homothorax (hth) is a Drosophila member of the Meis family of homeobox genes. hth function is required for the nuclear localization of the Hox cofactor Extradenticle (EXD). We show here that there is also a post-transcriptional control of HTH by exd: exd activity is required for the apparent stability of the HTH protein. In leg imaginal discs, hth expression is limited to the domain of exd function and this domain is complementary to the domain in which the Wingless (WG) and Decapentaplegic (DPP) signals are active. We demonstrate that WG and DPP act together through their targets Distal-less (Dll) and dachshund (dac) to restrict hth expression, and therefore EXD’s nuclear localization, to the most proximal regions of the leg disc. Furthermore, there is a reciprocal repression exerted by HTH on these and other DPP and WG downstream targets that restricts their expression to non-hth-expressing cells. Thus, there exists in the leg disc a set of mutually antagonistic interactions between proximal cells, which we define as those that express hth, and distal cells, or those that do not express hth. In addition, we show that dac negatively regulates Dll. We suggest that these antagonistic relationships help to convert the WG and DPP activity gradients into discreet domains of gene expression along the proximodistal axis.
The hedgehog (hh)-signaling pathway is required to promote limb development in mammals and flies (Basler and Struhl, 1994; Tabata and Kornberg, 1994; Chiang et al., 1996). In Drosophila, removal of hh function during imaginal development results in severe limb truncations because the cells that are fated to form adult appendages fail to divide (Basler and Struhl, 1994; Tabata and Kornberg, 1994). The major role of hh in the developing leg disc is to activate the transcription of two genes that encode signaling molecules, wingless (wg) and decapentaplegic (dpp) (Basler and Struhl, 1994; Diaz-Benjumea et al., 1994; Tabata and Kornberg, 1994). WG, a WNT family protein and DPP, a TGF-β homologue, act in concert to initiate the transcription of target genes necessary for limb development, such as Dll and dac, and to promote cell division and growth in the disc (Diaz-Benjumea et al., 1994; Lecuit and Cohen, 1997). In Dll hypomorphs, distal leg fates do not form and leg development is aborted (Cohen and Jurgens, 1989). Similarly, dac null flies display deletions and fusions of intermediate leg fates along the proximodistal (P/D) axis (Mardon et al., 1994).
Initial studies of the PBC gene exd focused on its role as a homeotic cofactor in embryonic development (reviewed in Mann and Chan, 1996). Clonal analysis of exd in adult leg development has shown its function to be limited to proximal regions, despite its ubiquitous transcription (Gonzalez-Crespo and Morata, 1995; Rauskolb et al., 1995). Immunocytochemical analysis of EXD protein distribution demonstrated that EXD is regulated at the level of its subcellular localization, such that it is only nuclear in a proximal domain in leg discs and cytoplasmic in distal cells (Mann and Abu-Shaar, 1996; Aspland and White, 1997). Further studies demonstrated that this nuclear localization depends on the presence of a homeodomain protein of the Meis family, Homothorax (HTH) (Rieckhof et al., 1997; Kurant et al., 1998; Pai et al., 1998). hth expression is restricted to a proximal domain in the leg disc which is coincident with nuclear EXD and corresponds to the regions that require exd function during development (Rieckhof et al., 1997).
Intriguingly, a proposition suggested by Snodgrass in 1935 concerning the evolution of the arthropod leg stated that the leg is subdivided into a proximal region (the coxopodite), which arose as an extension of the body wall, and a distal region (the telopodite), which represents the leg proper (Snodgrass, 1935). This hypothesis was supported in molecular terms by two recent studies (Gonzalez-Crespo and Morata, 1996; Casares and Mann, 1998). The earlier study showed that leg discs from which hh function is reduced during and after the second larval instar stage are unaffected proximally, or in the domain where EXD is nuclear, but have no Dll expression and are impaired in growth distally. These results suggest that the proximal and distal domains of the leg have distinct requirements for hh signaling during development. Similarly, reducing the function of wg or dpp after embryogenesis aborts leg development up to the proximal femur, but the most proximal fates remain unaffected (Diaz-Benjumea et al., 1994). These data suggest that, along the P/D axis, the requirement for dpp and wg signaling is much greater for the growth of distal and intermediate cells than it is for proximal cells. Moreover, misexpression of high levels of EXD, which overcome its nuclear import blockade (Gonzalez-Crespo and Morata, 1996), or misexpression of HTH (Casares and Mann, 1998) in the distal portion of the leg result in severe leg truncations similar to those seen upon the reduction of hh, wg or dpp function. One interpretation of these observations is that hh signaling and exd/hth function are incompatible with each other and that forcing exd or hth function in the distal domain interferes with hh signaling.
