Control of germ-band retraction in Drosophila by the zinc-finger protein HINDSIGHT

In metazoa morphogenetic cell shape changes and movements are important in the formation of complex three-dimensional body structures (Bard, 1994). The genetic approach — particularly in Drosophila — is beginning to yield important insights into the mechanisms that control and coordinate these morphogenetic processes. Such genetic strategies have previously illuminated the mechanisms by which cells are assigned positional values and are fated to contribute to particular tissue types (Lewis, 1996; Nüsslein-Volhard, 1996; Wieschaus, 1996). During Drosophila embryogenesis, several coordinated morphogenetic processes occur shortly after blastoderm cellularization is completed (Campos-Ortega and Hartenstein, 1985; Costa et al., 1993; Martinez Arias, 1993). First, mesoderm is internalized through the formation of the ventral furrow. The endoderm then invaginates from each end of the embryo as the anterior and posterior midgut. Concurrent with posterior midgut invagination, germ-band extension drives cells around the posterior tip of the embryo, converting it into a ‘U-shape’ folded upon itself dorsally. Several hours later, the germ band retracts back around the posterior tip, repositioning the caudalmost body parts at the posterior end of the embryo.

The cellular events and the genetic control of several of these morphogenetic processes are now under study. For example, it has been shown that mesodermal and posterior midgut invaginations occur in two phases (Kam et al., 1991; Sweeton et al., 1991). The first is slow and stochastic, with isolated cells within the presumptive mesoderm and endoderm initiating apical constriction. The second phase is rapid with all of the unconstricted cells in these domains simultaneously undergoing apical constriction (Kam et al., 1991; Sweeton et al., 1991). Two known loci — concertina (cta) and folded gastrulation (fog) — are required for the transition from the first to the second phase (Zusman and Wieschaus, 1985; Parks and Wieschaus, 1991; Sweeton et al., 1991; Costa et al., 1994). cta encodes a G_α-like protein that is produced maternally and deposited in the egg and early embryo (Parks and Wieschaus, 1991). fog encodes a novel putative secreted protein that is expressed in the invagination primordia in a pattern that precisely prefigures the pattern of apical cell constrictions (Costa et al., 1994). It has been suggested that these two gene products constitute part of a signaling pathway that coordinates the group behaviour of the cells undergoing morphogenetic alterations. Neither cta nor fog mutations affect the assignment of positional values to cells; rather they specifically disrupt coordinate cell shape changes.

In contrast to the mesodermal and endodermal invaginations, germ-band extension is not driven by cell shape changes but by cell rearrangements (Irvine and Wieschaus, 1994): extensive intercalation of cells results in a decrease in cell number along the dorsoventral (D-V) axis of the embryo with a concomitant increase in cell number along the anteroposterior (A-P) axis.

Germ-band retraction is attained through a combination of cell shape changes and cell rearrangements, with the former playing a more important role than the latter (Campos-Ortega and Hartenstein, 1985; Martinez Arias, 1993). Cell shape changes in the epidermal ectoderm account for a 40% reduction along the A-P axis and an 85% increase across the D-V axis relative to the unretracted state (Martinez Arias, 1993). Local cell rearrangements produce a further 10% decrease along the A-P axis with a concomitant 15% increase across the D-V axis (Martinez Arias, 1993). These cell shape changes and local cell rearrangements begin in the thoracic region and spread posteriorly (Martinez Arias, 1993).

Expression of six genes is required zygotically for germband retraction: the Drosophila homolog of the mammalian EGF receptor (variously called top, flb and DER and referred to here as Egfr) (Clifford and Schüpbach, 1989; Raz et al., 1991), the Drosophila homolog of the mammalian insulin receptor (encoded by the inr gene) (Fernandez et al., 1995), hindsight (hnt) (Wieschaus et al., 1984; Strecker et al., 1991, 1992), tailup (tup) (Nüsslein-Volhard et al., 1984), u-shaped (ush) (Nüsslein-Volhard et al., 1984) and serpent (srp) (Jürgens et al., 1984; Reuter, 1994). Four of these genes, the Egfr (Clifford and Schüpbach, 1992), hnt, ush and srp (Frank and Rushlow, 1996) are required for the maintenance of the differentiated amnioserosa.

In this report, we present our analysis of the hnt mutant phenotype and report the molecular cloning and analysis of the expression of the hnt gene. Our results indicate that hnt does not affect pattern formation or tissue specification prior to germ-band retraction. The sequence of the hnt cDNA and localization of HNT protein in nuclei suggest that it functions as a zinc-fingercontaining transcriptional regulator. Strikingly, hnt is not expressed in the epidermal ectoderm that undergoes the cell shape changes and movements that drive germ-band retraction. Rather, hnt is expressed in the endoderm and amnioserosa prior to, during and after retraction. Based on analysis of single and double mutants that eliminate expression of hnt in the regions that normally form midgut and/or amnioserosa, we argue that hnt expression in the amnioserosa is crucial for germ-band retraction. Two models are presented: a ‘physical’ model in which hindsight functions to maintain the differentiated amnioserosa which then controls retraction through direct physical interaction with cells of the germ band; and a ‘chemical’ model in which hindsight functions to maintain the amnioserosa, which then produces or activates a signal that is received by the germ band and coordinates germ-band retraction.

