Determination of wing cell fate by the escargot and snail genes in Drosophila

During development, groups of cells assume specific fates according to positional information. Two mechanisms, extrinsic induction and intrinsic determination, are required for these processes. Cells receiving inductive signals begin to express intrinsic determinants, acquire a specific cell fate, and differentiate to form specific patterns autonomously. One example is imaginal discs of Drosophila (Cohen, 1993). In embryos, each imaginal primordium is allocated to a specific position in the ectoderm and invaginates to form a sac-like imaginal disc. Subsequently, each performs a series of autonomous events to organize adult external structures.

The first sign of imaginal disc induction is the expression of the homeobox gene Distal-less (Dll) in the prospective leg imaginal disc (Cohen, 1990). In stage 11 embryos, the leg primordia, visualized by the Dll RNA expression, appears in clusters of cells that overlap the intersection between the dorsoventral row of cells expressing the segment polarity gene wingless (wg) and the anterior-posterior row of decapenta-plegic- (dpp) expressing cells (Cohen et al., 1993). Since Wg and Dpp are secreted signaling molecules (Padget et al., 1987; Rijsewijk et al., 1987), they were proposed to be the inductive signals that allocate the leg primordium within the ectodermal field established by the activities of segment polarity genes. In support of this idea, Simcox et al. (1989) used the embryo in vivo culture technique to show that wg is essential for the formation of imaginal discs. Using a temperature sensitive wg allele, Cohen et al. (1993) demonstrated that the Dll expression in the leg primordium requires wg activity at about 5 hours of development, roughly corresponding to the time when the rows of wg and dpp expression overlap. dpp has been shown to exert a strong organizing activity in the establishment of the embryonic dorsoventral axis and in patterning imaginal discs (Ferguson and Anderson, 1992; Zecca et al., 1995), but its role in imaginal disc induction needs to be further studied. The disc development continues after the overlap between Dll expression and the intersection of the wg and dpp rows are lost (Cohen et al., 1993), suggesting that the induced cells must activate an intrinsic determinant to irreversibly commit them to imaginal cell fate. Ventral leg and dorsal wing primordia appear to originate from a common imaginal primordium. The cell lineage tracing study has shown that in stage 12, the wing disc cells expressing vestigial (vg) segregate and move dorsally away from Dll-expressing cells (Cohen et al., 1993).

Although the intersection between wg- and dpp-expressing rows exist in all the trunk segments, wings and legs form only in the thorax. This was shown to be due to the negative regulation by homeotic genes. In the abdomen, genes in the bithorax complex repress leg and wing formation (Bate and Martinez Arias, 1991; Simcox et al., 1991; Vachon et al., 1992; Carroll et al., 1995) and in the first thoracic segment, wing formation was repressed by the Sex comb reduced gene in the Antennapedia complex (Carroll et al., 1995). In embryos mutant for Antenna-pedia, which is responsible for the identity of parasegment 4 and 5, formation of the leg and wing primordium was detectable (Mann, 1994; Carroll et al., 1995), suggesting that these appendages are formed as a default in the ‘ground state’ of segmental identity (Lewis, 1978). We must therefore seek a putative intrinsic determinant of imaginal disc formation outside the homeotic gene complex. The nuclear proteins, Dll and Vg are the earliest known markers for the leg and wing imaginal discs, and are required for pattern formation along the PD axis in the adult (Cohen and Jürgens, 1989; Cohen, 1990; Williams et al., 1991). However, their involvement in imaginal disc formation is not clear since imaginal discs are formed in the absence of Dll or vg (Williams et al., 1991; Cohen et al., 1993).

