Proximal to distal cell communication in the Drosophila leg provides a basis for an intercalary mechanism of limb patterning

Patterning in insect legs is organized along anteroposterior (AP), dorsoventral (DV) and proximodistal (PD) axes. In the case of Drosophila, AP and DV axes of the leg imaginal discs are established along the embryonic AP and DV axes that are set up based on maternal positional information. The PD axis, on the other hand, is zygotically specified by cellular interactions involving secreted signaling molecules Wingless (Wg) and Decapentaplegic (Dpp) (Cohen, 1990, 1993; Goto and Hayashi, 1997).

PD axis formation in the leg disc becomes first evident when cells expressing either Escargot (Esg) or Distal-less (Dll) are arranged in a circular pattern (Goto and Hayashi, 1997). Dll expression defines the central, distal domain. Esg-expressing cells become the proximal domain that surrounds the distal domain. The Meis family homeodomain protein Homothorax (Hth) is expressed in the proximal domain as well as in the surrounding body wall and regulates nuclear localization of another homeodomain protein, Extradenticle (Exd) (Rieckhof et al., 1997; Pai et al., 1998; Abu-Shaar and Mann, 1998). Exd is active in the nucleus but inactive in the cytoplasm (Mann and Abu-Shaar, 1996). The genetic requirement for Dll, Exd and Hth suggest that the distal domain gives rise to the majority of the adult leg including tarsus, tibia, femur and trochanter (Cohen and Jürgens, 1989; Cambell and Tomlinson, 1998) and that the proximal domain gives rise to the coxa and the ventral thoracic body wall (Gonzalez-Crespo and Morata, 1995; Rauskolb et al., 1995; Wu and Cohen, 1999). Initial PD subdivision in the embryonic leg disc becomes elaborated during larval stages by activation of additional genes such as dachshund (dac;Mardon et al., 1994) in a circular intermediate domain between the distal and proximal domains. dac is required for specification of the intermediate fate (Mardon et al., 1994).

The leg imaginal disc is also divided into a posterior compartment that expresses the secreted molecule Hedgehog (Hh) and an anterior compartment that responds to Hh by expressing Wg and Dpp along the AP compartment boundary (Campbell et al., 1993; Basler and Struhl, 1994). Mutual repression between Wg and Dpp limits Wg expression to the ventral side and Dpp expression to the dorsal side (Brook and Cohen, 1996; Jiang and Struhl, 1996; Penton and Hoffmann, 1996; Theisen et al., 1996). This spatial restriction of Wg and Dpp expression is essential for DV patterning of the leg. In addition, graded activities of Wg and Dpp are required for the expression of Dll and dac and repression of hth in the distal domain (Lecuit and Cohen, 1997; Abu-Shaar and Mann, 1998; Wu and Cohen, 1999). In the proximal domain, target gene activation by Dpp and Wg is inhibited by Hth and Exd, suggesting that the distal and proximal domains have distinct characters to respond to Dpp and Wg (Gonzalez-Crespo et al., 1998; Abu-Shaar and Mann, 1998; Wu and Cohen, 1999).

Based on the above observations, it was proposed that the circular patterns of gene expression along the PD axis in the distal domain are organized by the gradient of the combined activity of Dpp and Wg (summarized in Fig. 1A; Lecuit and Cohen, 1997; Abu-Shaar and Mann, 1998). In the central region where combined activity of Dpp and Wg would be high, Dll is activated and dac is repressed. An intermediate level of Wg and Dpp activities would allow dac expression in the intermediate domain. Gorfinkiel et al. (1997) showed that ectopic expression of Dll in the dorsal-proximal region induces wg, which is thought to interact with dpp to specify a new PD axis. These results suggest that the combination of Wg and Dpp constitute a ‘distalizing’ signal for the PD axis.

Although these results suggest that the combination of Wg and Dpp activities centered at the distal tip is essential for PD patterning, it is not known whether Wg and Dpp are sufficient to account for all aspects of PD positional information. In fact, the grafting and regeneration experiments using larval cockroach legs (reviewed by French et al., 1976) suggest that the reciprocal communication between distal and proximal parts of a leg segment promotes regeneration of the intermediate part. We thus speculated that a proximal to distal cell communication may also be used in PD patterning of the leg during development, and decided to test this idea by studying the functions of genes expressed in the proximal domain.