In this study, we extend the results of Gonzalez-Crespo et al. (1998) and explore the mutually antagonistic relationships between exd and hth in the proximal leg and the downstream events of hh signaling in the distal leg. We show that wg and dpp do not activate some of their target genes in the proximal domain, where hth is expressed. Further, we characterize the effects of loss of function and misexpression of hth on WG and DPP target gene expression. We also determine the effects of ectopic WG and DPP-signaling activity on hth expression and EXD’s subcellular distribution, as well as the effects on hth and EXD of removing the WG and DPP target genes Dll and dac. Our data suggest that there exists a mutually repressive relationship between the coxopodite genes hth and exd and the WG and DPP-signaling pathways mediated through their targets Dll and dac.
MATERIALS AND METHODS
Immunocytochemistry was carried out using antibodies previously described: Rabbit anti-EXD (Mann and Abu-Shaar, 1996), chicken anti-HTH (Casares and Mann, 1998), mouse anti-DLL (Cohen et al., 1993), mouse anti-DAC (Mardon et al., 1994), mouse anti-CD2 (Serotec), mouse anti-β-galactosidase (Promega) and rabbit anti-β-galactosidase (Cappel). Secondary antibodies coupled to FITC, Texas Red and Cy5 (all Jackson Immunoresearch) were used and imaging was carried out with Biorad MRC600 or MRC1024 confocal microscopes.
exdXP11 clones were generated using the FRT/FLP system 48-72 hours AEL with the hthP6 allele in the background. dac3 clones were generated by heat shocking 48-72 hour larvae of the genotype hsFLP; dac3FRT40A/armadillo-lacZ FRT40A. DllSA1 clones were generated from larvae of the genotype hsFLP; FRT42 DllSA1/FRT42 armadillo-lacZ M. Although Dll− clones usually sort into the hth domain, by using the Minute technique we observed several clones that did not sort and all of these derepressed hth. hth clones were derived from larvae of the genotypes hsFLP omb-lacZ/+; FRT82B ScrC1AntpNS+RC3hthP2/FRT82B CD2 M and hsFLP; H15/+; FRT82B ScrC1AntpNS+RC3hthP2/FRT82B CD2 M.
Ectopic GFP-hth-expressing clones were generated by heat shocking larvae of the genotype hsFLP actin>CD2>Gal4; UAS-GFP-hth/TM3 UAS y+ or hsFLP actin>CD2>Gal4; H15/+; UAS-GFP-hth. Ectopic MYC-MEIS-expressing clones were generated by heat shocking larvae of the genotype hsFLP omb-lacZ/hs FLP; armadillo-Gal4/+; UAS>CD2 y+> Myc-Meis/+ and hsFLP; UAS>CD2 y+>Myc-Meis/Cyo; C10/TM2.
UAS-NLS-lacZ-exd was made by using PCR to generate a version of the exd ORF with an NdeI site at the 5′ ATG, in frame with the NdeI site at the 3′ end of the NLS-lacZ gene (Riddihough and Ish-Horowicz, 1991). After fusing the Nde-exd ORF to the NLS-lacZ ORF, the chimera was placed into pUAST (Brand and Perrimon, 1993).