Flies were raised on standard medium at 25°C unless otherwise specified. The original hnt alleles, hnt^XE81 and hnt^X001, were isolated by Wieschaus et al. (1984). Two additional alleles, hnt^EH587 and hnt^EH704a, were obtained in a subsequent EMS mutagenesis screen (Eberl and Hilliker, 1988). We report here, that a putative fifth allele, l(1)EH275a (Eberl and Hilliker, 1988) also referred to as hnt^EH275a (Ray, 1993), is in fact not allelic to hnt. Most other mutations used in this study are described in detail in Lindsley and Zimm (1992); alleles used were: cact^A2, cact^HE9, cta^WU31, Egfr^f1, fog^S4, hkb², peb¹, pll⁰⁷⁸, pll³⁸⁵, sax^HB18, sax^WO18, srp^9L, tld⁹, tll¹, tor^PM51, tor^splc and zen^f62. tld^68-62 is described in Shimell et al. (1991). All zygotic lethal mutations were maintained over appropriate ‘blue balancer’ chromosomes harboring a P-element transgene that expresses β-galactosidase under the control of the ftz promoter: FM7Z (Kania et al., 1990), CyOZ (Raz and Shilo, 1993), TM3Z (S. Smolick-Utlaut and E. B. Lewis, personal communication). Flies carrying the temperature-sensitive rough-eyed mutation, pebbled, were raised at 28±1°C, the restrictive temperature.

Embryos were collected for 24 hours and allowed to age at least an additional 24 hours. In most cases, cuticles were prepared by clearing dechorionated embryos in mounting medium as described in Ashburner (1989), except that embryos were not fixed before mounting. In cases where the vitelline membrane was removed for analysis, unhatched embryos were dechorionated with 50% bleach for 2 minutes, transferred to methanol:heptane (1:1) and vortexed to remove vitelline membranes. Embryonic cuticles were then fixed, mounted and analyzed as previously described (Lamka et al., 1992). For quantitative analysis of the germ-band retraction defects, hatched versus unhatched embryos were counted and the latter class was then processed for cuticle analysis. The extent of germ-band retraction was measured by noting which abdominal segment was located at the posterior tip of the embryo; half-segment measures were used to increase resolution (see Results; Figs 1, 2). Complete retraction positions the telson at the tip.

Embryos were collected for 0.5 hour intervals from balanced heterozygous hnt females. The embryos were submerged in halocarbon oil and were filmed under bright-field illumination at 21±1°C for more than 12 hours, usually in groups of three to four at a time. Filming was done using a Dage-MTI CCD72 videocamera attached to a Nikon Diaphot-TMD inverted microscope and recorded with a Hitachi/GYYR time-lapse VCR.

Antibody staining of embryos was according to established procedures (Macdonald and Struhl, 1986; Patel et al., 1989). Primary antibodies were mouse anti-ABD-B monoclonal antibody (mAb) (1:2 dilution) (Celniker et al., 1989); rat anti-CUT (1:300) (Blochlinger et al., 1990); mouse anti-FKH (1:33) (Y. M. Kuo and S. K. Beckendorf, personal communication); rabbit anti-KR (1:200) (Gaul et al., 1987; M. Levine, personal communication); rabbit anti-LAB (1:75) (Diederich et al., 1991); mAbD₃(1:15) (68G5D3) (Giniger et al., 1993); mAb 22C10 (1:2) (Fujita et al., 1982); mouse anti-β-galactosidase (1:200 to 1:500) (Promega, Inc.); mouse monoclonal anti-HNT antibody 27B8 1G9 (1:10 to 1:20) (see below). Secondary antibodies were alkaline phosphatase-conjugated goat anti-mouse, goat anti-rabbit or donkey anti-rat antisera (Jackson Immunoresearch) used at 1:500-1:2000 dilution or goat anti-mouse or goat anti-rabbit antisera conjugated to horse radish peroxidase (HRP) (Jackson Immunoresearch) used at a dilution of 1:300. Since all zygotic lethal mutations were maintained over appropriate ‘blue balancer’ chromosomes (above) mutant embryos were identified by simultaneously staining batches of embryos with both anti-β-galactosidase and the particular antibody of interest.

Standard protocols were as described (Sambrook et al., 1989). Overlapping phage genomic DNA clones covering about 70 kb proximal to the proximal deficiency breakpoint of Df(1)rb⁴⁶(Pflugfelder et al., 1990) were hybridized to ³²P-labeled cDNA probes synthesized from 0-3 hour embryonic poly(A) RNA. A 4.5 kb BamHI fragment from phage clone X-59 (which substantially overlaps clone X-32) and located at +130 kb on the chromosomal walk (Pflugfelder et al., 1990) was identified as hybridizing to the cDNA and was then used to screen a cDNA library (Poole et al., 1985). We isolated a cDNA clone, designated E20, with a 2 kb insert and, using this as a probe, we screened a second embryonic cDNA library to isolate longer cDNA clones (Brown and Kafatos, 1988). Iterative screening using probes derived from the 5′-most portions of progressively longer cDNAs resulted in the isolation of long cDNA clones encompassing the entire open reading frame (e.g. NB701). Sequencing of cDNA inserts was done using the Sequenase kit (US Biochemical Corp.) or the Cycle Sequencing Kit (Applied Biosystems, Inc.).

Genomic DNA was isolated from hnt mutant embryos using at least 100 embryos for each preparation. The polymerase chain reaction (PCR) was used to amplify fragments of the hnt coding region from this genomic DNA; for each amplified fragment, triplicate PCR reactions were set up and processed simultaneously. Primers were: forward 5′-GCCAGTCTCTCGGAATCGGG-3′ derived from nucleotides 719-738 of the hnt cDNA sequence; reverse 5′-TCGCAGGCGGGACAACTTAG-3′ derived from nucleotides 1806-1787 of the hnt cDNA sequence. After PCR amplification, the triplicate reactions were combined and the amplifed fragment was purified using QIAquick PCR Purification Kit (Qiagen). Amplified DNA was then sequenced. The primers relevant to the point mutation in hnt^X001 were: forward 5′-TGCTATCCTCGGCTTCATCC-3′ derived from nucleotides 1110-1129 of the hnt cDNA sequence; reverse 5′-CTGC-GACTGTGACTATGTCCAC-3′ derived from nucleotides 1473-1452 of the hnt cDNA sequence.