Esg is a Snail (Sna)-type zinc finger transcription factor, and is expressed in all imaginal tissues (Hayashi et al., 1993). It is required for the maintenance of imaginal cell identity in the wing disc (Fuse et al., 1996) and in abdominal histoblasts (Hayashi et al., 1993). A requirement for Esg in leg development was suggested from the analysis of partial loss-of-function mutants of esg, which showed a defect in the proximal leg (Hayashi et al. 1993). In the tracheal system, Esg regulates expression of the cell adhesion molecule DE-cadherin to promote fusion of tracheal tubes (Tanaka-Matakatsu et al., 1996).

In this study, we investigated the roles of Esg in the patterning of the PD axis in the leg. We first show that Esg is expressed in the proximal domain throughout leg development. Next we show that ectopic expression of Esg and its activator Hth in the distal domain induce the intermediate fate in surrounding cells by inducing dac expression. We further show that Esg and Hth-expressing cells in the distal domain undergo a change in their adhesive property to sort out from surrounding cells. The proximal to distal inductive communication was unexpected from the model based on the graded activity of Dpp and Wg. Thus we propose an intercalary mechanism that elaborates the PD axis pattern of the leg (Fig. 1B).

esg^G66B (Kassis, 1994), esg^P3 (Hayashi et al., 1993), UAS-esg (Fuse et al., 1994), UAS-hth (Pai et al., 1998), Dll-lacZ (Dll⁰¹⁰⁹²; Goto and Hayashi, 1997), wg-lacZ (Kassis et al., 1992), and omb-lacZ (Sun et al., 1995) have already been described. esg^NP5130 is a Gal4 enhancer trap line that reproduces larval esg expression (S. G, S. H. and NP consortium, unpublished). Flip-mediated mitotic recombination and act5C promoter-Gal4 fusion technique were carried out as described by Xu and Rubin (1993); Ito et al. (1997). Relevant genotypes are: hsp-70-flp/+; esg^G66BFRT40A/hs2piM FRT40A for esg mutant analysis, hsp-70-flp/+; AYGal4 UAS-gfp/UAS-esg, omb-lacZ/hsp-70-flp; AYGal4 UAS-gfp/+; UAS-esg/+ for esg misexpression, hsp-70-flp/+; AYGal4 UAS-gfp/+; UAS-hth/+ or hsp-70-flp/+; AYGal4 UAS-gfp/esg^P3; UAS-hth/+ for hth misexpression. Recombination was induced by a 0.5-to 1-hour heat shock at 37°C applied at 48-72 hours of development.

Protein expression was detected by a combination of the following antibodies and reagents: rat anti-Esg (Fuse et al., 1994), mouse anti-Dac (Mardon et al., 1994), rat anti-Tsh (Zeng et al., 1993), rabbit anti-Exd (Mann and Abu-Shaar, 1996), rat anti-DE-cadherin (Oda et al., 1994), rabbit anti β-galactosidase (Cappel), mouse anti β-galactosidase (40-1a, Developmental study hybridoma bank, University of Iowa), Cy2- and Cy3-labeled secondary antibodies and Streptavidin (Amersham), and biotin-conjugated secondary antibodies (Jackson). Signals were sometimes enhanced by use of the TSA indirect system (NEN). Stained samples were observed with a Zeiss LSM410. Adult legs were mounted in Hoyer’s mountant and immediately photographed for GFP images, and then cleared for cuticle analysis.

The first sign of proximodistal axis formation in the leg imaginal disc was a circular arrangement of cells expressing either Esg or Dll during embryogenesis (Goto and Hayashi, 1997; Fig. 2A). As the disc grew in size and evolved circular folds that separated tarsus, tibia, femur, trochanter and coxa, the pattern of esg expression was maintained (Fig. 2B,C). At the late stage of the third instar, more Esg protein was detected in the proximal region corresponding to the coxa and trochanter (Fig. 2C). The distal most part of the esg-expressing domain partially overlapped with the Dac-expressing domain in the trochanter (Fig. 2D-F). The esg expression in the overlapping domain was weaker than that in the more proximal domain where only esg was detected. The domain of Esg expression appeared to overlap with the proximal domaindefined by expression of homothorax and teashirt, and nuclear localization of Extradenticle (Gonzalez-Crespo and Morata, 1996; Abu-Shaar and Mann, 1998; Gonzalez-Crespo et al., 1998).

