ABSTRACT
Boundaries of Ultrabithorax expression are mediated by long-range repression acting through the PBX or ABX control region. We show here that either of these control regions confers an early band of β-galactosidase expression which is restricted along the anteroposterior axis of the blastoderm embryo. This band is succeeded by a stripe pattern with very similar anteroposterior limits. Dissection of the PBX control region demonstrates that the two patterns are conferred by distinct cis-regulatory sequences contained within separate PBX subfragments. We find several binding sites for hunchback protein within both PBX subfragments. Zygotic hunchback function is required to prevent ectopic PBX expression. Moreover, the PBX pattern is completely suppressed in embryos containing uniformly distributed maternal hunchback protein. Our results strongly suggest that hunchback protein directly binds to the PBX control region and acts as a repressor to specify the boundary positions of the PBX pattern.
Introduction
Gap segmentation genes are expressed in broad and partly overlapping bands at different positions along the anteroposterior axis of the early Drosophila embryo (Knipple et al. 1985; Tautz et al. 1987; Nauber et al. 1988; Stanojevic et al. 1989; Pankratz et al. 1989). Their limits of expression are determined primarily by maternal genes (reviewed in Nüsslein-Volhard et al. 1987; see also Driever et al. 1989; Struhl et al. 1989). Gap gene products show homologies to transcription factors, mostly to zinc finger proteins (Rosenberg et al. 1986; Tautz et al. 1987; Nauber et al. 1988). They function regionally to control the activity of pair-rule segmentation genes (Carroll and Scott, 1986; Ingham et al. 1986; Frasch and Levine, 1987; Howard et al. 1988; Goto et al. 1989; Harding et al. 1989; Howard and Struhl, 1990), thereby ensuring the correct establishment of segment primordia within their functional domains (Nüsslein-Volhard and Wieschaus, 1980).
A second function that has been ascribed to the gap genes is the control of homeotic gene expression (reviewed by Akam, 1987; Ingham, 1988). This was first indicated by the phenotypes of gap gene mutants (Wieschaus et al. 1984; Lehmann and Nüsslein-Volhard, 1987; Bender et al. 1987). More direct evidence was provided by the findings that the distribution of homeotic gene products is altered in gap mutants (White and Lehmann, 1986; Harding and Levine, 1988; Irish et al. 1989a; Reinitz and Levine, 1990). Irish et al. (1989a) found, by focusing on the very first consequences of gap mutations, that these affect homeotic gene expression as early as they affect pairrule gene expression. Based on this, they suggested that the observed effects of gap mutations are likely to be due to direct interactions of gap gene products with regulatory regions of homeotic genes.
In an attempt to reconstruct the embryonic expression pattern of the homeotic gene Ultrabithorax (Ubx), we identified two separate control regions in this gene (called PBX and ABX) which confer β-galactosidase (βgal) patterns confined to the Ubx expression domain (Müller and Bienz, 1991). /Xs these patterns appear early in the embryo, we asked whether the PBX and ABX control regions might be activated directly by products of the segmentation genes. In particular, we were interested whether the limits of the βgal patterns along the anteroposterior axis might be determined by gap proteins. Here, we analyse the RNA patterns that are conferred by these control regions at the beginning of embryogenesis. We dissect the PBX control region and search it for different cis-acting control elements as well as for gap protein binding sites. We examine the PBX pattern in mutant embryos lacking individual gap proteins. We provide evidence that hunchback (hb) protein directly interacts with the PBX control region as a repressor to confer PBX expression boundaries.
Materials and methods
Fly strains and transformation
Embryos of a cn;ry42 strain were injected with the various constructs, and transformants were isolated and made homozygous as described (Bienz et al. 1988). The following mutant alleles were used: hb7M (Lehmann and Nüsslein-Volhard, 1987), Kr1 (Wieschaus et al. 1984), kniFC13 and osk346 (Tearle and Nüsslein-Volhard, 1987). Homozygous mutant embryos were obtained at a frequency of 1/4 in the case of gap mutations; all embryos obtained from homozygous osk mothers were mutant.
Plasmids
Minimal PBX and ABX fragments (Fig. 1, top two lines), cloned into HZ50PL (Hiromi and Gehring, 1987), were described before (Müller and Bienz, 1991). Four PBX subfragments (Fig. 1) were inserted separately, via a subcloning step into bluescript, as Xba1-Kpn1 fragments into HZ50PL.