A 6.3 kb SspI-NotI fragment containing the entire NB701 cDNA was cloned into the HpaI-NotI sites of the pCaSpeR-hs vector (Thummel and Pirrotta, 1991) and transformed into the germ line (Rubin and Spradling, 1982; Spradling and Rubin, 1982). Two transgenic lines were obtained: hs-hnt:E (X chromosome) and hs-hnt:M (2nd chromosome). For rescue experiments, virgin females from each of the four y hnt mutant stocks were mated to males from the hs-hnt:M transgenic line. The experimental crosses were: y hnt/FM7Z × w¹¹¹⁸/Y; hshnt:M/hs-hnt:M; control crosses were: y hnt/FM7Z × w¹¹¹⁸/Y. All progeny from the experimental crosses carry one copy of the hs-hnt transgene and were collected on yeasted grape juice-agar plates at 2 to 2.5 hour intervals. 3.5 or 4 hours after collection, embryos were heat shocked twice at 36.5°C for 0.5 hour per treatment with an interval of 0.5 hour at 25°C between heat shocks. Embryos were then allowed to develop at 25°C before cuticles were mounted for analysis. hnt mutant embryos were identified on the basis of the y marker.

Whole-mount RNA tissue in situ hybridization was based on previously published protocols (Tautz and Pfeifle, 1989; Ding et al., 1993). Digoxigenin (DIG)-labeled hnt DNA probe was synthesized from a 2 kb EcoRI fragment containing the 3′ end of hnt. DIG-labeled hnt antisense RNA probe was synthesized by in vitro transcription off the T7 promoter using cDNA NB701 as template. Both DNA and RNA probes gave identical results.

A 907 bp BglII-BamHI fragment from cDNA NB701 (nucleotides 2710 to 3617; codons 824 to 1125) encoding a 302 amino acid HNT polypeptide was cloned into the BamHI site of pGEX1 (Smith and Johnson, 1988). The GST-HNT fusion protein was expressed and purified from E. coli (Ausubel et al., 1987). To generate anti-HNT antibodies, 50 μg of purified GST-HNT fusion protein suspended in RIBI adjuvant (RIBI Biochem) was injected into three Balb/c mice (Simenson). Antisera from the mice were tested by ELISA against the fusion protein and on fixed Drosophila embryos. The blood from the mouse that produced the strongest response was saved as anti-HNT polyclonal antiserum. This mouse was then killed for monoclonal antibody production that followed standard procedures (Harlow and Lane, 1988) and used HL-1 myeloma cells (Hycor). Monoclonal supernatants were tested on fixed embryos. The monoclonal anti-HNT antibody used in most of our experiments is designated 27B8 1G9 and was used at a 1:10 to 1:20 dilution of supernatant.

Four embryonic lethal alleles of the X-linked gene hindsight (hnt) have been reported (Wieschaus et al., 1984; Eberl and Hilliker, 1988; Lindsley and Zimm, 1992). A fifth potential allele, l(1)EH275a (Eberl and Hilliker, 1988), is not allelic to hnt (see below). We demonstrate here and in a separate study that pebbled (peb) mutations are viable alleles of hnt (M. L. R. Y., Q. Sun, M. L. L. and H.D. L., unpublished data). The four embryonic lethal alleles hnt^XE81, hnt^X001, hnt^EH704a and hnt^EH587 display qualitatively similar phenotypes: hemizygous (hnt/Y) embryos fail to retract their germ bands (Figs 1, 2). All such embryos have the correct number of thoracic and abdominal segments which are patterned normally. As a consequence of failed germ-band retraction, the embryos are U-shaped with their posterior region folded onto the dorsal side (Fig. 1B-D). Additionally, mutant embryos show defects in head involution and often have a severely disrupted cephalopharyngeal skeleton (Fig. 1B-D) (Ray, 1993; Yip, 1995; Frank and Rushlow, 1996).

l(1)EH275a (Eberl and Hilliker, 1988) has previously been classified as a semi-lethal hnt allele and referred to as hnt^275a (Ray, 1993). Germ-band retraction occurs in most l(1)EH275a embryos, but these usually die with head defects and occasional dorsal holes. Normal looking l(1)EH275a escaper adult males can emerge but they are often sterile, sometimes exhibiting partially unrotated external genitalia. Thus it has not previously been possible to determine whether l(1)EH275a is indeed allelic to hnt (Eberl and Hilliker, 1988). We have recently used an autosomal duplication of the wild-type hnt locus to carry out inter se complementation tests among all putative hnt alleles (B. Reed and H. D. L., unpublished observations). l(1)EH275a complements embryonic lethal hnt alleles (hnt^XE81, hnt^X001, hnt^EH704a) (B. Reed and H. D. L., unpublished observations). l(1)EH275a also complements a chromosomal deletion, Df(1)ovoG6 (Pflugfelder et al., 1990), that removes most of the hnt locus (B. Reed and H. D. L., unpublished observations). In addition, l(1)EH275a complements the viable hnt allele, hnt^peb (M. L. R. Y. and H. D. L., data not shown; see below). We thus conclude that l(1)EH275a is not allelic to hnt.

We have identified an additional, viable hnt allele that was originally named pebbled (peb) because of its rough eye phenotype (Lindsley and Zimm, 1992). None of the four embryonic lethal alleles complements the pebbled rough eye phenotype (Table 1), suggesting that hindsight and pebbled are allelic. Detailed analyses of hnt functions during eye development will be reported separately (M. L. R. Y., Q. Sun, M. L. L. and H.D. L., unpublished data). We shall refer to peb henceforth as hnt^peb. Table 1 presents a summary of hnt mutant phenotypes.

