Mis-regulating segmentation gene expression in Drosophila

Embryonic pattern in Drosophila is initiated and refined through the expression of a hierarchy of segmentation genes (Niisslein-Volhard and Wieschaus, 1980; reviewed in Akam, 1987; Ingham, 1988). Initially, maternal genes establish coarse positional signals that define domains of gap-gene transcription (Driever and Niisslein-Volhard, 1988a; Driever et al. 1989; Niisslein-Volhard et al. 1987; Struhl et al. 1989). Overlapping gradients of the gap-gene proteins regulate the transcription of the pair-rule genes that are expressed in a series of stripes (Gaul and Jackie, 1989; Pankratz et al. 1989; Stanojevic et al. 1989). Pair-rule genes are expressed in different but overlapping sets of stripes (Fig. 1A) that expose individual blastoderm cells/ nuclei to different combinations of pair-rule proteins. Thus, the relative positionings of pair-rule stripes (pairrule ‘phasings’) define the even more precise domains of segment-polarity gene expression, such as the onecell-wide stripes of engrailed (en) and wingless (wg) that mark parasegmental embryonic metameric boundaries (DiNardo et al. 1985; Ingham et al. 1988; Lawrence et al. 1987).

We are particularly interested in the regulation and function of the pair-rule genes, whose correct expression underlies the establishment of metameric pattern. Initial pair-rule transcription is first detected in broad domains during nuclear cleavage stage 12, and evolves into stripes following the final blastoderm cleavage (stage 14). This occurs over a period of 15–30 min, consistent with the extreme instability of pair-rule transcripts and proteins (t₀.₅ ∼6min; Edgar et al. 1986; Weir and Kornberg, 1985). Pair-rule gene striping is predominantly transcriptionally controlled such that transcript levels are enhanced within stripe domains and diminished between them (i.e. are repressed in the ‘inter-stripes’).

The pair-rule genes have been classified according to their principle striping mechanism (Howard and Ingham, 1986; Ingham and Gergen, 1988). The ‘primary’ pair-rule genes (h, eve and runt) are thought to respond directly to gap-gene positional cues via extensive upstream promoters with independent regulatory elements (‘stripe elements’) for individual stripes (Howard et al. 1988; Pankratz et al. 1989; Stanojevic et al. 1989). Deletion of upstream h sequences leads to the loss of specific stripes, and upstream regions can drive striped expression of β-galactosidase (lacZ) reporter genes in individual h stripe domains (Howard et al. 1988; Pankratz et al. 1990; G. Riddihough and M. Lardelli, personal communication). Similarly, individual eve stripes are independently regulated (Goto et al. 1989; Harding et al. 1989). Less is known about the negative control of h and eve transcription, although individual stripe elements must include repressor sites to prevent interstripe expression.

In contrast, patterning of ‘secondary’ pair-rule genes (e.g. fiz) is largely a response to the striping of the primary pair-rule genes, fiz striping depends on a small upstream transcriptional control element (the ‘zebra’ element) that confers striped expression on a lacZ reporter gene, suggesting that all fiz stripes respond to a similar signal (Hiromi et al. 1985; Hiromi and Gehring, 1987; but see Dearolf et al. 1989a; Ueda et al. 1990). h and eve are both implicated in repressing fiz expression; indeed fiz stripe domains correspond to the cells that express neither h nor eve (Fig. 1A; Frasch and Levine, 1987; Carroll et al. 1988; Hooper et al. 1989; Ish-Horowicz et al. 1989). The sites through which h and eve act have not yet been defined although putative negative control elements have been defined within the fiz zebra element (Dearolf et al. 1989b).

fiz and eve are further subject to positive autoregulatory control, each promoter including a domain that activates positive feedback of transcription (Hiromi and Gehring, 1987; Goto et al. 1989; Harding et al. 1989). Ectopic fiz expression can transactivate endogenous fiz expression in specific cells (Ish-Horowicz et al. 1989). However, the normal role of such feedback may lie in ensuring persistent expression during germ-band extension to preserve the metameric boundaries that are initially defined by the anterior margins of eve and fiz expression (Lawrence et al. 1987).

The major task is to distinguish direct and indirect interactions between segmentation genes. The initial hierarchy was inferred from mutant cuticular phenotypes and from patterns of segmentation gene expression in mutant embryos. However, the large number of interacting genes makes it difficult to define direct genetic pathways, and has led to studies of ectopic segmentation gene expression using an inducible heat shock promoter to drive generalised segmentation gene expression during the blastoderm stage. Ectopic expression can be induced in precisely staged embryos, allowing immediate responses to be distinguished through their kinetics. In this manner, the effects of generalised fiz, h or hunchback (hb) expression at blastoderm have been explained in terms of direct effects on the expression of other segmentation genes (Struhl, 1985, 1989; Ish-Horowicz and Pinchin, 1987; Ish-Horowicz et al. 1989). For example, ectopic h leads to rapid extinction of fiz expression, consistent with It’s role as a primary repressor of fiz expression (Ish-Horowicz and Pinchin, 1987).

