Integration of the head and trunk segmentation systems controls cephalic furrow formation in Drosophila

During Drosophila gastrulation, the embryo is transformed from a regular array of morphologically identical cells, organized in a single layer, into a complex array of cell groups. These groups change cell shape and move together in a characteristic, predetermined manner (review by Costa et al., 1993). Two invaginations, the ventral and cephalic furrows, represent the first morphological manifestations of cell fate and differentiation programs. The ventral furrow (VF) invaginates along most of the ventral midline, bringing the mesoderm and possibly part of the anterior endoderm primordium into the interior of the embryo. Cephalic furrow (CF) invagination takes place laterally on both sides of the embryo, near its anterior end. Unlike VF formation, the CF is only transient. At the completion of germ-band extension, all cells of the CF slowly unfold back onto the surface of the embryo to contribute to the ectoderm (description in Campos-Ortega and Hartenstein, 1985; Foe, 1989; Costa et al., 1993).

Although the CF is a prominent morphological feature of the early gastrula, its developmental role remains enigmatic. No mutations have been isolated that affect only the CF and the mutations known to affect the VF and posterior midgut (PMG) invaginations leave the CF unaffected (Brönner et al., 1994; Costa et al., 1994; Reuter and Leptin, 1994). The cellular and genetic mechanisms that control its formation are also unknown. The absence of a CF in embryos derived from mothers mutant for bicoid (Frohnhöfer and Nüsslein-Volhard, 1986) and the reproducible shifts in its position and/or lateral extent observed in anteroposterior and dorsoventral pattern mutants (Zusman and Wieschaus, 1985; Driever and Nüsslein-Volhard, 1988; Struhl et al., 1989) indicates that the cell shape changes respond directly to positional information. However, these observations did not provide a specific clue to how that positional information is translated into specific changes in cellular morphology.

The CF arises in an interesting position in the embryo, at the juxtaposition of the patterning systems that define the head and trunk segments. These two systems involve separate groups of zygotically active genes and separate regulatory interactions (Ingham, 1988; Cohen and Jürgens, 1990; Jürgens and Hartenstein, 1993). Formation of the CF coincides with the anteriormost stripe of expression of the pair-rule gene even-skipped (eve), eve stripe 1; eve encodes a homeodomain protein that is essential for segmentation of the Drosophila embryo (Harding et al., 1986; McDonald et al., 1986). Just prior to cellularisation, the Eve protein is distributed in a series of 7 pair-rule stripes which foreshadow the position of odd-numbered parasegments PS 1 to 13 (Frasch and Levine 1987; Lawrence et al., 1987). The function of Eve stripe 1 in this process is unclear. Whereas Eve expression in stripes 2-7 has an instructive role in specifying expression patterns of the segmentpolarity genes engrailed (en) and wingless (wg), en expression in PS1 occurs normally in eve mutant embryos (McDonald et al., 1986; Ingham et al., 1988). An additional aspect that distinguishes Eve stripe 1 from other Eve stripes is the failure to identify cis-regulatory elements specific for that stripe, despite extensive studies on the mechanisms of regulation of eve expression (Goto et al., 1989; Harding et al., 1989).

In the experiments described below, we show that the CF is eliminated or abnormal in eve mutant embryos (see also Costa et al., 1993), suggesting a specific role of Eve stripe 1 in controlling this morphogenetic event. We further show that CF formation also depends upon activity of the head gap-like segmentation gene buttonhead (btd), which is required for formation of the antennal, intercalary and mandibular segments and part of the maxillary segment (Cohen and Jürgens, 1990; Wimmer et al., 1993, 1996). The transverse row of cells, which change shape and slip into the interior of the embryo during the first phase of CF invagination, express both btd and eve at the onset of gastrulation and eve expression in these cells is dependent upon btd. Our results suggest that the interaction between btd and eve integrates two systems of segmentation in controlling formation of the CF and patterning the border region that demarcates the head and the trunk.

