Gap gene properties of the pair-rule gene runt during Drosophila segmentation

The protein encoded by the Drosophila runt gene is a member of a newly identified family of transcriptional regulators that participate in a variety of important developmental processes. Work on runt was initiated due to its vital role in generating the segmented body pattern during the early stages of embryogenesis (Nüsslein-Volhard and Wieschaus, 1980; Gergen and Wieschaus, 1985). Subsequently it was found that runt also participates in sex determination and nervous system development (Duffy and Gergen, 1991; Duffy et al., 1991). In both the segmentation and sex determination pathways runt affects the transcription of other genes. The nuclear localization of the Runt protein (Kania et al., 1990) provided an initial suggestion that it may be directly involved in regulating transcription in these pathways.

The recent characterization of cDNAs for a subunit of the mammalian transcription factor PEBP2/CBF (Ogawa et al., 1993b) supports the idea that the Runt protein directly regulates transcription. This factor binds to a conserved core site in the enhancers of mammalian type C retroviruses and the polyoma virus and has roles in regulating transcription in embryonal carcinoma cells as well as in hematopoiesis (for review, see Kagoshima et al., 1993). PEBP2/CBF is a heterodimer of two unrelated protein subunits. One of these, the α subunit is encoded by a small family of related genes, all of which share a significant region of homology with the Drosophila runt gene. This homology region, termed the Runt-domain, is a 128 amino acid region that is responsible for the sequence-specific DNA-binding properties of all of these proteins, including the Drosophila Runt protein (Kagoshima et al., 1993). This strongly suggests that regulation by Runt will involve its binding to DNA. Demonstration of this requires the identification of genes that are directly regulated by runt and localization of the target cis-regulatory sequences responsible for mediating this regulation.

The segmentation pathway in Drosophila has been subject to extensive genetic and molecular analysis, making this an excellent model system for investigating runt’s regulatory functions. Segmentation depends on a hierarchical pathway that in several sequential steps converts broad gradients of maternally provided spatial information into segmentally repeated patterns of gene expression. Regulation of gene transcription plays a central role in this hierarchy. Accordingly, as a first step maternal positional cues are used to regulate gap gene transcription and thereby generate their different broad and overlapping expression domains (Gaul and Jäckle, 1987; Tautz, 1988; Driever and Nüsslein-Volhard, 1989; Finkelstein and Perrimon, 1990; Hülskamp et al., 1990; Eldon and Pirotta, 1991; Kraut and Levine, 1991a; Pignoni et al., 1992). One clear example of how maternal information is used is provided by the maternal anterior system. The primary output of the anterior system is a gradient of Bicoid protein (Bcd), a homeobox transcription factor that activates transcription in anterior regions of the embryo. Concentration-dependent, and thus position-dependent activation can be achieved by varying the number and affinity of Bcd-binding sites in the cis-regulatory regions of downstream target genes (Driever et al., 1989; Struhl et al., 1989; Ronchi et al., 1993).

The next step in the hierarchy involves the transfer of positional information from the gap genes to the pair-rule genes. The gap genes all appear to encode transcriptional regulatory proteins (Rosenberg et al., 1986; Tautz et al., 1987; Nauber et al., 1988; Dalton et al., 1989; Finkelstein et al., 1990; Pignoni et al., 1990; Capovilla et al., 1992). The phasing and overlaps in the expression domains of the gap proteins are used to establish of the expression patterns of the pair-rule genes. The transcriptional regulation that generates stripe 2 of the pair-rule gene even-skipped (eve) is a particularly elegant example of this. A discrete cis-element responsible for this stripe contains binding sites for four different transcription factors, the proteins encoded by the gap genes hunchback (hb), Krüppel (Kr) and giant (gt) as well as the maternal Bcd protein (Stanojevic et al., 1989). The stripe 2 element is activated by Bcd and Hb proteins and the stripe forms in a region of the embryo in which both are present. Conversely, the domains of Kr and Gt expression normally abut the posterior and anterior margins of eve stripe 2, respectively and act to define the boundaries of this stripe by repressing the activity of the stripe 2 element (Small et al., 1991, 1992; Stanojevic et al., 1991).

Runt is a pair-rule gene and is positioned downstream of the maternal and gap genes and upstream of the segment-polarity genes in the segmentation hierarchy. The identification of segment-polarity genes that are directly regulated by runt is complicated by the fact that the expression patterns of other pair-rule genes are altered in runt mutant embryos (Carroll and Scott, 1986; Frasch and Levine, 1987; Ingham and Gergen, 1988; Hooper et al., 1989; Gutjahr et al., 1993). Further, the identification of pair-rule targets of runt is complicated by the network of cross-regulatory interactions between these genes. For example, runt is interpreted as a positive regulator of the pair-rule gene fushi tarazu (ftz). However, the expression of the pair-rule genes eve and hairy, both of which are interpreted as negative regulators of ftz is also affected by runt. These previous observations are based on examining the consequences of reducing runt activity. Here we examine the consequences of over-expressing runt in order to identify transcriptional regulatory events that are most directly influenced by the presence or absence of the Runt protein.

The strategy that we used was to create transgenic Drosophila lines in which the expression of runt is under the control of the inducible hsp70 promoter. Transient over-expression of runt in blastoderm stage embryos dramatically alters the transcriptional regulation of other pair-rule genes. The effects on hairy, eve and runt itself are stripe specific, whereas ftz is more uniformly affected. Somewhat surprisingly, we find that the expression of upstream gap genes is also affected by runt. Evidence is presented that a subset of the effects on the gap genes, as well as the repression of hairy stripe 1 and eve stripe 2 are due to antagonism of Bcd-dependent transcription. Genetic experiments indicate that runt plays a normal role in preventing bcd-dependent gene expression in the central regions of the embryo. The implications of these results on the hierarchical view of Drosophila pattern formation and on our understanding of the function of the Runt protein are discussed.