While previous studies have carried out a survey of cuticular and/or amnioserosal defects, to date there has been no quantitative analysis of germ-band retraction in hnt mutant embryos (Eberl and Hilliker, 1988; Ray, 1993; Frank and Rushlow, 1996). We quantified germ-band retraction by examining populations of mutant embryos and plotting the distribution of abdominal segments lying at the posterior tip (Fig. 2; see Materials and Methods). The embryonic-lethal hnt alleles represent a phenotypic series with respect to failure of germ-band retraction (Figs 1, 2). When hemizygous, a deletion of the hnt locus, Df(1)bi^D3 (Banga et al., 1986; Oliver et al., 1988), results in 100% of the embryos with abdominal segment 4.5 to 6.0 at the posterior tip (Fig. 2A). The strongest hnt allele studied, hnt^XE81, is similar to the deletion in that the mode of the distribution lies at abdominal segment 5.0. However, this distribution is broader than that produced by the deletion; 12% of the embryos have abdominal segment 6.5 or greater at the posterior tip (Fig. 2B), a situation not seen with Df(1)bi^D3. Weaker still are hnt^EH704a and hnt^X001; respectively, 36% and 68% of the embryos have segment 6.5 or greater at the posterior tip (Fig. 2C,D). hnt^EH587is very similar in distribution to hnt^EH704a (data not shown).

We wished to determine whether the morphogenetic process of germ-band retraction failed in hnt mutants without significant defects in tissue specification or pattern formation. With the exception of the head, the cuticle of hnt mutant embryos does not show any pattern defects (Fig. 1) (Yip, 1995). In order to study the internal tissues, we examined embryos carrying the strongest mutant allele hnt^XE81 with a panel of antibodies that recognize proteins/epitopes with different temporal and spatial expression patterns and functions (Table 2). With one exception, all of these markers exhibited no obvious differences between hnt^XE81 mutant embryos and wild-type embryos prior to germ-band retraction. The exception, KRÜPPEL, which accumulates in wild-type embryos in the nuclei of amnioserosal cells, is absent from most but not all of these cells in stage 11 hnt mutant embryos (Fig. 3) (Ray, 1993).

Our examination of the external and internal tissues of hnt mutants revealed neither pattern abnormalities nor defects in tissue specification prior to germ-band retraction. However, it remained possible that, while cell position and tissue specification were normal, earlier morphogenetic events were abnormal and thus that failure of germ-band retraction was a secondary consequence of such defects. For example, if hnt mutants were defective in the timing or the spatial aspects of germ-band extension, germ-band retraction might fail as a consequence of inappropriate positioning of tissues to receive a ‘retraction signal’. In order to avoid the ambiguities inherent in analyses of fixed material, we used time-lapse video microscopy of living hnt^EH704a/Y mutant embryos to determine whether any morphogenetic defects were visible in the mutants at these earlier stages. Videotaped hnt mutant embryos were identified on the basis of their failure to retract their germ bands by 13 hour postfertilization and these embryos were then studied retrospectively for earlier morphogenetic events. +/+, hnt /+, +/Y and hnt/Y mutant embryos extend their germ bands at indistinguishable rates and with similar spatial relationships. The first difference appears after germ-band extension is complete: three quarters of the embryos (presumably hnt/+, +/+ or +/Y) initiate germ-band retraction normally while the remaining quarter of them (presumably hnt/Y) either did not initiate retraction or initiated premature retraction-like movements (4 hours after the beginning of germ-band extension in hnt/Y versus 7.5 hours in hnt/+, +/+ or +/Y siblings at 21±1°C) but failed to complete the process. The proportion of the embryos that underwent retraction-like movements correlated with the strength of the hnt^EH704aallele determined on the basis of cuticle analysis: 7/11 videotaped mutant embryos (57%) underwent some retractionlike movements, consistent with the fact that 59% of the hnt^EH704acuticles exhibited abdominal segment 6.0 or greater at the posterior tip (Fig. 2C). We conclude that the hnt gene is not required for early embryonic pattern specification, tissue specification or earlier morphogenetic events, such as germband extension.

Previous studies mapped hnt within polytene chromosome region 4C5/6 (Oliver et al., 1988; Lindsley and Zimm, 1992). In contrast to published data (Oliver et al., 1988), we found that the deficiency chromosome, Df(1)rb⁴⁶, complements the rough eye phenotype of hnt^peb, placing hnt proximal to the proximal breakpoint of Df(1)rb⁴⁶. A chromosomal walk that spanned the proximal breakpoint of Df(1)rb⁴⁶ had been conducted in this region (Pflugfelder et al., 1990). Using overlapping phage clones proximal to the breakpoint (kindly provided by Dr G. Pflugfelder), we initiated a search for early embryonic transcripts encoded by the region. One BamHI fragment at +130 kb on their walk was identified and used to screen two embryonic cDNA libraries (see Materials and Methods). Two cDNAs, denoted E20 (the first to be identified) and NB701 (one of the longest) were analyzed in detail.

Sequence analysis of NB701 and E20 revealed that E20 was a partial cDNA and that both clones were otherwise identical in sequence. NB701 contained a single large open reading frame (ORF) of 1920 codons with 241 bp of 5′-and 163 bp of 3′-untranslated regions. Conceptual translation of the ORF is shown in Fig. 4A. It encodes a protein with several characteristics of transcription factors [analyses were conducted using the BLAST algorithm (Altschul et al., 1990)]: 14 C₂H₂ type zinc fingers (Fig. 4B) in widely spaced clusters, multiple glutaminerich domains, proline-rich domains, serine/theronine-rich domains and acidic/charged domains (Fig. 4C).

