buttonhead does not contribute to a combinatorial code proposed for Drosophila head development

The Drosophila embryo has provided an excellent system for analyzing the formation of the metameric body pattern (Nüsslein-Volhard and Wieschaus, 1980). Most studies have focused on the segmentation of the trunk. They have led to the definition of an elaborate cascade of gene interactions (reviewed by Ingham, 1988). Maternally deposited determinants specify the domains of gap gene expression along the anteroposterior axis (reviewed by St Johnston and NüssleinVolhard, 1992). Gap genes then activate pair rule genes in repetitive patterns (reviewed by Pankratz and Jäckle, 1993). These, in turn, define the metameric expression patterns of segment polarity genes (reviewed by Martinez-Arias, 1993). In addition, the gap genes together with the pair rule genes define the spatial domains of the homeotic selector genes which assign segment identity (reviewed by Akam, 1987; McGinnis and Krumlauf, 1992).

Head and trunk segmentation differ in several respects. The gap genes acting in the trunk region, such as hunchback (hb), Krüppel or knirps, are expressed in adjacent domains with relatively small overlaps (reviewed by Pankratz and Jäckle, 1993). The corresponding expression domains of the head gap-like segmentation genes orthodenticle (otd), empty spiracles (ems) and btd are widely overlapping (Dalton et al., 1989; Finkelstein and Perrimon, 1990; Walldorf and Gehring, 1992; Wimmer et al., 1993). Whereas the gap genes in the trunk region regulate each other’s expression (reviewed by Pankratz and Jäckle, 1993), no cross-regulation between the head gap genes has been observed (Cohen and Jürgens, 1990; Wimmer et al., 1995; Gao et al., 1996).

Phenotypic analyses of mutations in the head gap genes indicate that they are required for the establishment of anterior head segments (Cohen and Jürgens, 1990; Schmidt-Ott et al., 1994, 1995). These segments do not depend on the activity of pair rule genes (Macdonald et al., 1986; Cohen and Jürgens, 1990; Lardelli and Ish-Horowicz, 1993). Therefore, a mechanism different from the one described for the trunk has to be proposed to account for segmentation of the anterior head region (Cohen and Jürgens, 1990). The three head gap genes otd, ems and btd are required in overlapping domains with their posterior margins out of phase by one segment. The phasing of the deletions in the head gap gene mutants suggests that these genes are responsible for metamerization in the developing head (Cohen and Jürgens, 1990). Moreover, no known homeotic selector genes are active in the anterior embryonic head (reviewed by Cohen and Jürgens, 1991; McGinnis and Krumlauf, 1992; Jürgens and Hartenstein, 1993). Consequently, it has been proposed that the overlapping domains of head gap gene activity might also specify the segment identity of the metameres (Cohen and Jürgens, 1990).

This combinatorial model predicts that misexpression of these genes in the early blastoderm stage embryo will interfere with metamerization and alter the identity of head segments. To test this model, we generated an inducible system which places the gene of interest under the direct control of a region specific promoter. Since such constructs are expected to be dominant embryonic lethal, a ‘flp-out cassette’ (Struhl and Basler, 1993) was inserted to prevent the transgene from being expressed. Excision of the flp-out cassette, which occurs in the male germ line, renders the fusion transgene active. Using this system, we show that ectopic expression of btd in a broad anterior domain provides functional btd activity, but does not disturb head development. The spatial limits of btd expression are therefore not informative for metamerization of the anterior head region. Moreover, according to the combinatorial code, ectopic expression of btd anterior to its normal expression domain should transform the ocular into an antennal segment anlage. Since the ectopic expression of btd does not affect the ocular segment, we conclude that btd does not contribute to a combinatorial code for head segment specification.

