The zygotic control of Drosophila pair-rule gene expression: II. Spatial repression by gap and pair-rule gene products

Analyses of terminal phenotypes and of segmentation gene expression patterns in various mutant embryos has defined the overall regulatory hierarchy of the segmentation genes. Generally, each class of segmentation genes interacts to specify the finer expression pattern of the next group of genes. Thus, the maternal coordinate genes affect each other (Frohnhöfer and Nüsslein-Volhard, 1987; Driever and Nüsslein-Volhard, 1988) and the gap genes (Gaul and Jackie, 1987), which interact (Jackie et al. 1986) to control pair-rule gene patterns (Carroll and Scott, 1986; Ingham et al. 1986; Frasch and Levine, 1987), which interact to specify the patterns of the segment polarity genes (DiNardo and O’FarreH, 1987; Martinez-Arias and White, 1988). Given the extent of these fundamental pattern-regulating gene interactions, it is a large task to determine the nature of the regulatory circuitry that operates between segmentation genes and to identify the trans- and cis-acting factors that are responsible for the pattern of gene expression.

The present approaches aimed at elucidating these factors include formal genetic analyses, studies on the effect of ectopic expression of putative regulatory proteins (Ish-Horowicz and Pinchin, 1987), analyses of cis-acting elements that control pair-rule gene expression (Hiromi et al. 1985; Hiromi and Gehring, 1987; Howard, 1988), studies on the effects of inhibiting segmentation protein synthesis (Edgar et al. 1986, 1989), and in vitro biochemical experiments (Hoey and Levine, 1988). This combination of approaches is expected to resolve the complex problem of how a crude pattern in the unfertilized egg is translated into a periodic pattern of segments.

In the accompanying paper, we showed that most zygotically required regulators of the pair-rule genes have apparently been identified and that the initial activation of the pair-rule genes does not depend upon other zygotic genes (Vavra and Carroll, accompanying paper). However, even with most of the key genes identified, it is difficult to demonstrate whether removing one gene from the system directly or indirectly perturbs the expression of others or if an observed interaction reflects positive or negative control. To address these difficulties we have analyzed pair-rule gene expression in selected single, double, and triple mutant embryos to uncover which zygotic genes are likely to act directly upon the hairy, eve and ftz genes. Our results, combined with other recent studies on pairrule gene interactions (Ingham and Gergen, 1988) and protein synthesis inhibition experiments (Edgar et al. 1989), support the view that certain pair-rule genes are extensively negatively regulated, i.e. specific maternal, gap and pair-rule proteins repress pair-rule genes. It appears that, at least for the ftz and hairy genes, their striped patterns are more the result of repression of gene expression in the interstripe regions than a regional activation of individual stripes.

We have examined pair-rule gene expression in whole-mount cellular blastoderm embryos by filtered fluorescence imaging (Karr and Kornberg, 1989; Carroll et al. 1988) after immunoperoxidase staining with polyclonal antibodies specific for the ftz (Carroll and Scott, 1985), eve (Frasch et al. 1987; antibody gift of M. Frasch and M. Levine) hairy (Carroll et al. 1988), and Krüppel (Gaul and Jackie, 1987; antibody gift from U. Gaul). This technique gives sharp images of protein localization and was used to double-label embryos to examine relative expression patterns or to unambiguously identify the genotype of an individual embryo derived from crosses that yield a variety of mutant progeny.

The null allele stocks used to generate the pair-rule double mutant embryos were: Df(l) runt^B57/FM6y⁺Y^mal102 (kindly provided by Peter Gergen), Df(2R) eve 1.27 cn bw sp/SM6a, and h^7h94 e^s/TM3.

Runt; eve and runt; hairy double mutant embryos were generated by mating runt males to heterozygous eve or hairy virgin females. The double heterozygous F₁ females were then mated to either heterozygous eve or hairy males, yielding one-sixteenth double mutant embryos.

Stocks used in gap mutant analysis were Kr^lcn bw/CyO (a null allele), and the double mutant kni^IID48hb^lmAScu sr e^sca/TM3 Sb Ser, generously provided by R. Lehmann and C. Nüsslein-Volhard. The triple gap mutant embryos were produced at a frequency of one-sixteenth from heterozygous hb, kni; Kr parents.

