Concentration-dependent patterning by an ectopic expression domain of the Drosophila gap gene knirps

In Drosophila, genetic screens have identified more than 30 genes that act in a regulatory cascade to subdivide the embryo into 14 contiguous segments (Nüsslein-Volhard and Wieshaus, 1980; Nüsslein-Volhard et al., 1985). The process of segmentation is initiated by gradients of maternal gene products such as bicoid (bcd) and nanos (nos), which establish anterior-posterior polarity and specify positional information based on their concentration (reviewed in Driever, 1993; St. Johnston, 1993). These maternal factors control a cascade of zygotic gene expression, which results in increasingly refined striped patterns along the anterior-posterior axis (reviewed in Akam, 1987; Ingham, 1988). The zygotic segmentation genes have been classified according to their mutant phenotypes and expression patterns. These classifications include the ‘gap’ genes, each of which is expressed in one or two broad overlapping domains, the ‘pair-rule’ genes, which are expressed in patterns of seven transverse stripes in the precellular blastoderm, and the ‘segment polarity’ genes, which are expressed in patterns of fourteen stripes about one-cell wide. The segment polarity stripes demarcate the boundaries of the presumptive segments.

A critical step in this process is the initial establishment of the reiterated expression patterns of the pair-rule genes. Genetic experiments suggest that these patterns are set up by the combined activities of maternal morphogens and the products of the gap genes, which activate expression of individual stripes, or repress expression to create borders between stripes and interstripes (reviewed in Ingham, 1988; Carroll, 1990). Furthermore, the detailed analysis of the regulatory regions of the pair-rule genes even-skipped (eve) and hairy (h) has led to the identification of separate enhancers that control the initial expression of individual pair-rule stripes (Howard et al., 1988; Goto et al., 1989; Harding et al., 1989; Howard and Struhl, 1990; Riddihough and Ish-Horowicz, 1991).

In the eve promoter, a 480 bp enhancer regulates the expression of stripe 2 of the seven-stripe pattern (stripe 1 is the most anterior stripe), and a separate 500 bp enhancer regulates stripes 3 and 7 (Small et al., 1992, 1996). Both enhancers contain multiple binding sites for maternal and gap proteins, suggesting that direct protein-DNA interactions are important for regulating these stripes (Stanojevic et al., 1989; Small et al., 1991, 1996). The stripe 2 enhancer is activated by the maternal morphogen bcd and the gap protein hunchback (hb), while its anterior and posterior borders are set by repression mediated by the gap proteins giant (gt) and kruppel (Kr), respectively. The stripe 3+7 enhancer is activated by a ubiquitous factor, D-stat/marelle, and the anterior and posterior borders of stripe 3 are set by the gap proteins hb and kni, respectively (Hou et al., 1996; Yan et al., 1996; Small et al., 1996). These two enhancers are positioned more than 1.5 kb away from each other in the eve regulatory region and function independently because of short-range repressive interactions that form the stripe borders (Small et al., 1993).

The eve stripes directed by these two enhancers are each about 4-5 cells wide and make up part of the initial seven-stripe pattern. This pattern is then refined via cross-regulatory interactions between the pair-rule genes until each stripe is 2-3 cells wide and exhibits anterior posterior polarity. The refinement process involves a separate regulatory element that contains binding sites for pair-rule proteins including eve itself and paired (prd) (Jiang et al., 1991a; Fujioka et al., 1996).

In contrast, similar studies of the regulatory region of another pair-rule gene fushi-tarazu (ftz) have only identified promoter elements that direct all seven stripes of expression (Hiromi and Gehring, 1987; Pick et al., 1990; Schier and Gehring, 1993). This has given support to the classification of ftz as a secondary pair-rule gene, whose periodic expression pattern is regulated by other pair-rule genes (Ingham and Martinez-Arias, 1986; Carroll, 1990). Specifically, the pair-rule gene runt (run) has been implicated in ftz activation, while hairy (h) may be involved in repression events that form the posterior borders of the ftz stripes (Tsai and Gergen, 1995, and references therein). run and h stripes overlap the anterior and posterior borders of each ftz stripe respectively, consistent with this hypothesis (Ingham and Gergen, 1988; Kania et al., 1990). However, the initial pattern of ftz stripes is correctly established in h and run mutant embryos (Ingham and Gergen, 1988; Yu and Pick, 1995), suggesting that these genes may only be important for maintaining and refining pattern. Thus, the initial ftz stripes may be regulated by aperiodic cues, which may include the gap genes (Yu and Pick, 1995).

Most of the experiments described above focus on cis-acting sequences that control pair-rule pattern formation. However, less is known about the mechanisms whereby the trans-acting factors (maternal and gap proteins) set up the pattern. There is good evidence that the maternal factors bcd and hb (hb^MAT) act as gradient morphogens to establish the positions of target gene expression in anterior regions of the embryo (Driever and Nüsslein-Volhard, 1988, 1989; Struhl et al., 1989, 1992; Hülskamp et al., 1990; Simpson-Brose et al., 1994). The best-characterized target gene is hb itself, which is activated zygotically in response to the bcd and hb^MAT gradients. The posterior limit of zygotic hb expression can be shifted by genetic manipulations that change the shape of the maternal bcd and hb gradients, consistent with the hypothesis that the exact concentrations of these proteins controls the on/off state of hb transcription.