The >y⁺> flp-out cassette was isolated from Act5C>y⁺>wg (pKB411; Struhl and Basler, 1993) as an NheI-fragment and cloned into pSL1180△RV-RI, a derivative of pSL1180 (Pharmacia) in which the polylinker region between the EcoRV and EcoRI sites had been removed. The yellow (y) gene was removed by a SalI cut and religa-tion to generate pSL>>, which was opened by a partial NotI cut and blunted by Klenow polymerase to insert the AUG-βgal (AB) gene from pCHAB△Sal (Wimmer et al., 1993, 1995) as a blunted 3.6 kb XbaI fragment. From the resulting construct pSL>AB>, the new 6.1 kb flp-out cassette was isolated by an SpeI and a partial NheI cut and cloned into the XbaI site of pCHAB△Xba and pCHABNot△Xba generating the basic constructs pC>AB>′ and pC>AB>, respectively (Fig 1). pCHAB△Xba and pCHABNot△Xba result from XbaI cuts and religa-tions of pCHAB△Sal and pCHABNot△Sal, which is derived from pCHAB△Sal by religating the cut and blunted NotI site. The 6.1 kb flp-out cassette was similarly cloned into the XbaI sites of pCbtdRV-2ndB△Xba and pChb(P2)△Xba (see below) to generate the constructs pCbtd>AB> and pChb(P2)>AB>, which provide expression in the btd head stripe domain or the anterior zygotic hb domain, respectively, and have left single restriction sites after the flp-out cassette for inserting different coding regions.

To generate transgenic fly lines, the constructs described below were injected together with the helper plasmid p△2-3 (Laski et al., 1986) into embryos of fly strain Df(1)w^67c23,y as reported by Rubin and Spradling (1982).

To generate pKSbtd covering all of the btd coding region, we fused in pBluescriptKS (Stratagene) the genomic 676 bp BamHI-PstI fragment, which contains upstream untranslated sequences, translation start site and encodes the first 209 amino acids (Wimmer et al., 1993), with a 1950 bp cDNA fragment, which was obtained by a partial PstI and complete NotI cut and encodes the rest of the btd protein (Wimmer et al., 1993). From pKSbtd we isolated the btd coding sequence by an EcoRV and SspI cut and cloned the resulting 2.2 kb fragment into the blunted NotI site of pSL>AB> (see above). From the resulting construct, pSL>AB>BTD, the 8.3 kb fragment containing the flp-out cassette and the btd coding region was isolated by an SpeI and a partial NheI cut and cloned into the XbaI site of pCbtdRV-2ndB△Xba generating pCbtd>AB>BTD. pCbtdRV-2ndB△Xba resulted from a XbaI cut and religation of the reporter gene construct btdRV-2ndB (Wimmer et al., 1993, 1995). pCbtd>AB>BTD was used for P-element transformation to generate the fly strains 850-43 and 850-89, which carry the interrupted btd transgene btd>AB>btd homozygous on the second chromosome.

The hb proximal promoter (Schröder et al., 1988) was cloned as a 0.7 kb SalI-XbaI fragment (SalI site blunted) into the XbaI and blunted EcoRI sites of pCHAB△Xba. Into the XbaI site of the resulting vector pChb(P2)△Xba, the 8.3 kb fragment containing the flp-out cassette and the btd coding region was inserted (derived from pSL>AB>BTD, see above) to generate pChb(P2)>AB>BTD. This construct led, after P-element transformation, to the isolation of five fly strains which carry the interrupted hb>AB>btd fusion transgene homozygous on the second (857-75, 857-79) or third chromosomes (857-35, 857-45, 857-83).

For identification of embryos carrying the induced fusion transgenes (Fig. 3), we crossed the β₂-tubulin-flp transgene (Struhl et al., 1993) into the background of several balancer chromosomes, which were marked by different lacZ reporter gene constructs (‘blue balancers’; Struhl et al., 1993). The following stocks carrying the β₂-tubulin-flp transgene homozygous on the X-chromosome were established:

DNA labelling and in situ hybridization have been performed as described by Hartmann and Jäckle (1995). DNA probes were prepared from otd (Finkelstein et al., 1990), ems (Dalton et al., 1989) and btd (Wimmer et al., 1993) cDNAs. Antibody stainings with rabbit anti-β-galactosidase (Cappel), mouse anti-en (Patel et al., 1989), mab22C10 (Fujita et al., 1982) and rat anti-otd antibodies (Wieschaus et al., 1992) of whole-mount embryos were carried out as described by Macdonald and Struhl (1986) using the Vectastain ABC Elite horseradish peroxidase system. For double stainings, alkaline phosphatase-conjugated goat anti-rabbit antibodies (Jackson ImmunoResearch Laboratories) were used to detect rabbit anti-β-galactosidase antibodies.