Several previous studies have analyzed the effect of individual pair-rule mutations on the expression of other pair-rule genes (Howard and Ingham, 1986; Carroll and Scott, 1986; Frasch and Levine, 1987; Ingham and Gergen, 1988). From these experiments, a general picture emerged of the pair-rule gene hierarchy that placed hairy, runt and eve at the top, with other pair-rule genes (e.g. ftz) being downstream from them. For example, ftz expression is altered by mutations in h, runt or eve, but ftz mutants have no impact on expression of h, runt or eve.

Recently, an extensive study was made of hairy, runt, eve and ftz RNA expression in certain double-mutant pair-rule combinations (Ingham and Gergen, 1988). The analysis of epistatic relationships helps to reveal which genes are likely to be directly involved in the regulation of other genes in the pathway. The observations presented here on pair-rule protein patterns overlap with the results of Ingham and Gergen (1988) on pair-rule RNA patterns. Because these interactions and the patterns of certain mutant combinations are critical to the subsequent discussion of gap gene control, we will present all of our results dealing with the regulation of h, eve and ftz protein expression and emphasize those details and mutant combinations that may differ from the previous observations of RNA patterns. We generally agree with the conclusions of Ingham and Gergen (1988) as to the nature of individual pair-rule regulatory interactions.

The wild-type hairy, ftz and eve protein patterns at the cellular blastoderm stage of embryogenesis consist of seven transverse stripes encircling the embryo and, in the case of the hairy gene, an additional dorsal anterior patch of expression (Fig. 1A-C). The ftz and eve stripes are in alternating domains while the hairy stripes are offset such that the six posterior ones transiently overlap the ftz stripes by about one cell and all seven hairy stripes overlap with each eve stripe (Carroll et al. 1988). Loss of runt⁺ activity changes h, eve and ftz expression (Fig. 1D-F). The hairy pattern partly expands with the first hairy stripe spreading posteriorly, while the interband between stripes 3 and 4 accumulates some protein, and stripes 6 and 7 are stronger and nearly fused (Fig. 1D). Note that the hairy pattern is still fairly periodic, so although loss of runt⁺ derepresses hairy expression, runt is only one of probably several negative regulators of hairy. We will symbolize these gene interactions with an arrow indicating positive regulation and a cross-hatch indicating negative control, i.e.

We also point out that eve is the earliest detectable pairrule protein (at cycle 12) and exhibits a broad early band of expression during cycle 14 over PS1-3 (Frasch et al. 1987; our unpublished observations). This early band of expression may be necessary for hairy and ftz accumulation in these parasegments (see Discussion).

Note that since runt and eve are the only known pairrule regulators of hairy, the pattern shown in Fig. 1J is very informative because it is the consequence of hairy regulation by genes that are above the pair-rules in the segmentation hierarchy, i.e. the gap genes (see below).

The ftz pattern is also strongly affected in run⁻ ; eve⁻ embryos; however, because of the strong influence of hairy on ftz, we cannot determine from this combination whether runt might regulate ftz more directly (Fig. 1K).

Ingham and Gergen (1988) have shown that hairy is required to repress runt.

Since pair-rule genes do not feed back upon the gap genes and our aneuploid screen turned up only one new zygotic locus so far, we have no evidence of any intermediary genes that could explain our observations; thus, we suggest that the interactions diagrammed above could represent direct regulation of pair-rule gene expression at either the DNA, RNA or protein levels.

It is critical to note that the basic periodic patterns of hairy and eve are only moderately perturbed by pairrule mutations (see for instance hairy expression in the eve^-/runt^- double mutant embryo in Fig. 1J, and eve expression in the runt^-/hairy^- embryo in Fig. 1O). This suggests that the genes above the pair-rule level, namely the zygotic gap genes, regulate their initial periodicity.