It has been suggested that the zygotic expression domains of the gap genes act as graded morphogens to establish sharp on/off patterns of pair-rule gene expression (Struhl, 1989; Gaul and Jäckle, 1990; Warrior and Levine, 1990; Pankratz et al., 1990; Kraut and Levine, 1991). Pair-rule stripes are severely disrupted in gap mutants, but these mutations also change the patterns of other gap genes, making the changes in pair-rule gene expression difficult to interpret. Similarly, heat-shock expression experiments have been used to overexpress gap proteins, causing severe effects on pair-rule expression patterns (Struhl, 1989; Kraut and Levine, 1991; Eldon and Pirotta, 1991; Hoch et al., 1992). However, these effects are also difficult to interpret since heat-shock induction results in ubiquitous expression that affects both gap and pair-rule expression domains.

In this paper, we use the eve stripe 2 enhancer and the yeast FLP-FRT system to create an ectopic expression domain of the gap gene kni in the blastoderm. We show that the strength and temporal profile of this ectopic domain can be manipulated by changing the enhancer copy number as well as the 3′ untranslated region of the misexpression construct. Relatively small changes in misexpression levels differentially affect individual eve, h, run and ftz stripes adjacent to the ectopic kni domain, suggesting unique mechanisms for the formation of each stripe. Increasing the levels of ectopic kni affects stripes positioned farther from the source of misexpression, suggesting that very low protein levels can significantly affect transcription. These results suggest that the ectopic kni domain acts as a source for a gradient of activity which represses or permits activation of individual pair-rule stripes.

The constructs shown in Fig. 1B contain one or two tandem copies of the 480 bp eve stripe 2 enhancer that contains five high-affinity bcd-binding sites (Arnosti et al., 1996). When two copies were used, they were separated by 22 bp of polylinker sequence. The BssHII site at the 3′ end of the stripe 2 enhancer(s) was fused to the PstI site at −42 with respect to the eve initiation codon. All constructs contained the eve basal promoter, as well as the eve leader sequence up to the HinfI site at +80 fused to a 2.3 kb fragment containing the 3′ termination signal from the hsp70 gene flanked by two FLP recombination targets (FRTs). This FRT cassette was fused at the 3′ end to the NruI site that lies 54 bp upstream of the kni translation initiation codon (Nauber et al., 1988). Therefore, when these constructs were activated, the 5′ UTR contained 80 bp of the eve leader, ∼100 bp of FRT sequence and 54 bp of the kni leader. At the 3′ end of the kni cDNA, two different 3′ UTRs were fused to the BamHI site located 240 bp downstream of the stop codon. The eve UTR consisted of a 1.5 kb BstUI-EcoRI genomic fragment containing the eve transcription stop. This BstUI site is located 5 bp downstream of the eve stop codon. The α-tubulin UTR consisted of an 800 bp fragment beginning at the ScaI site at position 1960 (Theurkauf et al., 1986). All constructs were generated by standard cloning procedures in the CaSpeR vector, which uses a mini-white gene as a selectable marker (Thummel et al., 1988). For each construct, between six and ten independent transgenic lines were generated in a yw⁶⁷ background using standard P-element transformation procedures (Spradling, 1986).

A transgenic line homozygous for a P(ry⁺), β2-tubulin-FLP insertion was a generous gift from Gary Struhl. This line was crossed with individual lines containing various misexpression constructs to obtain males that contained both constructs. Embryos were then collected from crosses between these males and yw females, and stained by in situ hybridization using antisense RNA probes as previously described (Jiang et al., 1991b). In experiments where levels of expression were compared, embryos were stained in parallel and photographed under identical conditions. All experiments were repeated at least three times. All embryos in this paper are oriented so that anterior is to the left and dorsal is up.

The major modifications used for the double-label experiments are as follows. Embryos were hybridized simultaneously with antisense probes for two different genes, one containing digoxigenin-UTP and the other fluorescein-UTP (labelled nucleotides, antibodies and alkaline phosphatase (AP) substrates were obtained from Boehringer Mannheim Corp., Indianapolis, IN). After washing off the unbound probes, embryos were incubated with an AP-conjugated anti-fluorescein antibody for 1-2 hours at room temperature, washed extensively in PBT (1× PBS, 0.1% Tween 80), then washed twice in AP staining buffer (100 mM Tris, pH 9.5, 100 mM NaCl, 50 mM MgCl₂, 0.1% Tween 80). The first colorimetric reaction was then performed at 25°C using Fast Red as a substrate in AP staining buffer. After staining, the embryos were washed several times in PBT to stop the reaction and were then incubated for 10 minutes in 0.1 M glycine pH 2.2, 0.1% Tween 80 to remove the first antibody. After several more washes in PBT, the embryos were incubated overnight at 4°C with an AP-conjugated anti-digoxigenin antibody. After washing as before, the second colorimetric reaction was performed using 4-nitroblue tetrazolium chloride (NBT) and 5-bromo-4-chloro-3-indolyl-phosphate (BCIP) as previously described (Jiang et al., 1991b). Staining patterns using NBT and BCIP generally appear purple or blue, but after the red reaction, this stain appears black. (Figs 8, 9). To preserve the morphology of the embryos, after washing several times with PBT, we performed an extra postfixation (25 minutes in PBT, 5% formaldehyde). After several more washes in PBT, the embryos were mounted on microscope slides using Aqua-Poly/Mount (Polysciences, Inc., Warrington, PA). Embryos were photographed using DIC optics on a Zeiss Axioscope.