To identify an intrinsic determinant of the imaginal disc, we studied the function of two closely linked genes, escargot (snail) and snail (sna). esg and sna encode transcriptional regulators with similar C2H2 type zinc finger domains (76% amino acid identity; Boulay et al., 1987; Whiteley et al., 1992). esg is expressed in most imaginal primordia found in the embryo and in the larva (Whiteley et al., 1992; Hartenstein and Jan, 1992; Hayashi et al., 1993; Younossi-Hartenstein et al., 1993) and has been shown to be required for the maintenance of diploidy of some imaginal cells (Hayashi et al., 1993; Fuse et al., 1994; Hayashi, 1996). sna is initially expressed and required in the prospective mesoderm (Simpson, 1983; Grau et al., 1984; Alberga et al., 1991). In later stages, sna is expressed in wing, haltere and genital discs at the same stage when esg is expressed in these discs (Alberga et al., 1991). In this work, we studied the function of esg and sna in the early stage of wing disc development and report that esg and sna act as intrinsic determinants of the wing cell fate.

esg^G66B has a complete deletion of the esg coding region with P[engrailed (en)-lacZ w⁺] inserted into the esg locus (Whiteley et al., 1992; Kassis, 1994). sna¹ has a small deletion within the coding region and is genetically null (Grau et al., 1984; Boulay et al., 1987). The esg sna double mutant chromosome was made by recombination between esg^G66B and sna¹. y w; esg^G66B/sna¹FRT40 neo^R females were crossed with y w; Gla/CyO males. Recombinants were selected as w⁺ and neo^R progenies, and then tested for complementation with sna and esg alleles. For analyses of embryonic phenotype, we used esg^G66BFRT40/CyO actin-lacZ, sna¹FRT40/CyO and esg^G66Bsna¹FRT40/CyO stocks. esg mutant embryos were identified by the lack of staining with anti-Esg antibody. sna mutant and esg sna double mutant embryos were identified by the sna phenotype. Other stocks are described by Lindsley and Zimm (1992).

Glutathione S-transferase (GST)-Esg and GST-Sna fusion proteins were prepared from E. coli as described by Ip et al. (1992); Fuse et al. (1994). The GST-Sna protein was purified from the inclusion body. The prepared proteins were >50% pure as judged by a Coomasie Brilliant Blue staining after SDS-polyacrylamide gel electrophoresis. Gel mobility shift assay was performed as described by Fuse et al. (1994). The double strand oligomers, which were used as a ³²P-labeled probe and unlabeled competitors, were 5′-GCGGCC-N₁₀-TTTG-3′ with a 5′-TCGA-3′ overhang on each strand. Sequences for N₁₀ are indicated in Fig. 1C. In a 10 μl reaction mixture, the probe DNA (Fig. 1C closed circle, final 1 fmol/μl) and GST-Esg or GST-Sna (approximately 10 pg/μl) were incubated with or without various concentration of competitors (10, 50 and 100 fmol/μl). The probe bound to the protein was separated from the free probe by electrophoresis and quantitated by Fuji BAS2000. GST alone did not bind the probe DNA (data not shown).

Immunostaining was performed as described by Hayashi et al. (1993). For double-label immunostaining, Oregon R embryos were incubated with rat anti-Esg (Fuse et al., 1994) and rabbit anti-Sna (a gift from Rolf Reuter) antibodies. Subsequently, the embryos were incubated with biotin-conjugated goat anti-rat IgG antibody (Jackson lab) and then with FITC-conjugated streptavidin (Vector lab) and Cy3-conju-gated anti-rabbit IgG antibody (Chemicon). The embryos were observed with a Zeiss Axioplan microscope. For analyses of mutant phenotypes, embryos were fixed and incubated with rabbit anti-Sna antibody, biotin-conjugated goat anti-rabbit IgG antibody (Jackson lab) and ABC complex (Vector lab), and then developed in diaminobenzidine. Alternatively, embryos were incubated with rabbit anti-Vg antibody (Williams et al., 1991), biotin-conjugated goat anti-rabbit IgG antibody (Jackson lab) and FITC-conjugated streptavidin (Vector lab). Stained embryos were examined using a confocal micro-scope (LSM410, Carl Zeiss). For a simultaneous detection of β-galactosidase (β-gal) and D-α-catenin, embryos were incubated with rabbit anti-β-gal (Cappel) and rat anti-D-α-catenin monoclonal antibody (Oda et al., 1993), with biotin-conjugated goat anti-rat IgG antibody, and finally, with a mixture of Cy3-conjugated anti-rabbit IgG and FITC-conjugated streptavidin.