More restricted spatial mis-expression would allow investigations of the finer mechanisms that must underly the precision of final blastoderm pattern, e.g. pair-rule stripe phasing. More precise disruptions could be achieved by using promoters that themselves display spatial regulation, i.e. segmentation gene promoters. Different patterns of mis-expression would arise according to the heterologous promoter chosen.

Previous experiments indicate that such misexpression leads to pattern disruptions and dominant lethality. For example, uncontrolled expression of fiz, h or runt causes pattern defects, indicating that expression in inter-stripe domains is deleterious and causes embryonic lethality (Struhl, 1985; Gergen and Wieschaus, 1986; Ish-Horowicz and Pinchin, 1987; Ish-Horowicz et al. 1989). Thus, the precise spatial and temporal patterns of segmentation gene expression are crucial in defining the embryonic body plan, and ectopic expression of segmentation gene products can redirect the fates of cells inappropriately expressing these genes. Nevertheless, several schemes might permit the recovery of flies transformed with predicted dominant-lethal constructs. For example, protein levels could be reduced by depressing translational efficiency. Alternatively, functional expression might be conditional on combining two constructs that are individually viable (e.g. nonsense mutation+tRNA suppressor; inducible promoter+trans-activator -cf. Kakidani and Ptashne, 1988; Webster et al. 1988).

As a preliminary to such experiments, we have generated three gene fusions that express the pair-rule genes, fiz, eve and h under the control of a gap-gene promoter (hb). These constructs should drive anterior mis-expression of the pair-rule proteins within the hb domain, the anterior half of the embryo. Such disruptions of segmentation domains should be dominant lethal although the exact effects on pattern will depend on the individual pair-rule gene and the degree of its mis-expression (i.e. the extents to which its endogenous domains overlap that of hb -Fig. 1A).

This paper describes and discusses our unexpected findings that flies transformed with such constructs are viable. We show that the fusion constructs are active and mis-express pair-rule genes in hb-like patterns, but that interstripe ectopic expression diminishes when pair-rule genes begin to stripe. These results illustrate the importance of timing in segmentation gene function, and indicate that pair-rule genes have transcriptional control regions downstream of their transcription start sites. We suggest that interstripe hb–fiz expression is repressed by eve. We also describe an unexpected effect of ectopic h expression on sex determination, which illustrates the potential dangers of analysing gene function through gene mis-expression (see also Parkhurst et al. 1990).

Flies were cultured on yeast, maize meal, molasses, malt extract, agar medium, at 25°C unless otherwise stated. The null alleles used in this study are: Df(2R)eve¹²⁷, cn sea bw sp/CyO, Df(3R)4Scb/TM3 (fiz), and Df(3R)h¹²², Ki roe [P/TMS. The FG2 fiz–lac Z transformant stock expresses a fiz–lacZ fusion protein that localises in the nucleus (Y. Hiromi, personal communication). The eve-lac Z transformant stock is described in Lawrence et al. (1987).

The 4.7 kb BamHI–XbaI fragment containing the hb promoter and all but 10 bases of the 5’ untranslated leader sequences was subcloned into the blunt-ended Sall site of pUChsneo (Steller and Pirrotta, 1985). This vector, hbneo, drives anterior zygotic expression of coding sequences inserted at unique BamHI or Smal polylinker sites adjacent to the hb promoter. For hb–fiz, the Avail (−75 bp) to HindIII (+2.5 kb) genomic fragment including all of the 5′ untranslated leader from pFKl (Hiromi et al. 1985) was subcloned into the blunt-ended Bam HI site of hbneo. For hb–eve, the 4.7 kb XhoI genomic fragment including all of the 5′-untranslated leader sequences of p48-X4.7 (Macdonald et al. 1986) was subcloned into the blunt-ended BamHI site of hbneo. For hb–h, the 6.5 kb Xbal genomic fragment including 230 bases of 5′ untranslated leader sequences (Rushlow et al. 1989) was subcloned into the blunt-ended BamHI site of hbneo. The appropriate orientation for all clones was determined by restriction analysis.