In precellular embryos, kni is normally expressed as a complex domain in the presumptive head region, and a broad posterior domain that encompasses parasegments 7 to 10 (Rothe et al., 1989). Since eve stripe 2 is positioned at parasegment 3, ectopic kni expression driven by the stripe 2 enhancer would create a third domain positioned between the two endogenous domains. Because this ectopic domain is likely to cause dominant embryonic lethality, we constructed silent transformation vectors in which the kni-coding region was separated from the stripe 2 enhancer by a transcriptional stop sequence flanked by two FLP recombination targets (FRTs; Fig. 1A). This stop sequence was removed by FLP-mediated recombination in the germ line of adult males that also carried a β?-tubulin-FLP construct (Struhl et al., 1993). The offspring from these males therefore contained a single copy of the activated kni transgene.

The eve stripe 2 enhancer is a 480 bp fragment that directs a broad domain of reporter gene expression in anterior regions early in nuclear cleavage cycle 14 (Small et al., 1992). By mid-cycle 14, this broad domain is refined to a stripe 4-5 nuclei wide, which persists until the beginning of gastrulation. The enhancer contains five DNA-binding sites for bcd protein, which are critical for enhancer activation (Stanojevic et al., 1991, Small et al., 1992). By increasing or decreasing the bcd-binding affinity of these sites, the level of reporter gene expression driven by this enhancer can be dramatically altered (Arnosti et al., 1996). To optimize expression of ectopic kni in these experiments, we have used a mutant version of the enhancer that contains high-affinity bcd sites at all five positions. We have augmented expression levels further by using two tandem copies of this enhancer (Fig. 1B). Moreover, we used the 3′ untranslated regions (UTRs) from α-1 tubulin (Theurkauf et al., 1986) and eve itself to direct significantly different developmental profiles of expression.

Embryos containing activated stripe 2-kni constructs were assayed for ectopic expression by in situ hybridization using an anti-sense kni RNA probe. Each construct shown in Fig. 1B was capable of directing a third kni expression domain between the two endogenous domains (Figs 2, 3). Double staining experiments using antisense kni and eve probes determined that the ectopic domain was centered over the position of eve stripe 2 (data not shown).

We examined the kni staining pattern in embryos from several independently transformed lines for each construct. Ectopic expression levels varied minimally among individual embryos from a single transgenic line. However, we observed a range of levels among lines carrying the same construct, probably due to position effects at different genomic insertion sites (Fig. 2). For example, among seven independent lines that carried the 22FKE construct, we could discern at least four distinct levels of ectopic expression by comparing photographs of more than 20 embryos per line, and using the endogenous kni domain as an internal control for staining intensity. The highest levels of ectopic expression were very similar to the levels of endogenous kni RNA (Fig. 2G,H), and the ectopic stripe in these lines was wider than in the weaker lines, spanning an area about 6-7 cells wide even after refinement (Fig. 2H). The strongest of the six lines carrying the single copy enhancer construct (2FKE) displayed expression levels similar to the weak 22FKE lines (Fig. 2C,D).

The differences in expression levels caused by changing the enhancer copy number and the genomic location of the construct were apparent throughout the genesis of the ectopic domain (Fig. 2). However, changing the 3′ untranslated region (UTR) altered the temporal profile of ectopic expression. In cleavage cycles 12, 13, and early 14, the 22FKE construct directed high expression levels in a broad anterior domain (Fig. 3B) whereas in the 22FKT lines, we could detect only weak, scattered expression at this time (Fig. 3A). However, by mid-cycle 14, the intensity of the kni domain driven by the two constructs appeared very similar (Fig. 3C,D). Later in cycle 14, kni RNA possessing the eve 3′ UTR diminished more rapidly (compare Fig. 3F with 3E). Furthermore, the quality of the ectopic domain was also affected by changing the 3′ UTR. For example, the 2FKT construct, which contains the α-tubulin 3′ UTR, expressed at a low level in patches of nuclei in the eve stripe 2 area (data not shown), in contrast with the uniform stripe generated with the 2FKE construct (Fig. 2C). These results suggest that the eve 3′ UTR allows for early and uniform activation of the transgene and confers a rapid turnover rate in cycle 14.

In summary, analysis of the kni staining patterns suggests that we have created different types of localized gradients that vary in their shape and amplitude. By correlating these gradients with consistent perturbations in gap and pair-rule gene expression, we have attempted to determine whether ectopic kni has the properties of a morphogen.

In situ hybridization experiments using RNA probes for the gap genes gt, Kr and hb were performed on stripe 2-kni embryos. We observed no effect on the expression patterns of gt and Kr at any level of kni misexpression (data not shown). However, the ectopic kni expression did significantly alter the hb pattern. Normally, zygotic hb is activated throughout the anterior half of the embryo (Fig. 4A) in response to bcd and hb^MAT, and, starting in mid-cycle 14, is refined into a broad stripe at parasegment 4 (Fig. 4B). It has been shown that a regulatory region upstream of the bcd response element forms this stripe (Margolis et al., 1995) and that its expression is essential for the correct establishment of neighboring parasegment 5 (Hülskamp et al., 1994). Stripe 2-kni, centered on PS3, completely prevents the expression of the PS4 hb stripe (Fig. 4C,D). This repression occurs even in embryos that contain the lowest levels of ectopic kni.