To make the HSsna construct, sna ORF flanked by XbaI and EcoRI sites was amplified by RT-PCR and inserted into the pCaSpeR-hs vector (Thummel and Pirrota, 1991). Transformants were established by the germ line P-transformation method (Rubin and Spradling, 1982). HSesg, HSΔzf (Fuse et al., 1994) or HSsna was introduced into the esg sna double mutant chromosome by recombination. Recombinants were checked for heat shock induced ubiquitous Esg or Sna expression (data not shown). To rescue the double mutant phenotype, 10- to 11.5-hour old embryos were given four or five heat shocks at 37 °C for 20 minutes every 90 minutes. After the last heat shock, the embryos were incubated at 25 °C for 70 minutes, and were stained with anti-Vg or anti-Sna antibody.

esg and sna encode transcriptional regulators with similar zinc finger domains (Boulay et al., 1987; Whiteley et al., 1992). Using recombinant proteins produced in E. coli, we compared DNA binding specificities of Esg and Sna. GST fusion proteins containing the zinc finger domain of each protein were produced and used for gel mobility shift assays (Ip et al., 1992; Fuse et al., 1994). Both proteins bound to ³²P-labeled DNA containing the E2 box (Fig. 1C closed circle). Addition of excess DNA containing the E2 box or its derivatives competed with this binding. Competition profiles of the five competitor DNAs (Fig. 1C) were indistinguishable for the two proteins (Fig. 1A,B), indicating that Esg and Sna have similar DNA binding specificities. Next, expression patterns of Esg and Sna were compared in embryos by double-label immunostaining (Fig. 2; Alberga et al., 1991; Whiteley et al., 1992). In early embryos, Esg expression starts in the dorsal side of the embryo whereas Sna is expressed in prospective mesoderm on the ventral side (Fig. 2A). In stage 13, the wing and haltere primordia that have segregated out from the leg primordia begin to express Esg and Sna (Fig. 2B). Expression of the two proteins in the wing and haltere primordia continues in stage 15 embryos (Fig. 2C) when they invaginate to form imaginal discs (Fig. 3F, data not shown) and Esg expression in most of the other tissues disappears. In these discs, a strict cell to cell correspondence of Esg and Sna expression was observed (Fig. 2D). In leg discs that express Esg, only a subset of the cells coexpressed Sna (Fig. 2E).