The btd^XG, svb and the hb>btd transgenic lines were provided by Ernst Wimmer, the Rockefeller University, New York and described in Wimmer et al. (1997). The btd¹, btd^XG btd^XA, eve^R13and eve^IIR59 mutant strains were obtained from the Tübingen Stock Center and the stocks carrying chromosomal translocations from the Bowling Green Stock Center (Bowling Green, OH). Blastoderm-stage embryos defective for CF formation were selected by observation of living embryos under Voltalef 3S halocarbon oil, using a stereomicroscope. The segregation of the translocation chromosomes in males was tested by mating them to C(1)DXywf; 1E females. The segregants can be easily identified by the presence or absence of the folded gastrulation phenotype (for reference, see Wieschaus and Sweeton, 1988; Müller and Wieschaus, 1996). Time-lapse recording of optical sections of individual embryos at higher magnification (×200 or ×400) was used to confirm the absence of CF formation in eve, btdand btdmutant embryos.

Whole-mount in situ hybridisation to embryos was performed with digoxigenin-labeled RNA prepared with the Genius kit from Boehringer Mannheim, using the procedure from Tautz and Pfeifle, 1989, except for the absence of proteinase K treatment. For Arm and Eve immunostaining, embryos were fixed using the heat-methanol and paraformaldehyde procedures, respectively (see Müller and Wieschaus, 1996). Staining of F-actin and Eve was performed on cellular blastoderm and early gastrula embryos devitellinized by hand-peeling according to Wieschaus and Sweeton (1988). The purified rabbit anti-Eve antibody (1/2000) was from Manfred Frasch (Mount Sinaï University, New York). The monoclonal anti-Sxl antibody was provided by Paul Schedl (Princeton University) and used 1/50. Goat anti-rabbit and anti-mouse biotin secondary antibodies (Vector Laboratories) were used 1/500. For immunofluorescence and rhodamine-phalloidin staining, embryos were mounted in Mowiol containing DABCO (1,4-diazabicyclo[2.2.2]octane) as an anti-bleaching agent and observed under an MRC 600 confocal microscope (BioRad Labs). Images were merged using Confocal Assistant (CAS) and processed with Adobe Photoshop TM software.

Transgenic lines expressing a nuclear Green Fluorescent fusion Protein (nlsGFP) were provided by Ilan Davis and P. O’Farrell (UCSF, San Francisco). The nlsGFP 34H; 34M strain (Davis et al., 1995) was used to establish the eve^R13; nlsGFP34M and btd^XGnlsGFP 34H; 34M strains. nlsGFP expression was observed on living embryos selected at the blastoderm stage and mounted in halocarbon oil on a permeable membrane with a glass coverslip. Images were collected every 3 minutes with the confocal microscope. Each image was obtained by merging four independent scans corresponding to sections separated by 2 μm (6 μm total depth).

Previous analyses of time-lapse cinematography and sectioned material suggested a sequence of cell shape changes and nuclear migrations underlying CF formation (Turner and Mahowald, 1977 and E. Wieschaus, unpublished data). First, cells in a single transverse row, designated henceforth as initiator cells, shift their nuclei basally and slip into the interior, forming a barely visible cleft in the embryo’s surface. This movement seems to be associated with cell shortening along the apical-basal axis, as illustrated on confocal images of fixed embryos stained with fluorescent phalloidin to highlight cell shape (Fig. 1A). Subsequently, the apices of the cells immediately anterior and posterior to this initial cleft roll over its edge and follow the first row of cells into the interior (Fig. 1B,C). While a thorough description of CF formation will be reported elsewhere, we have focused our attention here on the antero posterior position of the CF relative to the segmentation process by double staining embryos with phalloidin and antibodies against the Eve protein. At the end of cellularisation when gastrulation is initiated, each Eve stripe is about three cells wide with a sharply defined anterior border. The cells that first undergo the change in shape that initialises CF formation were consistently found to be the second row of cells in stripe 1. During early stages of Eve expression, these cells express the highest levels of Eve in stripe 1. The position of the CF relative to Eve expression was maintained when the CF was shifted anteriorwards by gradually decreasing the amount of Bicoid gene product, using mutations either in bicoid (bcd) (Driever et al., 1988; Struhl et al., 1989) or serendipity delta, a bcd transcriptional activator (Payre et al., 1994; data not shown). In no instance did we observe formation of a cephalic fold in embryos when Eve stripe 1 was missing due to altered maternal genotypes.