Embryos were prepared and analysed as described by Wieschaus and Niisslein-Volhard (1986). Immunohistochemical detection of h, fiz, eve, en and fi-gal was performed essentially as described by Macdonald and Struhl (1986), using biotinylated secondary antibodies and avidin-biotin-HRP complexes (Vector Laboratories, Inc.). The antibodies used in this study were generously provided by: H. Krause, rabbit anti-/tz antibodies (Krause et al. 1988); M. Frasch, rabbit anti-eve antibodies (Frasch et al. 1987); M. Wilcox, monoclonal anti-en antibodies (Patel et al. 1989); D. Tautz, rabbit anti Ab antibodies (Tautz, 1988); K. Hooper, rabbit anti-h antibodies (Hooper et al. 1989) and H. Durbin, 4C7 monoclonal anti-/3-ga/ antibodies (Imperial Cancer Research Fund). All secondary antibodies were obtained from Jackson ImmunoResearch Labs (West Grove, PA). The stained embryos were dehydrated in 100% ETOH and mounted under a coverslip in methacrylate mounting medium (JB-4, Polysciences) that was polymerised under CO₂ for 1–2 h at room temperature.

Immunohistochemical whole-mount in situ hybridisation was performed according to the protocol of Tautz and Pfeifle (1989). The probes used for the random priming were: the 4.7 kb Xhol genomic fragment for eve (p48-X4.7; Macdonald et al. 1986), the 3.5 kb EcoRI genomic fragment for fiz (pFKl; Hiromi et al. 1985), and the 1.8kb EcoRI cDNA fragment of h (ThΔl; Rushlow et al. 1989).

bw;st embryos were transformed by injection with a mixture of recombinant plasmid (500 μg ml ⁻¹) and helper plasmid (100 μg ml ⁻¹), as described by Spradling (1986). The bw;st G_o adults were outcrossed to wild-type and selected on standard medium supplemented with Geneticin G418 Sulphate (Gibco -1.5 mg ml^-1 but varied according to batch). G_tbw/+; st/+ progeny were mapped by back-crossing crossed to bw;st on G418 food. This retested their drug resistance and assigned the insert to a specific chromosome, allowing construction of homozygous or balanced stocks. All neo-resistant transformants were confirmed by Polymerase Chain Reaction (Erlich, 1989) using primers specific to neo portion of the P-element transformation vector.

We measured the position (anterior margin) of the ectopically expressed stripes in hb –eve; eve⁻ embryos and compared them to the endogenous fiz stripe 1 and eve stripes 1 and 2 in wild-type embryos. Ten embryos were measured for each stripe and the results are expressed below in percentage egg length, where 0% is the posterior end:

Thus, the two ectopically expressed hb-eve stripes are coincident with the endogenous eve stripes.

We analysed all three fusion gene constructs for dominant interactions (eg., hb –fiz/ +; Kr⁻/+) with the gap alleles Kr¹, hb^PTXIS and kni^11D4S and for dominant as well as recessive (eg., hb –fiz/ +; h⁻/h⁻) interactions with the pair-rule deletions Dfffleve¹’, Df(3)4Scb (fiz⁻), Df(3)h¹²² and Df(l)run^lllB. The only interaction identified was hb –fiz/+; eve⁻/ +. (The recessive interaction (hb–fiz/+; eve⁻/eve⁻) could not be tested due to the lethality of the transheterozygotes.)

Adult males were etherised, mounted on metal discs with double-sided tape, sputter-coated with ionised gold, then viewed with a Phillips 515 scanning electron microscope.

We used the hb promoter to examine the effects of misexpressing pair-rule genes in the anterior of the embryo. Three fusion genes – hb – fiz, hb – eve, and hb – h – were generated by fusing a 4.7 kb hb promoter fragment to genomic coding sequences from the fiz, eve and h genes (Fig. 1C). The three fusion genes retain most of the hb and pair-rule gene 5′-untranslated leader sequences, while excluding the 5’ flanking sequences of the pair-rule genes that are known to function in striping (Fig. 1C; see Materials and methods). Reporter gene constructs indicate that these hb sequences should be sufficient to mis-express pair-rule genes in the anterior 45 % of the embryo, through about 2 pair-rule stripes (Fig. 1A,B – Driever et al. 1989; Hülskamp et al. 1989). Anterior zygotic hb expression derives from the proximal of two promoters that is first active at stage 11/12, preceding pair-rule expression by about two cleavage-cycles (Fig. 1C – Tautz et al. 1987; Schroder et al. 1988).

The 4.7 kb hb fragment also includes part of the distal promoter that is first expressed during oogenesis, depositing maternal transcript into the oocyte (Fig. 1A. Tautz et al. 1987; Schroder et al. 1988). This hb promoter is also zygotically active during blastoderm stage 14 in two major stripes, one abutting the anterior zygotic domain, and one in the posterior of the embryo (Fig. 1A,B -Tautz and Pfeifle, 1989). The posterior hb stripe overlaps and extends posterior to h/eve stripes 7 (which share approximately similar phasings), fiz domains are reciprocal to those of eve (Frasch and Levine, 1987) so the posterior hb stripe also overlaps fiz interstripe 6/7. The 4.7 kb fragment drives maternal expression, but previous experiments have not revealed whether this fragment is sufficient for the posterior stripe expression.