To examine the role played by the spatial domains of gap gene expression in embryonic head development, we developed a system to direct spatially restricted gene expression under the control of promoters active at early blastoderm. This system allows for immediate and targeted gene expression and will be referred to as the IT-system. Heat shock misexpression tends to cause phenocopies when applied in early embryos (Walter et al., 1990). The GAL4 system (Brand and Perrimon, 1993) cannot be used early in embryogenesis because of the lag in activation. The ideal system would be one in which a region-specific promoter is directly linked to the gene of interest. This arrangement, however, may interfere with normal development and cause dominant lethality. In order to avoid isolating only fly strains that carry low expressing or partially suppressed fusion transgenes, it is necessary to control transgene expression in an inducible manner (Parkhurst and Ish-Horowicz, 1991). To provide this function, we made use of the flp-out system developed by Struhl and Basler (1993) that allows the temporary separation of a region specific promoter from the coding region of a gene. We modified the flp-out cassette (Fig. 1) which is flanked by direct repeats of flp recombinase target sites (FRTs) and inserted it between promoter and coding region. Thus, induction of flp recombinase allows removal of the flp-out cassette and the desired fusion transgene is created.

To initially prevent expression of the coding region of interest, the flp-out cassette contains a transcriptional stop signal (hsp70) and a special chromatin structure (scs) element (Vasquez and Schedl, 1994) providing enhancer blocking activity (Fig. 1). To mark for the presence of the flp-out cassette, we inserted a lacZ gene (β-gal; Fig. 1). The lacZ gene serves also as a reporter gene that monitors the efficiency of the region-specific enhancer. This allows the selection of the transgenic fly lines for the ectopic expression experiments (Fig. 2F-H). To facilitate rapid cloning and germline transformation of different fusion genes, we generated the basic P-element transformation vector pC>AB> (Fig. 1). To efficiently remove the flp-out cassette and activate the fusion transgene, we used fly strains carrying the flp recombinase gene under the control of the β₂-tubulin promoter (Struhl et al., 1993). This promoter functions exclusively in maturing spermatocytes (Michiels et al., 1989). Male flies carrying both the ectopic expression construct with the flp-out cassette and the β₂-tubulin-flp gene will transfer to their progeny the induced fusion transgene (Fig. 3). The direct control of the region specific promoter then allows expression of the fusion transgene without delay.

The intact btd transgene has been shown to rescue the btd mutant head phenotype (Wimmer et al., 1996). To test the inducibility of the IT-system, we placed the flp-out cassette (>AB>) between the btd promoter and btd coding sequences (Wimmer et al., 1993, 1995). The resulting btd>AB>btd transgene mediates β-galactosidase (β-gal) expression in the btd head stripe domain (Fig. 2F), but it does not rescue the btd mutant phenotype (Fig. 4A,B). This demonstrates that the flp-out cassette renders the btd transgene inactive. When transmitted through the male germ line in the presence of the β₂-tubulin-flp transgene, the flp-out cassette is removed and a functional btd transgene is generated (Fig. 4C). Analyzing the proportion of rescued progeny indicates that excision of the cassette occurred in more than 80% of fertile sperm. The IT-system therefore presents a very efficient way for expressing a fusion transgene in an immediate and targeted manner.

The expression patterns of the head gap genes otd, ems and btd (Fig. 2A-E) are consistent with the idea that their spatial limits have a direct input into metamerization of the anterior head region (Cohen and Jürgens, 1990). This idea can be tested by expanding the expression domain of one of these genes, while keeping the others constant. To ask whether the spatial limits of btd expression are instructive for head metamerization, we expanded its expression domain by placing the btd coding sequence under control of the well defined proximal hb promoter (Schröder et al., 1988; Driever et al., 1989; Struhl et al., 1989). We selected several transgenic fly lines, which all carry the hb>AB>btd transgene, but show different levels of lacZ expression in the anterior half of the blastoderm embryo (Fig. 2G,H). Excision of the flp-out cassette allowed btd expression anteriorly and posteriorly to its normal expression domain (Fig. 2I). The ectopic btd expression is detectable until the late blastoderm stage. There are no indications that the expression mediated by the hb promoter is suppressed prior to its normal fading, as had been observed in an ectopic expression study of pair-rule genes (Parkhurst and Ish-Horowicz, 1991). Embryos carrying the induced hb>btd fusion transgene were identified by the absence of β-gal expression (Fig. 3).