The expression of the two transcription factors with similar DNA binding specificities in these imaginal discs raised the possibility that Esg and Sna cooperate to play important roles in the early stage of wing and haltere development. To test this possibility, we compared the phenotypes of esg sna double mutant embryos with those of esg and sna single mutant embryos. We describe below the situation in the wing disc, but exactly the same observations were made for the haltere disc. We used two null mutations, esg^G66B and sna¹ for this study (Materials and Methods). Since sna¹ has a small deletion within the coding region (Grau et al., 1984; Boulay et al., 1987), a truncated Sna protein, which does not function, is detectable in sna¹ mutant embryos by anti-Sna antibody (Figs 3C, 6E). We used Sna expression as a marker for the wing disc. In esg mutant embryos, the wing disc expressed Sna normally (Fig. 3B). The sna¹ mutation results in the malformation of the whole body due to the requirement of sna for mesoderm development. But the truncated Sna protein was expressed in invaginated wing discs (Fig. 3C). The number of Sna-expressing cells in the imaginal discs appears to be lower than that in the control and esg mutant embryos (see also Fig. 6E). It is not clear whether this is caused by the sna mutation itself or by the secondary effect of body malformation. Such Sna expression was abolished in esg sna double mutant embryos (Fig. 3D). esg^G66Bsna¹ placed in trans to the deficiency Df(2L)A48 (Ashburner et al., 1982) results in the same phenotype as esg^G66Bsna¹ homozygotes (data not shown), indicating that no other recessive mutation contributes to the phenotypes. In the embryos which have only one dose of the wild-type esg or sna gene (esg^G66Bsna¹/sna¹ or esg^G66Bsna¹/esg^G66B, respectively), the Sna marker was expressed in wing discs (data not shown). These results indicate that both esg and sna are required for proper Sna expression in wing discs and that the loss of one of the genes can be compensated by one copy of the other gene. To confirm the double mutant phenotype, we also examined the expression of two additional markers for wing disc. The second wing disc marker, Vg was expressed in the wing disc in esg or sna mutant embryos (Fig. 4A,B) as well as in the control embryos (data not shown). However, in esg sna double mutants, the Vg expression was lost (Fig. 4C). esg^G66B used in this study has a complete deletion of the esg coding region with P[en-lacZ] inserted into the esg locus (Whiteley et al., 1992; Kassis, 1994). Since en-lacZ is regulated by endogenous esg enhancer, in esg^G66B embryos, β-gal was expressed in the wing disc as well as in leg discs, anterior spiracle, tracheal pits and most of the other tissues which normally express esg (Whiteley et al., 1992; Fig. 5B). Therefore, we would like to designate the P[en-lacZ] inserted into the esg locus as esg-lacZ, and use the expression of esg-lacZ as a third marker for the wing disc. In both esg^G66B and esg^G66Bsna¹/sna¹ embryos, esg-lacZ were expressed in the invaginated wing disc as well as in the control embryo (Fig. 5A-C). Similarly to the above observations, such esg-lacZ expression was undetectable in esg^G66Bsna¹ double mutant embryos (Fig. 5D). Thus, the expression of the three wing disc markers were abolished in esg sna double mutant embryos, while they were detected in both the single mutants, indicating that overlapping activities of esg and sna are required for proper wing disc development.

In wild-type embryos, after the onset of esg and sna expression, the wing primordial cells constrict their apical surface, and then invaginate basally to form the disc structure (Bate and Martinez Arias, 1991; Fig. 3F). To determine whether the wing primordium invaginates normally in esg sna double mutant embryo, we examined embryos double-labeled with anti-D-α-catenin and anti-β-gal antibodies. β-gal expressed from the esg-lacZ was distributed uniformly within the cell body, and D-α-catenin was localized in the adherence junction along the apical circumference of the cells (Oda et al., 1993; Fig. 5A). Apical constrictions of wing primordium cells were detected as condensed yellow staining (Fig. 5 arrowhead), and the invaginated cell bodies were seen in esg^G66B and esg^G66Bsna¹/sna¹ mutant embryos (Fig. 5B,C) as well as in the control embryos (Fig. 5A,E). However, such apical constriction did not occur in esg^G66Bsna¹ double mutant embryos (Fig. 5D,F). These results indicate that the wing disc is absent in the double mutant, and that esg and sna are required for the wingspecific cell shape change.

In contrast to the situation in the wing disc, leg discs formed normally in esg sna double mutant embryo as well as in both single mutants (Fig. 5G,H; data not shown). This result indicates that the genetic interaction between esg and sna observed in the wing disc does not occur in the leg disc where expression of the two genes overlaps only in a few cells (Fig. 2E).