Each of these constructs were introduced into the fly germ-line (see Materials and methods). Unexpectedly, transformed lines were readily recovered for each construct; 10 hb–fiz; 6 hb–eve; 16 hb–h, indicating that these fusion genes do not give rise to significant degrees of dominant lethality. The viability of the transgenic flies suggested either that the constructs do not cause pair-rule gene mis-expression (e.g. because of an inactive hb promoter), or that such mis-expression is either tolerated or repressed. We therefore examined expression of the fusion genes and determined the effects of ectopic segmentation gene expression on embryonic pattern.

We shall consider each construct in turn.

We analysed four independent hb–fiz stocks (in the presence of their endogenous fiz genes) and all show the hb–fiz gene directs ectopic fiz expression. Very weak overall fiz expression is first seen at stages 10/11, before the onset of zygotic hb transcription (Fig. 2A). We have not investigated this phase of fiz protein expression further, but appropriate genetic crosses for hb-eve (below) predict that the overall fiz misexpression is derived from the distal (maternal) hb promoter. Thereafter, hb-fiz embryos express ectopic fiz protein zygotically in the anterior hb domain during nuclear cycle 12, two cleavage cycles before the endogenous fiz protein is normally seen (Fig. 2B). At the beginning of cycle 14, endogenous fiz expression begins which is superimposed on the ectopic hb–fiz pattern (Fig. 2C,D). The anterior domain othb protein expression extends into fiz stripe 2 (Fig. 2C-E), and the posterior hb–fiz stripe results in continuous fiz expression between stripes 6 and 7 (Figs 1A, 2D,E).

Surprisingly, fiz mis-expression fails to persist through the blastoderm stage, although hb expression is detectable until the onset of gastrulation (Tautz and Pfeifle, 1989). Most ectopic fiz staining decays during blastoderm stage 14, the time at which endogenous fiz striping becomes prominent (Fig. 2C–F).fiz expression is reduced in interstripe domains (i.e. between stripes 1/2, and 6/7 and anterior of stripe 1 -Fig. 2E,F). By the end of the blastoderm stage, ectopic fiz expression is restricted to a novel stripe, 3–4 cells anterior to fiz stripe 1, that does not correspond to a normal hb domain (Fig. 2F).

Most hb–fiz lines are completely homozygous viable and show no significant embryonic cuticular pattern defects (not shown). Nevertheless, hb–fiz encodes an active protein. ∼20% of homozygous hb–fiz adults lack external genitalia that derive from anlagen of segments A8-11 (Schüpbach et al. 1978; Tautz et al. 1987 -Fig. 3A–C). Although the missing structures derive from within the posterior hb stripe domain, we cannot unambiguously demonstrate ectopic fiz expression in this region. The altered pattern in such embryos indicates that the hb–fiz gene encodes functional fiz protein.

Indeed, hb-fiz partially rescues the pattern defects of fiz⁻ embryos. The T1 ‘beard’ is restored, as are various chitinised mouthpart structures (Fig. 3E). This implies that the hb promoter is still active during cycle 14 when fiz is needed for patterning. Nevertheless, ectopic fiz is no longer expressed between fiz stripes, indicating that zygotic expression from hb–fiz is negatively regulated in the head and fiz interstripe regions where it would cause pattern defects. The remaining fiz mis-expression in the head region appears not to cause significant pattern abnormalities.

We analysed eve in three transformed lines with autosomal insertion sites and all behave similarly, eve protein expression in hb–eve embryos is first detectable as generalised nuclear staining at blastoderm stage 10/11 (Fig. 4A). This protein derives from maternal transcript as it is only seen in embryos from hb–eve mothers, but not from wild-type mothers. Such maternal staining is only transitory, soon being replaced by the zygotic hb pattern of expression.

Strong ectopic anterior eve protein expression is first evident at blastoderm stage 12/13, and persists until stage 14 when it overlaps endogenous eve stripes 1 and 2 (Fig. 4B–D). A weak posterior eve stripe is seen during early stage 13/14 (not shown), but its expression is soon masked by the endogenous eve stripe 7. During blastoderm stage 14, hb–eve expression decays until, by the end of blastoderm, mostly endogenous protein expression exists with very low level ectopic expression in a small anterior cap (Fig. 4D,E). eve is thought to act at the late blastoderm stage to regulate segmentpolarity gene domains and to define the odd-numbered parasegmental boundaries (Lawrence et al. 1987; Ingham et al. 1988). The lack of ectopic eve expression at this stage explains the viability of heterozygous hb–eve embryos.