In situ hybridisations and antibody stainings
Embryonic sections were prepared and hybridised with uniformly 35S-labelled RNA probes as described (Müller et al. 1989; βgai probe as in Müller et al. 1989; Ubx probe: Stul-Xhol fragment from Ubx cDNA; Gonzales-Reyes et al. 1989). For antibody stainings, whole embryos were fixed and incubated with a polyclonal rabbit serum against βgal protein (Cappell) and/or a rabbit serum against eve protein (Frasch et al. 1987) as described (Lawrence et al. 1987; Tremml and Bienz, 1989).
Footprinting analysis
Extracts containing gap proteins (overexpressed in E. coli with T7 RNA polymerase; Rosenberg et al. 1987) were prepared as described by Kadonaga et al. (1987). Various DNA fragments were end-labelled with Klenow enzyme and used for DNAase I footprinting according to Kadonaga et al. (1987). Both strands of the minimal PBX control region (Pst1/BarnH1 fragment; Müller and Bienz, 1991) were screened for hb, Kr and kni footprints.
Results
We previously described the βgal staining patterns conferred by the PBX and ABX control region in embryos at the extended germ band stage (Müller and Bienz, 1991; Fig. 1). These patterns are limited to parasegments (ps) 6–12 in the case of PBX and to ps5–13 in the case of ABX. βgal staining appears in even-numbered (PBX) or odd-numbered (ABX) parasegments in stripes with sharp anterior margins, coinciding with ps boundaries; the stripes fade towards posterior.
We determined the earliest time point in development at which these patterns become detectable. We hybridised parallel sections of PBX and ABX transformants with radioactive RNA probes to detect βgal transcripts and to compare the distribution of these transcripts to that of endogenous Ubx transcripts. In both cases, we found broad bands of βgal RNA expression at the syncytial blastoderm stage (Fig. 2). The PBX RNA pattern is very strong and appears to precede Ubx expression; the ABX RNA pattern is weaker and becomes detectable at about the same time as Ubx RNA. The margins of the broad βgal bands are not sharp, and their positions are therefore difficult to determine. Also, the posterior limits appear to shift somewhat towards posterior with time, due to progressively increasing accumulation of transcripts. Nevertheless, from measurements of several different blastoderm embryos, we estimate that )3gal RNA is detectable between ∼21 and 47 % egg length in the case of PBX, and between ∼40 and 55 % egg length in the case of ABX (in each case ±2%; the posterior pole corresponds to 0% egg length). Soon after, and before the onset of gastrulation, the broad bands of βgal RNA expression appear to resolve into stripes and thus undergo a pair-rule modulation. At the same time, the anterior margins of βgal expression begin to sharpen. The stripe patterns evidently correspond to the βgal staining patterns as described (Müller and Bienz, 1991). The PBX RNA pattern resembles the distribution of mature Ubx transcripts in the early embryo: the latter extends between ∼20 and 50 % egg length and shows a similar pair-rule modulation (Akam and Martinez-Arias, 1985).
We chose the PBX control region for further analysis because of its resemblance to Ubx RNA expression. Also, PBX-mediated expression is strong and can be visualised at the blastoderm stage by βgal staining, if constructs are used that contain an hsp70 RNA leader instead of the Ubx RNA leader (there appear to be sequences in the Ubx RNA leader that delay early translation; S.Hoppier and M.B., unpublished).
Dissection of the PBX control region
We made four constructs containing different PBX subfragments linked to the hsp70 promoter and the βgal gene (Fig. 1) in order to locate and dissect cis-regulatory sequences required for the early PBX pattern. Several transformant fines were obtained in each case, and their patterns were analysed by βgal antibody staining.