During the first 20 minutes of cycle 14, a seven-stripe pattern of eve expression is formed (MacDonald et al., 1986; Frasch et al., 1987; Fig. 5A,B). These stripes appear in a particular temporal order, suggesting that the initiation of each stripe depends on a different mechanism. As outlined above, several lines of evidence suggest that gap proteins set the borders of eve stripes 2 and 3 by binding to discrete cis-elements and repressing transcription. It is likely that kni functions as a repressor to set the posterior border of stripe 3: in kni mutants, there is a posterior expansion of reporter gene expression driven by the stripe 3 enhancer (Small et al., 1996). Thus we expected that ectopic kni produced by the stripe 2 enhancer, two parasegments distant, might cause a repression of the endogenous stripe 3.

In wild-type embryos, eve stripe 3 is initiated after stripes 1 and 2, but before the resolution of stripes 4, 5 and 6. Embryos containing low levels of ectopic kni did not form stripe 3 at the correct time (Fig. 5D), and this delay was also apparent in slightly older embryos, whose stripe 3 transcription level was reduced relative to other stripes (Fig. 5E). Later in cycle 14, stripe 3 appeared nearly normal. Intermediate levels exacerbated the delay in stripe 3 initiation (Fig. 5G), but a stripe finally did appear at the correct position (Fig. 5H). In embryos with the highest levels, eve stripe 3 was completely abolished early in cycle 14 (Fig. 5J), but reap-peared later in cycle 14 in a more posterior position (Fig. 5K).

To test whether the early repression of stripe 3 was mediated through the previously characterized stripe 3 enhancer, stripe 2-kni constructs were crossed with a line carrying a lacZ reporter gene under the control of both the eve stripe 2 and stripe 3 enhancers. This reporter gene contains a 300 bp spacer sequence between the two enhancers (Small et al., 1993) and directs similar levels of both stripes (Fig. 5C). Ectopic kni specifically represses stripe 3 lacZ expression driven by this construct in a dose-dependent manner (Fig. 5F,I,L). These results support the hypothesis that kni may be a direct repressor of the eve stripe 3 enhancer.

Compared with stripe 3, other eve stripes were not as severely affected by ectopic kni. No effect was detected on stripe 1 at any level. Also, low and intermediate levels had no effect on the expression of eve stripe 2 (Fig. 5D-F,G-I). These results confirm that individual eve stripes are controlled by distinct mechanisms and that the enhancers that control the initiation of these stripes function independently (Small et al., 1993). However, the highest levels of ectopic kni caused a mild but consistent reduction in the levels of stripe 2 in some embryos (Fig. 5J,L). It is unlikely that this repression is direct, since there are no high affinity kni-binding sites in the stripe 2 enhancer and the high levels of kni do not interfere with the expression of the stripe 2-kni transgene. Interestingly, although there are strong repressive effects on stripes 2 and 3 early, these stripes recover by the time gastrulation starts, although not always in the correct position (Fig. 5K). We suggest that this recovery is due to the activities of other pair-rule genes that normally refine and stabilize the eve striped pattern, and that ectopic kni does not significantly affect these later patterning mechanisms (see Discussion).

In wild-type embryos, run stripes are formed that overlap the posterior half of each eve stripe. In early cleavage cycle 14, individual run stripes also appear in a stereotypic sequence. A transient pattern is detected at this time that is composed of stripes 1, 2, 3 and 6, which evolves into a mature seven-stripe pattern (Kania et al., 1990; Fig. 6A,B). Previous genetic experiments suggest that run is a direct target of gap gene regulation (Klingler and Gergen, 1993), although stripe-specific enhancers such as those in the eve and h promoters have not yet been identified.

Stripe 2-kni constructs caused disruptions of run stripes 2 and 3, but had no effect on stripe 1. Early in cycle 14, low levels of ectopic kni repressed run stripe 2 quite strongly, but stripe 3 only mildly (Fig. 6C). At the same age, higher levels increased repression of both stripes: stripe 2 is absent and stripe 3 is severely reduced (Fig. 6E,G). These results suggest that a similar threshold of kni-mediated repression may exist for both stripes. Nuclei in more posterior positions cross this threshold in embryos with higher levels of ectopic kni. However, a simple concentration-dependent repression mechanism cannot explain the late cycle 14 run patterns in these embryos. In the presence of low levels of ectopic kni, the early repression of run stripe 2 is alleviated, and the mature stripes 2 and 3 appear nearly normal (Fig. 6D). In contrast, the repression caused by intermediate levels persists longer, resulting in reduced levels of both run stripes 2 and 3 (Fig. 6F). In embryos containing the highest levels, run stripe 3 is more severely reduced, while expression near the position of stripe 2 recovered to near wild-type levels (Fig. 6H).

run stripes 2 and 3 respond differently to changing the levels of kni at the position of eve stripe 2. The repression of stripe 3 increases in proportion to the level of ectopic kni, a response similar to that seen for eve stripe 3. For stripe 2, all levels cause repression early in cycle 14, but there is a restoration of expression in this region that increases with the level of ectopic kni. Since these nuclei are adjacent to the source of misexpression, they must contain higher levels of kni protein than stripe 3 nuclei. Thus, it is possible that low concentrations of kni mediate repression, while higher levels permit activation. Alternatively, the activation of run in this region may be caused by the combination of higher concentrations of kni with other regulatory factors (see Discussion).

