.e., the acron (defined here as the brain and parts of the head skeleton) and the telson (defined here as the dorsal portion of the eighth abdominal segment and all structures posterior to it) (Nusslein-Volhard et al., 1987).

The pattern deletions observed in these maternal effect mutants are essentially non-overlapping, leading to the view that the anterior and terminal systems are largely independent of each other (Nusslein-Volhard et al., 1987). An exception to this generalization is provided by an additional feature of the bcd phenotype: in embryos from bcd females, the acron is replaced by a telson (Frohnhofer and Nusslein-Volhard, 1986). Therefore, while the maternal terminal system suffices to establish a telson at either pole and bcd the gnathal and thoracic regions, the combined activity of both the maternal terminal system and the bcd morphogen are required to establish the acron (Frohnhofer and Nusslein-Volhard, 1986).

The anterior domain of the early Drosophila embryo can be viewed as a field in which positional information is provided by a gradient of the bcd protein morphogen. The bcd protein is a homeodomain transcription factor distributed in a monotonic gradient with its highest concentration at the anterior pole; different concentrations of the bcd protein along the anterior-posterior axis lead to expression of different zygotic genes (Driever and Nusslein-Volhard, 1988a and b; Tautz, 1988; Driever and Nusslein-Volhard, 1989; Struhl et al., 1989; Driever et al., 1989; Gaul and Jackle, 1989; Dalton et al., 1989; Hülskamp et al., 1990; Finkelstein and Perrimon, 1990).

The maternai terminal system also appears to establish a system of graded positional information (Casanova and Struhl, 1989). The terminal genes specify the components of a signal transduction path-way that is thought to function as follows. A uniformly distributed receptor tyrosine kinase, encoded by the torso (tor) gene, is activated at the poles of the embryo by a locally released ligand controlled by the torsolike (tsl) gene (Sprenger et al., 1989; Casanova and Struhl, 1989; Stevens et al., 1990). The tor receptor then activates the serine/threonine kinase encoded by the l(1)polehole (l(1)ph) gene, the Drosophila homolog of the vertebrate proto-oncogene c-raf-1 (Mark et al., 1987; Nishida et al., 1988; Ambrosio et al., 1989a and b). Ultimately, the tor/D-raf pathway activates one or more as yet unidentified transcription factors that control the expression of zygotically expressed, terminal-specific genes.

The requirement for both the terminal system and bcd in the anterior terminal domain indicates that the combined activity of these two maternal systems is necessary to establish the specific program of gene expression required for the development of the acron. While shared zygotic targets have been identified, in these cases bcd and the terminal system have been found to act independently of one another and as antagonists, the former activating and the latter repressing transcription of the same gene (Finkelstein and Perrinton, 1990; Tautz, 1988; Driever and Niisslein-Volhard, 1989; Struhl et al., 1989). The regulation of the zygotic terminal gene tailless (tll) exemplifies a different mode of interaction between these two maternal systems: bcd and the terminal system function together to activate and to repress tll expression in specific domains in the anterior of the Drosophila embryo.

The requirement for tll gene function in the terminal regions is shown by the lack of acron derivatives at the anterior and the absence of the telson at the posterior of tll mutant embryos (Strecker et al., 1988). While initially activated in two symmetrical caps at both poles, expression of the tll gene becomes restricted to a dorsolateral stripe at the anterior of blastoderm stage embryos (Pignoni et al., 1990). We show here that the early tll expression in two terminal caps is mainly under the control of the maternal terminal system, while the anterior tll stripe is generated in response to complex interactions between the anterior, terminal and dorsal-ventral systems. Our analysis of the regulation of tll expression by the anterior and terminal systems provides the first evidence that two positional information systems are required together to both activate and repress transcription of a specific gene.

Whole-mount in situ hybridization was carried out with single-stranded DNA probes labeled by unidirectional PCR (Nipam Patel, personal communication), or with double-stranded DNA probes labeled by random priming with digoxigenin-dATP (Genius kit, Boehringer Mannheim), as describcd previously (Pignoni et al., 1990). Embryonic stages are those of Campos-Ortega and Hartenstein (1985); nuclear cycles were determined as in Foe and Alberts (1983).