We found that transformants containing the subfragment pbxSB (0.6 kb) show a βgal expression pattern similar to the PBX pattern as described (Müller and Bienz, 1991). A characteristic feature of this pattern are the four stripes in ps6, 8, 10 and 12 which are recognisable in pbxSB transformant embryos at the extended germ band stage (Fig. 3B), apparently with some βgal staining extending into the anteriormost regions of adjacent odd-numbered parasegments (ps5, 7, 9, 11 and 13). We also observe a somewhat lower level of continuous βgal staining between the stripes in the epidermis as well as throughout ps6–13 in the mesoderm. We believe this continous staining to be a remnant of the early βgal expression (see below). The stripe pattern is preceded by a broad band of βgal expression, first visible at the mid-blastoderm stage (Fig. 3A), with somewhat fuzzy boundaries. The limits of this broad βgal band appear to be the same as the limits of early βgal PBX-mediated RNA expression (Fig. 2; the βgal protein band is widened posteriorly compared to the βgal RNA band, probably reflecting a higher sensitivity of protein compared to RNA detection). Double-labelling with βgal as well as evenskipped (eve) antibody (Frasch et al. 1987) confirms that the anterior boundary of the broad βgal band is somewhere within ps6 (βgal staining is clearly detectable in eve stripe 4, or ps7, but undetectable in eve stripe 3, or ps5; Fig. 5A). Posterior βgal staining is approximately co-extensive with eve staining in psl3.
The shorter fragment pbxPB (0.3 kb) confers βgal expression in a broad band similar to the one in pbxSB transformants, although βgal staining is strong only at the ventral side (the presumptive mesoderm), but weaker at the dorsal side (the presumptive epidermis; Fig. 3C). Accordingly, we observe weak continuous βgal staining in the epidermis at later stages, but quite strong continuous staining in the mesoderm (Fig. 3D). The margins of the continuous βgal staining have not sharpened; they are within ps6 and within psl3 in the epidermis. Mesodermal staining extends through ps7–13. Based on the anteroposterior extent of the continuous βgal staining and its relative intensities in the two germ layers as well as within their primordia, we believe that this staining reflects βgal expression from the preceding blastoderm stage. As in pbxSB transformants, we observe some βgal staining in odd-numbered ps of pbxPB transformants; however, the strong βgal staining blocks in even-numbered ps are completely missing.
In contrast, transformants of the near-complementary fragment pbxAS (0.6 kb) show exclusively βgal stripes in ps 6, 8,10 and 12 in the epidermis of extended germ band embryos (Fig. 3F). This pattern is reminiscent of the pattern generated by fushi tarazu (ftz) – βgal fusion genes (Hiromi et al. 1985; Lawrence et al. 1987), suggesting that it may be a consequence of ftz activity. There is no other βgal staining in the ectoderm or in the mesoderm of pbxAS transformants. The ftz-like stripe pattern is first detectable after gastrulation (Fig. 3E); there is no βgal expression prior to this stage.
Finally, transformants of the subfragment pbxAP (0.3kb) do not show any βgal expression, except for some staining of the midline at later embryonic stages (not shown). We conclude that we can separate cis-regulatory elements conferring the early band of βgal expression and the subsequent ftz-like stripe pattern. Although these two patterns are completely different, their limits along the anteroposterior axis are nearly the same.
The PBX pattern in mutant embryos
We previously found that the PBX and ABX control regions contain target sites for repressors mediating Ubx expression boundaries at advanced embryonic stages (Müller and Bienz, 1991). We wondered whether the limits of the PBX pattern are due to repressors already at the onset of expression. The hb gene product, required for the anterior Ubx expression boundary (White and Lehmann, 1986; Irish et al. 1989a), is a good candidate for such a repressor. The PBX expression pattern is flanked by two zygotic hb expression domains (Tautz et al. 1987; Tautz, 1988): hb protein forms a anteroposterior gradient occupying the anterior half of the embryo, but is also found in a stripe near the posterior pole (located between ∼10 and 20 % egg length). We therefore asked whether the PBX pattern is altered in mutant hb embryos.