Dll induces distal leg development when expressed ectopically in the proximal domain (Gorfinkiel et al., 1997). To ask if any of the proximal genes have an organizing activity analogous to that of Dll, we induced esg ectopically using the flip-out technique (Ito et al., 1997). In the adult, Esg-positive clones marked by GFP were found as vesicles inside the leg cuticle and were often associated with malformation (Fig. 3B-D). In the region proximal to the clones, the bristles and epidermal hairs, which normally point distally, were often reversed (Fig. 3E,F). These bristles and hairs were genetically wild type, suggesting that the polarizing activity of Esg is non-cell-autonomous.

In the third instar leg disc, dac is expressed in a partially overlapping manner with the expression of Dll and esg in an intermediate ring that corresponds to the proximal tarsus, tibia, femur and trochanter (Mardon et al., 1994). When esg expression was induced during the second instar, clones in the distal tarsal region showed compact morphology; and many of them were associated with ectopic dac expression in cells within and surrounding the clone (Fig. 4A,B). The ectopic dac expression resulted in a local reversion of the proximal-distal order of the gene expression (Fig. 4A), which prefigured the change in the cell polarity in the adult leg (Fig. 3B-F). The esg-positive clones in the coxa spread normally and did not show induced dac (Fig. 4A, arrowhead).

The non-cell-autonomy of the Esg function could be due to a modulation of known secreted molecules controlling anteroposterior and dorsoventral patterning. However, the expression patterns of the Hh target genes wg and dpp, and optomoter-blind (omb;Brook and Cohen, 1996), a target gene of Dpp, were unaffected by misexpression of Esg (Fig. 4D-F). Moreover, the non-cell-autonomous induction of dac by Esg showed no strong bias for orientation (Fig. 4B) and position with respect to the compartment boundary and dorsoventral position. These observations suggest it to be unlikely that the inductive activity of Esg involves modulation of hh, wg or dpp; rather, it is probably mediated by an as yet unidentified signaling mechanism.

We generated esg^G66B null mutant clones to assess the requirement of Esg for dac expression. esg^G66B is a derivative of an enhancer trap and lacks the coding region of esg but retains the lacZ gene that reproduces the expression pattern of esg (Kassis, 1994). esg mutant cells are marked by the loss of Myc antigen (Fig. 4G-I) or by the high expression of β-gal produced from the two copies of the lacZ gene (Fig. 4J-L). dac expression was frequently lost in clones induced at the late second instar larval stage. The partial loss of dac expression in large clones may have been due to a non-cell-autonomous rescue by esg⁺ cells next to the clones. The clones were sometimes associated with ectopic fold formation (data not shown). Taken together with the gain-of-function analysis, these data suggest that Esg is necessary and sufficient for dac induction.

Proximal cell identity is, at least in part, controlled by the homeodomain protein Hth that regulates nuclear localization of Exd. When expressed ectopically in the tarsal region, Hth caused non-cell-autonomous induction of dac expression (Fig. 5A-C) and reversal of bristle and cell polarity (Fig. 3I,J); phenotypes very similar to those caused by Esg. Unlike esg-expressing clones, which secrete a smooth cuticle, hth-expressing clones in the distal part of the leg sometimes formed thick socketted bristles without bracts, which are characteristics of the bristles in the proximal part of the leg (data not shown). We found that Hth strongly activated a reporter gene under the control of the esg enhancer in the distal domain, but did not or only weakly did so in the proximal domain (Fig. 5D,E). This effect was cell-autonomous, suggesting that Hth may directly regulate transcription of esg. On the other hand, neither a loss nor a gain of esg expression affected the activity of Hth/Exd as assessed by the expression of Hth and nuclear localization of Exd, or by the expression of another proximal gene, teashirt (Gonzalez-Crespo and Morata, 1996; data not shown). These results suggest that Esg acts downstream of Hth/Exd to regulate proximodistal patterning.

The esg- or hth-expressing clones in the distal region were round in shape with smooth borders and often invaginated basally to form vesicles in the adult legs and in the larval discs (Figs 3D,H, 4-6). In contrast, control clones expressing non-functional esg, which lacked the zinc-finger domain, and esg-expressing clones located in the coxa and trochanter had ragged borders (Fig. 4A,C). The epithelial-type homophilic cell adhesion molecule DE-cadherin (Oda et al., 1994) was expressed throughout the leg discs (Fig. 6A-C) and its apical localization was maintained normally in esg-expressing clones (Fig. 6D-F), suggesting that these cells kept their epithelial character. These results of ectopic expression studies, together with the loss of function studies on hth (Abu-Shaar and Mann, 1998; Wu and Cohen, 1999; data not shown), indicate that Hth and Esg regulate a cell surface property that distinguishes the proximal and distal domains.