The hs/runt transgene was constructed in the CaSper P-element trans-formation vector (Pirotta, 1988). Site directed mutagenesis was used to create a BamHI recognition site 9 basepairs upstream of the initiator ATG for the runt protein (Zoller and Smith, 1984). The sequence GGATACGAGATG was altered at the two underlined positions to the sequence GAATTCGAGATG. The initiator ATG for the runt open-reading frame (ORF) is italicized. Site directed mutagenesis was done on a composite cDNA insert (ED, nucleotides 30 to 2492 in Kania et al., 1990) subcloned into the EcoRI site of M13mp19. The integrity of clones containing the desired BamHI site was verified by DNA sequence analysis, and then the 2.248 kb BamHI + EcoRI fragment containing the entire runt ORF was subcloned into the EcoRI site of the plasmid vector BMKS (Stratagene) in a three way ligation with a 450 basepair EcoRI + BamHI fragment base containing the Drosophila HSP70 promoter (courtesy of V. Pirotta, University of Geneva). A portion of the runt cDNA insert extending towards the 3′ end of the mRNA from an SphI site at nucleotide 1625 in the map of Kania et al. (1990) to a HindIII site in the flanking polylinker of BMKS was then replaced by an overlapping 1.8 kb SphI fragment of runt genomic DNA. The resulting construct thus contains the heat-shock promoter with 207 nucleotides of hsp70 5′ untranslated leader, 8 nucleotides of the modified runt 5′ untranslated leader, the entire runt ORF and 3′ untranslated trailer, followed by 1 kb of 3′ flanking runt genomic DNA. This entire fusion gene was removed as an EcoRI + KpnI fragment and cloned into EcoRI + BamHI digested CaSper P-element transformation vector. The KpnI and BamHI sites were both filled in with T4 DNA polymerase, thus recreating a BamHI site in the final construct. In the final construct, the white and hs/runt genes are divergently transcribed. Germ-line transformants were obtained by injecting CaSper:hs/runt DNA at a concentration of 1 mg/ml with 100 mg/ml of the P-transposase helper plasmid pπ25.7 wc (Karess and Rubin, 1984). Of 596 embryos injected with these DNAs, 85 successfully completed embryogenesis and of these 19 developed to adult stages. One transformant line, hs/runt^a was obtained from the fertile surviving G_o adults. Other inserts of the hs/runt transgene were obtained by mobilizing the hs/runt transgene in this line with the Δ2-3 transposon (Robertson et al., 1988). Three different hs/runt lines were tested for their effects on the expression of ftz, eve and gt. They gave similar results after the standard heat-shock treatment.

The following stocks were used as sources for different segmentation mutations. gt: Df(1)62g18/ FM6; Kr: y; Dp (1;2)B89, y⁺Kr⁻/SM1; runt: y w f runt^LB5/FM7c/y⁺Ymal¹⁰² (Lindsley and Zimm, 1992). The strain used as the wild-type control for the effects of heat-shock treatment was the parental y w^67c23 strain used as the host for the establishment of the hs/runt line.

We used Drosophila lines containing the following reporter gene constructs: For hairy stripes 1 and 5, h^KB/lacZ (Pankratz et al., 1990); for eve stripes 2, 3 and 7, eve^5.2/lacZ (Harding et al., 1989; Stanojevic et al., 1991); for the eve autoregulatory, or pair-rule element, eve^4E/lacZ (Jiang et al., 1991); for ftz, ftz/lacC (Hiromi et al., 1985); for a bcd-dependent reporter, HSBG [bcd] (Ronchi et al., 1993), for a bcd + hb-dependent reporter, HSBG [hb+bcd] (J. Treisman and C. Desplan, Rockefeller University).

Embryos were collected on apple juice plates for periods of up to 4 hours at 25°C. Embryos were washed off plates and collected on a nytex net mounted on the end of polypropylene cylinder. In order to insure uniform and efficient heat-shock treatment, the density of embryos was controlled such that they formed a monolayer on the net. Heat-shock treatments were carried out by resting the nets containing the embryos on the top of prewarmed blocks in a 37°C water bath for 20 minutes. After heat-shock treatment, the embryos in the net were removed to room temperature, aged for another 30-40 minutes, and then processed for in situ hybridization. The in situ hybridization protocol used the non-radioactive digoxigenin-based system (Boehringer-Mannheim) essentially as described in Klingler and Gergen (1993) with the exception that the two phase fixation solution contained: 0.1 ml 1 M EGTA, 0.4 ml 37% formaldehyde, 3.5 ml 1× PBS, and 6.0 ml heptane.

Embryos collected and processed in this manner represent a mixture of developmental stages. The text refers to the stage of the embryos at the time of fixation, i.e. approximately 50 to 60 minutes subsequent to the time at which they were initially collected for heat-shock treatment. The expression patterns described are representative of the majority of the embryos of a particular developmental stage. All heat-shock experiments have been repeated on at least three separate occasions with similar results. Less severe, as well as more severe alterations are sometimes observed in other embryos, presumably due to subtle differences in their developmental stage and/or in their heat-shock response.