The expression of Eve relative to the position of CF invagination raised the possibility that eve is specifically required for this morphogenetic event. CF defects were indeed recently noticed in eve mutant embryos (Costa et al., 1993). We therefore re-investigated the eve gastrulation phenotype in more detail by examining live embryos during gastrulation. Roughly one quarter of embryos from an eve^R13 mutant stock (an amorphic eve allele, Nüsslein-Volhard et al., 1984) showed no CF formation at the expected time, i.e., concomitant with mesodermal invagination. Cuticle preparations of the CF-defective embryos hand-selected at stage 6 confirmed they were the eve mutant embryos. Fig. 2A and B show wild-type and eve embryos at the early gastrula stage (transition between stages 6 and 7), stained for Armadillo to highlight cell shape changes. armadillo encodes a Drosophila β-catenin homolog, which is a component of the adherens junction complex and accumulates at the sites of cell-cell contact (Peifer, 1993). At that stage, the cephalic furrow is clearly visible in wild-type embryos (Fig. 2A). In homozygous eve mutant embryos, there is no indication of CF formation while the other major morphogenetic movements, the VF and proctodeal invagination, have progressed normally (Fig. 2B). Time-lapse recording of live eve embryos (sagittal sections) between the beginning of cycle 14 and stage 7 did not show any signs of initiator cell shortening (data not shown).

In order to identify other genes involved in CF formation, we chose to search for CF defects in living embryos deficient for overlapping chromosomal regions. We started with the X-chromosome, using attached X-chromosomal females and males carrying XY translocations, a strategy previously successful in identifying cellularisation and cell adhesion genes (Wieschaus and Sweeton, 1988; Müller and Wieschaus, 1996). A single X-chromosomal region, 8F-9C, defined by the translocations T(1;Y) B115 and T(1;Y) B152, was identified as being absolutely required for initiation of CF formation. Among the genes already known to map in this region, an immediate candidate for a ‘CF gene’ was buttonhead (btd, cytological position 9A), a head gap gene that encodes a zinc finger transcription activator and likely a direct Bcd regulatory target (Wimmer et al., 1993, 1995). Mutations in btd lead to an incomplete head involution (Wieschaus et al., 1984) with the absence of the antennal, intercalary, mandibular and part of the maxillary segments (Cohen and Jürgens, 1990; Wimmer et al., 1993, 1996). Although no defect in CF formation was previously ascribed to mutations in btd, we decided to re-investigate this possibility by examining mutant embryos carrying either an amorphic or a strong hypo-morphic allele of btd, btd^XA and btd^XG, respectively (Wimmer et al., 1993). In both cases, examination of live and fixed btd embryos revealed the absence of CF (Fig. 2C). During early gastrulation, stages 6 and 7, the eve and btd mutant phenotypes looked morphologically identical (compare Fig. 2B and C), i.e., they displayed a lack of initiator cell activity.

Although no initiator cell behavior is observed during early gastrulation in eve embryos, an irregular anterior fold was frequently observed in later stage embryos, at a variable position within mitotic domain 2 (Fig. 3, see Foe, 1989 for description of mitotic domains). Although these folds arise in the region of the embryo that would normally form the CF, they do not appear to result from the normal CF program. Fig. 3 shows a time-lapse analysis of gastrulation in embryos expressing a nuclear form of the Green Fluorescent Protein (nlsGFP) (Davis et al., 1995). Whereas, in wild-type embryos, CF formation preceded mitosis 14 (Fig. 3A, see Foe et al., 1989), in eve embryos, the irregular anterior fold that forms arose only after cells in MD2 had entered mitosis (Fig. 3B,C). This fold eventually resolved into a complete transversal furrow (Fig. 3D). This late, deep fold observed in eve embryos was never seen in btd embryos. In btd embryos, cells in MD2 do not undergo mitosis, because expression of string, the Drosophila cdc25 homolog that drives the G₂/M transition during embryonic cycle 14, is dependent upon btd activity in this domain (Edgar et al., 1994). On the contrary, MD2 is expanded posteriorly in eve mutant embryos, to encompass more cells than in wild type (Fig. 3D). Thus, it is possible that the difference in late furrow formation between eve and btd embryos is due to their differential control of patterned mitosis. Finally, it should be noted that neither btd nor eve hypomorphic alleles (btd¹ and eve^IIR59, respectively) affect the cephalic fold (not shown), indicating that this morphogenetic movement is a low sensitivity phenotype contrary to the case in either head involution or trunk segmentation. This low sensitivity might explain why CF defects, which we only observed in amorphic btd or eve mutant embryos, have not been reported previously.