Embryos were examined and photographed using a Zeiss Axiophot microscope with differential interference contrast optics Measurements were made from photographs in which all views were lateral Measurements were made midlaterally for the early anterior and posterior caps in syncytial blastoderm embryos, and for the posterior cap in cellularizing embryos; the position of the resolved dorsolateral stripe in cellularizing embryos was determined on the dorsal side. Ten or more embryos of the appropriate stage were measured in each case (with the exception of cellular blastoderm stage exu vas embryos, of which five were measured). All measurements were converted to percentage EL Sample mean and standard deviations (between 1 and 2%) were calculated; differences of less than 2% EL were considered insignificant The wild-type standard was taken to be the pattern of tll expression in Oregon-R embryos stained separately or mixed with mutant embryos.

A construct expressing a β-gal-tll fusion protein was made by inserting the 1150 PstI-HindIII fragment (this encodes amino acids 38-421 of the tll protein) from the cDNA N4 (Pignoni et al., 1990) into the vector pTRBO (Burglin and DeRobertis, 1987). A construct expressing a glutathione-S-transferase-tll fusion protein was made by adding EcoRI linkers to a 1200 bp TaqI fragment (this encodes amino acids 25-425 of the tll protein) from the cDNA N4, digesting with EcoRI and HindIII, and inserting this piece into a pGEX vector (Smith and Johnson, 1988) that had been modified to contain a HindIII site. The β-gal-tll fusion protein was induced with IPTG and purified by SDS-PAGE; a homogenized gel slice containing the fusion protein was used to immunize rats. Initial injections were with 100 μ g of fusion protein in Freund’s adjuvant (complete for the first injections, incomplete for subsequent boosts). The final boost was with 500 μg of the glutathione-tll fusion protein that was isolated as inclusion bodies according to the method of Hoey (1989), and sonicated prior to injection.

Embryos were fixed and stained as describcd by Hartenstein and Campos-Ortega (1986), with the following modifications. The anti-tll antiserum, biotin-SP-AffimPure anti-rat antibodies and streptavidin-HRP (the latter two from Jackson ImmunoResearch) were diluted 1:10 and preadsorbcd to 3-20 hour wild-type embryos, then used at final dilutions of 1.400, 1:2000 and 1:2000, respectively The staining reaction was carried out in the presence of N1Cl₂ using the Vector Laboratories DAB substrate kit for HRP. After they were stained, embryos were washed with PBST (PBS + 0.3% Triton X-100), dehydrated in ethanol and acetone, and mounted m Poly/Bcd P12 (Polysciences) Approximately one-quarter of the embryos collected from a Df(3R)tll^PGX stock (Df(3R)tll^PGX deletes the entire tll gene, Pignoni et al, 1990) are not stained by the anti-tll antiserum, consistent with the notion that this antiserum stains only the protein encoded by the tll locus in early embryos.