Antisense RNA probes for in situ hybridization were as follows:

Immunohistochemical detection of the Bcd and Runt proteins was done as in Kania et al. (1990). The anti-Runt antibody was a cocktail of four different monoclonal antibodies, each at a final concentration of 50 mg/ml. The anti-Bcd antibody was prepared from supernatants of monoclonal line 733.3 obtained from D. Jay (Harvard University). Ammonium sulfate precipitation was used to purify IgG fractions from this supernatant. These concentrated antibody preparations were used at a final concentration of 100-250 mg/ml after pre-absorbing them against aged (>4 hours) Drosophila embryos.

To investigate the regulation of segmentation gene expression by runt, we generated transgenic lines of Drosophila where runt expression is under control of the inducible hsp70 promoter. Heat-shock treatment leads to a high level of runt mRNA and protein accumulation in all somatic cells of blastoderm stage embryos homozygous for the hs/runt transgene (data not shown). The failure of the hs/runt transgene to be expressed in pole cells is not understood, but is similar to results obtained with other hs transgenes (e.g. Capovilla et al., 1992). Whole mount in situ hybridization was used to examine the RNA expression patterns of other segmentation genes and thus determine the regulatory consequences of this ectopic expression. As shown below, there are several dramatic pattern alterations in hs/runt embryos. The severity of these alterations is influenced by the duration and degree of the heat shock. All of the experiments described below used a 20 minute heat shock at 37°C, followed by a 30 to 40 minute recovery period before the embryos were processed for in situ hybridization. This protocol gives clear and reproducible effects in hs/runt embryos. Importantly, this treatment does not alter the patterns in genetically wild-type embryos that are processed identically. A second important factor for interpreting these experiments is the developmental stage of the affected embryos. In the text below, the developmental stage is described based on the morphology at the time of fixation. As all experiments followed the same protocol, this method allows for accurate comparison of embryos from different experiments. Finally, the expression patterns described below are representative of the majority of embryos at a particular developmental stage.

As expected, the pair-rule expression patterns of ftz, eve and hairy are altered in hs/runt embryos. From analyses of expression patterns in runt mutant embryos, we anticipated that eve and hairy would be repressed, and that ftz would be activated by hs/runt treatment. Aspects of the eve and hairy patterns are repressed in hs/runt embryos, but unexpectedly these effects are markedly stripe specific. For hairy, the most notable difference is the repression of stripes 1 and 6 (Fig. 1B). Stripes 2 and 5 are also reduced in intensity. In contrast to these repressive effects, hairy stripes 3 and 4 appear to become more intense in hs/runt embryos. In the case of eve, the most dramatic alteration is the elimination of stripe 2 (Fig. 1F). The intensity of eve stripes 4 and 7 is also consistently reduced and stripes 5 and 6 are not well resolved in hs/runt embryos. These stripe-specific effects of hs/runt on eve and hairy are most apparent at the mid-blastoderm stage when the pair-rule patterns are normally well formed, but similar effects are also observed on both younger and older blastoderm stage embryos.

In contrast to the above results, ftz is activated by hs/runt and in a much more uniform manner (Fig. 1J). This activating effect is readily apparent in mid-blastoderm stage embryos, but again similar effects are also observed in both older and younger embryos. Expression of ftz in the interstripe regions is de-repressed and there is an almost uniform, high level of ftz mRNA accumulation within an interval extending from 15% to 65% egg-length (EL, 0%=posterior pole). This spans from positions where the anterior margin of the first ftz stripe and the posterior margin of the seventh stripe would form in normal embryos. This increased expression is consistent with previous interpretations that runt acts as a positive regulator of ftz. Furthermore, as hs/runt does not lead to the uniform repression of eve and hairy within this region, runt must have regulatory effects on ftz that are independent of its effects on the expression of these other two pair-rule genes.

The above results demonstrate that runt acts as a regulator of pair-rule gene expression but do not show that this is due to transcriptional regulation. Previous work has identified cis-regulatory elements that participate in the transcriptional regulation of ftz, eve and hairy during segmentation. We examined the expression patterns of reporter gene constructs containing regulatory elements from these genes in order to determine if hs/runt treatment alters their transcriptional regulation. The ftz/lacC reporter gene contains three major regulatory elements from the ftz upstream region (Hiromi et al., 1985). This reporter gene is expressed in a characteristic 7-striped pattern in late blastoderm stage embryos (Fig. 1K). In hs/runt embryos, this reporter gene shows a broad band of expression similar to that observed for the endogenous ftz gene (Fig. 1L). This result confirms that runt regulates ftz transcription, and indicates that cis-elements within the ftz/lacC reporter gene mediate this regulation. Further work is needed to determine if this is due to a direct interaction between the Runt protein and these cis-elements.

The eve upstream region contains several stripe-specific cis-elements that are regulated by the gap genes, as well as an autoregulatory element that can direct the formation of all 7 stripes, but only in the presence of wild-type eve function (Goto et al., 1989; Harding et al., 1989; Jiang et al., 1991; Small et al., 1991). Previous work has indicated that runt acts through the autoregulatory element to repress eve expression (Goto et al., 1989). However, the stripe-specific alterations that are observed in hs/runt embryos suggest that in this case runt may be acting through the stripe-specific elements. In fact, the expression of reporter genes containing the eve autoregulatory element was not significantly altered in our regularly treated hs/runt embryos (not shown). Conversely, a reporter gene that contains regulatory elements for eve stripes 2, 3 and 7 and that lacks the autoregulatory element is affected in a manner identical to the endogenous eve gene, i.e. stripe 2 is eliminated and stripe 7 is reduced in intensity (Fig. 1G,H). The fact that stripe 3 is not affected indicates that the repression of stripes 2 and 7 expression is not due to a generalized reduction in the transcription of the reporter gene construct in these embryos. The results obtained with reporter genes for hairy are similar to those obtained for eve. The h^KB/lacZ reporter gene contains cis-regulatory elements for hairy stripes 1 and 5 (Fig. 1C; Pankratz et al., 1990). In hs/runt embryos, the expression of this reporter gene in the region corresponding to hairy stripe 1 is eliminated and the expression of stripe 5 is reduced (Fig. 1D). The results obtained with these eve and hairy reporter gene constructs provide further evidence that ectopic runt expression alters the transcriptional regulation of pair-rule genes. These results also strongly suggest that stripe-specific cis-regulatory elements of eve and hairy respond to the regulatory effects of runt.