We confirmed that this cDNA is encoded by the hnt gene in three ways. First, we made an hsp70 promoter-hnt cDNA transgene using the NB701 clone and produced transgenic flies using P-element-mediated transformation (Rubin and Spradling, 1982; Spradling and Rubin, 1982). This transgene rescues germ-band retraction in embryos mutant for any of the four embryonic lethal alleles (Table 3). On average, 50% of the heat-shocked genotypically y hnt embryos carrying one copy of the transgene fully retracted their germ bands (‘T’ at posterior tip). In contrast, <5% of control hnt^XE81and hnt^X001embryos complete germ-band retraction (Table 3 cf. Fig. 2). The heat-shock regimen used did not rescue the hnt mutant head defects, nor did it induce any dominant pattern or morphogenetic defects. Second, we identified the molecular lesion caused by the EMS-induced allele hnt^X001. Sequence analysis of fragments PCR-amplified from hnt^X001genomic DNA revealed a C to T transition that introduced a premature stop codon at amino acid residue 348 in place of a glutamine (CAG to TAG) (Fig. 4A,C). The truncated protein is predicted to contain only 3 of the 14 zinc fingers. Third, none of the four lethal hnt alleles shows detectable immunostaining when examined with our anti-HNT antibodies (data not shown). We conclude that we have cloned the hnt gene and that it encodes a putative zinc-finger transcription factor.

The embryonic expression of hnt was determined using wholemount tissue in situ hybridization to visualize hnt RNA (data not shown) and anti-HNT antibodies to visualize HNT protein (Fig. 5). HNT protein is localized to nuclei as expected of a transcription factor. No hnt mRNA or protein is detectable before stage 5 of embryogenesis, consistent with the fact that analysis of embryos produced by homozygous clones of hnt cells in the germ line revealed no requirement for hnt expression or function during oogenesis (Wieschaus and Noell, 1986). hnt mRNA accumulation begins in the cellular blastoderm (stage 5) in a posteriorterminal domain corresponding to the posterior midgut primordium and dorsally in the presumptive amnioserosa (stages are according to Campos-Ortega and Hartenstein, 1985). HNT protein appears in these cells slightly later, at stage 6 (Fig. 5A). During stage 7, dorsal expression expands to cover the entire presumptive amnioserosa from the cephalic furrow to the posterior midgut primordium (Fig. 5B). Anteroventral staining, corresponding to the anterior midgut primordium, is first detected at stage 8 (Fig. 5C). Accumulation in these tissues continues as gastrulation proceeds (Fig. 5D). Commencing at stage 11, accumulation is also detectable in the cells of the emerging larval peripheral nervous system and the tracheal system (Fig. 5E-G). Expression also occurs in peripheral and CNS glial cells (M. L. L. and H. D. L., unpublished data). The complex expression pattern of hnt suggests that it may have multiple functions during embryogenesis. Functions during the development of the embryonic nervous system (M. L. L. and H. D. L., unpublished data) and during adult eye development (M. L. R. Y., Q. Sun, M. L. L. and H. D. L., unpublished data) will be reported elsewhere. Here we focus on hnt functions in germ-band retraction since the first detectable morphogenetic defect in hnt mutants is failure of this process. Strikingly, hnt is absent from the epidermal ectodermal cells that undergo the cell shape changes and movements that drive germ-band retraction (Martinez Arias, 1993). Since hnt RNA and HNT protein accumulate in the endoderm and amnioserosa prior to germ-band retraction, it seemed likely that hnt expression in one or both of these tissues plays a role in the control of this morphogenetic process (see below).

We carried out a survey of the genetic control of hnt expression in order to enable us to begin to dissect which aspects of hnt expression are important for germ-band retraction. hnt expression in the midgut is controlled by the maternal and zygotic members of the torso-mediated ‘terminal’ pathway as suggested by our previous genetic studies (Strecker et al., 1989, 1991, 1992; Strecker and Lipshitz, 1990). Embryos produced by homozygous torso loss-of-function mutant females lack HNT expression in the posterior midgut which lies within the domain of tor function. Instead of extending their germ bands dorsoanteriorly, most such embryos form spiralled germ bands (Fig. 6A) (Schüpbach and Wieschaus, 1986, 1989). Reciprocally, embryos from homozygous torso^splcgain-of-function mothers lack dorsal expression (i.e. in the presumptive amnioserosa) consistent with conversion of central cell fates to more terminal ones (Schüpbach and Wieschaus, 1989; Strecker et al., 1989, 1991, 1992; Strecker and Lipshitz, 1990). These embryos also show expanded expression of HNT in the enlarged posterior midgut primordium and twisted gastrulation (Fig. 6B) (Yip and Lipshitz, 1996). Two genes, tailless (tll) and huckebein (hkb), have been identified as key zygotically expressed components of the torso-mediated terminal pathway (Klingler et al., 1988; Strecker et al., 1989; Weigel et al., 1990). tll mutant embryos have an abnormal acron anteriorly, and lack abdominal segments 8-10, hindgut and Malpighian tubules posteriorly (Strecker et al., 1986, 1988; Pignoni et al., 1990). hkb is required for the formation of endodermal midgut and stomodeum (Weigel et al., 1990; Brönner et al., 1994). tll mutations have little effect on HNT expression (Fig. 6C); from analysis of hkb tll double mutant embryos (Fig. 6E), it is clear that the only loss of HNT expression in tll mutants occurs in the region from which the Malpighian tubule primordia originate, consistent with the reported role for tll and hnt in the development of these structures (Strecker et al., 1986, 1988; Harbecke and Lengyel, 1995). hkb mutant embryos lack HNT expression in the regions from which the anterior and posterior midgut normally arise; expression remains only in the presumptive ureter of the Malpighian tubules (Fig. 6D). In hkb tll double mutant embryos, HNT is not expressed at all in the domains that would form anterior and posterior midgut and Malpighian tubule primordia; expression does, however, occur in the amnioserosa (Fig. 6E). Germ-band retraction occurs in tll or hkb single mutants (see Fig. 7B) as well as in hkb tll double mutants, suggesting that midgut expression of HNT is not necessary for germ-band retraction.