The observation that both btd and eve are required for CF formation suggested that the two were expressed in the CF initiator cells. In situ hybridization to eve and btd transcripts, or hybridisation to btd RNA and staining with Eve antibodies on embryos at early cycle 14, indicated an overlap of btd expression and eve stripe1 in one or two cells (Fig. 4 and data not shown). This overlap was clearly seen at high magnification (Fig. 4C), because the btd and eve transcripts display different subcellular localisations, basal and apical, respectively (Fig. 4A,B). The CF initiator cells, the cells at the center of Eve stripe 1 (see also Fig. 1A), are the posterior-most btd-expressing cells at that stage. Together with the CF phenotype of btd and eve mutant embryos, the btd/eve overlap in expression suggested at least two possible alternatives. Either Btd and Eve proteins act together in regulating the expression of a downstream target gene(s), or one may regulate the other, the two possibilities not being mutually exclusive. To test the two possibilities, we examined expression of eve and btd in mutant backgrounds. Consistent with the gap-like gene properties of btd, no change of its expression was seen in eve⁻ embryos. In contrast, Eve stripe 1 was found to be dependent upon btd activity. In btd embryos at the early cellularisation stage, Eve stripe 1 was reduced to a single line of cells expressing low levels of Eve (Fig. 5A,B). The position of these cells relative to the stripe of btd expression or the anterior border of Eve stripe 2 indicated that they correspond to the posterior edge of stripe 1 (Fig. 5C,D). During early gastrulation (stage 6), Eve expression in wild-type embryos becomes restricted to a 2-cell-wide stripe laterally and about a 3-cell-wide stripe ventrally. In btd embryos, residual Eve expression was detected only in a few scattered cells mainly located on the ventral side of the embryo (Fig. 5D, see also Fig. 6D). Loss of the last row of Eve-expressing cells in the gastrula btd embryo presumably results from the normal sharpening of the Eve stripes by other pair-rule genes that occurs at this stage. Identical results were obtained by in situ hybridisation to eve RNA in place of Eve antibodies (not shown), indicating that btd is required, directly or indirectly, for eve transcription in parasegment 1 during cycle 14.

A conditional hunchback>buttonhead (hb>lacZ>btd) transgene in which the btd-coding region is separated from the hunchback proximal promoter region by a flp-out cassette containing a lacZ gene has recently been constructed (Wimmer et al., 1997). When the flp-out cassette is removed during spermatogenesis, the hb>btd transgene allows btd expression in the anterior half of the embryo, both anteriorly and posteriorly to its normal expression domain (compare Fig. 6A and 4A). hb>btd expression completely overlaps the position of Eve stripe 1 and most of the stripe 1-stripe 2 interstripe. Expression of this transgene rescues the btd mutant head phenotype to wild-type (Wimmer et al., 1997). We tested whether it also was able to restore normal CF formation by examining live embryos at the early gastrulation stage. In the control btd^XG81, svb/FM7 inter se cross, roughly 25% of the embryos (over 200 embryos individually examined) lacked a cephalic furrow and showed a typical btd, svb double mutant phenotype in cuticular preparations. When btd, svb females were mated to males carrying both the hb>lacZ>btd and b2-tubulin>FLP transgenes, the fraction of progeny lacking a cephalic furrow lowered to around 15% (48 embryos out of a total 333 individual embryos examined). This result, consistent with an excision of the lacZ cassette in more than 80% of fertile sperm (Wimmer et al., 1997), indicated that the hb>btd transgene is able to rescue formation of the CF. This was confirmed by examining the larval cuticles of CF⁺ embryos hand-selected at the early gastrulation stage. The expected fraction showed both normal head skeleton structures and a svb phenotype, indicating that they were rescued btd, svb mutant embryos (data not shown). Finally, we looked at Eve stripe 1 in this class of embryos. The different embryonic phenotypes were distinguished by either triple immunostaining for Eve, Sxl and β-Gal or in situ with a lacZ probe followed by immunostaining for Sxl and Eve (Fig. 6 B-D and data not shown). In btd, hb>btd embryos, the anteroposterior position and width of Eve stripe 1 and its placement in relation to CF position were restored to normal (Fig. 6B,C). This suggests that, while eve stripe 1 is dependent upon btd activity, its position involves repressor elements acting independently of btd.