The strongest available mutant alleles were selected, based on previous phenotypic analyses of the homozygous and hemizygous combinations (L. Ambrosio personal communication; Klingler, 1989, Sprenger et al., 1989) One terminal mutant allele is available for each of the fs(1)N and fs(1)ph loci: fs(1)N²¹¹ and fs(l)ph¹⁹⁰¹ are hypomorphic, as they are the only alleles that do not result in the more severe collapsed egg phenotype (Degelmann et al., 1986 and 1990; Perrimon et al., 1986). Germline clones of 1(1)ph^EA75 were generated by using the dominant female sterile technique (Pemmon, 1984, Pemmon et al., 1984). Other alleles used were tor^XRI, a 9.5 kb deletion of the locus (Sprenger et al., 1989), and tor^PM51, a strong allele (Schiipbach and Wieschaus, 1986); tsl⁴⁹¹ and tsl¹⁴⁶, both strong alleles, and tsl^MK, the weakest tsl allele (Klingler, 1989), trk^HD, a strong allele (Schupbach and Wieschaus, 1986); l(1)ph^EA75 and l(1)ph^11-29, both strong alleles (Perrimon et al, 1985; Ambrosio et al., 1989a); bcd^E1, a small deletion of the locus (Berleth et al., 1988), exu^PJ, a strong allele (Schupbach and Wieschaus, 1986); and d1¹, a strong allele (Roth et al., 1989). Females carrying four copies (bcd 4+) and six copies (bcd 6+) of the wild-type bcd gene were obtained from the 5.8/FM7 stock (Dnever and Ntisslein-Volhard, 1988b). Double mutant chromosomes used were tor_WK trk_RA (Schiipbach and Wieschaus, 1986); bcd^E1tsl¹⁴⁶, both strong alleles (Lehmann and Frohnhofer, 1989); and exu^PJvas^PD(Schüpbach and Wieschaus, 1986). The tor^XR1,tsl⁴⁹¹, tsl¹⁴⁶, tsl^MK, and trk^HD stocks were provided by R. Tearle and C. Nusslein-Volhard; the 1(1 )ph^EA75 stock by L. Ambrosio; the tor^WKtrk^RA stock by J. Casanova, the bcd^E1tsl¹⁴⁶ stock by R. Lehmann; the fs(1)N²¹¹, fs(1)ph^19m, bcd^E1, and tor^PM51 stocks by T. Strecker; the multiple-bcd⁺ stock by N. McGinnis; the exu^PJ and exu^PJ vas^PD stocks by T Schupbach; the dl¹ and ovo^D1 stocks by K. V Anderson. The stock P75.2-3, carrying extra copies of the wild-type tll gene, was obtained from an original transformant line that had two inserts, one on the second and one on the third chromosome (Baldarelli, 1990).

We have increased the sensitivity of our detection of tll transcripts and have extended our analysis of tll localization to the protein level, tll mRNA was detected by in situ hybridization to whole embryos using a full-length cDNA clone (Pignoni et al., 1990), while tll protein was detected with a rat polyclonal anti-i// antibody (see Materials and methods).

Transcripts of the tll gene are first detected during nuclear cycle (NC) 9 as two dots of stain in nuclei in the terminal regions of the syncytial blastoderm embryo (Fig. 1A, and data not shown). Staining is stronger in the most terminal nuclei, which appear labeled throughout, and progressively less intense in subterminal nuclei, appearing as two dark dots and then two very faint dots of stain (Fig. 1A). During NC10, 11 and 12, staining increases first over the nuclei (Fig. IB) and then in the surrounding cytoplasm, forming two solid caps of staming at both termini (Fig. 1C).

After the formation of the terminal caps, the RNA expression pattern resolves into smaller domains at both anterior and posterior termini (Fig. 1D-F). By the beginning of celhilarization, the posterior cap has retracted from approximately 20 to 16% egg length (EL), while expression at the anterior pole has undergone a senes of dramatic changes. Soon after anterior cap formation, staining is lost, first from the extreme anterior tip progressing toward the posterior (Fig. ID), and then from the ventral midfine progressing laterally (Fig. IE); at the same time, expression on the dorsal side extends posteriorly by a few per cent EL. As a result, by the end of the syncytial blastoderm stage, the anterior tll domain has become a horseshoe-shaped stripe located between 75 and 88% EL and extending about two-thirds of the way towards the ventral midline (Fig. IF). Although activation appears to occur simultaneously at both poles, staining in the anterior domain (cap or stripe) is generally weaker than staining in the posterior domain.

The tll protein appears m the same pattern as the RNA. Two caps of tll protein are seen in late syncytial blastoderm stage embryos (Fig. 1G), while the resolved pattern is visible in cellular blastoderm stage embryos (Fig. 1H). As was seen for the RNA staining, the intensity of antibody staining in the posterior cap is greater than that at the anterior. The pattern of tll expression detected in cellular blastoderm stage embryos corresponds to the domains deleted in tll mutants as projected onto the blastoderm fate map (Pignoni et al., 1990).

In what follows, the domains of tll expression in syncytial blastoderm stage embryos are referred to as anterior and posterior caps (Fig. 1C) while the anterior tll domain in cellular blastoderm stage embryos is referred to as the tll stripe (Fig. IF). The repression of tll extending from the anterior tip towards the posterior will be referred to as anterior repression (Fig. ID), while the repression extending from the ventral midline laterally will be referred to as ventral repression (Fig. IE).