However, all five autosomal hb–eve lines are homozygous and trarty-heterozygous lethal indicating that two doses of hb–eve are unconditionally lethal. A sixth line, in which hb–eve is X-linked, is weaker and can be made homozygous. In hb–eve balanced stocks, about 25 % (presumably the homozygous embryos) show a consistent cuticular phenotype including fusions of T2/3, Al/2 and A3/4, and loss of the A6 denticle band (Fig. 5A,B). Strikingly, pattern abnormalities arise outside the hb domain where little or no ectopic eve is expressed. These could be due either to non-autonomous action of zygotic eve protein, or to generalised maternal expression from the distal promoter (see Discussion).

hb–eve also exerts a weak dominant effect on segmentation and on viability. Most hemizygous (single-copy) hb–eve embryos survive, but about 20% (37/183) die with weak and occasional fusions of adjacent denticle bands (Fig. 5C,D). The frequency and character of the defects is independent of whether the hb–eve gene is maternally or paternally inherited, indicating that they are due to zygotic, not maternal, eve mis-expression.

Cross-regulatory interactions among primary pairrule genes indicate that eve regulates the pattern of other pair-rule genes (Ingham and Gergen, 1988) and implies that the hb–eve pattern defects may be due to ectopic eve affecting expression of other segmentation genes. We therefore analysed the patterns of h, fiz and en expression in hb–eve embryos.

fiz and h patterns are indeed affected by hb–eve. We analysed embryos from balanced hb–eve stocks in which one half of the eggs contain a single hb–eve copy and one quarter are homozygous for hb–eve. Homozygous hb–eve embryos (24/79) show partial or complete fusions of fiz stripes 3 to 6 (Fig. 5E). Hemizygous embryos show a weaker phenotype in which stripes 3 to 6 are present but compressed (Fig. 5F).

hb–eve also disrupts h expression, stripes 1,2,3 and 7 becoming stronger and broader relative to the other bands (Fig. 5G,H). More strikingly, anterodorsal h expression (stripe ‘O’ in Fig. 1A) is completely missing in 35 % (21/60) of these embryos. This suggests that the ectopic eve suppresses the regionalised activation of h stripe 0. h stripes 1 and 2 are not eliminated, indicating that eve is interfering with regional signals specific for stripe 0. The cis-regulatory region for h stripe 0 has not yet been characterised, but may be responding directly to elevated levels of bicoid and/or dorsal morphogens (Driever and Niisslein-Volhard, 19886; Steward et al. 1988).

Initial metameric patterning is roughly normal in hb–eve as judged by the earliest pattern of en expression (not shown). However, eve expression at gastrulation, which should be similar to that of en, is somewhat abnormal in homozygous hb–eve embryos. 25 % of hb–eve embryos, show 14 eve stripes whose domains appear correctly positioned but whose anterior margins (which parallel those of en) appear less well defined (Fig. 4E,F). eve expression is also weaker than wild-type. This altered pattern is found in all three lines examined as well as in trarw-heterozygous lines containing two different hb–eve copies.

Drastic effects on metameric patterning become apparent about 1 h later when some embryos begin to show low-level en expression in all cells. By 6–7 h postfertilisation, the generalised en expression becomes stronger and seen in 37% (41/112) of embryos (Fig. 51). In some embryos, the endogenous en stripes are still visible above the generalised expression and are disorganised in about half such embryos (Fig. 5J). Thus, the pattern defects in homozygous hb–eve embryos are due to an inability to maintain metameric subdivisions.

Fig. 6A shows ectopic expression of h in the anterior region of hb–h embryos at about nuclear cycle 12. This expression begins to clear during early cycle 14 (Fig. 6B-D), although the hb promoter remains active in a stripe of cells anterior to stripe 1 (Fig. 6E). hb–h embryos show no obvious segmentation defects, consistent with the cessation of ectopic h expression before it would inhibit fiz expression. Ectopic h expression in a h⁻ background partially rescues h pattern defects in the anterior of the embryo, indicating that hb–h is active while metameric pattern is being established. Mouthparts become more organised, and anterior structures including the maxillary sense organs and T1 denticle band are restored (Fig. 6G).

Unexpectedly, hb–h interferes with sex determination, a process in which h does not normally function. hb–h males are fully viable and fertile whereas more than 99 % of hb–h females die as embryos whose head defects correlate with the domain of h mis-expression. See Parkhurst et al. (1990) for a detailed examination and explanation of this phenotype.

The above results show that homozygous hb–fiz, hb–h and most hemizygous hb–eve transformants can tolerate early ectopic expression of the respective pair-rule gene, but that later expression does not lead to pattern defects. This is not merely due to lack of promoter activity as hb–fiz and hb–h partially rescue embryos lacking endogenous fiz and h, respectively. (The variable hb–eve cuticular phenotype prevents unambiguous identification and analysis of hb–eve; eve⁻ embryos.) Rather, interstripe expression is eliminated before it can affect pattern, either by regulation of transcript levels or by inhibition of translation of the hybrid mRNAs.