Larvae of the genotypes hsFLP; tsh-Gal4/UAS>CD2 y+>Nrt-flu-wg and hsFLP; tsh-Gal 4/+; UAS>CD2 y+>tkvQD/hthP6 were heat shocked 24-48 hours prior to dissection to generate ectopic tethered WG and TKVQD-expressing clones. tsh-Gal4 and Nrt-flu-wg have been described previously (Calleja et al., 1996; Zecca et al., 1996).
dac null discs were generated by crossing flies with the dac1 allele to dac3FRT40A flies. Transheterozygote larvae were identified by the reduction in the size of the eye imaginal discs (Mardon et al., 1994).
A mutual requirement for EXD and HTH
In the leg disc, hth is expressed in a peripheral ring of cells corresponding to the most proximal cell fates (Rieckhof et al., 1997; Casares and Mann, 1998) while the exd transcript is present throughout the disc (Gonzalez-Crespo and Morata, 1995; Rauskolb et al., 1995). In contrast to its transcription, the subcellular localization of EXD protein is regulated such that it is nuclear in those cells that express hth and cytoplasmic elsewhere (Fig. 1A,B). We showed previously that HTH is both necessary and sufficient for EXD’s nuclear localization (Rieckhof et al., 1997; Casares and Mann, 1998). That is, removal of hth from embryos or imaginal discs resulted in the cytoplasmic localization of EXD in places where it is normally nuclear, and ectopic expression of HTH or its mammalian homolog MEIS-1B resulted in the nuclear localization of EXD in places where it is normally cytoplasmic (Fig. 1C-E).
To test if nuclear HTH requires exd, we examined the localization of HTH protein in mitotic clones mutant for exd. Strikingly, we were unable to detect HTH protein in exd− clones (Fig. 1F-H). The lack of HTH protein in these clones could be because exd is required for hth transcription. To determine whether the lack of HTH in exd− clones was a result of transcriptional or post-transcriptional regulation, we examined the expression of the hth enhancer trap hthP6. In contrast to HTH protein, lacZ expression from hthP6 is maintained (Fig. 1F,I) and in many cases upregulated (data not shown) in exd− clones. This suggests that the loss of HTH protein occurs post-transcriptionally, perhaps by protein degradation. Thus, the activities of hth and exd are intimately associated with each other: removing hth function results in cytoplasmic and presumably non-functional EXD, and removing exd function results in the loss of detectable HTH protein.
Defining domains of gene expression in the leg
We examined the expression of several targets of the signaling molecules WG and DPP in relation to the hth expression domain. dpp expression in the leg disc at the early third larval instar stage consists of a sector that originates at the center of the disc, extends dorsally to the periphery and shows extensive overlap with HTH (Fig. 2A) (Masucci et al., 1990). omb, a target of the DPP-signaling pathway (Brook and Cohen, 1996; Grimm and Pflugfelder, 1996; Lecuit et al., 1996; Nellen et al., 1996), is expressed in a dorsal sector that, in contrast to dpp, extends dorsally only to abut, but not overlap with, the hth domain (Fig. 2B). wg expression consists of a ventral sector of cells that extends from the center to the periphery of the disc (Fig. 2C), whereas H15, an enhancer trap line that requires wg signaling for its activation (Brook and Cohen, 1996), is largely not transcribed in the hth domain (Fig. 2D). The restriction of these WG and DPP target genes to non-hth-expressing cells suggests that hth restricts signaling by these two molecules. By the late third larval instar stage, there is a small degree of overlap between hth and omb expression as well as between hth and H15 (Fig. 2D). This expression corresponds to the trochanter domain where gene activation can occur independently of the WG- and DPP-signaling pathways (Diaz-Benjumea et al., 1994).
Unlike omb and H15, the Dll and dac genes require input from both the DPP and WG signal transduction pathways to be activated in leg discs (Lecuit and Cohen, 1997). Dll encodes a homeodomain protein present in the central portion of leg discs, and its activation requires the highest concentrations of WG and DPP. dac encodes a nuclear protein and a putative transcription factor whose expression is repressed by high concentrations and activated by intermediate concentrations of WG and DPP (Lecuit and Cohen, 1997).