The establishment of the hairy and eve expression patterns depends on the regulation of stripe-specific cis-regulatory elements by maternal and gap gene regulatory products (Pankratz et al., 1990; Riddihough and Ish-Horowicz, 1991; Small et al., 1991, 1992; Stanojevic et al., 1991). One explanation for the stripe-specific effects described above is that hs/runt interferes with the expression or function of the gap genes. To address this possibility, we examined gap gene expression patterns in hs/runt embryos. Five different gap genes, Krüppel (Kr), knirps (kni), giant (gt), hunchback (hb), and tailless (tll) contribute to the regulation of pair-rule gene expression within the segmented region of the embryo (hereafter referred to as the presegmental region). The expression patterns of all of these, except for tll (not shown), are clearly altered in hs/runt embryos (Fig. 2). Both positive and negative effects on gap gene regulation are observed, depending on the particular domain of expression and its position along the anteroposterior axis. In order to interpret these effects, it is important to take into account the differences in the regulatory cues that are used to generate gap gene expression domains in different regions of the embryo. Thus, the following description considers the gap gene expression patterns in hs/runt embryos on a region by region basis.

The stimulation of the central domain of Kr expression is the major positive regulatory effect observed within the presegmental region of the embryo. In wild-type embryos that have just completed the 13th and final syncitial nuclear division cycle, there is a central domain of Kr expression extending from 45 to 60% EL (Fig. 2A). In hs/runt embryos of this same developmental stage, this centrally located expression domain is expanded in both the anterior and posterior directions (Fig. 2B). The boundaries of this Kr domain normally are defined by repression from the maternal anterior and posterior systems (Gaul and Jäckle, 1987). The expansion of Kr’s central domain suggests that repression of Kr by these two maternal systems is antagonized in hs/runt embryos.

The enlarged central domain of Kr expression was of particular interest because it seemed likely that it could account for the repression of eve stripe 2. Kr is normally expressed posterior to this stripe and acts to define its posterior boundary by repressing stripe 2 element-dependent transcription (Small et al., 1991). The enlarged Kr domain observed in hs/runt embryos extends from the posterior margin of eve stripe 1 to the anterior margin of eve stripe 6 and thus spans the region in which eve stripe 2 should form. This alteration could readily explain the repression of eve stripe 2 in hs/runt embryos. However, in hs/runt embryos that were also mutant for Kr, eve stripe 2 is still eliminated (Fig. 3B). Thus the anterior expansion of Kr does not account for the specific elimination of this stripe in hs/runt embryos.

The gap genes kni, gt, hb and tll all have expression domains that are posterior to the central Kr domain and that fall within the presegmental region of the embryo (Fig. 2C,E,G). The posterior domain of kni is normally juxtaposed and posterior to the central Kr domain. In early blastoderm stage hs/runt embryos, the level of kni expression within this domain is reduced in intensity (Fig. 2D). A similar, but stronger effect is observed on gt. The posterior expression domain of gt is essentially eliminated in early blastoderm stage hs/runt embryos (Fig. 2F). This aspect of gt expression is subject to repression by Kr (Kraut and Levine, 1991b; Eldon and Pirotta, 1991), and thus could be indirectly due to the posterior expansion of Kr that occurs in hs/runt embryos. However, the repression of gt’s posterior domain occurs in hs/runt embryos that are also mutant for Kr (Fig. 3D). Thus the negative effect of runt on gt is not due solely to runt’s positive effects on Kr. The more posterior stripes of hb and tll are not significantly affected in hs/runt embryos, although a slight anterior expansion of hb is sometimes observed. It is notable that the differences in the sensitivity of these different domains of gap gene expression to hs/runt correlate with the primary regulatory cues used in their establishment. The posterior stripes of hb and tll depend on the activity of the maternal terminal system, and are relatively unaffected by hs/runt. Conversely, the posterior domains of kni and gt, which are negatively affected by hs/runt, are activated by the maternal posterior system. The repression of these two domains suggests that ectopic runt expression antagonizes the activity of the maternal posterior system. This putative antagonism is consistent with the effects of hs/runt on Kr that were described above.

There are multiple effects of hs/runt expression in regions anterior to the central Kr domain. The gap gene hb is expressed just anterior to Kr. In fact, there are two phases in the transcriptional regulation of hb in the anterior end of the embryo, both of which result in expression domains that border on the central Kr domain. The first phase is the activation of the proximal hb promoter during nuclear division cycle 11 (Schröder et al., 1989). This promoter is activated in direct response to the maternal bcd gradient (Driever and Nüsslein-Volhard, 1989; Struhl et al., 1989), and produces a broad band of mRNA accumulation extending from 100% EL almost to the middle of the embryo. There is no obvious difference in this aspect of hb expression in hs/runt embryos. In contrast to this, there are significant changes in the hb pattern in hs/runt embryos that have completed the 13th nuclear division. At this stage in wild-type embryos the distal hb promoter is activated and generates two stripes of expression (Fig. 2G). In hs/runt embryos the anterior hb stripe is broadened and shifted anteriorly (from 56% to 60% EL, Fig. 2H). There is also increased expression of hb in regions anterior to this stripe, especially in older blastoderm stage hs/runt embryos (Fig. 2J). Presumably the broadened anterior stripe of hb expression is due to mis-regulation of the distal hb promoter. The ectopic expression in more anterior regions could also be due to mis-regulation of this promoter, or may reflect a de-repression of the proximal hb promoter in the anterior-most regions of late blastoderm stage hs/runt embryos.