We excluded the latter explanation by showing that transcript patterns mirror those of the mis-expressed pair-rule protein. In hb–fiz embryos, fiz transcripts initially accumulate in the anterior half of the embryo, but are then repressed between the normal stripes, leaving only a band of ectopic transcripts anterior to fiz stripe 1 (Fig. 7A–D). eve and h transcription in hb–eve and hb–h embryos, respectively, mimic the patterns of protein accumulation (Fig. 7E–L), showing that negative regulation of interstripe expression from the hb fusion genes is transcriptionally/post-transcriptionally (but not translationally) regulated. The different expression patterns of the three fusion genes shows that the regulation must act through pair-rule sequences present in the fusion constructs.

Analysis of the hb–fiz; fiz⁻ and hb–eve; eve⁻ embryos also show that the fusion gene constructs retain control sequences that repress their expression in the interstripe regions. We find that the initial fusion-genestaining patterns are not altered in fiz⁻or eve⁻ mutant embryos. All stage 13 embryos in a balanced hbβfiz; fiz⁻ stock show high-level anterior fiz expression, including the 25 % of embryos that must lack endogenous fiz activity (not shown). Similarly, all embryos from a balanced hb–eve; eve⁻ stock, including those lacking an endogenous eve gene, show ectopic anterior eve expression that can only derive from the hb–eve fusion gene. During blastoderm stage 14, hb–eve; eve⁻ embryos are distinguished by their lack of endogenous striped expression. Such embryos still express eve in the anterior hb domain except within two domains that correspond in position to the two overlapping eve interstripes -anterior to stripe 1 and between stripes 1 and 2 (Fig. 8A,B; see Materials and methods). By the late blastoderm stage, hb transcripts from the distal hb promoter accumulate in two anterior stripes, the more posterior of which corresponds to fiz stripe 1 (Tautz and Pfeifle, 1989); however, these expression domains do not overlap with those of eve ectopic expression in hb–eve; eve⁻ embryos. Thus, the control elements mediating such negative control must exist within the eve sequences included in the hb–eve construct.

It is more difficult to visualise interstripe repression of hb–fiz, as endogenous fiz domains overlap domains of late zygotic hb expression. However, the interaction between hb–fiz and eve suggests that fiz also retains a repressor element that imposes its negative control on the hb promoter (see below), hb–h may also retain downstream negative regulatory elements that clear ectopic h expression between h stripes 1 and 2, but the female lethality of hb–h embryos has prevented our demonstrating this directly.

Generalised fiz expression from the inducible heatshock promoter causes pattern defects by autoregulat-ory activation of the chromosomal fiz gene (Hiromi and Gehring, 1987; Ish-Horowicz et al. 1989). Similarly, eve can autoregulate its own expression (Harding et al. 1989). However, initial fusion-gene-staining patterns are not altered in fiz⁻ or eve^- mutant embryos, suggesting that hb–fiz and hb–eve are independent of endogenous fiz or eve activity. We confirmed this by using lacZ fusions to the fiz and eve promoters to monitor endogenous promoter activities (Hiromi et al. 1985; Lawrence et al. 1987). hb–fiz; fiz–lacZ embryos display no ectopic lacZ expression (not shown), indicating that endogenous fiz transcription is not autoactivated by the Aô-encoded ectopic fiz protein. Similar results are obtained using hb–eve and a eve–lacZ fusion gene (not shown). Thus, hb– fiz and hb–eve ectopic expression do not autoactivate endogenous expression. The female lethality associated with the hb– h construct precludes our analysis of hb–h; h⁻ embryos, but the h gene appears not to be autoregulated (Hooper et al. 1989).

As interstripe repression is likely to be mediated by other segmentation genes, we analysed phenotypic interactions between the fusion genes and gap or pairrule mutations, hoping to identify such repressors. Although most mutant combinations show no dominant interactions, hb–fiz and eve trans-heterozygotes (hb–fiz/+; eve /+) are lethal (0/187 adult progeny), dying as embryos with substantial pattern defects (Fig. 8C).

Fig. 8D shows that interstripe fiz expression persists in hb–fiz/+; eve⁻/ + embryos, suggesting that eve is responsible for repressing such interstripc expression.

In contrast, hb–fiz shows no dominant interactions with h, the other well-characterised fiz repressor, either because hb–fiz is /r-independent, or because 50% of wild-type h levels is sufficient to regulate hb–fiz-We favour the former explanation as anterior fiz expression is still regulated in embryos that completely lack h. fiz expression is still repressed anterior to stripe 1 and between stripes 1 and 2 in hb–fiz; h⁻ embryos, suggesting that h does not regulate hb–fiz (Fig. 8E), although the broadened endogenous fiz expression in h⁻ embryos (Carroll and Scott, 1986; Howard and Ingham, 1986; Hiromi and Gehring, 1987) precludes detection of low-level interstripe expression. This indicates that eve is likely to be a major repressor of fiz expression.