By performing triple stains for the dacP-lacZ reporter gene, and DLL and HTH proteins at the early third larval instar stage (approximately 72 hours AEL), we found that the leg disc is defined by three non-overlapping domains of gene expression (Fig. 3A,F). The distal-most domain of the leg disc contains DLL protein (the Dll domain). Dorsal and dorsolateral, but not ventral, to the Dll domain are cells that express dac (the dac domain) (Fig. 3A,F). The proximal-most cells of the disc, which surround the dac and Dll domains, express hth (the hth domain). At the mid 3rd larval instar stage (∼96 hours AEL), the distal-most cells express only Dll and are surrounded by a ring of cells that express both Dll and dac (Fig. 3B,G). Also at this stage, there is a dorsal patch of cells that express dac but not Dll (Fig. 3B,G). hth expression remains limited to the proximal-most cells of the disc and shows no overlap with dac or Dll. By the late 3rd larval instar stage (∼120 hours AEL), hth is still not co-expressed with dac or Dll, with the exception of a thin band of cells corresponding to the trochanter domain, where all three genes are co-expressed (Fig. 3C,H). Gene expression in the trochanter domain is likely to represent secondary patterning events, because it is not dependent on WG- or DPP-signaling (Diaz-Benjumea et al., 1994; Lecuit et al., 1996). Also at this stage dac expression surrounds and partially overlaps the Dll expression domain (Fig. 3C,H).
We propose that the Dll and dac domains, where hth transcription is off and EXD is cytoplasmic, are DPP- and/or WG-responsive domains, as demonstrated by the ability of these cells to respond to these signals by activating the target genes Dll, dac, omb and H15. In contrast, the hth domain, where hth is active and EXD is nuclear, is a WG- and/or DPP-non-responsive domain, where these signals are present but cannot activate these targets (Gonzalez-Crespo and Morata, 1996; Gonzalez-Crespo et al., 1998). This model is tested below.
HTH limits WG and DPP signaling
As described above, expression of the DPP and WG targets omb and H15 is restricted to those cells that do not express HTH. To determine if hth inhibits target gene activation by DPP and WG, we either removed hth from its endogenous domain or misexpressed a GFP-HTH fusion protein or the murine hth homolog MEIS-1B in the distal portion of the leg disc.
Removing hth function resulted in the expansion of wg and dpp target gene expression. Dorsally situated hth− clones resulted in the expansion of omb expression, as marked by the omb-lacZ reporter gene (Fig. 4A-C). As expected from this result, exd− clones in the dorsal leg also derepress omb (Gonzalez-Crespo et al., 1998). Ventrally situated hth− clones showed derepression of H15 (Fig. 4D-F). For both omb and H15, the ectopic expression domains were continuous with their normal domains of expression, suggesting they still depend on dpp and wg, respectively. hth− clones outside the normal DPP or WG response sectors had no effect on these targets (data not shown).
Conversely, randomly generated clones of cells expressing a GFP-HTH fusion protein or a MYC epitope-tagged MEIS-1B protein repressed omb-lacZ (Fig. 5A-C) and H15-lacZ (Fig. 5D-F). Thus, we conclude that the presence of HTH is incompatible with DPP and/or WG activation of these, and perhaps other, target genes.
We also determined if hth repressed Dll and dac. Similar to removing exd function (Gonzalez-Crespo et al., 1998), when we examined hth loss-of-function clones, we found that dac was only partially derepressed, and that derepression was more likely to occur in clones that arise near the endogenous dac expression domain (Fig. 4G-I and data not shown). hth− clones had no effect on Dll expression, regardless of where they were situated. However, when we generated clones of GFP-HTH- or MYC-MEIS-expressing cells, we found that both Dll and dac could be repressed (Fig. 5G-L). These results suggest that the expression of Dll and dac requires two conditions to be satisfied: first, the absence of HTH and second, sufficient activities of the DPP and WG pathways. Interestingly, misexpression of a constitutively nuclear form of EXD (containing the exd ORF fused to a nuclear lacZ gene at its 5′ end), or high levels of wild-type EXD are unable to repress Dll (Gonzalez-Crespo et al., 1998; and data not shown). Thus, by these criteria, hth is more active than EXD, alone, suggesting that HTH does more than simply translocate EXD into the nucleus.