Gap genes have strong effects on the expression oí ftz (Carroll and Scott, 1986; Ingham et al. 1986), hairy (Ingham et al. 1986; Carroll et al. 1988; Howard, 1988) and eve (Frasch and Levine, 1987). However, it has been difficult to distinguish which effects could be direct versus those that may be indirect. It has been shown that, in Krüppel⁻(Kr) and knirps⁻(kni) embryos, ftz expression is in a pattern that is complementary to that of hairy expression, while in hunchback^-(hb) embryos ftz does not strictly follow hairy in the posterior of the embryo (Carroll et al. 1988; Ingham et al. 1986; Howard, 1988). From these observations, we had concludedthat ftz regulation is mostly independent of the three gap genes, hb, Kr and kni.

In order to better understand the role of these three gap genes in establishing the periodic patterns of hairy and eve, we have examined their expression in certain combinations of gap mutants. In Kr⁻ embryos, eve (Fig. 2A) and hairy (Fig. 2B) expand into very similar patterns with the middle of the embryo expressing two large blocks of each protein (about 10-12 nuclei in width) separated by a band of unlabelled cells (about 3-4 nuclei in width). The ftz pattern is complementary to the hairy pattern (Fig. 2C). While the expansion of the eve and hairy domains could indicate a basic negative control of these genes by Kr, there are many alternative explanations to consider that are best explored after examining mutant combinations.

In hb⁻, kni⁻ embryos, hairy is not expressed over most of the anterior segment primordia (about 40–65 % egg length, Fig. 2E) but has spread out on the posterior part of the kni⁺ domain (about 20–35% egg length). We infer from this pattern that kni⁺ is required to keep hairy off (either directly or indirectly via the runt gene) in this posterior region of the embryo and that some gene(s) still keeps hairy off in the unstained middle third of the embryo. The best candidate for this latter activity is Kr⁺ which, while normally expressed in PS 4-7, spreads anteriorly in hb⁻ embryos and posteriorly in kni⁻ embryos, approximately that region where hairy is not expressed in a hb⁻kn^−- embryo (Jäckle et al. 1986; Box C diagrammed in Fig. 3). To determine whether Kr⁺ is responsible for keeping hairy off over the middle third of the hb⁻kni⁻ embryo,, we constructed triple mutants that were h⁻; kni⁻ ; Kr^-. In these embryos, hairy is expressed across most of the posterior two-thirds of the embryo; that is, the additional removal of Kr⁺ has derepressed hairy (compare Fig. 2H with Fig. 2E; Box C in Fig. 3). Therefore, Kr⁺, in the absence of the hb⁺ and kni⁺ gene products, appears to negatively regulate hairy.

The eve pattern in hb⁻, kni⁻ embryos differs considerably from the hairy pattern in that it extends from pàrasegments 1–13 (about 15–70% egg length, Fig. 2D) and is at much lower than wild-type levels. And, in contrast to hairy, the additional removal of Kr⁺ has no discernible effect on eve expression in a hb⁻ ; kni⁻ background (compare Fig. 2G with Fig. 2D), hb⁺ and kni⁺ appear to be required for spatial repression of eve into stripes, while Kr⁺ appears to have little influence on eve under these circumstances. Also, it appears that hb⁺ and kni⁺ may be necessary for the proper level of eve expression since eve protein does not accumulate to normal levels in h⁻ ; kni⁻ or the triple mutant embryos. We cannot tell from these experiments whether the control of eve by hb⁺ and kni⁺ involves repression or activation, or through combinatorial interactions, both mechanisms (see Discussion).

Surprisingly, all three pair-rule genes are expressed in the triple gap mutant embryos despite the elimination of these key gap gene functions. (The ftz patterns in both the hb⁻, kni⁻ embryos and the Kr⁻; hb⁻, kni⁻ embryos are complementary to the hairy patterns.) From this observation, we conclude that the Kr⁺, hb⁺, and kni⁺ activities are not absolutely required to activate pair-rule gene expression. Rather, Kr⁺ (and perhaps hb⁺ and kni⁺) appears to repress hairy expression, while hb⁺ and knC could be involved in both the spatial repression and the activation of eve.