In contrast to the strong disruptions observed in the eve and run expression patterns, we could not detect any disruption of h expression in lines that expressed low or moderate levels of ectopic kni. However, the highest levels caused a subtle effect on the h pattern late in cycle 14. These embryos showed a significant delay in the separation of h stripes 3 and 4, which is one of the last events in the evolution of the h pattern (data not shown). It has been proposed that the separation of these stripes is regulated by run-mediated repression (Hartmann et al., 1994). As shown above, the highest levels of ectopic kni also cause a reduction of run stripe 3, consistent with this hypothesis.

Although ftz has been classified as a secondary pair-rule gene, the sequence of initiation of individual stripes is unique among the pair-rule genes (Yu and Pick, 1995). In early cycle 14, the first stripes formed are 1, 2 and 5, followed by stripes 3, 6 and 7, and finally stripe 4 (Fig. 7A,B). The mature seven-stripe pattern is exactly reciprocal to eve; thus, the ectopic kni domain directed by the eve stripe 2 enhancer is positioned between ftz stripes 1 and 2.

Different levels of ectopic kni caused disruptions of ftz stripes 2 and 3, but had no effect on the expression of ftz stripe 1. Early in cycle 14, even low levels caused a dramatic repression of ftz stripe 2 (Fig. 7C). This repression of ftz stripe 2 was also observed in embryos containing intermediate and high levels (Fig. 7E,G). Stripe 3 expression was unaffected at low levels, mildly delayed at intermediate levels, and strongly repressed by the highest levels (Fig. 7C,E,G). The repression of ftz stripe 3 represents the posterior-most effect that we could detect resulting from the highest levels of ectopic kni driven by the eve stripe 2 enhancer.

The early repression of ftz stripe 2 mediated by low levels of ectopic kni was still evident later in cycle 14 (Fig. 7D). These embryos showed only a very narrow stripe of ftz RNA expression about 1-2 cells wide. Similarly, embryos containing high levels continued to show a significant reduction of ftz stripe 3 (Fig. 7H), while expression was strongly activated near the normal position of ftz stripe 2. The late ftz expression pattern detected in embryos containing intermediate levels showed significant reductions in the levels and widths of both stripes 2 and 3 (Fig. 7F).

In summary, in a situation similar to that observed for run, changing the levels of ectopic kni creates a continuum of effects on the formation of ftz stripes 2 and 3. There is a strong correlation between high levels of ectopic kni at eve stripe 2 and the repression of ftz stripe 3. In contrast, after the initial repression of ftz stripe 2, expression is activated near the original position of stripe 2. This expression increases with the levels of ectopic kni.

The preceding experiments suggest that the observed changes in ftz expression are due to direct effects of ectopic kni protein. Alternatively, it is possible that these effects are indirect and may be mediated through other segmentation genes. It has previously been shown that run and h are important for maintenance and refinement of the ftz pattern (Kania et al., 1990; Tsai and Gergen, 1995), and thus are good candidates for intermediate factors involved in disrupting the ftz pattern. If this is so, their altered expression patterns should parallel those observed for ftz. By this criterion, it is unlikely that h plays any role in the ftz disruptions, since the early h expression pattern is unperturbed in embryos containing ectopic kni. However, the disruptions of the run expression pattern (Fig. 6) seem very similar to those observed for ftz (Fig. 7). To test whether the effects on these two genes coincide, we performed double-label experiments to simultaneously visualize both expression patterns. From these experiments, it is clear that the effects on run can be uncoupled from those on ftz (Fig. 8). For example, a relatively low level of kni that was sufficient to repress ftz stripe 2 had very little effect on the expression of run stripe 2 (Fig. 8B). Conversely, a higher level caused a severe reduction of run stripe 3, while ftz stripe 3 was not significantly affected (Fig. 8C).

The uncoupling of the effects on individual run stripes from those on ftz stripes, and the fact that h expression is virtually unaffected in stripe 2-kni embryos suggests that the observed effects on ftz are not mediated through these other pair-rule genes. Rather, it seems more likely that the ftz effects are caused by direct activity of the kni protein driven by the eve stripe 2 enhancer. These results support the hypothesis that kni, and possibly other gap genes, are important aperiodic cues that establish the ftz pair-rule stripe pattern (Yu and Pick, 1995).

The highest levels of ectopic kni caused a significant repression of ftz stripe 3, which is located eight to ten nuclei away from the position of eve stripe 2. This long-range effect could be the result of diffusion of kni protein from the position of eve stripe 2 to the position of ftz stripe 3. Alternatively, since the ectopic expression domain driven by the eve stripe 2 enhancer is very broad before being refined to a stripe, it is possible that effects on ftz stripe 3 reflect activity of kni protein that is produced before refinement. The early patterns generated by the strongest lines extend quite far into the middle regions of the embryo (e.g. Fig. 3B), consistent with the latter model.