High levels of WG and DPP signaling repress the nuclear localization of EXD by repressing hth transcription
The direct action of both the WG- and DPP-signaling pathways is required to specify cell fates along the P/D axis (Lecuit and Cohen, 1997). High levels of WG and DPP signaling are required to activate Dll, a determinant of distal cell fates, and to repress expression of dac, a determinant of intermediate fates along the P/D axis. At intermediate levels of WG and DPP signaling, dac, but not Dll, is activated. The distal edge of hth expression coincides with the proximal edge of dac expression, suggesting that the threshold of DPP and WG signaling required to activate dac is similar to that required to repress hth. To test this idea, we elevated WG or DPP signaling in the hth expression domain by generating clones of cells that express either a membrane-tethered form of WG (Zecca et al., 1996) or an activated DPP receptor, ThickveinsQD (TKVQD) (Lecuit et al., 1996; Nellen et al., 1996). We expressed these proteins using a combination of the ‘flip-out’ and GAL4 methods, using a teashirt (tsh)-GAL4 driver line (see Methods). Because tsh is a proximally expressed gene with a nearly identical expression pattern to hth (Gonzalez-Crespo and Morata, 1996), this method generated clones that were primarily, if not exclusively, in the hth domain.
When we generated WG-expressing clones dorsally, where endogenous WG levels are low but where DPP is present at high concentrations, we saw a loss of HTH protein and a shift of EXD protein to the cytoplasm (Fig. 6A-E). When we examined lateral or ventral clones of WG-expressing cells, there was no effect on HTH or EXD (Fig. 6A,F-I). This suggests that sufficient levels of both WG and DPP signaling are required to repress HTH.
Consistent with this explanation were the behavior of clones expressing TKVQD (Fig. 6J-R). In this experiment, we observed repression of HTH ventrally, where endogenous levels of WG are high but where DPP levels are presumably limiting (Fig. 6J,O-R). Misexpression of TKVQD had no effect on EXD or HTH when the clones were situated dorsally or dorsolaterally, where the endogenous WG concentration is presumably limiting (Fig. 6J-N). The effects of both ectopic WG and TKVQD presented here are in agreement with those on the localization of EXD protein described elsewhere (Gonzalez-Crespo et al., 1998).
To determine if the absence of HTH protein was due to repression of hth transcription or due to a post-transcriptional mechanism, we examined the effect of ectopic TKVQD expression on the hth enhancer trap (hthP6). When we generated ventral clones of TKVQD-expressing cells, we found that, like HTH, lacZ expression from hthP6 is repressed (Fig. 7A-C). Thus, high levels of WG and DPP signaling affect HTH and EXD, at least in part, by repressing hth transcription.
dac and Dll mediate WG and DPP mediated repression of hth
The demonstration that WG and DPP signaling repressed hth transcription and EXD’s nuclear localization was surprising, because these two signaling molecules induce EXD’s nuclear localization in the endoderm of the embryonic midgut (Mann and Abu-Shaar, 1996). We investigated the possibility that the repression of hth by WG and DPP was indirect and perhaps mediated by dac and Dll, which are not expressed in the midgut. As described above, we generated TKVQD-expressing clones and examined HTH, DLL and DAC. Although we were unable to see a complete correlation between the absence of HTH and the activation of DAC alone or DLL alone (Fig. 7D-K), after examining a large number of clones (>60), we were able to account for every cell in which hth was repressed by the presence of DLL or DAC protein (Fig. 7L-O). These results support the idea that WG and DPP’s ability to repress hth is indirectly mediated by Dll and dac.
To verify that this was a normal function of DAC and DLL in leg discs, we generated loss of function clones of Dll and dac. When we generated Dll– clones before ∼72 hours of development, we found that hth was derepressed and EXD was nuclear (Fig. 8A-D) (Gonzalez-Crespo et al., 1998). However, clones generated after ∼72 hours had no effect on hth or EXD (data not shown), suggesting that there is an alternative mechanism for maintaining hth repression. When we generated early Minute+ clones that eliminate Dll activity from most of the disc, we found that the remaining cells express either HTH or DAC proteins, but not both (Fig. 8E-H). Thus, like DLL, DAC appears to have the capacity to repress hth.