Even if one assumes that the regulatory circuit diagrams presented in the previous sections are correct, given the known spatial relationships between all of the segmentation genes discussed, it is still not possible to accurately predict the pattern of a particular pair-rule gene in a particular mutant embryo. Why, for instance, if runt⁺ is a negative regulator of hairy doesn’t hairy expression expand throughout a runt⁻ embryo? Why is eve expression fairly normal initially in a runt⁻ embryo? The clues to these questions lie in the asynchronous kinetics of stripe formation and in how different regulators may be present in different amounts at different stages during cycle 14.

At any given moment in the pair-rule stripe formation sequence, the degree of gap, pair-rule and auto-regulatory input may vary. The best evidence for this variation is twofold. First, stripes do net form uniformly. As shown for the ftz (Karr and Kornberg, 1989) and eve proteins (Frasch et al. 1987) and hairy mRNA (Howard, 1988), the intensity and width of the pair-rule stripes changes dramatically during cycle 14. Second, the range of novel stripe phenotypes induced by injection of cycloheximide becomes more restricted as the embryo nears the cellular blastoderm stage (Edgar et al. 1986; 1989).

The asynchronous resolution of pair-rule stripes explains why the removal of an individual regulatory protein may not result in a predictable symmetrical change in the target gene pattern along the entire embryo. We can illustrate this point for the runt⁺- and eve⁺-hairy interactions by comparing the early stages of hairy protein accumulation in a wild-type embryo with the eventual hairy protein pattern in runt⁻ and eve⁻ embryos. The accumulation of hairy protein is not uniform across the cycle 14 embryo. The earliest stripe to appear is a broad first stripe (Fig. 4A), followed by a very broad stripe in the position of what will be the third and fourth stripes (Fig. 4B). The seventh, second, fifth and sixth stripes follow shortly afterward. The narrowing of the first stripe and the separation of the third and fourth stripes, which occurs shortly before the entire seven stripe pattern is completed (Fig. 4C) depends upon runt since, as described earlier, the hairy pattern is not resolved in these regions in a runt⁻ embryo (Fig. 4D). Similarly, the delayed appearance of the second hairy stripe may reflect the temporal requirement for the eve⁺ product to act upon hairy (Fig. 1G). Taken together, the kinetics of wild-type stripe formation and the effects of the runt⁻ and eve⁻ mutations indicate that there is a temporal sequence to stripe resolution and that runt⁺ and eve⁺ functions are downstream from the gap genes that initially refine the hairy pattern. These results explain the frequent lack of correspondence between the place where a regulatory gene is expressed and the effect of its removal on a target gene pattern. Even though each hairy⁺ stripe is eventually overlapped by an eve⁺ stripe and each interstripe contains a runt⁺ stripe, the effects of these latter two genes are restricted to certain regions of the embryo by the preceding set of gap regulatory activities.

From this analysis of pair-rule gene expression in known segmentation mutant embryos, we have drawn three main conclusions about the genes that establish the periodic patterns of pair-rule gene expression. First, several previously described pair-rule interactions are probably indirect, indicating that there are fewer transacting regulators of individual pair-rule genes than studies of single mutants may have suggested. Second, those pair-rule gene interactions that may be direct often involve repression of other pair-rule genes. Finally, the three key gap genes hb, Kr and kni are not essential to activate genes such as hairy, on the contrary, when expressed alone in a cell, Kr⁺ appears to repress hairy expression. These observations have several implications for understanding how pair-rule genes are regulated by trans-acting factors.

We have described the logic behind the epistasis tests of pair-rule regulatory interactions. Our interpretations support the conclusions of Ingham and Gergen (1988) and extend those of several previous studies (Howard and Ingham, 1986; Carroll and Scott, 1986; Ingham et al. 1986; Frasch and Levine, 1987) that have dissected the pair-rule regulatory circuit. All evidence suggests that most interactions involve negative regulation with hairy⁺ acting as a negative regulator oíftz (Howard and Ingham, 1986; Carroll and Scott, 1986; Ish-Horowicz and Pinchin, 1987) and runt (Ingham and Gergen, 1988), runt⁺ as a negative regulator of eve (Frasch and Levine, 1987) and hairy (Ingham and Gergen, 1988), and eve⁺ as both an early activator of hairy, ftz and runt, and a late repressor of ftz and runt (Ingham and Gergen, 1988). The net regulatory effects of each gene are not equivalent. Based upon the severity of the gene expression pattern perturbations in each mutant, hairy⁺ appears to be a stronger or earlier-acting repressor of runt and ftz than runt⁺ is of hairy and eve.