To test whether the early kni prepattern might be responsible for the distant effects on pair-rule patterning, we compared a strong misexpression line containing the eve 3′ UTR (22FKE) with one containing the α-tubulin 3′ UTR (22FKT). These two lines generate indistinguishable domains of RNA expression midway through cycle 14 (Fig. 3C,D), from which similar amounts of kni protein might diffuse. Earlier, however, the 22FKE construct directs a much stronger prepattern of expression (compare Fig. 3B with 3A). Conversely, the domain produced by the 22FKT construct persists significantly longer (compare Fig. 3F with 3E). If these different expression profiles change the effective gradient of ectopic kni protein, the resulting pattern disruptions of pair-rule genes should be significantly different.

We simultaneously visualized eve and ftz mRNAs in embryos from these two lines. While both constructs caused a repression of eve stripe 3 early in cycle 14 (Fig. 9A,C), only the 22FKE construct caused a repressive effect on ftz stripe 3 (Fig. 9A), which lies farther from the misexpression source. This suggests that the early prepattern generated by 22FKE is very important for repression of ftz stripe 3. However, the repression of eve stripe 3 caused by the 22FKT construct persisted significantly longer compared to the 22FKE construct (compare Fig. 9B with 9D). Since the 22FKT domain lasts longer (Fig. 3), this later repression may reflect the activity of kni protein diffusing from the nuclei at the position of the mature eve stripe 2. Together these results suggest that both the early prepattern of expression and protein diffusion contribute to the effective gradient of ectopic kni protein.

A regulatory relationship between the gap and pair-rule classes of segmentation genes was originally inferred by examining pair-rule pattern disruptions in gap mutants. Ubiquitous expression of gap proteins also caused severe disruptions of pair-rule patterns, suggesting that the localized expression of gap genes was required for their patterning activity (Struhl, 1989; Eldon and Pirrotta, 1991; Kraut and Levine, 1991; Hoch et al., 1992). However, it was not clear that gap proteins function as gradient morphogens in this process because it was not possible to spatially control the distribution of ectopic protein in these experiments. In this paper, we have created an ectopic domain of kni mRNA using the eve stripe 2 enhancer.

Since the ectopic activation of kni might cause dominant lethality, the yeast FLP-FRT system was used in our experiments. In this system, silent constructs containing a FRT-Stop-FRT cassette were activated by FLP recombinase in the male germ line (Struhl et al., 1993). This activation before fertilization permits the generation of high expression levels during early blastoderm development. Thus, this approach may be significantly more effective than other indirect strategies for enhancer driven ectopic expression. One such strategy is the Gal4-UAS system (Brand and Perrimon, 1993), in which the enhancer drives expression of the yeast transcriptional activator Gal4, which then activates expression of a target gene through Gal4-binding sites. This strategy has worked well in several studies later in development, but may not be effective in the early blastoderm because of the delay between activating the enhancer and expressing the target gene.

We have also shown that the developmental profile of ectopic expression varies significantly depending on the 3′ UTR used in the misexpression construct. Early in cycle 14, constructs containing the eve 3′ UTR directed higher levels of expression than those containing the α-tubulin 3′ UTR. The accumulated levels of RNA driven by the two types of constructs were very similar in mid cycle 14, but transcripts containing the eve 3′ UTR disappeared before those containing the α-tubulin 3′ UTR. It has been previously shown that the UTRs of several patterning genes, including eve itself, control RNA localization to a particular region of the embryo, or to a particular subcellular region (Davis and Ish-Horowicz, 1991). Our results suggest that sequences in the UTR may be important for RNA stability and possibly for early activation of transcription as well. Also, we found that the levels of transcription directed by the stripe 2-kni constructs vary to a significant degree according to the position of insertion in the genome. These factors have permitted us to generate a series of lines that contain qualitatively different ectopic kni domains at the position of eve stripe 2.

The PS4 stripe of hb expression was abolished by stripe 2-kni (Fig. 4), even in embryos containing very low levels of ectopic expression. This suggests a mechanism of activation for this stripe that is very sensitive to repression by kni. In contrast, no changes were observed in the expression patterns of gt or Kr even in embryos containing high levels of ectopic kni, suggesting that the observed repressive effects are specific for hb among the gap genes.

The absence of an effect on the Kr pattern apparently contradicts previous experiments that suggested a direct role for endogenous kni in setting the posterior border of the Kr domain (Hoch et al., 1992). These experiments focused on a 16 bp Kr regulatory sequence that contains overlapping bcd- and kni-binding sites. In P-element transformation experiments, six copies of this sequence activated lacZ expression in the anterior part of the embryo. This activation was repressed by inducing ubiquitous kni expression using a heat-shock approach. There are several explanations for these conflicting results. First, the levels of ectopic expression driven by the stripe 2 enhancer may be insufficient for effective repression of Kr. This seems unlikely because the strongest stripe 2-kni lines, which direct levels similar to the endogenous kni gene (Fig. 2H), do not show any detectable Kr repression. Alternatively, perhaps the activities mediated by the 16 bp sequence represent only a small part of the in vivo mechanism of Kr regulation. Much larger fragments (∼700 bp) are required to drive lacZ expression in the central Kr domain (Hoch et al., 1990). In summary, it seems that kni probably plays only a minor role in setting the posterior Kr border in vivo. This view is supported by the observation that the endogenous Kr and kni domains overlap to a significant extent (Pankratz et al., 1989; D. K., unpublished data). This overlap would not be possible if a strong negative interaction existed between these two genes (Kraut and Levine, 1991).