Several lines of evidence suggest that the initial activation of tll transcription at both poles of the embryo is controlled by the maternal terminal system. Phenotypic analysis of loss-of-function mutants shows that the terminal regions deleted in tll embryos are large subdomains of those deleted in tor or trunk (trk) embryos (embryos from homozygous maternal effect mutant mothers will be referred to by the maternal genotype in what follows) (Strecker et al., 1988). Gain-of-function mutations in the tor gene give an opposite phenotype in which the terminal domains are expanded; this expansion depends on the ectopic expression of tll that occurs in these mutant embryos (Klingler et al., 1988; Strecker et al., 1989; Steingrfmsson et al., 1991).

To examine the role of the maternal terminal pathway in activation of tll transcription, we analyzed tll expression in the strongest alleles available of each of the six describcd maternal terminal genes: tor, trk, tsl, l(1)ph,fs(1)Nasrat(fs(1)N), and fs(1)polehole (fs(1)ph) (Schüpbach and Wieschaus, 1986; Stevens et al., 1990; Ambrosio et al., 1989a; Perrimon et al., 1985, 1986; Degel man n et al., 1986, 1990). Essentially the same pattern was obtained when the spatial distribution of either tll RNA or protein was examined: at the posterior, little or no tll expression can be detected in either syncytial or cellular blastoderm embryos; at the anterior, the early cap does not appear while an anterior stripe, although abnormal, appears by the late syncytial blastoderm (Fig. 2A-D and data not shown). These results are seen even in null tor^XRI embryos and in tor trk embryos, mutant in two components of the terminal pathway (Fig, 2A,B and E). The maternal terminal system, therefore, is necessary to activate tll expression in the terminal caps; at the anterior, however, an abnormal tll stripe appears in the absence of a functional terminal system.

Since the bcd morphogen is active in the anterior, and is required to make the distinction between anterior and posterior termini (Frohnhofer and Nüsslein-Volhard, 1986), it seemed likely that bcd might be involved in activating tll at the anterior. Consistent with this hypothesis, bcd tsl embryos, which lack both anterior and terminal gene functions, show no tll expression at either pole (Fig. 2F). We conclude that bcd. is responsible for the activation of tll transcription at the anterior pole in terminal mutant embryos.

We can therefore use bcd embryos to examine the activation of tll transcription by the terminal system alone. In bcd embryos, tll transcription is activated at both poles in two terminal caps (Fig. 3A,B). Two features of these early caps differ from those seen in wild-type embryos. First, the level of staining of the caps is approximately equal, in contrast to the wild-type situation in which staining is more intense at the posterior. The presence of bcd protein, therefore, appears to have a negative effect on the anterior cap of tll expression. Second, the caps are less symmetrical and, in some embryos, considerably larger than those seen in wild-type embryos, extending up to 45% EL from the posterior pole and up to 72% EL from the anterior pole (compare Fig. 1C with Fig. 3A). By the beginning of cellularization, these caps have generally retracted towards the termini, extending from 0 to 17%

EL in the posterior and from 87 to 100% EL in the anterior (see Fig. 3C for the RNA and 3D for the protein). The anterior cap does not resolve into the stripe seen in wild-type embryos, but mimics the behavior of a normal posterior cap of tll expression (Fig. 3E), disappearing from the ectopic anterior amnioproctodeal invagination during germ band extension (Fig. 3F). The presence of caps at both poles of bcd embryos and the absence of caps in terminal mutant embryos indicate that the maternal terminal system activates tll expression in these domains.

When activation is only by bcd (in tor embryos for example) tll RNA is first detected during NC 12 in the dorsal nuclei at the anterior tip of the embryo (Fig. 4A); this domain of expression then spreads posteriorly and ventrally (Fig. 4B). By the beginning of cellularization, the most anterior expression is lost, and a broad dorsolateral stripe is seen between 80 and 96% EL (Figs 2A and 4C). This domain is expanded and shifted anteriorly compared to the tll stripe seen in wild-type embryos (compare Figs IF and 2A). That this stripe is activated by bcd in a concentration-dependent fashion can be demonstrated by altering the bcd concentration gradient (Driever and Nüsslein-Volhard, 1988b). In embryos from tor mutant females bearing 4 wild-type copies of the bcd gene (tor bcd 4+), the anterior domain extends more posteriorly than in tor embryos containing the normal dosage of bcd, i.e. from 70 to 94% EL (Fig. 4D).