HNT expression was assayed in four mutants that are defective in germ-band retraction: serpent (srp), u-shaped (ush), tailup (tup) and the EGF receptor (Egfr) (Nüsslein-Volhard et al., 1984; Clifford and Schüpbach, 1989). HNT expression is not affected in ush, tup and Egfr mutants (Fig. 6G shows an Egfr mutant; data not shown for ush and tup). These results suggest that hnt either resides upstream of these three genes in the same hierarchy or one or more of these genes functions in a parallel pathway involved in germ-band retraction. In contrast, endodermal expression of HNT is missing in srp mutant embryos (Fig. 6F). This last result is consistent with the fact that srp is required to establish the identity of the endodermal midgut; loss-of-function mutations in srp result in transformation of the endoderm into ectoderm (Reuter, 1994).

HNT expression in the amnioserosa is regulated by the dorsoventral pathway. Dorsal HNT expression is reduced in genetically ventralized embryos such as those produced by saxophone (sax) or cactus (cact) females (Fig. 6H shows an embryo from a cact female; data not shown for sax) or homozyous for zen or tolloid (tld) (Fig. 7C for tld; data not shown for zen). Reciprocally, dorsal HNT expression expands ventrally in dorsalized embryos (Fig. 6I shows an embryo from a pelle (pll) female). Anterior midgut expression of HNT is affected by the dorsoventral pathway (e.g. Fig. 6I) since anterior midgut development requires inputs from both terminal and dorsoventral pathways (Reuter and Leptin, 1994; Yip and Lipshitz, 1996). Since dorsoventral mutants fail to undergo germ-band extension, the role of amnioserosal expression of HNT in germ-band retraction could not be assayed in embryos singly mutant for any these genes (but see below).

The concertina (cta) and folded gastrulation (fog) genes are required for proper ventral furrow formation and posterior midgut invagination (Zusman and Wieschaus, 1985; Parks and Wieschaus, 1991; Sweeton et al., 1991; Costa et al., 1994). HNT expression is unaffected by either mutation (Fig. 6J shows an embryo from a cta female; data not shown for fog).

hkb mutants eliminate midgut cell fates and expression of HNT in the presumptive midgut regions at the embryonic termini (Figs 6D, 7C). However hkb mutant embryos undergo normal germ-band extension and retraction (Fig. 7C,D) (Reuter et al., 1993). Thus, neither the midgut per se nor HNT expression in the termini is necessary for germ-band retraction, while dorsal HNT expression in the amnioserosa is sufficient for retraction.

Strongly ventralizing dorsoventral pattern mutants such as zen and tld eliminate amnioserosa cell fates and dorsal expression of HNT (Fig. 7E; tld⁹/tld⁶⁸⁻⁶²). However, assaying the function of the amnioserosa and dorsal HNT expression in germ-band retraction is not possible using these mutants because they fail to undergo germ-band extension (Fig. 7E,F).

hkb²tld⁹ double mutant embryos lack HNT expression dorsally as well as in the termini (Fig. 7G) but undergo germband extension (Fig. 7G,H), thus allowing us to assay germband retraction in the absence of both midgut and amnioserosa. Strikingly, hkb²tld⁹double mutant embryos fail to undergo germ-band retraction (Fig. 7H). These results suggest that the amnioserosa per se — and likely HNT expression in this tissue — is necessary for germ-band retraction (see Discussion).

We have shown that hnt gene function is required for germband retraction and that hnt mutations do not cause defects in this morphogenetic process as an indirect consequence of misspecification of cellular positional values or tissue identity. hnt encodes a large zinc-finger protein that accumulates in the nuclei of several tissues prior to germ-band retraction. Strikingly, hnt is not expressed in the epidermal ectoderm, the tissue that undergoes the morphogenetic alterations that drive retraction. Analyses of hnt expression and germ-band retraction in mutants that lack specific tissues in which HNT normally accumulates suggests that hnt expression in the extraembryonic amnioserosa is necessary and sufficient for retraction.

Two lines of evidence support the conclusion that failure of germ-band retraction is a primary effect of hnt mutations. First, hnt mutant embryos undergo normal pattern and tissue specification in the germ band prior to retraction. Second, the temporal and spatial aspects of germ-band extension are indistinguishable in wild-type and in hnt mutant embryos. Thus, there are no detectable defects either in tissue specification or in morphogenesis prior to germ-band retraction. Of eight molecular markers assayed, one – KRÜPPEL – was abnormal in hnt mutants prior to germ-band retraction. KRÜPPEL is absent from many but not all amnioserosal cells by stage 11 (similar observations have been reported by Ray, 1993). Absence of KRÜPPEL at this stage correlates with the recently reported premature apoptosis of the differentiated amnioserosa in hnt mutants (Frank and Rushlow, 1996). The amnioserosal cells of hnt^XE81mutant embryos display similar morphology to those of the wild type until apoptosis commences, leading us to suspect that the lack of KRÜPPEL in the amnioserosa of hnt mutants is an indirect consequence of premature apoptosis, rather than that the Krüppel gene is a regulatory target of HNT. The absence of any retraction defects in Krüppel mutants is consistent with this suggestion.