Genetic and molecular analyses of the biological process of patterning during Drosophila development have defined elaborate cascades of gene interactions that sequentially subdivide the embryo into an array of different types of tissues and specialised segments (review by St Johnson and Nüsslein-Volhard, 1992). Among the earliest manifestations of this patterning are the movements of gastrulation, which involve successive changes in the shape of epithelial cells at defined positions (Costa et al., 1993; Leptin, 1994). The initial step in CF formation is a change in shape and apical positioning of a single row of cells. The position of these initiator cells may be defined by the overlapping expression of the segmentation genes btd and eve, both of which are required for the initiator cells to form. The control of Eve expression by btd in these cells reveals a new level of integrated regulation, which interfaces the head and trunk segmentation systems. We propose that the CF in Drosophila transiently separates two morphogenetic fields and possibly reflects the existence of an early organiser region analogous to the mid-hindbrain border in vertebrates (see Bally-Cuif and Wassef, 1995 and Joyner, 1996, for reviews).

The best known process of Drosophila gastrulation is the formation of the VF, which involves the apical flattening of cells in the ventral plate, a domain about 18 cells wide, followed by the apical constriction of cells in the central region. This ultimately results in the formation of a shallow groove or furrow along the ventral midline (Sweeton et al., 1991). A different sequence of cell shape changes occurs during CF formation. First, a single row of lateral cells shorten along the apicobasal axis. We define these cells as the initiator cells. The presence of a single row of initiator cells distinguishes the invagination of CF from that of VF and PMG invagination, in which a sheet of cells undergo a coordinated cell shape change. Control of the timing and extent of mesodermal invagination by the transcription regulators twist, snail and huckebein, now well established, involves the expression of at least one gene, folded gastrulation, in a region prefiguring the VF (Costa et al., 1994; Reuter and Leptin, 1994). Neither the upstream regulators nor the downstream effectors of VF play a role in CF formation. Further understanding of CF formation at the cellular level will therefore require the identification of such putative effectors.

The Bicoid gradient specifies a unique positional identity to each cell along the anteroposterior axis by activating target genes, including btd and eve (stripes 1 and 2) in a concentration-dependent manner (Driever and Nüsslein-Volhard, 1988; Small et al., 1991; Wimmer et al., 1995). The CF is either missing or shifted anteriorwards in embryos laid by bcd mutant females. The same relative shift is observed for btd expression and eve stripe1. Together with the observation that both btd and eve are required for CF formation, it suggests that the trigger driving the shape change of initiator cells is the combinatorial action of Bcd, Eve and Btd in these cells. It was recently shown that early Eve stripes act as morphogenetic gradients at the single-cell level (Fujioka et al., 1995). Since the cell with the highest level of Eve expression at the early cellularisation stage corresponds to the second cell of Eve stripe 1, it is possible that a high level of Eve expression prefigures the initiator cell. However, this behavior cannot be purely quantitative as hypomorphic alleles of eve still form a CF even though the initiator cell would presumably have a lower level of Eve activity than the non-initiator cells in Eve stripe 1 of wild-type embryos. Thus, it may be that, within the stripe, relative levels of Eve expression, rather than absolute values, determine which cell displays initiator activity.

The finding that eve stripe 1 expression itself requires btd raises the possibility that absence of CF in btd mutant embryos is solely a consequence of the lack of Eve stripe 1. Since a CF forms only at the position of stripe 1, some additional factor must distinguish this stripe of eve expression from the other six. Although a direct role of Bcd concentration in the process of demarcating Eve stripe 1 from the rest cannot be formally excluded, we suspect that a combination of zygotically active genes may provide a more precise designation of CF cells. One candidate for this function might be Btd itself. However, a hb>btd chimaeric gene can rescue the embryonic btd phenotype (Wimmer et al., 1997) despite a domain of btd expression shifted both anteriorly and posteriorly, therefore completely overlapping Eve stripe 1 and possibly extending into Eve stripe 2. In rescued embryos, Eve stripe 1 expression and CF formation are at a normal position. These results indicate that the position and size of Eve stripe 1 is determined by repressor elements acting downstream of Bcd and overriding btd-dependent activation. It strengthens the conclusion that btd might be a ‘generic transcriptional activator’ required for transcriptional activation of specific target genes, such as eve and collier (col) (this report and Crozatier et al., 1996) but whose limits of expression are not instructive for head development (Wimmer et al., 1997).