To investigate the relationship between the bcd-dependent tll domain seen in terminal mutant embryos (Fig. 2A-E) and the tll stripe seen in wild-type embryos (Fig. IF), we tested the effect of altering either the bcd gradient or the activity of the terminal system on anterior tll expression.

Evidence that the position of the tll stripe in wild-type embryos is under the control of the bcd morphogen comes from experiments in which the bcd gradient is altered by changing the number of bcd copies. The tll stripe, which lies between 88 and 75% EL in wild-type embryos, is narrower and shifted anteriorly in bcd 1 + embryos, while it is shifted posteriorly in bcd 4+ embryos and becomes broader in bcd 6+ embryos (Fig. 5E; compare Fig. IF with Fig. 5A-C for the RNA pattern, and Fig. 1H with Fig. 5D for the protein pattern). These results suggest that bcd plays a critical role in the formation of the tll stripe. Since the shape and position of this domain is abnormal whenever terminal system activity is absent (Figs 2A-E and 4D), the maternal terminal pathway must also be involved in the correct establishment of the stripe domain. Further manipulation of the terminal and anterior activities was carried out to elucidate the roles played by these two systems in forming the tll stripe.

The domain in which the terminal system is active can be reduced by using hypomorphic alleles of the tsl gene, which is believed to control the ligand for the tor receptor (Stevens et al., 1990). Embryos mutant for the weak allele tsl^MK, although showing no cuticular defects at the anterior (i.e. acron development is normal), show posterior phenotypes ranging from an almost complete telson to a lack of all structures posterior to A7 (Klingler, 1989).

In tsl^MK embryos, the initial activation of tll in two symmetrical domains is not seen. By cellular blastoderm stage, a smaller than normal posterior cap (extending up to 7% EL from the posterior pole) can be seen in approximately one third of the embryos (Fig. 6B); the tll stripe, on the other hand, narrower by a few per cent EL but otherwise normal in shape and location, is present in all tsl^MK embryos (Fig. 6A,B). These results indicate that when the activity of the terminal pathway is reduced to a point where activation of tll in terminal caps is no longer detected (Fig. 6A) an essentially normal tll stripe can stlll be formed. Since all tsl^MK embryos develop normal heads, the stripe seen in these embryos apparently provides sufficient tll activity for normal acron development. Thus the requirement for the terminal system in estabfishing the stripe in the correct position (Fig. 2A-E) and in forming the acron (Schiipbach and Wieschaus, 1986; Nüsslein-Volhard et al., 1987) can be satisfied by a relatively low level of terminal system activity (i.e., that present in tsl^MK embryos).

A shallower bcd gradient, i.e. lower than normal at the anterior tip but higher than normal at more posterior positions, is obtained in exuperantia vasa (exu vas) embryos (Driever and Nüsslein-Volhard, 1988b; Struhl et al., 1989). In these embryos, tll expression is initially activated in two, smaller than normal, terminal caps (Fig. 6C). By the beginning of cellularization, the anterior cap has retracted to a smaller domain while a thinner than normal tll stripe appears between 86 and 92% EL (Fig. 6D). This stripe, in contrast to that seen in strong terminal class mutants, behaves hke a wildtype stripe in that it separates into two lateral anterior domains late in the cellular blastoderm stage (data not shown). The anterior cap, on the other hand, behaves like the posterior cap in that it disappears early during gastrulation (Fig. 6E). The lower level of bcd protein at the anterior of exu vas embryos presumably results in the anterior shift of the tll stripe. The position of this stripe is consistent with the expression of another bcd-regulated gene empty spiracles (ems); ems is expressed just posteriorly to the tll stripe in bcd 1 +, 2+ and 4+ embryos, and appears at about 84% EL in exu vas embryos (Dalton et al., 1989; Figs IF, 5A,B, and 6D). The absence of the tll stripe in exu vas tor embryos (Fig. 6F) is consistent with the lack of tll expression posterior to 80% EL in strong terminal mutant embryos and indicates that the level of bcd protein at the anterior of exu vas embryos is not sufficient to activate tll in the absence of terminal gene activity.