The anterior expression of gt is also affected in hs/runt embryos. In normal embryos the initial band of anterior gt expression extending from 80 to 60% EL splits into two stripes during the early blastoderm stage. The posterior of these two lies just anterior to the hb stripe and represents the most anterior domain of gap gene expression within the presegmental region of the embryo. This gt stripe is uniformly expressed around the dorsoventral circumference of the embryo (Fig. 2E). The other, more anterior stripe of gt lies outside of the presegmental region in an area that will give rise to the head. A distinctive characteristic of this stripe is its reduced expression in the ventral regions of the embryo. In early blastoderm stage hs/runt embryos, these two stripes are replaced by a single broad stripe with uniform expression on the dorsoventral axis (Fig. 2F). The posterior margin of this stripe (68% EL) is anterior to the posterior limits of gt expression in this region of wild-type embryos (62% EL). This anterior shift is similar to that described above for the anterior stripe of hb expression. One interpretation of the changes in these hb and gt domains is that hs/runt treatment is causing an anterior expansion of the presegmental region at the expense of the more anterior, precephalic region. Note that the previously described anterior expansion of Kr’s central domain in hs/runt embryos is also consistent with this interpretation.

There are several other significant changes in gap gene expression patterns in the precephalic region of hs/runt embryos. Kr normally is not expressed in this region in early blastoderm stage embryos. In hs/runt embryos, there is a broad cap of Kr expression at the anterior end (Fig. 2B). The effect on kni is reciprocal to this. In wild-type embryos there is a broad patch of ventral expression that extends from the anterior pole to 75% EL (Fig. 2C), and at this position a circumferential stripe of expression forms. In hs/runt embryos this complex pattern is replaced by a small cap of low-level expression that is restricted to the anterior pole (Fig. 2D). These several observations strongly suggest that hs/runt treatment interferes with the regulatory program that normally operates in the head region of the embryo. In order to investigate this further, we examined the effect of hs/runt treatment on empty spiracles (ems) and orthodenticle (otd), two genes that have been classified as head gap genes based on their early roles in anterior pattern formation (Cohen and Jürgens, 1990). The expression of both of these genes is altered in early blastoderm stage hs/runt embryos. The band of otd expression that normally extends from 75 to 95% EL is replaced by an anteriorly shifted domain with a reduced level of expression (Fig. 4A,B). In contrast, the expression of ems is almost completely suppressed in the anterior end of hs/runt embryos (Fig. 4D). These results provide strong evidence that ectopic runt expression disrupts the normal regulation of gap gene expression in the head of the Drosophila embryo.

Proper expression of gap genes in the anterior end of the embryo is dependent on the maternal anterior system. The principle output of this system is a morphogenetic gradient of Bcd protein with the highest levels at the anterior pole. Several of the results presented above suggested that runt antagonizes bcd activity. First, the anterior expansion of Kr’s central domain as well as the shift in the anterior band of gt expression in hs/runt embryos are similar to effects observed when bcd is reduced (Gaul and Jäckle, 1987; Eldon and Pirotta, 1991; Kraut and Levine, 1991a). Similarly, the anterior expression domains of the head-gap genes ems and otd, both of which are negatively affected by runt, are also both bcd dependent (Dalton et al., 1989; Finkelstein and Perrimon, 1990). Further evidence comes from the effects of hs/runt on the more anterior pair-rule stripes. The proper expression of eve stripe 2 depends on activation by bcd and hb and repression by gt and Kr (Small et al., 1992). As described above, the elimination of this stripe in hs/runt embryos is not due to the expanded expression of Kr. This stripe is also eliminated in hs/runt embryos that are mutant for gt (not shown). Thus the elimination of eve stripe 2 is presumably due to a regulatory interaction between runt and the activators of this stripe. In this context, it is notable that hairy stripe 1, which is eliminated in hs/runt embryos, is also bcd dependent (Riddihough and Ish-Horowicz, 1991). In fact, in addition to bcd, runt is the only other segmentation gene shown to regulate this stripe. Taken together, these observations strongly suggested that interference with bcd-dependent regulation could account for many of the defects in the anterior ends of hs/runt embryos.