For the pair-rule genes, different levels of ectopic kni caused disruptions of specific eve, run and ftz stripes, but there was no detectable effect on the initial h pattern. The absence of an effect on h was unexpected since individual h stripes are thought to be directly regulated by gap proteins (Pankratz et al., 1990; Howard and Struhl, 1990; Riddihough and Ish-Horowicz, 1991). In particular, the hb PS4 stripe has been previously shown to be required for the expression of h stripe 3 (Hülskamp et al., 1994), and this stripe is missing in stripe 2-kni embryos. The results reported here suggest that, even in the absence of the PS4 hb stripe, h stripes are formed that are indistinguishable from the endogenous pattern. The exact mechanism for how this occurs is not clear at present, but one possibility is that the ectopic kni may be capable of replacing some activities normally mediated by PS4 hb (see below).

The effects of stripe 2-kni on eve, run and ftz patterning can be summarized as follows. First, there are significant effects on individual stripes that lie posterior to eve stripe 2, but no effects on anterior stripes. This suggests that the regulatory mechanisms governing the expression of stripes anterior and posterior to eve stripe 2 must be significantly different. Second, the level and profile of the ectopic kni domain determines which pair-rule stripes will be affected. Low levels cause disruptions of pair-rule stripes near the source, while higher levels cause disruptions further away (Fig. 10A). Third, changing the strength and profile of the ectopic domain causes qualitatively different effects on individual pair-rule stripes. For example, embryos with very high levels of ectopic kni exhibit repression of ftz stripe 3 and run stripe 3, but these genes are both activated near their normal stripe 2 positions, which lie closer to the source of misexpression.

Previous studies indicated that kni acts as a repressor to set the posterior border of eve stripe 3 (Small et al., 1996). The identification of five kni-binding sites in the stripe 3+7 enhancer suggested that this interaction may be direct. This hypothesis is supported by the finding here that ectopic kni at eve stripe 2 caused a significant repression of eve stripe 3 early in cycle 14. The repression mediated by the ectopic kni was transient, however, suggesting that the complete abolishment of this stripe might require relatively high kni concentrations for a longer period of time. This is supported by the observation that the 22FKT construct, which directs ectopic expression longer than the 22FKE construct, also repressed eve stripe 3 for a longer period (Fig. 9).

Later in cycle 14, stripe 3 expression was restored, even in embryos carrying the 22FKT construct. This recovery is probably due to the activities of other pair-rule genes that are normally involved in maintenance and refinement of the early stripes. It has been previously suggested that h, prd and eve itself are involved in positive regulation to maintain eve expression, while run represses from the posterior to refine the stripes (Jiang et al., 1991a; Fujioka et al., 1996). Embryos containing the highest levels of ectopic kni exhibit a posterior shift of the recovered stripe. In these embryos, there is also a severe reduction of run stripe 3 (Fig. 6H), and a posterior expansion of prd stripe 3 (D. K., unpublished results), which could account for the observed shift.

Several previous experiments suggest that eve function is normally required for the late striped pattern. For example, in eve mutant embryos, all seven stripes fail to refine, and are prematurely lost (Frasch et al., 1988; Lawrence and Johnston, 1989). Furthermore, there are two eve-binding sites in a minimal eve autoregulatory element that directs reporter gene expression in seven late stripes (Jiang et al., 1991a). Mutating these sites causes a reduction in reporter gene expression, suggesting a direct role for eve in autoregulation. In the studies presented here, embryos containing intermediate and high levels of ectopic kni exhibit a recovery of eve stripe 3 expression in the absence of the early stripe 3 response (Fig. 5H,K), apparently contradicting this earlier work. However, since the ectopic kni-mediated repression of stripe 3 is transient, it is possible that low levels of eve protein are produced slightly later. Perhaps these low concentrations of eve, together with positive inputs from prd and h, may then activate the stripe maintenance response.

Previous genetic experiments suggested that hb and kni act as repressors to set the anterior and posterior borders of eve stripe 3, as well as run stripe 3, and ftz stripe 3. There are expansions of the anterior borders of eve 3 and run 3 in mutants that lack zygotic hb (Klingler and Gergen, 1993; Small et al., 1996). Furthermore, the hb PS4 stripe has been shown to be important for establishing the interstripe between ftz stripes 2 and 3, suggesting a role for hb in forming the anterior border of ftz stripe 3 (Hülskamp et al., 1994). In mutants that lack kni function, the posterior borders of all three of these stripes fail to form (Carroll and Scott, 1986; Klingler and Gergen, 1993; Small et al., 1996). However, the positions of these three stripes are offset by about two cells along the anterior posterior axis in wild-type embryos (Fig. 10A). We suggest that the different positions of the three stripes reflect their different sensitivities to the repressive activities of hb anteriorly and kni posteriorly. In this model, eve stripe 3 would be most sensitive to repression by kni, followed by run stripe 3 and then ftz stripe 3. The repression of all three of these stripes by high levels of ectopic kni is consistent with a model that kni-mediated repression normally forms the posterior border of these stripes. Intermediate levels repress eve stripe 3, whereas higher levels are required to repress run stripe 3 and ftz stripe 3.