The results obtained with tsl^MK and exu vas embryos indicate that the activation of the tll stripe requires the activities of both bcd and the terminal system. In the absence of the maternal terminal system, higher levels of bcd are required to activate tll than in wild-type. One apparent exception to this is that the tll stripe extends as far posteriorly in bcd 4+ tor^XRI embryos as it does in bcd 4+ embryos (70% EL; compare Figs 4D and 5B). Since terminal system activity does not extend much further than 20% EL from the poles (Casanova and Struhl, 1989; Figs 1C and 5B,C, see Repression below), activation of tll by bcd around 70% EL in bcd 4+ embryos is probably independent of terminal gene function.

The resolution of the tll domain from a cap into a stripe at the anterior, but not at the posterior, and the lack of this resolution in bcd embryos, indicate that anterior repression requires bcd function. Since the domain in which tll is repressed at the anterior expands as the number of wild-type copies of the bcd gene is increased from 1 to 4 (Figs 5A, IF and 5B), the extent of anterior repression is dependent on the concentration of bcd protein.

As shown by the anteriorward expansion of the tll stripe in all terminal system mutants (Fig. 2A-E), anterior repression also requires terminal gene activity. The requirement for the terminal system in anterior repression can be observed in a bcd 2+ background (compare the wild-type embryo in Fig. IF with the tor embryo in Fig. 2A) as well as in a bcd 4+ background (compare the bcd 4+ embryo in Fig. 5B with the tor bcd 4+ embryo in Fig. 4D). That the anterior repression domain extends as far posteriorly as 78% EL but not beyond when the number of bcd gene copies is increased from 4 to 6 (Fig. 5B,C) can also be explained by the dependence of anterior repression on terminal gene activity.

The retreat of the tll stripe from the ventral side of the embryo (Fig. 1E,F) suggests a role for an additional patterning system in regulating tll transcription. Positional values along the dorsal-ventral axis are initially established by the maternal dorsal group genes. These genes encode the components of a signal transduction pathway that controls the graded nuclear localization of a transcription factor, the dorsal (dl) protein (Steward 1987, 1989; Rushlow et al., 1989; Roth et al., 1989; reviewed by Nüsslein-Volhard, 1991).

Examination of tll expression in dl embryos reveals that the dl gene product is required for ventral repression. While in a wild-type cellular blastoderm stage embryo the tll stripe extends about 240° around the circumference of the embryo (Figs 7A and IF), in dl mutant embryos the stripe extends completely around the circumference of the embryo (Fig. 7B,C). Repression by dl appears to require bcd function, since ventral repression does not occur either at the posterior of wild-type embryos or at the anterior of bcd mutant embryos.

The main conclusion to be derived from the results presented here is that, while the terminal system, as predicted, activates the tll gene in symmetrical domains at the embryonic poles, both the bcd protein and the terminal system act together to establish the anterior tll stripe. The postulated interaction between the anterior and terminal activities in establishing the acron is thus shown to occur at the level of regulation of a single gene, tllt both maternal activities are required together for both activation and repression of this gene. A summary of the genetic control of tll expression by these two activities, as well as by the dorsal-ventral system, is provided in Fig. 8.

The initial pattern of tll activation depends mainly on the terminal system (Figs 1A-C and 8A) and is consistent with the proposed model for the functioning of the terminal pathway (see Introduction). The activation of the zygotic genes huckebein (hkb) and tll in caps of different sizes, and the observation that the posteriormost region of the embryo is most sensitive to loss of terminal gene activity, suggest that there is a gradient of terminal system activity, highest at the poles and decreasing toward the center of the embryo (Pignoni et al., 1990; Weigel et al., 1990; Casanova and Struhl, 1989). Our data on the pattern of tll expression in exu vas embryos also support this hypothesis. The presence of both a cap and a stripe in exu vas embryos (Fig 6D) can be explained if the lower level of bcd protein present at the anterior of these embryos is sufficient to repress the activation of tll transcription around 90% EL, but not sufficient to override the tll activation caused by a higher activity of the terminal system at the very tip. The pattern of tll RNA staining at the posterior of wild-type embryos provides the most direct evidence for graded terminal pathway activity. As longer staining periods are used to reveal lower levels of tll RNA by in situ hybridization, the size of the domain where tll RNA is detected increases by several per cent EL (data not shown). Moreover, the intensity of tll RNA staining in the nuclei of very early embryos decreases from terminal to subterminal positions (Fig. 1A). The maternal terminal genes can therefore be viewed as establishing a coordinate system of positional information (Wolpert, 1989).