Several lines of evidence indicate that the Bcd protein is directly responsible for activating transcription of genes expressed in the anterior regions of Drosophila embryos. Two convincing examples are the activation of eve stripe 2 (Small et al., 1991, 1992), and the activation of the proximal hb promoter (Driever and Nüsslein-Volhard, 1989; Struhl et al., 1989). Interestingly, the former is extremely sensitive to hs/runt inhibition whereas the latter is not. One significant difference between these is that the activation of hb occurs during the syncitial nuclear division cycles whereas eve stripe 2 is most strongly expressed subsequent to the final syncitial nuclear division cycle during the blastoderm stage. This difference in stage of expression could account for the differential sensitivity to runt. Alternatively, these different responses may be due to differences in the proximal hb promoter and the eve stripe 2 element. In order to investigate this, we examined the effects of hs/runt on two different reporter genes with well-defined cis-regulatory control regions. The first, HSBG[bcd], contains three copies of the bcd recognition sequence upstream of a minimal promoter. This reporter gene produces an anterior cap of lacZ mRNA expression in early blastoderm stage embryos (Fig. 5A). This expression is bcd dependent (Ronchi et al., 1993). We also examined the expression of a related reporter gene, HSBG[bcd+hb], that contains three copies of the recognition sequence for the hb protein in addition to the three bicoid-binding sites (J. Treisman and C. Desplan, personal communication). In normal embryos, the expression of this gene is significantly stronger and extends more posteriorly than the expression of HSBG[bcd] (Fig. 5C). The expression of both reporter genes is affected in early blastoderm stage hs/runt embryos. The expression of HSBG[bcd] is almost completely eliminated (Fig. 5B), whereas the expression pattern of HSBG[bcd+hb] is both reduced and anteriorly shifted in hs/runt embryos (Fig. 5D). These results provide strong evidence that bcd-dependent transcriptional activation is antagonized in blastoderm stage hs/runt embryos. Both of these reporter genes are initially activated during the syncitial stages prior to formation of the blastoderm. As HSBG[bcd] is only expressed at a very low level during these earlier stages, it is difficult to determine whether this expression is repressed by hs/runt treatment. The stronger expression of the HSBG[bcd+hb] reporter may be slightly reduced in preblastoderm hs/runt embryos, but not as significantly as at later stages (not shown). This suggests that the effect of runt on bcd-dependent transcription is dependent on the developmental stage of the embryo.

The bcd mRNA is synthesized and then localized to the anterior pole of the oocyte during oogenesis (Frigerio et al., 1986; Berleth et al., 1988). The negative effects on bcd-dependent expression in hs/runt embryos are due to ectopic expression of a transgene in the early blastoderm stage. Therefore this effect cannot be due to an alteration in bcd transcription. No significant difference was observed in the bcd mRNA and protein expression patterns in hs/runt embryos (data not shown). These results indicate that hs/runt treatment is not affecting the synthesis, stability or localization of the Bcd protein. They further suggest that runt is responsible, either directly or indirectly, for interfering with the ability of the Bcd protein to activate transcription during the blastoderm stage of embryogenesis.

The above observations are based on ectopic expression of runt. Many of these effects are observed in early blastoderm stage embryos (prior to nuclear elongation). Based on the experimental protocol used (see Materials and Methods), these embryos were either in the 12th or 13th syncitial nuclear divisions at the time runt expression was induced by heat-shock treatment. In wild-type embryos during these stages, runt has not yet developed its characteristic pair-rule expression pattern, but instead is expressed in a central broad field (Klingler and Gergen, 1993). Therefore we speculated that many of the effects observed in early blastoderm stage embryos might reflect regulatory functions normally provided by runt’s early broad field.

Based on the ectopic expression experiments, one obvious prediction is that there would be overexpression of bcd-dependent genes in runt mutant embryos. For example, the posterior margin of the anterior domain of hb expression is established at a point on the bcd gradient where the level of activity drops below the threshold value required for activation of the hb promoter (Driever et al., 1989). In the simplest model, this hb expression domain would extend more posteriorly in the absence of runt activity. However, this is not observed (data not shown). The anterior expression domain of hb is also subject to repression by Kr, which is normally expressed just posterior to and abutting this domain (Jäckle et al., 1986). Therefore Kr could be responsible for preventing the posterior expansion of hb in runt mutant embryos. We tested this possibility by examining hb expression in embryos mutant for both runt and Kr. There is posterior expansion of the anterior hb domain in these double mutant embryos (Fig. 6E). This effect is stronger than the slight posterior expansion of hb observed in embryos mutant only for Kr (Fig. 6C). Notably, the posterior margin of this expanded domain is more diffuse in the double mutant embryos than in either single mutant, and continues to extend more posteriorly as the embryos progress to later stages. In fact, the anterior and posterior domains of hb become fused on the dorsal side of double mutant embryos by the completion of cellularization. This result suggests that formation and maintenance of the normally sharp border of hb’s anterior expression domain is due to repression by both runt and Kr. There is also an effect on hb’s posterior pattern. This domain expands anteriorly towards the central region of these double mutant embryos. However, even more striking is the significant expansion of the posterior domain of gt into the central regions of runt, Kr double mutant embryos (Fig. 6F). Again, these effects are much more pronounced than the changes that are observed in embryos that are mutant for only runt or Kr. These results provide strong evidence that runt plays a role in suppressing the expression of these two gap genes in the central regions of the embryo. These alterations are observed in early to mid-blastoderm stage embryos, i.e. prior to the establishment of runt’s characteristic pair-rule expression pattern. Therefore the earlier expression and activity of runt is involved in the position-specific regulation of gap gene expression in the Drosophila embryo.

A surprising finding of this work is the regulation of gap gene expression by runt, a pair-rule gene downstream of the gap genes in the conventional view of the segmentation hierarchy. Many of the effects that we have described result from the expression of a hs/runt transgene. As pointed out by others (Parkhurst and Ish-Horowicz, 1991; Jarman et al., 1993), it is important to use caution when interpreting the results of ectopic expression experiments such as these. However, the alteration of gt and hb expression patterns in embryos where runt is eliminated indicates that the regulation of gap genes by runt is a component of the normal regulatory circuitry of embryonic pattern formation. The effects on gt and hb are observed in stages preceding the establishment of runt’s pair-rule expression pattern. Further evidence that runt is active prior to its expression as a pair-rule gene comes from work on sex determination. Drosophila females mutant for runt fail to express the Sex-lethal gene within a broad domain that roughly corresponds to the presegmental region of the embryo (Duffy and Gergen, 1991). The gap in the Sex-lethal expression pattern is also observed in early blastoderm stage embryos, prior to the establishment of runt’s pair-rule expression pattern. There is expression of runt mRNA within a central broad field of the embryo during the syncitial nuclear division cycles that precede blastoderm formation (Klingler and Gergen, 1993). This expression correlates extremely well, both spatially and temporally with the early requirements for runt in sex determination and gap gene regulation.