We crossed a pbxSB transformant line into a strain carrying a loss-of-function hb mutation (Lehmann and Nüsslein-Volhard, 1987) and found that a quarter of the offspring embryos (the putative hb-homozygotes) show an early βgal pattern with strikingly blurred anterior expression boundaries (Fig. 4). Measurements of 11 and 9 embryos, presumed to be wild type or hb−, show that the anterior βgal expression limit is located on average between 45 and 48 % egg length (wild type) or between 54 and 58% egg length (hb−). The most anterior position at which βgal staining becomes maximal is also different in the two classes of embryos (39 % versus 45 % egg length). The posterior expression limit, harder to determine because of the shape of the embryo, is located approximately at 11 % (wild type) or 6% (hb−) egg length; in the same sample of embryos, maximal βgal staining is observed at ∼23 % (wild type) or at ∼14% (hb−) from the posterior pole. By the beginning of germ band extension, we observe extensive ectopic βgal staining in the putative hb− embryos outside of the normal PBX expression domain, most noticeable in the mesoderm where staining extends into the head region (Fig. 4D). Double-labelling with βgal and eve antibody confirms that the embryos with widened βgal bands are the ones lacking zygotic hb function (Fig. 5C). We conclude that both limits of the PBX pattern, as early as this pattern becomes detectable, are determined by the activity of a repressor, and that this repressor is identical with or dependent on the hb protein.
A prediction from this is that the PBX-mediated expression might be suppressible by moderately high levels of uniformly distributed hb protein in the early blastoderm embryo. This situation is created in fertilised eggs laid by homozygous mutant osk mothers (Tautz, 1988). We crossed male pbxSB transformants to homozygous osk females and stained the offspring embryos with βgal antibody. As expected, every embryo from this cross was mutant, as judged by the lack of pole cells (Lehmann and Nüsslein-Volhard, 1986). We could aot detect any βgal protein whatsoever in these embryos, not even at advanced stages (Fig. 4E and F). Thus, the PBX pattern is completely suppressed in embryos in which hb protein is uniformly distributed throughout. While this lack of the PBX pattern could reflect a failure of PBX-mediated activation (due to absence of an activator, e.g. knirps product, as a consequence of the osk mutation; Struhl, 1989; Hüls-kamp et al. 1989; Irish et al. 19896), we think it more likely that it reflects hb protein acting as a repressor on the PBX control region (see below).
We also monitored the PBX pattern in embryos homozygous for the gap mutations Krüppel (Kr) or knirps (kni). Double-labelling with βgal and eve antibody served to identify the homozygous mutants as well as to stage the embryos (Frasch et al. 1987; Frasch and Levine, 1987). Among the blastoderm embryos which show the first unambiguous signs of an altered eve stripe pattern, due to lack of Kr or kni function, we found several with a normal βgal staining band (Fig. 5B, D). In both types of mutants, this band of βgal staining remains normal throughout the early stages. Alterations of the βgal pattern do not become apparent until after the onset of germ band extension (not shown); however, these are restricted to those regions of the embryo in which parasegment primordia are disrupted by Kr or kni mutation (Carroll and Scott, 1986; Frasch and Levine, 1987). These alterations are likely to reflect indirect consequences of these mutations. Therefore, there is no evidence that Kr or kni function is required directly for the PBX pattern.
Gap protein footprints in the PBX control region
We asked whether any of the gap proteins might bind directly to sequences within the PBX control region. We tested this by footprint analysis, using protein overproduced in E. coli. We screened both DNA strands of the PBX fragment (Fig. 1) for putative gap protein binding sites. We found 8 short sequence stretches protected by hb protein as well as 2 sequence stretches each protected by Kr or kni protein (Fig. 6; the Kr footprint shown in Fig. 6A is weak compared to the one found in the pbxAS subfragment; see Fig. 1). The hb and Kr footprint regions each contain a consensus sequence as previously determined (Stano-jevic et al. 1989; Treisman and Desplan, 1989; Pankratz et al. 1989). The regions protected by kni protein show sequence similarities with 6 additional kni footprint regions found elsewhere in the PBX control region (downstream of the minimal PBX fragment; Fig. 6C). From these, we deduce a kni binding consensus sequence T/A AATGG A/G A/C C.
Discussion
The early band of PBX-mediated expression is positioned between the two hb protein expression domains (Tautz et al. 1987; Tautz, 1988). Its widening anteriorly as well as posteriorly due to absence of zygotic hb function led to our main conclusion that hb protein acts as a repressor to position the PBX expression boundaries. The widened PBX pattern is still limited along the anteroposterior axis; however, the altered boundaries are most probably due to the maternal hb protein in the anterior region (Tautz, 1988; see also Struhl, 1989; Hülskamp et al. 1989; Irish et al. 1989b) and perhaps to tailless protein in the posterior region of the embryo (Reinitz and Levine, 1990). Our conclusion is reinforced by the observation that complete suppression of the PBX pattern is observed in a situation where hb protein is distributed uniformly throughout the embryo. A similar boundary-determining activity of hb protein with respect to the expression domains of other gap genes has been demonstrated previously (Hülskamp et al. 1990).