By the cellular blastoderm stage, the anterior tll cap has been replaced by a dorsal stripe. While repression of the early cap plays a role in the appearance of the stripe (see below), transcription in the stripe appears to be largely independent of that in the anterior cap. Thus in strong terminal mutant and in tsl^MK embryos, the early caps of tll expression are not seen, but the stripe does appear by late syncytial blastoderm (Figs 2A-E and 6A,B). Also, the posterior border of the stripe extends beyond the border of the anterior cap. Finally, the dependence of the position of the stripe on the bcd gradient (Figs IF, 4D, 5A-D) shows that tll transcription in this domain is activated by specific concentrations of the bcd protein (as indicated in Fig. 8B). The anterior shift of the tll stripe in strong terminal mutant embryos (Fig. 2A-E) shows that the terminal system is necessary for establishing the stripe in the correct position. The appearance of a relatively normal stripe in tsl^MK embryos (where terminal activity is reduced) suggests that a low level of terminal system activity is sufficient for activation of the stripe in the correct position.

The anterior repression domain (indicated in Fig. 8B) is defined as the region of repression seen at the anterior of dl mutant embryos (Fig. 7C). Anterior repression by bcd displays a stricter requirement for terminal gene function than does bcd-dependent activation, since even the highest levels of bcd protein present in wildtype embryos seem to be insufficient to repress tll in the absence of terminal gene function. Thus in tor dl embryos, which have a normal complement of bcd protein but lack the dorsal and terminal systems, there is a complete lack of anterior repression, i.e., the anterior domain appears as a cap (data not shown). The failure of the anterior repression domain to extend beyond 80% EL, as the number of wild-type bcd copies is increased from 4 to 6 (Fig. 5B,C), is also consistent with the notion that terminal activity is required for anterior repression.

That dl acts to repress the tll stripe ventrally (as indicated in Fig. 8B) is shown by the extension of this stripe to the ventral midline in dl embryos (Fig. 7). Since ventral repression is absent at the posterior of wild-type embryos and at the anterior of bcd embryos, repression on the anterior ventral side depends on both dl and bcd. In exu vas embryos, however, ventral repression occurs around 90% EL (where the stripe forms) but not at the anterior tip (where a small cap persists) even though somewhat higher levels of bcd are present at the pole (Fig. 6D; Struhl et al., 1989). This could be explained if dl represses the bcd-dependent activation of tll (the stripe), but not the terminal system-dependent activation of tll (the cap). This hypothesis also provides an explanation for the observation that, although nuclear localization of the dl protein is detected as early as NC 10 (Roth et al., 1989; Rushlow et al., 1989; Steward, 1989), ventral repression is not seen in wild-type embryos untll late syncytial blastoderm (around NC 13), after the onset of anterior repression and after the time when activation by bcd is seen in terminal mutant embryos.

The early (detectable by NC 9) activation of tll transcription at both poles by the terminal pathway argues for a direct activation of tll by one or more maternally provided transcriptional activators. The later activation and repression involving bcd, on the other hand, might be achieved indirectly through zygotic genes which in turn regulate tll transcription. Mutations in genes such as orthodenticle (otd), hunch-back (hb) and hkb (Finkelstein and Perrimon, 1990; Driever and Nusslein-Volhard, 1989; Struhl et al., 1989; Weigel et al., 1990), however, do not significantly affect the anterior expression of tll (FP, unpublished). Thus, if there are zygotic targets of bcd that regulate anterior tll expression, these have not yet been identified.