Comparing otd expression in wildtype embryos and in embryos with the ectopically expressed btd gene (Fig. 5A,B) confirms the previous observation that head gap genes do not regulate each other’s expression (Cohen and Jürgens, 1990; Wimmer et al., 1995; Gao et al., 1996). The ectopic expression of btd therefore provides an experimental situation where the expression domain of one head gap gene is altered, and the domain of another is kept in its normal spatial limits. To monitor head metamerization in this situation, we used the expression pattern of the segment polarity gene engrailed (en). en expression provides a marker for identifying the number and type of head segments (Fig. 5C; Schmidt-Ott et al., 1994). Expanding the expression domain of btd does not alter the number or shape of the en stripes (Fig. 5D). This indicates that the spatial limits of btd expression are not instructive for metamerization of the head region.

The hb>btd fusion transgene expresses btd at lower levels than the endogenous gene (Fig. 2I). To rule out that this expression is below the threshold level necessary for btd function, we asked whether it provides sufficient btd activity to rescue a btd null mutation. Using a marked, btd mutant chromosome (Wimmer et al., 1996), we could show that the hb>btd fusion transgene rescues the embryonic btd head phenotype (Fig. 4D). Moreover, the fusion transgene rescued males carrying the hemizygous lethal alleles btd^XA and btd^XG (Wimmer et al., 1993) to adulthood like a wild-type btd transgene (Wimmer et al., 1996). Independent transgenic fly lines which mediate different levels of ectopic transgene expression (Fig. 2G,H) are functionally equivalent. This clearly indicates that the hb>btd fusion transgene provides sufficient btd activity in the anterior half of the blastoderm embryo to support normal head development.

btd has been proposed to have a function in head segment specification (Cohen and Jürgens, 1990). otd, ems and btd are each required for the formation of the antennal segment (Fig. 6A; Cohen and Jürgens, 1990). At early blastoderm, btd is normally not expressed in more anterior segments (Fig. 2C,E). Thus, the combination of otd and ems, without btd, has been thought to give rise to the ocular segment (Fig. 6A; Cohen and Jürgens, 1990). According to this model, ectopic expression of btd in the anlage of the ocular segment should change its fate to an antennal one. This should result in the loss of the ocular segment and either in the formation of two adjacent antennal segments or in the fusion of the two segments forming one expanded antennal segment (Fig. 6B).

To examine the effect of ectopic btd expression on segment specification, we analyzed head structures (Jürgens et al., 1986) in cuticle preparations of btd mutant embryos rescued by the hb>btd fusion transgene (see above). The normal array of the btd dependent structures in the antennal, intercalary and mandibular segments was found (Fig. 4D). Furthermore, those structures which derive from the ocular segment anlage (Schmidt-Ott et al., 1994) and the posterior maxillary segment anlage (Gonzalez-Reyes and Morata, 1991) were normal (Fig. 4D).

Since the ocular segment anlage also gives rise to structures not scoreable in cuticle preparations, we monitored internal sensory structures by the neuronal specific marker mab22C10 (Fujita et al., 1982). btd mutants lack a number of head sensory organs including the dorsal organ (do) and the lateropharygeal organ (lpo) (Fig. 7A,B; Schmidt-Ott et al., 1994). In addition, btd mutant embryos show an abdominal peripheral nervous system (PNS) phenotype (Wimmer et al., 1996), which is not rescued by the hb>btd fusion transgene. The abdominal btd PNS phenotype was used as an internal marker to identify btd mutant embryos. The presence of the rescued do and lpo indicates that the hb>btd fusion transgene was fully functional. In such embryos, we analyzed the sensory organs derived from segments adjacent to the btd domain: the Bolwig organ (bo) for the ocular segment and the terminal organ (to) for the posterior part of the maxillary segment. Both organs could be identified and their innervation pattern was normal (Fig. 7C). Thus, the ocular and maxillary segments are specified correctly when btd is ectopically expressed in the anterior half of btd mutant embryos. This indicates that ectopic btd expression covering the anlagen of head and thorax segments does not interfere with specification of segment identity. Fusion transgenes expressing different levels of btd have identical properties (not shown). This observation argues that btd does not contribute to a combinatorial code responsible for specifying head segment identity.