To further verify the role of esg and sna in wing development, we examined whether Esg or Sna supplied from transgenes can rescue the esg sna double mutant phenotype. We introduced hsp70-esg (HSesg; Fuse et al., 1994) and hsp70-sna (HSsna) into the double mutant chromosome. The ubiquitous Esg or Sna expression induced by heat shock restored Vg expression in the wing primordium of the double mutant (Fig. 4E,F). Such rescues were dependent on heat shock treatment (data not shown), and the heat shock alone without the transgenes did not have such an effect (Fig. 4D). The ubiquitous expression of Esg also restored the Sna expression in the wing primordium (Fig. 3E). Such cells partially invaginated to form wing discs (Fig. 3G). Among 76 heat shocked esg^G66Bsna¹HSesg embryos stained with anti-Sna, 39 (51%) had restored Sna expression in at least one wing or haltere primordium. In 24 of them (32%), the Sna-positive cells were found to be invaginated. As a control, esg^G66Bsna¹HSΔzf (zinc finger domain deleted esg derivative; Fuse et al., 1994) embryos were similarly treated. None of them (n=58) expressed Sna in the wing primordium (data not shown). These results demonstrate that esg and sna have overlapping functions that are essential for wing disc formation.

The response to ubiquitous Esg and Sna expression were confined to cells in a small region where the wing disc normally forms. This suggests that such cells have already acquired a potential to respond to Esg and Sna. We therefore examined the mutant embryos when Sna expression was first detectable in the wing primordium. In stage 13 embryos, Sna expression was detectable in the wing primordium within the ectoderm in the control, sna and esg mutant embryos (Fig. 6A,B, data not shown). The level of Sna expression varied from cell to cell. The Sna expression continued and became stronger and more uniform during stage 15 (Fig. 6D,E, data not shown). However, in esg sna double mutants, the initial Sna expression was detected in stage 13 (Fig. 6C), but was not detectable in stage 15 when the wing primordium normally would have completed invagination (Fig. 6F). To determine whether the loss of Sna expression is due to the death of the wing primordium, we examined cell death by the in situ nick translation method (Hay et al., 1994). In the double mutant embryo, no sign of cell death was detectable in the prospective wing disc, although massive cell death in the endoderm due to the sna mutation was clearly visible (data not shown). These results suggest that in the double mutant, the wing primordium begins initial development until stage 13, but fails to maintain the fate in later stages.

Esg and Sna are also coexpressed in the genital disc (Fig. 2C). Sna expression in the genital disc was lost in the esg sna double mutant as in the case of the wing disc (Fig. 3D) and was restored by repeated induction of HSesg (data not shown). These results suggest that esg and sna are required for the proper development of the genital disc as well. However, at present, we do not know which step of genital disc development is regulated by the two genes.

We have shown that esg and sna are required for early wing development. When both of the genes were absent, the wing imaginal disc failed to develop. One copy of either of the genes was sufficient for the wing primordium to express the marker genes and to invaginate. The ubiquitous expression of either of the genes was able to restore the marker gene expression and cell invagination in the wing primordium. Thus, the functions of Esg and Sna in wing disc development are interchangeable. Similar DNA binding specificities of the two proteins suggests that they control the same set of genes required for wing development.

Overlapping or ‘redundant’ activity of related basic helixloop-helix transcription factors during neurogenesis and myogenesis have been reported (reviewed by Jan and Jan, 1993). Given the result of this study, other aspects of development are also likely to be regulated by overlapping activity of multiple transcription factors belonging to the same family. Therefore results from genetic experiment in which a single gene is disrupted should be interpreted with caution.

We have shown that in the esg sna double mutant, Sna expression initiated normally. We also observed that Vg was transiently expressed in wing primordium of the double mutant (our unpublished observation). Although we were unable to detect such transient expression of esg-lacZ in the double mutant, probably due to the low level of the lacZ expression, it is very likely that the initial stage of wing development proceeded normally in the absence of esg and sna until stage 13. However, the following three points demonstrate that wing development is discontinued in later stages. First, the expression of the three marker genes (sna, vg and esg-lacZ) was abolished. Second, the apical constriction of the primordium cells was not observed. Third, the cell death of the primordium did not occur. These observations support the idea that the absence of esg and sna allows wing primordium to transform into larval epidermis. Larval epidermal cells have a characteristic microvilli structure over their apical surface and secrete cuticle in late stage embryos (Martinez Arias, 1993). We did not examine these properties of the larval epidermis in wing primordia of double mutant embryos because we were unable to positively identify the primordium after it lost the marker gene expression.