In this paper, we show that ectopic expression of fiz, h or eve under the control of the hb promoter does not necessarily result in pattern defects. Embryos containing the hb fusion genes initially express pair-rule genes in the anterior hb domain, but such ectopic expression largely ceases during blastoderm stage 14 (Figs 2, 4, 6). This is partially due to a decline in activity of the proximal hb promoter, as well as to its regulation by residual pair-rule sequences. Several lines of evidence demonstrate that the hb promoter in these fusion genes is still active at the time of pair-rule gene function and is regulated in the head and interstripe regions. First, the fusion gene constructs partially rescue the mutant pairrule phenotype (Figs3B and 6G). Second, at blastoderm stage 14, there are ectopic stripes of eve in a hb–eve; eve⁻ background that overlap the positions of the endogenous eve stripes (Fig. 8A,B). Third, although the same promoter is used for all three fusion gene constructs, each is expressed ectopically in head domains that differ between constructs (Figs 2F, 4E, 6E and 7C,G,K). Finally, interstripe expression is no longer regulated in hb–fiz/+; eve/+ embryos (Fig. 8D), and persists through the time of pair-rule gene expression.

For all three fusion genes, the patterns of transcript and protein localisation are similar, indicating that the lack of interstripe expression is not due to a failure of translation (Fig. 7). Clearing of interstripe transcripts through their differential stability is also unlikely as segmentation gene transcripts are extremely unstable (Edgar et al. 1986, 1989). Although post-transcriptional mechanisms cannot be excluded, lack of interstripe transcripts is most likely due to repression of interstripe transcriptional initiation.

The most likely candidate for a fiz interstripe repressor is eve. The lethality of hb–fiz/+; eve/+ embryos shows that reduced eve levels exaggerate the effects of ectopic fiz expression, i.e. that eve normally acts to inhibit fiz-This is consistent with previous suggestions that eve repression defines the anterior boundaries of each /tz stripe (Ish-Horowicz et al. 1989), and would be mediated, at least in part, through downstream eve-responsive elements. In contrast, hb–fiz shows no interactions with h, the other characterised fiz repressor, hb–fiz expression is repressed between fiz stripes 1 and 2, even in hb–fiz embryos lacking h (Fig. 8E), suggesting that h does not act on hb-fiz and that h regulation of fiz striping might operate through upstream fiz sequences.

We do not know which genes repress eve and h expression in hb–eve and hb–h. The best candidate is the runt pair-rule gene whose stripe domains are roughly complementary to those of eve and h (Gergen and Butler, 1988; Ingham and Gergen, 1988). Although there is no direct evidence for such regulation, we consider it more likely that the downstream control regions react to a single pair-rule regulator than to alternative combinations of differing gap-genes.

The timing of pair-rule gene expression is crucial: embryos are unaffected by generalised anterior misexpression during blastoderm stages 10 to 13. Only late in cleavage cycle 14 do the pair-rule genes act to regulate segment-polarity gene expression and metameric pattern, by which stage expression from the fusion genes is restricted to functionally irrelevant domains.

Further indications of the importance of timing in patterning the early embryo comes from the temporal specificity of pair-rule autoregulation. Neither hb–fiz nor hb–eve autoactivate their endogenous genes, despite the presence of autoregulatory elements within each promoter. In contrast, late blastoderm and early gastrula-staged embryos are susceptible to fiz autocatalytic activation (Struhl, 1985; Ish-Horowicz et al. 1989). Autoregulation appears to be important for persistent expression during gastrulation and germ-band extension, but not during the earlier phases when pair-rule domains are being established (in contrast to reaction–diffusion models for pair-rule striping – Meinhardt, 1982). We note that h is not autoregulated and that its expression decays immediately following blastoderm (Hooper et al. 1989).

Although all three constructs are viable, each has specific effects on development.

Homozygous hb–fiz embryos show no obvious embryonic cuticular defects, but a proportion of adult flies lack terminal structures. The exact basis for this pattern abnormality is unclear, although it might be due to a weakly expressed posterior stripe of ectopic fiz expression, which extends into the A8–11 genital primordia. Only occasional cells can be affected as most adults are viable and the embryonic en pattern appears normal.

Unlike the other two constructs, hb–eve causes significant pattern abnormalities, with two copies being almost completely lethal and leading to metameric instability and subsequent segmentation defects, eve stripes during gastrulation are weak and irregular, and a high proportion of older hb–eve embryos display a generalised pattern of en expression in which clear en boundaries are lacking (Fig. 51). Nevertheless, hb–eve embryos retain considerable metameric organisation (Fig. 5A,B), indicating that parasegmental boundaries can be maintained even in the absence of clear en boundaries. We presume that métamérisation initially requires en stripes, but thereafter other segmentpolarity genes can contribute to intrasegmental patterning.