We have established an inducible system for region-specific and immediate ectopic gene expression. We developed the IT-system because the most commonly used ectopic expression systems are not suitable for gene expression in the early blas-toderm embryo. The heat shock method (Struhl, 1985) bears several disadvantages: first, the expression cannot be regionally restricted; second, the level of expression varies, is difficult to control and measure, and might exceed physiological levels; third, the heat shock treatment itself can cause phenocopies, especially when applied during the syncitial blastoderm stage, when the gap genes are active (Walter et al., 1990). The GAL4 system was developed to overcome the potential problems of ectopic gene expression causing dominant phenotypes (Brand and Perrimon, 1993). However, it has the disadvantage that it requires two rounds of transcription and translation before the gene of interest is ectopically expressed. This delay in expression excludes the GAL4 system for studying the ectopic expression of gap genes, which represent the first zygotically active genes with a relatively short phenocritical period (Rothe et al., 1992). In contrast, the system presented here allows the coding sequence of interest to be expressed under the direct control of a region-specific promoter, which is temporarily separated by a flp-out cassette. After flp-out, immediate ectopic expression can occur. We show that the IT-system enables the study of gene functions active very early in development. Moreover, it represents a general tool that controls both position and timing of gene expression, which may be critical for studying other developmental processes such as neurogenesis and organogenesis.

We used the IT-system to test the proposed role of the gene btd in head development. Neither metamerization of the head, nor segment specification are affected when btd is expressed in regions outside its normal expression domain in blasto-derm embryos. This finding indicates that the spatial limits of btd expression are not instructive for head development. The factor encoded by btd might be a ‘generic’ transcriptional activator, like its vertebrate homologues Sp1 and Sp4 (Wimmer et al., 1993; Hagen et al., 1995; Supp et al., 1996). As an activator, btd is probably necessary for the expression of several target genes, like cap’n’collar (Mohler, 1993) or collier (Crozatier et al., 1996), whose expression patterns in the head are even more regionally restricted than the btd domain. The expression of these putative btd target genes must therefore be further restricted within the btd domain by other factors. This might explain why ectopic expression of btd outside its normal expression domain has no consequence. The btd target genes would still be restricted by adjacently acting repressors overriding btd-dependent activation. This, however, raises the question, why btd is normally expressed in a regionally restricted pattern that covers exactly the anlagen of the segments affected in btd mutants (Cohen and Jürgens, 1990; Wimmer et al., 1993, 1996).

The hb>btd fusion transgene does not have any detrimental effects on head development. It shows, however, that btd expression in the early hb domain is sufficient to rescue btd mutants to adulthood. Since the fusion transgene does not restore btd function in the abdominal PNS (Fig. 7C), the reduced numbers of scolopidia in the chordotonal organs of btd mutants (Wimmer et al., 1996) must still be sufficient for larval survival. Moreover, the proximal hb promoter confers btd expression only in the blastoderm embryo and does not persist through grastulation. Thus, except for its role in the PNS, btd expression seems to be only required in the earliest of its expression domains. This probably reflects the proposed redundant function of D-Sp1, another Sp1 homologue in Drosophila. D-Sp1 probably substitutes for the lack of btd activity in all the other postblastodermal expression domains of btd (Wimmer et al., 1996).