Our study revealed the interesting regulatory mechanisms of Esg and Sna expression in the wing primordium. The Sna expression seen in the esg mutants was not seen in the esg sna double mutants (Fig. 3B,D), indicating that Sna expression is activated by sna itself. Similarly, Sna expression seen in the sna mutants was not seen when the esg activity was additionally removed (Fig. 3C,D), indicating that esg activates Sna expression. The activation of Sna by esg and sna itself are likely to account for the accumulation of Sna in stage 15 embryos (Fig. 6D). Furthermore, it is likely that the expression of Esg is also regulated by similar mechanisms. The esg-lacZ expression seen in the esg mutant embryo was not seen in the esg sna double mutant (Fig. 5B,D), indicating that sna activates the esg enhancer. Similarly, the esg-lacZ expression seen in esg^G66Bsna¹/sna¹ embryos was not seen when one copy of wild-type esg was additionally removed (Fig. 5C,D), indicating that the esg enhancer is activated by esg itself. Since the esg-lacZ expression appears to reflect endogenous esg expression, these results suggest that Esg expression in the wing disc is regulated by esg itself and sna. Taken together, our observations suggest that Esg and Sna expression in wing discs is maintained by their auto- and crossactivation. Sna expression in prospective mesoderm is also regulated by sna itself in the early embryo (Ray et al., 1991). Autoactivation was found for several genes such as fushi-tarazu (Hiromi and Gehring, 1987), even-skipped (Frasch et al., 1988), Ultra-bithorax (Bienz and Tremml, 1988) and en (DiNardo et al., 1988), and was implicated for stable, sometimes heritable maintenance of gene activity. In the case of esg and sna, the crossactivation by each other in addition to the autoactivation is likely to give an additional advantage to maintain their stable expression.

The finding that Esg and Sna activate gene expression is in sharp contrast to the previous findings that Sna functions as a direct repressor of rhomboid transcription in the embryo (Ip et al., 1992; Gray et al., 1994), and Esg acts as a repressor of transcription in tissue culture cells (Fuse et al., 1994). One explanation may be that the function of Esg and Sna is to repress genes that determine larval character, thereby allowing expression of wing disc-specific genes. In such cases, the roles of Esg and Sna in promoting wing disc development are indirect and permissive. However, in the heat shock rescue experiment, the level of Esg or Sna expression that was sufficient to rescue the esg sna double mutant phenotype did not have adverse effects on the development of larval tissue (Figs 3 and 4). Therefore, if Esg and Sna function by repressing larval cell fate, such a function must be restricted to the cells that were induced to differentiate as wing imaginal disc. Thus, the putative wing inductive signal has two effects. One promotes wing differentiation and the other represses larval development. We consider this explanation complicated and less likely. An alternative, more simple explanation is that Esg and Sna function both as an activator and as a repressor, depending upon the target promoter context. For example, a cellular coactivator protein may mediate transcriptional activation by an otherwise silent DNA binding protein. Another possibility is that Esg and Sna interfere with a transcriptional repressor, thereby allowing transcriptional activation by an activator binding at a different site. A case of a single protein that functions both as an activator and as a repressor of transcription has been reported for the C2H2 type zinc finger transcription factor Krüppel (Kr). In tissue culture cells, a low concentration of Kr activates a target promoter containing a Kr binding site and at high concentration, Kr repress the same promoter (Sauer and Jäckel, 1991). Since our arguments are based solely on genetic data, any interaction involving Esg and Sna could be indirect. To resolve this issue, a molecular analysis of interaction between Esg, Sna and their target promoters is necessary. Analysis of the esg or sna enhancer responsible for expression in the imaginal discs is a key to understanding the function of Esg and Sna at a molecular level.