The hnt gene encodes a large protein with fourteen C₂H₂-type zinc fingers. This type of zinc finger is found in many transcription factors and has been shown to function as a DNA-binding domain (Pabo and Sauer, 1992). The arrangement of the HNT zinc fingers is unusual in that twelve of the fourteen fingers occur in widely spaced clusters, each of which includes two or three tandemly arranged zinc fingers. Only two of the fourteen zinc fingers — the ninth and twelfth — are unclustered. The fifth zinc finger does not have the second conserved histidine residue and therefore may be non-functional. If so, then four clusters of zinc fingers alternate with isolated functional zinc fingers throughout the length of the protein. In addition to the zinc fingers, the HNT protein contains several structural domains commonly found in transcriptional regulators, including multiple glutamine-rich, proline-rich, serine/theronine-rich and acidic/charged domains. Each of these types of domains has been shown to function in trans to mediate protein-protein interactions important for transcriptional control (Courey and Tjian, 1988; Mermod et al., 1989; Ptashne and Gann, 1990; Stringer et al., 1990; Tanaka and Herr, 1990; Dynlacht et al., 1991; Lin and Green, 1991; Lin et al., 1991; Madden et al., 1991; Han and Manley, 1993). These structural motifs, combined with the nuclear localization of HNT protein as revealed by antibody staining, make it highly likely that HNT protein functions as a transcription factor. The fact that the hnt^X001 mutation results in a protein product truncated after the first three zinc fingers but is only a weak allele (Fig. 2D), suggests that partial function can be conferred by the first cluster of three zinc fingers.

The hs-hnt transgene rescues the morphogenetic process of germ-band retraction without causing any dominant gain-of-function defects, strongly suggesting that the role of HNT is permissive rather than instructive; that is, that HNT is involved in the spatial and/or temporal coordination of germ-band retraction, rather than in the implementation of the morphogenetic cell shape changes and movements. The fact that HNT is a zinc-finger protein that is expressed in a distinct set of tissues from those that undergo the morphogenetic alterations, is consistent with this possibility.

Expression of hnt in the amnioserosa dorsally versus in the posterior midgut primordium is regulated, respectively, by the dorsoventral and the terminal gene hierarchies, while expression of hnt in the anterior midgut primordium receives input from both of these pathways (Yip and Lipshitz, 1996). hnt is positioned downstream of the genes that specify endodermal (midgut) identity, such as hkb. In contrast, mutations in the early morphogenetic control genes cta and fog do not affect hnt expression. This last result suggests that distinct pathways control the early morphogenetic movements that result in internalization of the mesoderm and endoderm (controlled by cta and fog) versus the later morphogenetic movements that drive germ-band retraction (controlled by the U-shaped class of genes that includes hnt).

HNT expression is normal in embryos mutant at any of four other loci required for germ-band retraction: ush, tup, Egfr (this study) and inr (M. L. L. and H. D. L., unpublished data). Furthermore, the hs-hnt transgene is unable to rescue ush, tup and Egfr mutants under experimental conditions that rescue hnt (M. L. R. Y. and H. D. L., unpublished data). These data are consistent with ush, tup, inr and Egfr residing either downstream of hnt or in a parallel pathway that regulates germ-band retraction. In contrast, endodermal expression of HNT is missing in a fifth mutant that affects germ-band retraction, srp. This last result is consistent with the fact that srp is required to establish the identity of the endodermal midgut and is expressed in both the anterior and posterior midgut primordia through stage 9 (Reuter, 1994; Rehorn et al., 1996). Not surprisingly, then, HNT expression is absent in srp mutants since the endoderm is absent. The fact that germ-band retraction is normal in hkb mutants that lack midgut but fails in srp mutants in which the midgut is converted into foregut and hindgut, suggests that loss of HNT from presumptive midgut per se does not result in failure of germ-band retraction; rather the failure of retraction in srp mutants must have some other cause.

The most striking result of the present studies is the absence of hnt expression in the epidermal ectodermal cells that undergo the morphogenetic alterations that accomplish germband retraction, and the presence of hnt expression in several other tissue types — notably the endoderm and amnioserosa — prior to and during germ-band retraction. The question thus arises as to the role of these tissues in germ-band retraction. Definitive conclusions will require analyses of embryos genetically mosaic for hnt (M. L. L. and H. D. L., unpublished data). However, at this point, several tentative conclusions can be reached on the basis of genetic and phenotypic studies.

The tissue actually responsible for executing germ-band retraction is the epidermal ectoderm. As described in the Introduction, morphological and anatomical analyses of developing embryos show that epidermal ectodermal cells undergo extensive shape changes as well as local rearrangements during germ-band retraction (Martinez Arias, 1993). These processes initiate in the thoracic region and proceed posteriorly, driving germ-band movement around the posterior pole of the embryo (Martinez Arias, 1993). In contrast, the mesoderm is dispensable for germ-band retraction since embryos lacking mesoderm undergo normal germ-band retraction (Leptin et al., 1992). The fact that germ-band retraction is normal in hkb mutants that lack midgut, and thus lack HNT expression in the termini (Fig. 7C,D) suggests that neither the midgut per se nor HNT expression in the termini is essential for germ-band retraction. Further, embryos in which the migration and morphogenesis, but not specification, of the endoderm are disrupted can still undergo germ-band retraction (Reuter et al., 1993; Tepass and Hartenstein, 1994). Thus, neither endodermal nor mesodermal cells per se, nor the migration and morphogenesis of these tissues, are necessary to direct germ-band retraction.

In contrast to the endoderm and mesoderm, a crucial role for the extraembryonic, amnioserosal cells in programming germband retraction is likely. First, since the amnioserosa is present in hkb mutants, which successfully complete germ-band retraction, the amnioserosa is sufficient to program retraction in the absence of midgut. Second, reduction in the size of the amnioserosa (in embryos mutant for weakly ventralizing alleles of zygotically active dorsoventral pattern genes) results in a U-shaped phenotype (Arora and Nüsslein-Volhard, 1992). This suggests that retraction is sensitive to the size of the amnioserosa, consistent with the amnioserosa being necessary for retraction. Third, our double mutant analysis (Fig. 7G,H) confirms that the amnioserosa is necessary for germ-band retraction.