Our results indicate that runt and Kr cooperate to prevent expression of the gap genes gt and hb in the central regions of the embryo. Another similarity between runt and Kr is that the primary regulation involved in generating their central broad domains of expression involves pole-region-dependent repression (Gaul and Jäckle, 1987; Klingler and Gergen, 1993). For example, in embryos from bcd mutant mothers, the early runt and Kr domains are both expanded significantly towards the anterior pole. Thus, at one level bcd is acting to repress the expression of runt and Kr. Yet, in those regions where they are able to escape from this repression they appear to antagonize the effects of bcd (Fig. 7). A similar situation exists at the posterior end of the embryo, but in this case the interactions involve two of the maternal systems. The posterior margins of the central Kr and runt domains are defined by the maternal posterior and terminal systems, respectively (Gaul and Jäckle, 1987; Klingler and Gergen, 1993). Similarly, the posterior domains of gt and hb, both of which expand towards the center of runt, Kr double mutant embryos, are also activated by the maternal posterior and terminal systems, respectively (Kraut and Levine, 1991a; Eldon and Pirotta, 1991; Tautz, 1988). Thus the underlying principle is similar to that at the anterior end. In regions where runt and Kr escape from repression, they appear to antagonize the action of their repressors. A net result of such reciprocal negative regulatory interactions is a sharpening of the positional information provided by the different maternal systems at specific boundary points in the embryo. Thus, runt’s gap activity seems to help define the central region of the blastoderm embryo and sharpen the distinction between this and the flanking anterior and posterior regions.

The genetic pathway that generates the segmented body pattern in Drosophila has been organized into a hierarchy based on the spatial and temporal expression patterns of the segmentation genes and their genetic regulatory interactions. In a strict interpretation, a ‘downstream’ pair-rule gene runt regulating the expression of ‘upstream’ gap genes contradicts this hierarchical organization. However, in our interpretation, runt’s activity as a pair-rule gene is still downstream of the gap genes, and the gap-like regulatory function is due to the earlier broad field of runt expression. Therefore, this view does not contradict the hierarchical organization of segmentation genes. It does raise the possibility that alternative, ‘non-classical’ activities of other segmentation genes may play important roles in this developmental pathway. For example, evidence suggests that a gap-like activity is associated with early expression of the pair-rule gene eve. There is a broad band of eve expression that overlaps with and extends anterior of runt’s early broad field (Harding et al., 1986; our unpublished observations). This expression correlates well with the stripe-specific defects in the expression of several pair-rule genes that are observed in eve mutant embryos (Carroll and Scott, 1986; Ingham and Gergen, 1988). Importantly, these defects are observed at the earliest stages of stripe formation, prior to the stages when most other pair-rule gene regulatory interactions become apparent. It will be interesting to determine if this earlier activity of eve is similar to that of runt in its ability to influence gap gene expression. The striped patterns of other pair-rule genes are also preceded by broader, gap-like expression domains (Kilchherr et al., 1986; Grossniklaus et al., 1992). It remains an open question whether these are also functionally relevant, or merely reflect an early phase in the evolution of their expression patterns.

The expression patterns of all of the pair-rule genes that have been examined to date are altered in runt mutant embryos (Carroll and Scott, 1986; Frasch and Levine, 1987; Ingham and Gergen, 1988; Hooper et al., 1989; Gutjahr et al., 1993). One aim of these ectopic expression experiments was to identify which of these regulatory effects are most direct. In addition to eve, hairy and ftz, we have also examined the expression of the endogenous runt gene and the paired (prd) gene in hs/runt embryos. The regulation of all of these genes is affected, thus confirming runt’s importance as a pair-rule regulator. Some insight on pair-rule regulatory circuitry is provided by the different responses of these genes to runt. Stripe-specific defects are observed for eve, hairy and runt in hs/runt embryos. These three genes have other features in common. They cross-regulate each others expression and are involved in regulating the expression of the other pair-rule genes (i.e. ftz and prd). Conversely, the initial pair-rule patterns of eve, hairy and runt are not altered in ftz and prd mutant embryos (Frasch and Levine, 1987, Hooper et al., 1989; Klingler and Gergen, 1993). Based on these regulatory interactions eve, hairy and runt have been classified as primary pair-rule genes, whereas ftz and prd have been described as secondary or tertiary pair-rule genes.

As primary pair-rule genes, eve, hairy and runt are thought to respond ‘directly’ to the positional cues provided by the gap genes. The identification of gap-responsive, stripe-specific elements in the cis-regulatory regions of each of these three genes strongly supports this view (Harding et al., 1989; Goto et al., 1989; Riddihough and Ish-Horowicz, 1991; M. Klingler, J. Soong and JPG unpublished observations). It seems likely that these stripe-specific cis-elements are involved in mediating the regulatory responses of the primary pair-rule genes to hs/runt treatment. The stripe-specific repression of reporter gene constructs containing cis-elements for hairy stripe 1 and eve stripe 2 in hs/runt embryos supports this. The eve and hairy patterns observed in runt mutant embryos provides a further suggestion that runt regulates these genes through their stripe-specific elements. In both cases there are non-uniform effects that can be described as stripe-specific and there are also abnormalities in stripe spacing (Frasch and Levine, 1987; Ingham and Gergen, 1988; Hooper et al., 1989). Finally, it is notable that hs/runt treatment fails to repress a reporter gene containing the eve autoregulatory element. This suggests that the altered regulation of this element in runt mutants (Goto et al., 1989) is indirect and due to effects on the expression of other genes.