The ability of DAC to repress hth expression was confirmed by generating dac– clones. These clones were primarily observed in the dorsal region of the dac domain, where dac expression does not overlap with Dll (see Fig. 3). These clones fell into four classes, based upon their location (schematized in Fig. 9H). First, dac– clones in the hth domain (class 1), where dac is not expressed, had no affect (Fig. 9D-G). Class 2 clones are those that were located dorsally in the dac domain, close to the hth domain. These clones displayed ectopic HTH (Fig. 9A-C) and nuclear EXD (data not shown), confirming that dac contributes to hth repression in this region of the leg disc. The other two categories of clones were unexpected: class 4 clones resulted in a derepression of Dll but had no effect on hth (Fig. 9D-G). These clones were located just dorsal to the Dll domain and suggest that dac normally represses Dll in these cells. Class 3 clones did not derepress Dll or hth (Fig. 9D-G). In general, these clones were located medially, in between class 2 and class 4 clones. These dac− clones suggest that there might be other regulators of hth in addition to dac and Dll (see Discussion).
Completely removing dac function results in viable animals that have deletions along the P/D axes of their legs (Mardon et al., 1994). We therefore examined the expression patterns of HTH and DLL in leg discs that were entirely devoid of dac function. In these discs, the hth domain appears expanded distally and the Dll domain appears to be expanded proximally, consistent with the idea that dac normally represses both hth and Dll (Fig. 9I). However, in these dac− discs, we also observed a dorsal-medial group of cells that did not express either Dll or hth (Fig. 9l). We suggest that these cells represent the same cells that generated the class 3 dac− clones (see above).
EXD and HTH are interdependent partner proteins
In the first part of this study, we establish that both HTH and EXD proteins require each other for their stable nuclear localization. In all imaginal discs that we have examined, EXD is nuclear only when HTH is present and is cytoplasmic in the absence of HTH. Conversely, the stability of HTH protein requires EXD. The mutual requirement for both EXD and HTH for their nuclear localization has also been documented in embryos (Rieckhof et al., 1997; Kurant et al., 1998), and thus may be a general feature of their regulation. Because exd function correlates with where EXD is nuclear (Gonzalez-Crespo and Morata, 1995; Rauskolb et al., 1995), these results imply that HTH and EXD require each other to function. Thus, although the transcription of exd and hth are regulated independently, their protein products are interdependent.
What might be the significance of this HTH-EXD interdependence? One possibility is that the biologically active forms of EXD and HTH are, almost always, when they are bound to each other, perhaps as a 1:1 heterodimer. Consistent with this idea, HTH and EXD proteins are able to physically interact (Rieckhof et al., 1997) and their vertebrate homologs, PBX and MEIS, are found associated with each other in cells (Chang et al., 1997; Knoepfler et al., 1997). According to this view, HTH degradation would be a mechanism to maintain the correct HTH:EXD stoichiometry. The amount of active HTH-EXD complex would depend on the amounts of both HTH and EXD: when HTH was limiting, the correct amount of EXD would be stably transported into the nucleus, leaving excess EXD in the cytoplasm. In contrast, when HTH was in excess all of the EXD would be stably transported to the nucleus, and excess HTH would be degraded. In this way, the ratio of EXD to HTH in the nucleus would remain constant.
Maintaining a constant stoichiometry between EXD and HTH may be important for the function of these proteins. When EXD is expressed at high levels in the distal domain of leg discs, some of it can enter nuclei and interfere with leg development (Gonzalez-Crespo and Morata, 1996). However, in this situation EXD cannot repress Dll (Gonzalez-Crespo et al., 1998). In contrast, expressing HTH in this domain, which induces the nuclear localization of EXD, represses Dll (this work and H.-D. Ryoo and R. S. M., unpublished observations). These results demonstrate that, together, EXD plus HTH have different properties than EXD alone. Thus, altering the EXD:HTH stoichiometry may alter their activities and, consequently, may interfere with development.