Based upon double mutant patterns, we also conclude that the requirements for hairy in eve regulation (Frasch and Levine, 1987) and runt in ftz regulation (Carroll and Scott, 1986; Frasch and Levine, 1987) are probably indirect and are mediated via runt and hairy in these two cases, respectively.

The most important conclusion drawn from our analysis of hairy gene expression in different gap mutant combinations is that some gap genes, particularly the Kr⁺ product, repress hairy expression. Three pieces of evidence support this claim. First, hairy expression in a runt⁻; eve⁻ embryo shows several gaps in the pattern, and since runt and eve are the only known pair-rule regulators of hairy, the gap genes must be responsible for the remaining spatial restriction of the hairy pattern (Fig. 1J). Second, in the absence of hb⁺ and/or kni⁺, loss of hairy expression occurs in the region where the Kr⁺ domain expands (Fig. 2E and Fig. 3). Finally, removing Kr⁺ along with hb⁺ and kni⁺ derepresses hairy expression over the posterior two-thirds of the embryo (Fig. 2H); that is, in the absence of these three gap genes, the hairy pattern is nearly uniform and the gene is strongly active. We do not believe that these interactions are indirect (for example, mediated via the runt gene) because of the close correspondence of ectopic Kr⁺ expression with hairy repression and because the hairy pattern is strongly modulated even in the absence of runt⁺ and eve⁺ (Fig. 1J). These results are significant because they challenge some current notions about pair-rule gene regulation that emphasize transcriptional activation by gap genes.

There are two different sets of experiments that support a gap-gene-driven region-specific activation mechanism for pair-rule genes. The first involves the characterization of a series of alleles of the hairy gene that express only a subset of the normal seven hairy blastoderm stripes (Howard et al. 1988). Four alleles were described that consisted of progressively larger deletions of the 5′ DNA flanking the hairy promoter. As the amount of 5′ flanking sequence was reduced, certain hairy stripes disappeared from the spatial mRNA pattern. Apparently, the lost cis-acting elements are required for the expression of various stripes in specific regions of the embryo and respond to regionalized cues in the blastoderm nuclei. The simplest explanation would be that gap proteins acted upon these elements to activate the different hairy stripes. The second set of experiments involve demonstrations that certain elements of the eve stripe pattern can be generated by placing 5’ flanking DNA of the eve gene upstream of the B-galactosidase reporter gene, and that different deletions within this DNA result in deletion of different eve-Bgal stripes (Harding et al. 1989; Gato et al. 1989). These studies suggest that different 5′ elements respond to regional cues that activate eve transcription.

It is also possible that gap genes could modulate pairrule gene expression by transcriptional repression. In this model formulated by B. Edgar and G. Odell (personal communication), there are two possible modes for repression to operate. In combinatorial repression, two different gap gene products work together to repress pair-rule transcription, while in competitive repression individual gap gene products act as repressors but their repressive effects are offset by competition from those gap genes that overlap them. Each gap gene domain then has a peak of repression that is narrower than the whole domain.

One set of experiments that led to this model involved cycloheximide injection into blastoderm embryos, which demonstrated that the polar and periodic repression of ftz (Edgar et al. 1986), hairy, eve and runt (Edgar et al. 1989) mRNA expression could be blocked by inhibiting protein synthesis, as was the turnover of their normally very short-lived mRNAs. This indicated that the short-lived regulators of pair-rule genes might be repressors and not necessarily activators of transcription. The combination of spatial repression and rapid RNA degradation is still a relatively simple explanation for the seven-stripes pattern, but in the case of the hairy-regulated ftz gene, this appears to be the central mode of its spatial regulation.