The stripe-specific regulation of the three primary pair-rule genes is qualitatively different from the more uniform response of the secondary pair-rule gene ftz. The pair-rule gene prd responds similarly to ftz and becomes uniformly expressed within the presegmental region of hs/runt embryos (C. Tsai, unpublished). Given the differences in their responses, the uniform activation of ftz and prd observed in hs/runt embryos cannot be due solely to runt’s effects on the expression of the primary pair-rule genes. These observations suggest that the most direct regulatory effects of runt on these secondary pair-rule genes involve cis-element(s) that normally contribute to regulating their expression throughout the presegmental region. Identification of these cis-elements and their mode of interaction with runt is important to our understanding of the segmentation pathway and should also provide further insight on the function of the Runt protein.

One effect of ectopic runt expression that we have characterized in detail is the antagonism of bcd-dependent activation, an effect obtained on synthetic promoter reporter gene constructs containing three copies of a Bcd recognition sequence. From our in vivo data, this antagonism is apparently not due to an effect on the synthesis or stability of the Bcd protein, but instead reflects interference with its function as a transcriptional activator. The Runt protein contains a novel type of DNA-binding motif referred to as the Runt-domain (Kagoshima et al., 1993). The effects of Runt could be readily accounted for if the protein were to recognize to the same binding site as the Bcd protein and thus prevent Bcd-dependent activation by a simple competition mechanism. The consensus binding site for the mammalian Runt-domain proteins is PuACCPuCA (Melnikova et al., 1993; K. Shigesada, personal communication). This is not similar to the consensus Bcd-binding-site (TCTAATCCC, Driever and Nüsslein-Volhard, 1989) that is present in the HSBG[bcd] reporter gene construct used in this study (Ronchi et al., 1993). The mammalian Runt-domain proteins as purified from cells are in a heterodimeric complex with an unrelated partner protein referred to as PEBP2/CBFβ (Ogawa et al., 1993a; Wang et al., 1993). This partner protein increases the DNA-binding affinity of the Runt-domain proteins, but does not appear to alter the sequence specificity (Melnikova et al., 1993; K. Shigesada, personal communication). The DNA-binding affinity of the Drosophila Runt protein is increased by addition of murine PEBP2/CBFβ. However, even under these conditions, we have not been able to demonstrate an in vitro interaction between Runt and the Bcd recognition sequence (C. Tsai, unpublished). Thus, although we cannot rule out the possibility that Runt may directly interact with the Bcd-recognition sequence in vivo, we do not favor this model.

A second simple model to explain this antagonism is that the Runt protein causes the Bcd activation domain to be masked, perhaps through a direct protein-protein interaction. Bcd acts as a transcriptional activator both in cultured Drosophila cells and in yeast (Driever and Nüsslein-Volhard, 1989; Driever et al., 1989; Struhl et al., 1989). In neither of these heterologous systems did the runt protein interfere with bcd-dependent activation (C. Tsai, S. Lin, J. Till and T. Chu, unpublished observations). Although there are many potential explanations for such negative results, they strongly suggest that other factors play a role in the antagonism of bcd-dependent activation by runt. In this context it is important to note that several of the other regulatory events observed in hs/runt embryos (increased expression of Kr, ftz) indicate that runt does not act solely by sequestering factors that are generally required for transcriptional activation.

As noted above, there are stage-specific differences in the effect of hs/runt on bcd-dependent expression. These differences are easily accounted for if the interaction between runt and bcd depends on other developmentally regulated factors. Several other observations suggest that the regulatory functions of runt are influenced by the developmental context. The relatively uniform expression of ftz in the presegmental region of hs/runt embryos provides evidence that runt may act to directly stimulate ftz transcription. However, the fact that this occurs only in the presegmental region indicates that other spatially regulated factors influence the response to runt. Future work should reveal whether runt is interacting with other activating factors that are present only in the presegmental region, or alternatively whether the activation by runt is not strong enough to overcome factors that suppress expression at the poles. The diverse specificity of runt’s many regulatory functions during Drosophila embryogenesis strongly suggest that interactions with other factors are central to its function. Several of the regulatory effects described here provide an excellent starting point for further investigating the mechanism of action of this novel transcriptional regulatory protein.

The following individuals and labs generously provided transgenic fly lines and/or cloned segmentation gene probes for these experiments: Yash Hiromi; Ulrike Nauber, Mike Pankratz and Herbert Jäckle; Steve Small, Jin Jiang and Mike Levine; Jim Posakony; Beth Eldon and Vincenzo Pirotta; Bill McGinnis; Robert Finkelstein; Jessica Treisman and Claude Desplan. We appreciate the availability of the tissue culture line for the anti-bcd monoclonal from Dan Jay and the monoclonal facility at Harvard University. Cell supernatants from this line were generated by the cell culture facility in the Department of Microbiology, SUNY at Stony Brook. This manuscript was greatly improved by comments from Claude Desplan, Martin Klingler, Sid Strickland and Henry Krause. We also thank Henry Krause and Armen Manoukian for communicating results they obtained with hs/runt flies prior to publication. This work is supported by a grant from NIH (HD22448) and a Faculty Research Award (FRA-428) from the American Cancer Society.