The existence of two different mechanisms that underly head and trunk segmentation, might reflect the evolutionary history of insects. In primitive insects, embryogenesis proceeds in two phases. The head segments are established first, then trunk segments are subsequently added (Sander, 1976). In contrast, segmentation of head and trunk occurs simultanously in Drosophila which may have acquired a special mode of trunk segmentation in order to support the very rapid mode of embryogenesis. The head segmentation mechanisms are likely to be conserved in all insects and even other arthropods (Cohen and Jürgens, 1991), while trunk segmentation appears to have diverged much more extensively (Patel, 1994). Moreover, vertebrate homologs of Drosophila head segmentation genes are expressed during brain development, which suggests an evolutionary conservation of their functions across the animal kingdom (Bally-Cuif and Boncinelli, 1997; Li et al., 1996; Wimmer et al., 1996; and references therein). Our results show that the spatial limits of btd expression are not instructive for head metamerization and raise the question again, of how segmentation in the head region is established. In the anterior head anlage, the blastodermal expression domains of the segment polarity genes wingless and hedgehog each depend specifically on only one head gap gene (Mohler, 1995). Initially these expression domains overlap widely. Due to the mutual exclusiveness of these two genes, interference might generate their segmentally iterated expression pattern and metamerize the anterior head region (Mohler, 1995).

The results reported here argue that btd activity does not con-tribute to the proposed combinatorial code for specification of head segment identity. It might be possible that head segments are not specified by a combinatorial code, but rather by ‘phe-notypic suppression’. This phenomenon which is based on a hierarchy of gene functions had been noted in the context of Hox gene activities in the trunk (González-Reyes and Morata, 1990). Phenotypic suppression, also referred to as ‘posterior prevalence’, could explain why ectopic expression of posterior Hox genes overrides the function of more anteriorly expressed Hox genes in both flies and vertebrates (Duboule and Morata, 1994). The molecular explanation for phenotypic suppression proposes competition and differential affinities of the involved homeodomain proteins for similar binding sites (González-Reyes et al., 1990). A related phenomenon in opposite direction (‘anterior prevalence’), could explain why the ectopic expression of btd in the anterior head regin is functionally overridden and therefore of no consequence. However, btd encodes a zinc finger transcription factor and otd and ems code for homeodomain proteins with different DNA binding specificities (Treisman et al., 1992; Wimmer et al., 1993). Thus, if phenotypic suppression exists in the head, it must be mediated by different molecular mechanisms than proposed for the trunk. The fact that btd is not part of a combinatorial code, does not necessarily exclude a combinatorial model for specifying different head regions. For the gnathal segments, it has been suggested that genes downstream of gap genes control segment identity in a combinatorial manner (Mohler et al., 1995). It is still possible that the homeobox-containing head gap genes otd and ems (Dalton et al, 1989; Finkelstein et al., 1990; Walldorf and Gehring, 1992) provide such a function directly without involving downstream genes. This would imply, however, that after having participated in head metamerization they serve this second function by contributing to a combinatorial code in conjunction with other head genes such as, for example, sloppy paired (Grossniklaus et al., 1994). Our findings clearly show that the zinc finger protein encoded by btd (Wimmer et al., 1993) does not contribute to a combinatorial code that specifies head segments. Thus, btd might act more like the canonical type of gap genes in the trunk, which have no direct role in specifying segment identities (Pankratz and Jäckle, 1993). Therefore, the results presented here contradict the combinatorial model which proposes that the gap-like genes are required in a simultaneous process resulting in head metamerization and specification of head segments (Cohen and Jürgens, 1990).

We thank K. Basler for the original flp-out cassette, M. Hoch for the hb promoter, S. Small for pointing out and G. Struhl for providing the β₂-tubulin-flp flies, R. Finkelstein for otd antibodies, D. Schmucker for mab22C10 antibodies, A. Goriely for en antibodies, S. Sciani-manico for help in generating transgenic fly lines, C. Hartmann and V. Morel for help with stainings, the members of the Jäckle and Desplan laboratories for stimulating discussions and especially T. Turner for all her support. This work was supported by the EMBL (S. M. C.), the Max-Planck Society (H. J.), the Howard Hughes Medical Institute (C. D.), and the Human Frontiers Science Program Organization (H. J., C. D.). E. A. W. was the recipient of an EMBO fellowship.