We propose that the wing disc formation can be separated into two steps. The first step is determination by an extrinsic signal which induces vg, esg and sna transcription (Fig. 7A). Such an inducer could be the combined activity of Dpp and Wg or a transient intrinsic gene activity induced by an external signal. Since the transcriptions are dependent on an external signal source, the determination is not fixed in this stage. In the second step, esg and sna initiate an intrinsic program of auto- and crossactivations to stabilize their own expression (Fig. 7B). It should be noted that the ubiquitous esg or sna expression restored wing disc formation only in the wing primordium of the double mutant. This point highlights the second role of the putative inductive signal in establishing the competence of the wing primordium to respond to esg and sna. Thus, only those cells that transiently express esg and sna, and have acquired the competence can turn on the auto- and crossregulatory circuits of Esg and Sna, and enables further wing development. We speculate that this two-fold restriction is important for defining a sharp border between imaginal and larval cells. Esg and Sna then control subordinate genes essential for wing development, such as vg (Fig. 7B). Since the wing cell fate was not maintained in the esg sna double mutant, the auto- and crossactivations by esg and sna are likely to be responsible for irreversible and autonomous fate commitment of the wing primordium. The wing primordium begins to express vg in stage 12 before the onset of esg and sna expression (Williams et al., 1991; Kassis, 1994; our unpublished observations). It is likely that the expression of vg is independent of esg and sna in step 1 (Fig. 7A). Subsequently, vg expression falls under the control of esg and sna in step 2 (Fig. 7B).

esg and sna are coexpressed and are required for Sna expression in the genital disc. Although we did not examine expression of other markers in the genital disc, it is possible that esg and sna are also involved in the fate maintenance of the genital disc. Esg is also expressed in other imaginal cells such as eye-antennal, labial, leg and humeral imaginal discs and abdominal histoblasts in the embryo (Whiteley et al., 1992; Hartenstein and Jan., 1992; Hayashi et al., 1993; Younossi-Hartenstein et al., 1993). The role of esg in the development of these cells during embryogenesis is not clear because these cells were apparent in strong esg mutant embryos (Hayashi et al., 1993; Fig. 5G). Given the result of this study, it is conceivable that another gene with a role similar to sna in the wing disc is active in these imaginal cells to supplement the function of esg. Such a gene may be found in several copies of sequences homologous to the snail-type zinc finger domain identified in Drosophila (Whiteley et al., 1992; Roark et al., 1995). esg has another function during the larval stage of imaginal development. Our previous results indicate that esg expressed in imaginal cells maintains diploidy by inhibiting endoreplication, while larval cells that do not express esg repeat endoreplication to become polyploid. Upon reduction of esg activity, some imaginal cells entered endocycle, became polyploid, and resembled larval epidermal cells (Hayashi et al., 1993; Fuse et al., 1994; our unpublished observation). Esg was shown to positively regulate the cell cycle regulator Cdc2, which in turn inhibits endoreplication (Hayashi, 1996). Thus, esg has multiple functions during imaginal development, and each is related to the maintenance of imaginal cell identity.

We thank Rolf Reuter, Tadashi Uemura and Sean Carroll for antibodies, Michael Ashburner, Judith Kassis and Bloomington Stock Center for fly stocks, Michael Levine and Carl Thummel for plasmid DNA. We are grateful to Jun-ichi Tomizawa, Sonia Chalfin, John Wakeley, Lee Ann Younger, Jeffery Porter and Philip Beachy for critically reading the manuscript. N. Fuse is grateful to all members in the laboratories and colleagues in the Institute for suggestions and discussions. N. Fuse was the recipient of the postdoctoral fellowship from the Japan Society for the Promotion of Science. This work was supported by the Ministry of Education, Science and Culture of Japan.