Alternatively, the bcd protein might activate and repress tll transcription directly. As its concentration increases to a threshold level, the bcd protein might activate transcription of tll by binding to high affinity binding sites in the tll promoter. The similarity between the spreading from the anterior tip of newly synthesized bcd protein and of tll RNA expression in terminal mutant embryos (Fig. 4A-C; Driever and Nüsslein-Volhard, 1988a) supports such a direct activation model. As the bcd protein concentration continues to increase at the anterior pole, binding to low affinity binding sites might repress tll transcription. Models in which the bcd protein both activates and represses have also been proposed for the control of the gap genes Kruppel and giant (Gaul and Jâckle, 1989; Hülskamp et al., 1990; Eldon and Pirrotta, 1991; Kraut and Levine, 1991). While the bcd protein has been shown to directly activate transcription of the hb gene (Driever and Nüsslein-Volhard, 1989; Struhl et al., 1989), a repressor function has not been demonstrated directly.

Since the dl protein can act as a repressor of transcription (Ip et al., 1991) and the tll stripe appears normal in embryos mutant for the ventral zygotic genes twist and snail (Thisse et al., 1987; Boulay et al., 1987; FP, unpublished), ventral repression of tll by dl is probably direct.

The absence of the posterior tll cap from terminal mutant embryos, and the requirement for both the terminal system and the tll gene to form the telson, demonstrate that the posterior cap of tll expression is required to establish the telson. Consistent with this conclusion, the persistence of the anterior cap of tll expression into the cellular blastoderm stage, as is seen in bcd and exu vas embryos, is correlated with the formation of ectopic posterior structures at the anterior. Thus an ectopic posterior midgut invagination is seen in bcd and exu vas embryos (Fig. 3E,F; Schüpbach and Wieschaus, 1986) and an ectopic telson differentiates in bcd embryos (Frohnhôfer and Nüsslein-Volhard, 1986).

The early anterior cap of tll expression seen in wildtype embryos, however, does not appear to be necessary for proper acron development. Thus tsl^MK embryos, even thought they lack this early cap, develop a normal acron. The anterior tll stripe, however, does appear to be required, as it is always present when acron development is normal. That an anterior stripe alone is not sufficient is shown by the fact that the thin stripe present in exu vas embryos and the broader stripe seen in strong terminal mutant embryos do not result in normal acron formation. Since tll tsl double mutant embryos do not differ in either cuticular or nervous system morphology from tsl mutants (Strecker et al., 1988; data not shown), the tll protein present at the anterior of terminal mutant embryos (Fig. 2B) is unable to provide all functions necessary for acron development. Presumably other genes controlled by bcd and the terminal system are also required for this process.

The terminal system and the bcd protein gradient are each independently capable of defining a coordinate system of positional information. Thus the terminal pathway in the posterior and the bcd protein in the prospective gnathal and thoracic regions activate different sets of target genes (Pignoni et al., 1990; Driever and Nusslein-Volhard, 1989; Struhl et al., 1989; Gaul and Jackle, 1989).

That bcd and the terminal system are required together for the formation of the acron suggests that, in addition to their independent functions, the two systems also have common targets. One such common target is the head gap gene otd. In this case, however, the anterior and terminal systems provide separate and independent regulatory inputs: bcd activates otd as a large cap, while the terminal system represses otd in a smaller region at the anterior tip (Finkelstein and Perrinton, 1990). The regulation of the tll stripe is the first describcd example in which two maternal systems work together to regulate both activation and repression of a particular gene. Thus both terminal activity and bcd are necessary to activate transcription in the anterior stripe of tll, as well as to establish anterior repression of tll. Our discovery of these interactions provides an explanation for the dependence of positional values at the anterior of the Drosophila embryo on both maternal systems. Other zygotic genes expressed at the anterior, such as hkb (Weigel et al., 1990) and other yet unidentified loci, might be controlled by both systems in a similar fashion.

We thank L. Ambrosio for advice on generating germline clones and the donation of l(I)ph^11-29 embryos, A. Ip and M Levine for advice on antibody staining, A. Mahglig and L. Ramirez for help with the stocks, and K. V. Anderson, A. Courey, E. DeRobertis, M Levine and L. Zipursky for helpful comments on the manuscript. This work was supported by NIH and NSF grants to J A.L. and Ursula Mandel scholarships to F.P. and E.S.