The interdependence of hth and exd is reminiscent of a similar relationship exhibited by a set of bHLH-PAS proteins encoded by the genes tango (tgo), trachealess (trh) and single-minded (sim) (Ward et al., 1998). In this case, TGO is a general factor expressed throughout the embryo. It is cytoplasmic in all cells except for those that express its binding partners TRH and SIM. TRH and SIM also require the presence of TGO for their nuclear localization. Thus, the relationships between HTH and EXD documented here may be common among transcription factors that function as multiprotein complexes.
Defining proximal versus distal leg identities
The domains of gene expression for HTH, DAC and DLL, as well the regulatory interactions between them, suggest that the leg is functionally divided into two major domains. The first is a proximal domain, which expresses hth, has nuclear EXD and does not express at least some of the potential target genes of the WG- or DPP-signaling pathways. The second is a distal domain, which does not express hth, has EXD localized to the cytoplasm, and expresses the targets of WG, DPP and WG+DPP signaling. Together with previous results (Gonzalez-Crespo and Morata, 1996; Gonzalez-Crespo et al., 1998), these data suggest that the proximal domain is what Snodgrass (1935) referred to as the coxopodite, or an extension of the body wall, and is distinct from the distal domain, the telopodite.
Our results, together with those of Gonzalez-Crespo et al. (1998), suggest that hth expression and nuclear EXD in the coxopodite restricts the ability of the WG and DPP signals to activate their target genes (Fig. 10). This idea is consistent with the observation that these two domains differ with respect to their requirement for HH signaling: unlike the telopodite, which exhibits severe truncations upon the reduction of hh function, the coxopodite is less severely affected (Basler and Struhl, 1994; Tabata and Kornberg, 1994). These two domains also appear to have different cell surface properties; cells from one domain prefer not to mix with cells from the other domain. For example, Dll mutant clones almost always relocalize to the hth-expressing domain (G. Campbell and A. Tomlinson, personal communication; our observations) and hth mutant clones frequently sort into distal regions of the leg disc (data not shown). This phenomenon is not observed in the wing disc, where hth and Dll are restricted to outside and within the wing pouch, respectively: hth or Dll mutant clones are positioned randomly in this tissue (M. A.-S. and R. S. M., unpublished observations). Finally, the mutant phenotypes displayed by the loss of coxopodite gene function are qualitatively different from those displayed by the loss of telopodite gene function. Removal of coxopodite gene functions such as exd (Gonzalez-Crespo and Morata, 1995; Rauskolb et al., 1995) results in either ‘nonsense’ or proximal to distal cell fate transformations whereas removal of telopodite gene functions such as Dll (Cohen and Jurgens, 1989) and dac (Mardon et al., 1994) results in deletions of the appendage. In summary, the data support the idea that the proximal and distal regions of the leg have independent origins and differ from each other primarily due to the expression of hth, which limits or alters the ability of proximal cells to respond to WG and DPP signaling (Fig. 10).
WG and DPP signaling results in different outcomes in the embryonic midgut and the leg disc
It is an apparent paradox that WG and DPP repress hth in the leg disc while these same signals activate hth expression and nuclear EXD in the midgut endoderm (Mann and Abu-Shaar, 1996; Rieckhof et al., 1997). This may be explained because in the leg, WG and DPP repress hth indirectly, by activating the hth repressors Dll and dac. In the absence of Dll or dac, hth is derepressed in the leg disc, even in cells that receive high levels of the WG and DPP signals. In contrast, in the embryonic endoderm, dac and Dll are not activated by WG and DPP, nor are any other known hth repressors, allowing hth to be activated in these cells.