The finding that dac is induced by esg-expressing cells in the tarsal region was unexpected from the model (Lecuit and Cohen, 1997) proposing that a combination of Wg and Dpp produces an instructive gradient of a distalizing signal that is highest at the distal tip where dac is repressed (Fig. 1A). This model assumes that a response to a certain threshold of a combination of Wg and Dpp directly leads to a circular pattern or that cell proliferation and migration of responding cells lead to a circular arrangement. However, no proof has been given to support the idea that dac or other genes respond directly to Wg and Dpp in a circular pattern. Clonal analysis suggested that cells in the leg disc generally migrate in a PD direction (Bryant and Schneiderman, 1969), not along the circumference as the model would predict. We therefore propose a revised model of leg patterning in which the circular arrangement of the proximal cells provides patterning information that regulates the distal gene expression in conjunction with Wg and Dpp (Fig. 1B). The circular arrangement of proximal and distal cells of the leg disc is set up by late embryogenesis (Goto and Hayashi, 1997; Gonzalez-Crespo et al., 1998) by a mechanism that is not well understood. Once established, the circular pattern is maintained by repression of hth in the distal domain by wg and dpp (Abu-Sharr and Mann, 1998; Wu and Cohen, 1999), and by activation of esg by hth in the proximal domain (this study). During the transition from the second to third instar, dac expression is induced by a combination of a signal from proximal cells, and Wg and Dpp signaling from the AP compartment boundary. The range of each signaling limits dac expression to the intermediate domain.

The cell autonomous sorting of Esg- or Hth-expressing cells in the distal domain suggests that Esg and Hth regulate cell adhesive properties in the proximal cells. Similar sorting behavior was reported for cells that lost Dll or Hth (Gorfinkiel, 1997; Wu and Cohen, 1998; Campbell and Tomlinson, 1998). Differential cell affinity controlled by the proximal and distal genes may contribute in part to the separation of the proximal and distal domains into a circular pattern.

An intercalation mechanism of the proximodistal patterning was suggested (French et al., 1976) from the results of grafting and regeneration experiments utilizing larval cockroach legs. The experiments showed that joining of normally non adjacent positions within a leg segment resulted in intercalary regeneration of the intermediate structures, suggesting that the abutment of distal and proximal positional value can induce an intermediate value. The combination of a graft from a distal level with a proximal level host resulted in a normally oriented regenerate. However, combining a proximal graft with a distal level host produced a regenerate with reversed proximodistal polarity, which is reminiscent of the phenotype caused by ectopically induced Esg in the distal domain. This phenotype was difficult to interpret assuming that the distal tip is the only source of proximodistal positional information. A suggestion that such an intercalary mechanism is also used in normal development came from the work by Schubiger (1974), who showed that the presumptive most proximal (thorax and coxa) and distal (claws) parts were the first to commit to differentiation when premature discs from larvae of 74-80 hour of development were immediately subjected to metamorphosis. Intermediate structures (trochanter, femur, and tibia) were recovered from discs from older larvae. The proximal to distal signaling dependent on Esg and Hth described here may provide a molecular basis for the intercalation mechanism.

Proximodistal axis patterning in the vertebrate limb requires interaction between the apical ectodermal ridge (AER) and mesenchyme. Dlx (Ferrari et al., 1995), a homolog of Dll, and Slug (Ros et al., 1997), a homolog of esg, are expressed in the AER and mesenchyme, respectively. Nuclear localization of mouse Exd homolog Pbx1 in limb mesenchyme is modulated in a similar manner to Exd in Drosophila leg discs (Gonzalez-Crespo et al., 1998). Thus the proximodistal patterning in the vertebrate and insect limb (French et al., 1976) may share a common genetic basis for an intercalary mechanism.

We thank H. Sun, Y. Hiromi, T. Uemura, R. Mann, G. Mardon, D. Andrew, Developmental Study Hybridoma Bank and Bloomington Stock Center for reagents, and Yash Hiromi and Tadashi Uemura for critical comments on the manuscript. This work was supported by grants from the Japan Society for Promotion of Science (Research for the Future, S.H.) and from the Ministry of Education, Sports, and Culture (S.G.).