Even though the PS4 hb stripe may be involved in setting the anterior border of ftz stripe 3, repressing the hb stripe with ectopic kni does not significantly derepress ftz transcription. One explanation for this result is that the ectopic kni domain may be able to substitute for some hb functions in the absence of the PS4 stripe. If this is so, perhaps both the anterior and posterior borders of ftz stripe 3 are set by kni in embryos that contain the ectopic domain. Consistent with this view, a one nucleus anterior expansion of ftz stripe 3 can be detected in some embryos containing low levels of ectopic kni (data not shown). Increasing the levels of ectopic kni first turns off the expression in the one cell expansion, and then represses the stripe itself.

A morphogen was originally defined as a ‘form-producing’ substance that creates pattern by diffusion within an embryonic field (Turing, 1952). A localized source of such a morphogen would establish a concentration gradient that might specify different cell fates at different positions (reviewed in Slack, 1987), possibly by specifically affecting the transcriptional state of individual target genes. Based on these criteria, the ectopic kni domain may generate a gradient of morphogenetic activity. The hb PS4 domain is sensitive to repression by very low levels of ectopic kni, while the expression of Kr is unaffected. Also, different levels are required to repress the expression of individual pair-rule genes (see above).

It is especially interesting that embryos containing the highest levels of ectopic kni show severe reductions of run stripe 3 and ftz stripe 3, but activate these genes in regions closer to the source of misexpression (summarized in Fig. 10C). This suggests that low concentrations of ectopic kni repress transcription of these genes, while higher concentrations activate, or at least permit activation. Several mechanisms could account for the observed concentration-dependent change in kni activity. One possibility is that kni protein might exist in different forms (e.g. monomers versus dimers) depending on concentration and that these forms have different activities. In transient cotransfection assays, it has been shown that the gap proteins hb and Kr can change their transcriptional activities depending on concentration (Zuo et al., 1991; Sauer and Jäckle, 1991). However, the in vivo relevance of these studies is not clear. Alternatively, interactions with other proteins may be involved in changing the transcriptional activity of kni in different positions along the anterior posterior axis. Since increasing kni levels at the eve stripe 2 position causes more overlap with the endogenous Kr domain, one possibility is that some interaction between kni and Kr may change the transcriptional state of ftz. The simplest model might involve a Kr-kni heterodimer that could activate ftz transcription. This possibility is supported by the demonstration that Kr and kni can form a heteromeric complex in vitro (Sauer and Jäckle, 1995).

Mutations in the gap genes cause disruptions in embryonic pattern outside the regions where their protein products are detectable by histochemical analysis. In the case of kni, the posterior domain of RNA expression encompasses about four presumptive parasegments at the beginning of cycle 14 (Rothe et al., 1989), but strong kni mutants show pattern disruptions of at least seven contiguous segments (A1 to A7; Nauber et al., 1988). The discrepancy between expression pattern and mutant phenotype suggests that sufficient kni protein for pattern formation exists a significant distance away from the region where its mRNA is detectable. There are at least two major factors that contribute to the effective kni gradient. Since kni and the other gap genes are initially expressed in very wide domains that are refined by interactions between the gap genes, the initial prepatterns could be the source of activity that is undetectable later. The second factor is continuous protein diffusion from the refining domain.

The expression domain directed by the eve stripe 2 enhancer is initially very broad and then refined to form a discrete stripe, and thus serves as a good model for an endogenous gap expression domain. Furthermore, since changing the 3′ UTR of the stripe 2-kni constructs changes the developmental profile of expression, it is possible to separate the effects of the early prepattern from the effects of protein diffusion. The 22FKE construct, which directs a strong prepattern, represses eve stripe 3 early, as well as run stripe 3 and ftz stripe 3. In contrast, the 22FKT construct, with a very weak prepattern, represses eve stripe 3, but not ftz stripe 3 (Fig. 9). However, the RNA generated by this construct lasts longer, resulting in a longer delay in the activation of eve stripe 3. Since the ectopic kni domain is well-refined at this time, and is four to five cells anterior to the position of eve stripe 3, the most likely source of kni protein for the extended repression is protein diffusion. For repression of ftz stripe 3, however, it seems clear that a strong early prepattern is required. Interestingly, once ftz stripe 3 has been repressed, the stripe cannot fully recover, as is observed for the eve stripes. This may reflect differences in the mechanisms of maintenance and refinement for these two genes.

We thank Gary Struhl for the β?-tubulin-FLP stock, and a plasmid containing the FRT-STOP-FRT cassette. We also thank Ueli Gross niklaus for advice on double staining embryos with two different RNA probes, Gary Struhl and Stuart Newman for critical comments on the manuscript, and Xue Lin Wu, Vikram Vasisht, and Rajesh Vakani, for stimulating discussions and moral support. This work was supported by the Keck foundation, and research grant #IBN-9513550 from the National Science Foundation.