Previous promoter fusion studies have established that at least some of the Eve stripes are regulated by separate enhancers present in the eve promoter (Goto et al., 1989; Harding et al., 1989), but the Eve stripe 1 enhancer has not been characterised. Our results suggest that activation of Eve stripe 1 is similar to that of Eve stripe 2 (Small et al., 1992; Arnosti et al., 1996), with bcd, possibly hb, and btd representing ‘generic’ activators working synergistically. A striking difference, however, is the mosaic nature of the stripe 1 enhancer since the posterior-most row of Eve stripe 1 cells is still present in btd mutant embryos. This correlates with the observations that btd is not expressed at significant levels in this row of cells and that part of the maxillary anterior compartment (posterior PS1) remains in btd mutants (Gonzales-Reyes and Morata, 1991; Wimmer et al., 1996).

Recent genetic and molecular data from Drosophila support the view that the process of segmentation of the head is fundamentally different from that of the trunk (Cohen and Jürgens, 1990; Finkelstein and Perrimon, 1991; Wimmer et al., 1997). The questions of the interface between the two segmentation systems and the potential role of the CF have remained, however, largely unaddressed. These are important issues since the cephalic furrow marks the boundary between the procephalon and the segmented germ band (Gonzales-Reyes and Morata, 1991), although it does not correspond to any boundary within the later head. The control of eve by btd in anterior cells of PS1 is the first indication of an hierarchical relation between head gap genes and trunk pair-rule genes. Interestingly, Engrailed expression in PS1 requires btd but does not require eve, contrary to the situation in parasegments 2 to 14 (McDonald et al., 1986; Cohen and Jürgens, 1990). Despite having no effect upon En expression in PS1, eve may have a role in patterning this parasegment, as its expression and placement in relation to En is conserved between long germband and short germ-band insects (Patel et al., 1994). Moreover, recent data on the activation of col, a novel transcription factor, in PS0/MD2 (Crozatier et al., 1996), suggest a possible mechanism by which btd and eve cooperate to pattern PS1. Activation of col requires btd. Conversely, in the absence of eve, col expression is expanded posteriorly to overlap a region roughly corresponding to PS1, indicating that Eve acts as a repressor of col in this parasegment (Crozatier, M. and Vincent, A. unpublished). Likewise, expression of string in MD2, which also requires btd (Edgar et al., 1994), is expanded posteriorly in eve mutant embryos (Fig. 3 and data not shown). Our current working model is that the activation of eve by btd in anterior PS1 cells allows for differential gene expression between PS0 and PS1 (Fig. 7). In addition to the control of CF formation, the btd/eve interaction may thus assign separate gene expression and mitotic programs to cells on either side of the pro-cephalon/posterior head border. It remains to be understood what possible function the CF plays in this demarcation process. In embryos laid by bcd females, the anterior part of the germ band on the ventral side extends further anteriorly, suggesting that the CF may serve to anchor anterior (ventral) cells such that the germ band extends only posteriorly (Costa et al., 1993). The CF may therefore be viewed as transiently separating two morphogenetic fields. Whether the foremost expression of Engrailed in PS1 in Drosophila and at the midbrain-hindbrain boundary in vertebrates reflects the existence of an ancestral organiser region, of which the CF might constitute a specialised evocation, is an interesting question for future consideration.

We are grateful to Michèle Crozatier and Ernst Wimmer for communication of unpublished results and Manfred Frash, Ilan Davis, Paul Schedl and Ernst Wimmer for antibodies and transgenic lines. We would like to thank Arno Müller and Laure Bally-Cuif for helpful discussions, Julian Smith for comments on the manuscript and Joseph Goodhouse for training and assistance with confocal microscopy. This study was supported by the National Institutes of Health (Grant 5RO1HD22780), the CNRS, the Ministère de la Recherche et de l’Enseignement Supérieur, and the Human Science Frontier Organisation.