Our new data support a gap gene repression mechanism for hairy stripes over a regional activation mechanism because of three key observations. First, because hairy expression expands at high levels across a hb⁻, knC; Kr⁻ embryo, the idea that these genes are essential activators of hairy expression must be incorrect. Second, Kr⁺ expression in the absence of hb⁺ and/or kni⁺ is always associated with repression of hairy expression. And third, the restriction of the hairy pattern in an embryo lacking the specific hairy regulatory genes eve⁺ and runt⁺ reflects spatial repression in several regions of the embryo that must be due to gap genes. We interpret the pattern (Fig. 1J) of hairy expression in the middle of a runt⁻/eve⁻ embryo as suggesting competitive repression between hb⁺ and Kr⁺, and kni⁺ and Kr⁺. This is because the stripes that do form (a weak third and fourth) appear to be positioned near the edges of the Kr domain, overlapping with the hb⁺ domain on the anterior edge and the presumed kni⁺ domain on the posterior edge and separated by an interstripe towards the center of the Kr domain. Not all individual region-specific regulators necessarily act negatively upon hairy, the function of eve⁺ may involve activation of the second hairy stripe.

If spatial repression by gap genes is occurring, how does one explain the patterns of the 5’ cis-acting mutants of hairy (Howard et al. 1988) or of the eve-Bgal gene fusions (Gato et al. 1989; Harding et al. 1989)? In the case of the hairy civ-acting deletion mutants, we can offer an alterative explanation for the loss of hairy stripes besides the lack of positive regulation. In h^m3/ h^m3 embryos, for example, the third and fourth hairy stripes, which form over the normal Kr domain, are missing. While this might be due to the deletion of a Xr⁺-driven enhancer element, the lack of expression could also be due to the loss of gap regulatory elements that would modulate Kr repression. That is, perhaps the h^m3/h^m3 embryos lack stripes three and four because of Kr repression, not because of a failure to activate these stripes.

We can find evidence for similar dual gap gene control of eve stripes/interstripes in the experiments described by Gato et al. (1989) (see Fig. 5 therein) in that when these authors examined the pattern of the relevant eve-Bgal stripes in hb⁻ ; Kr⁻; or even g⁻ mutants, no stripe disappeared, the patterns merely shifted. If a given stripe required activation by a single gap gene, then removing that gene should delete the stripe. This was not observed. At present, we cannot tell which genes repress or activate eve expression. The spreading of eve⁺ in hb⁻ ; kni⁻ ;embryos suggests some level of spatial repression but the accompanying lower level of expression may suggest a requirement for hb⁺ and kni⁺ in eve activation. We believe that the very different effects of the hb⁻ ; kn⁻ and triple gap mutant backgrounds on eve and hairy, and the demonstration that eve is auto-regulated (Frasch et al. 1988; Harding et al. 1989; Gato et al. 1989), provide evidence that the regulatory wiring of hairy and eve may not be quite so similar as their spatial overlap may have suggested.

In order to determine how hairy and eve are regulated, it will be crucial to determine whether gap proteins interact with sequences upstream of the pairrule genes and if there are any functional associations between the gap proteins or their target sites. To understand the entire program of pair-rule gene regulation, the regulatory effect of each trans-acting protein will need to be determined in the context of the other regulatory proteins that may function competitively or combinatorially on the different pair-rule genes.

We thank Bruce Edgar for helpful discussions and he and Garry Odell for communicating their ideas prior to publication; Bruce Thalley for preparations of the hairy antibody and the data on hairy transition patterns; Allen Laughon for many important suggestions; Peter Gergen for the runt stocks; Ruth Lehmann, Christiane Nüsslein-Volhard and Eric Wieschaus for additional stocks; Ulrike Gaul for Kr antibody; and Manfred Frasch and Mike Levine for their eve antibody. We also thank Bruce Thalley and Allen Laughon for their critical review of the manuscript; Leanne Olds for artwork; and Pat Hanson and Carmen Huston for typing the manuscript. S.H.V. is a predoctoral trainee supported by a NIH training grant to the Department of Genetics. This research was supported by NSF grant DCB-8801814, a Basil O’Connor Starter Scholar Award no. 5-666 from the March of Dimes, and an NSF Presidential Young Investigator Award to S.B.C.