Consistent with an important role for HNT expression in the amnioserosa, and for the amnioserosa per se, it has recently been shown that hnt, srp and ush mutations all result in premature death of the differentiated amnioserosa (Frank and Rushlow, 1996). Indeed, the timing of initiation of the premature germ-band retraction-like movements that we observed in a subset of hnt mutant embryos, correlates well with the stage at which the amnioserosa cells have been reported to undergo premature apoptosis in these mutants (Frank and Rushlow, 1996). Why amnioserosal death should result in premature initiation of this morphogenetic process cannot yet be explained. However, it suggests that the amnioserosa plays both a negative and a positive role in coordinating germ-band retraction: it prevents premature retraction movements while promoting retraction at the appropriate stage of embryogenesis.

Two classes of models – not mutually exclusive – could explain the role of the amnioserosa in germ-band retraction. The first, or ‘physical’ model, proposes that physical contact between the amnioserosa and the epidermal ectoderm is important for retraction. For example, cell shape changes in the amnioserosa may be necessary to drive, or to allow, retraction. The second, or ‘chemical’ model, suggests that a (possibly diffusible) signal from the amnioserosa to the epidermal ectoderm directs or coordinates the ectodermal cell shape changes that drive germ-band retraction. These models differ in several respects. Most importantly, the former predicts that the amnioserosa must be present for retraction to occur, while the latter predicts that the amnioserosa per se may be dispensible as long as the appropriate signals to the ectoderm are provided.

Several lines of evidence make us supect that the amnioserosa produces or activates signal(s) that coordinate the morphogenetic alterations in the ectoderm during germ-band retraction. Among the loci required for germ-band retraction, hnt, srp, inr and Egfr are the only four for which information regarding the molecular nature of the gene products and expression patterns is available (this study; Raz and Shilo, 1993; Fernandez et al., 1995; Rehorn et al., 1996). Expression and function of the zinc-finger-containing GATA-like factor, SRP, has been considered above. Both the EGFR (M. L. L. and H.D.L, unpublished data) and the INR (Fernandez et al., 1995) are expressed throughout the embryo with the exception that the INR is never present in the amnioserosa and the EGFR is absent from the amnioserosa after stage 10 (M. L. L. and H. D. L., unpublished data). Based on these expression patterns and the fact that the products of these last two genes are transmembrane receptor tyrosine kinases, it is possible that coordinating signals from the amnioserosa are received in the ectoderm by the INR and/or EGFR and transduced into the shape changes and local cell rearrangements that drive germ-band retraction. The coordinating signal(s) produced by the amnioserosa may be the ligands for these receptor(s). Alternatively they could be an activity or activities that process or activate the ligand(s). Or they could function more indirectly through effects on the extracellular matrix. Future genetic and molecular analyses will focus on specific tests of these models.

Independent of whether the ‘physical’ or the ‘chemical’ model is correct, it will be important for future analyses to distinguish alternative mechanisms by which hnt functions in the amnioserosa to program the germ-band retraction process. It is possible that hnt’s only function in the amnioserosa is to prevent premature apoptosis (Frank and Rushlow, 1996). In hnt mutants physical contact between the amnioserosa and epidermal ectoderm, or chemical signaling from the amnioserosa to the epidermal ectoderm, would then be disrupted as a secondary effect of loss of the amnioserosa. Alternatively, hnt may play a dual role in the amnioserosa: on the one hand, hnt might function to promote survival of the amnioserosal cells (preventing, for example, premature retraction movements) while, on the other hand, hnt might play a direct role in regulating production or activation of amnioserosa-to-ectoderm signal(s) that coordinate germ-band retraction.

HNT is expressed in several tissues other than the amnioserosa in the developing embryo. These include the midgut, the ureter of the developing Malpighian tubules, the developing tracheal system and glial cells in the nervous system (Fig. 5). Taken together with the fact that HNT is likely to be a transcription factor, it is possible that HNT controls additional morphogenetic events during embryonic and postembryonic development by transcriptionally regulating sets of genes that function to coordinate these processes. Future genetic and molecular analyses will address this possibility.

Special thanks to G. Pflugfelder for kindly providing the chromosomal walk in the vicinity of omb that enabled us to clone hnt. Drosophila stocks were provided by the Bloomington Drosophila Stock Center as well as S. Celniker, P. Gergen, A. Hilliker, M. O’Connor, W. Pak, S. Parkhurst, S. Parks, T. Schüpbach and B.-Z. Shilo. Antibodies were kindly provided by S. Beckendorf, S. Benzer, K. Blochlinger, S. Celniker, T. Kaufman, M. Krasnow, Y. Kuo, M. Levine and E. B. Lewis. Thanks to E. Wieschaus for suggesting the hkb tld experiment, S. Ou of the Caltech Monoclonal Antibody Facility for carrying out the hybridoma fusions involved in anti-HNT monoclonal antibody production, K. Zinn and E. B. Lewis for the use of laboratory equipment after the majority of the Lipshitz laboratory moved to Toronto. We thank G. Boulianne, S. Egan, C.c. Hui, R. McInnes and L. Rochwerger for comments on the manuscript. M. L. R. Y. was supported in part by a predoctoral fellowship from the Howard Hughes Medical Institute and M. L. L. by a postdoctoral National Research Service Award from the USPHS (GM16541). This research was supported by research grants to H. D. L. from the American Cancer Society (DB-14) and the National Cancer Institute of Canada (#7447) with funds from the Canadian Cancer Society.