Although our data demonstrate that WG and DPP signaling are compatible with hth expression in the leg disc, they do not address if these signals are required for hth expression in the leg disc. An experiment that would address this question is to remove WG and DPP signaling in leg discs by using a temperature-sensitive allele of hh. When hh function was reduced during the second larval instar stage, Dll expression was lost (Diaz-Benjumea et al., 1994) but EXD remained nuclear in the proximal most cells of these discs (Gonzalez-Crespo and Morata, 1996). These data suggest that WG and DPP signaling are not required for nuclear EXD or hth expression in the coxopodite. However, in this experiment EXD did not become nuclear throughout the entire disc as would be expected if the hth repressors DLL and DAC were both eliminated (Gonzalez-Crespo and Morata, 1996). Because DAC was not monitored in this experiment, it is possible that its expression remained, and was sufficient to repress hth. Alternatively, it is possible that, in the telopodite, HH signaling (mediated by WG+DPP) is required to activate hth expression. Further experiments are required to distinguish between these possibilities.
Generating additional subdomains during leg development by mutual repression
In addition to providing evidence for antagonistic proximal and distal domains, this work illustrates that the leg disc is further divided into subdomains that result from cross regulation between hth+exd, dac and Dll (Fig. 10). In particular, our results suggest that both dac and Dll contribute to repression of hth in the telopodite, and that dac also represses Dll in a subset of the disc. We suggest that this regulation is important for converting the WG and DPP activity gradient into distinct domains of gene expression (Fig. 10). However, our results further suggest that this regulation occurs in at least three temporal phases.
In the first phase, only Dll is required for repressing hth in the telopodite. This conclusion is based on three observations: first, during embryogenesis, dac is not expressed in the leg primordia. Second, by late embryogenesis, the leg disc primordia consists of cells that express either hth or Dll, but not both (Casares and Mann, 1998; Gonzalez-Crespo et al., 1998). Third, in Dll− embryos, the entire leg disc primordia contains nuclear EXD (Gonzalez-Crespo et al., 1998).
The second phase begins when dac expression initiates during larval development. Two results suggest that, at this time, dac participates in hth repression. First, in leg discs in which nearly all Dll expression is eliminated, dac expression remains, and these dac-expressing cells do not express hth and have cytoplasmic EXD. Second, dac− clones that arise in the proximal portion of the dac domain (class 2) derepress hth and have nuclear EXD (Fig. 9). Thus, before dac is expressed, Dll appears to be solely responsible for repressing hth but, after dac expression initiates, it represses hth in these more proximal cells. We also propose a role for dac in repressing Dll at this stage, because some dac− clones (class 4) derepress Dll. These regulatory relationships are summarized in Fig. 10.
In the third phase, WG and DPP signaling is no longer required for the maintenance of Dll or dac expression; instead, these genes appear to be positively autoregulated (Lecuit and Cohen, 1997). In this phase, we also propose that the antagonism between dac and Dll no longer exists, because there is overlap in the expression of these two genes by the mid 3rd larval stage. In contrast, with the exception of the trochanter region, there is no overlap between hth and dac even at these later stages, suggesting that the mutual antagonism between these two genes could still be intact.
Other regulators of hth?
In dac null discs, there exists a group of dorsal-medial cells that do not express either Dll or hth. The behavior of these cells is reminiscent of the class 3 dac− clones, which do not derepress hth despite the absence of either of the known repressors, dac or Dll. One interpretation of these results is that there is another, as yet unidentified, repressor of hth in leg discs. Such a repressor would be expected to be expressed in these dorsal-medial cells. An alternative explanation is that dac or Dll is responsible for repressing hth in these cells, but that hth repression is maintained, even if these repressors are removed. This second model supposes the existence of a memory mechanism to maintain hth repression that may be analogous to how the Polycomb group of genes maintain homeotic gene repression during development (reviewed in Paro and Harte, 1996). Additional experiments are required to distinguish between these two mechanisms, and to further establish how this domain is set up and maintained during leg development.
We thank Gerard Campbell, Graeme Mardon, Gines Morata, members of the Struhl laboratory and the Bloomington Stock Center for fly stocks; Steve Cohen and Gerry Rubin for antibodies; Gerard Campbell and Fernando Casares for helpful discussions and comments on the manuscript; Calvin Tomkins for technical assistance and the Jessell laboratory for the use of their confocal microscopes. This work was supported by grants from the NIH and HFSP to R. S. M., who is a Scholar of the Leukemia Society of America.