The major surprise is that the pattern abnormalities in homozygous hb–eve embryos are not restricted to the hb domain, i.e. to the domain of eve mis-expression. Thus, eve stripes at gastrulation are disrupted throughout the embryo (Fig. 4F; see also Fig. 5A,B). Such nonautonomy is unexpected as eve encodes a nuclear homoeobox protein whose direct actions should be local, i.e. restricted to the hb domain. Such defects are seen (albeit rarely and more weakly) in heterozygous embryos (even when from wild-type mothers), indicating that they are due to zygotic expression, presumably in the posterior domain. Although we do not directly detect such expression, we note that the domains of gap-gene action extend into domains where protein levels are immunologically undetectable (Gaul and Jackie, 1989; Pankratz et al. 1989, 1990; Stanojevic et al. 1989; Hülskamp et al. 1990).

Although hb–h causes female lethality by interfering with sex determination (Parkhurst et al. 1990), the viability of hb–h males shows that the ectopic h does not cause segmentation defects (even in two doses, unpublished observations). This is unexpected as h and eve are both primary pair-rule genes that can affect each others patterning (Ingham and Gergen, 1988). Indeed, hb–eve affects h patterning, although eve pattern is normal in hb–h embryos. This may indicate that h’s role in embryonic patterning is subsidiary to that of eve. Alternatively, the embryo could be less sensitive to ectopic h because h’s targets are under more redundant control (see below).

Previous experiments have demonstrated roles for upstream sequences in regulating pair-rule striping. The fiz zebra element is able to direct striped expression of reporter genes, albeit predominantly in the mesoderm (Hiromi and Gehring, 1987), and putative negative regulatory elements have been defined within this region, including potential eve binding sites (Dearolf et al. 1989b).

h and eve striping appear to be regulated differently from/tz. Upstream domains appear to control specific individual stripes, presumably by sensing regionalised spatial cues (e.g. gap genes; Howard et al. 1988; Struhl, 1988). Upstream h elements can confer striped expression on a reporter gene construct, indicating that they include both positive and negative elements (Pankratz et al. 1990; G. Riddihough and M. Lardelli, personal communication). Our experiments indicate that there are also repressor sites downstream of the transcription start, i.e. that stripe repression is under redundant control. A likely reason for redundant control is to achieve the necessary precision of striping to allow precise phasing between different pair-rule genes, h and eve show similar stripe domains except that each h stripe is 1–2 cells anterior of each eve stripe (Carroll et al. 1988). Such displaced phasings could arise because h and eve sense similar positional cues, but with slightly differing affinities for their signals. Such striping would involve upstream repressor elements acted upon by gap genes proteins. The final stripe phasings would be achieved by the action of other pair-rule genes acting, at least in part, through downstream elements. Our analysis of h striping patterns in embryos mutant for other segmentation mutations has suggested that both mechanisms may operate, i.e. that the control is redundant (Hooper et al. 1989). Such redundancy might be required to define stripes with precise phase relationships.

Two further messages come from these experiments. First, the unexpected viability of the fusion genes indicates the difficulty of predicting the outcome of simple mis-expressing constructs, and the need to test them before embarking on more complex strategies. Second, mis-expression experiments can give extremely mis-leading impressions of wild-type function. The female lethality of hb–h arises despite h’s playing no normal role in sex determination. Although hb–h has proved very valuable in studying helix-loop-helix proteins and mechanisms of sex determination (Parkhurst et al. 1990), the results would have been misinterpreted without previous genetic evidence of wildtype h function. Mis-regulation experiments in genetically less-well-characterised systems (e.g. vertebrates and cultured cells) should be interpreted with caution unless wild-type function is assayed independently.

We thank Diethard Tautz for sharing unpublished information and dealing patiently with our many questions; Suki Parks for suggestions and many helpful discussions. We are grateful to Paul Martin and Barbara Luke for help with and use of the scanning electron microscope. We also thank Helen Durbin, Manfred Frasch, Ulrike Gaul, Kate Hooper, Herbert Jäckle, Henry Krause, Diethard Tautz and Michael Wilcox for antibodies, and Peter Gergen, Yash Hiromi, Herbert Jäckle, Gary Struhl and Diethard Tautz for fly stocks and DNA clones used in this study. We are very grateful to Karen Downs, Peter Gergen, Yash Hiromi, Howard Lipshitz, Suki Parks and Detlef Weigel for critical reading and comments on the manuscript. We thank our colleagues in the lab (Kate Hooper, Sheena Pinchin, Mark Wainwright, Michael Lardelli, Guy Riddihough and Ilan Davis), in the ICRF Developmental Biology Unit, at Caltech and in the fly groups at Princeton for suggestions and stimulating discussions during the course of this work. S.M.P. was supported by a postdoctoral fellowship from the Helen Hay Whitney Foundation.