We found a cluster of 7 hb binding sites within a small fragment of the PBX control region that confers a PBX-like expression pattern in the blastoderm embryo. These binding sites are evenly distributed between two hardly overlapping subfragments (3 on the pbxAS, 4 on the pbxPB subfragment) either of which confers a βgal expression pattern with essentially the same boundaries, although the patterns themselves are completely different from each other. This, together with the fact that the boundaries in both cases depend on hb function, suggests strongly that hb protein exerts its boundary-determining function by binding as a repressor directly to the PBX control region. Recently, an expression pattern conferred by an intronic Ubx control region was described which resembles the PBX pattern and whose anterior boundary is determined directly by hb protein (Qian et al. 1991).
The PBX pattern resembles Ubx expression in early embryos with respect to its expression boundaries and its pair-rule modulation (Akam and Martinez-Arias, 1985). As in the case of PBX-mediated expression, Ubx expression limits are dependent on zygotic hb function (White and Lehmann, 1986; Irish et al. 1989a), and Ubx expression is prevented in embryos containing uniformly distributed maternal hb protein (Irish et al. 1989a). Because of these parallels between PBX-mediated expression and early Ubx expression, it is very likely that the same molecules and mechanisms are responsible for both expression patterns.
Early PBX-mediated expression is not affected by the absence of Kr function, nor is the pattern reported by Qian et al. (1991), although initial Ubx expression at the blastoderm stage becomes undetectable under these conditions (Irish et al. 1989a). As expected from our result, the one strong Kr binding site that we found in the PBX control region is located within the pbxAS subfragment which is not capable of mediating early blastoderm expression. Early PBX-mediated expression must therefore be conferred by an activator other than Kr, perhaps by a general activator which is uniformly distributed in the embryo. In any case, initial activation at the blastoderm stage appears irrelevant for the generation of expression boundaries: the subfragment pbxAS is capable of conferring PBX expression boundaries without mediating expression prior to gastrulation (Fig. 1). Similarly, boundaries of Ubx expression can be generated in Kr mutant embryos although they lack initial Ubx expression (Irish et al. 1989a). Clearly, the two processes are separable and independent of each other. Whereas boundaries of Ubx expression are essential for the developing embryo (Gonzales-Reyes et al. 1990; Mann and Hogness, 1990), it is unclear whether the initial Ubx expression at the blastoderm stage is functionally significant.
We previously found that a mechanism of long-range repression mediated by the PBX control region acts to confer PBX expression boundaries at advanced stages, and that this long-range repression is dependent on Polycomb function (Müller and Bienz, 1991). We now find that the PBX expression boundaries themselves are generated by a repression mechanism. This however, does not require Polycomb function (Müller and Bienz, 1991) and may not act at a distance as we have not been able to separate PBX-mediated repression from PBX-mediated activation. It thus appears that PBX expression boundaries, and probably also Ubx expression boundaries, are generated in a two-step process involving two types of repression mechanisms. All current evidence suggests that hb protein is the primary repressor responsible for the first step of repression. Interestingly, this initial repression can be maintained throughout embryogenesis, as suggested by our result with osk mutant embryos, far beyond the developmental stage at which hb protein becomes undetectable (Tautz, 1988). It is possible that the second step of repression (Müller and Bienz, 1991) is dependent on the first one and thus on the same primary repressor, hb protein may therefore assume a pivotal role in the generation of stable and heritable Ubx expression boundaries.
ACKNOWLEDGEMENTS
We would like to thank Manfred Frasch for the eve antibody and Jani Nüsslein-Volhard for mutants. This work was supported by the Ernst Hadom Stiftung (fellowship to C.-C.Z.) and by the Swiss National Science foundation (grant nr. 31–26198.89 to M.B.). The part of the work done at the University of München was supported by EMBO (short term fellowship to C.-C.Z.), by the Deutsche Forschungsgemein-schaft (grant nr. Ja 312–4 to H.J.) and by the Fonds der Chemie.