Formation and specification of distal leg segments in Drosophila by dual Bar homeobox genes, BarH1 and BarH2

The formation of compartments or domains specified by the region-specific expression of transcription factors may be essential for the body plan of insects and vertebrates (e.g. for Drosophila, see Azpiazu et al., 1996; Lawrence and Struhl, 1996; Sato et al., 1999). Drosophila leg development may provide a good system for studying region-specific expression and compartment formation, since legs are simple in structure and their formation encompasses various developmental processes.

Drosophila legs comprise the segmental units, from proximal to distal, coxa, trochanter, femur, tibia, tarsal segments 1-5 and pretarsus. Leg formation occurs through concentric folding and subsequent segmentation of monolayered epithelia of leg discs invaginated from the epidermis during embryogenesis (Cohen, 1993). According to Lecuit and Cohen (1997), decapentaplegic (dpp) and wingless (wg) expressed dorsally and ventrally, respectively, along the anteroposterior compartment boundary (AP boundary) are essential for the concentric expression of Distal-less (Dll) and dachshund (dac) in leg discs. Dll, encoding a homeodomain protein, is expressed in the distal region of leg discs and is required for the development of all distal structures other than the coxa (Cohen et al., 1989; Diaz-Benjumea et al., 1994). dac, encoding a novel nuclear protein, is expressed in the middle leg region and is essential for the formation of intermediate portions of legs (Mardon et al., 1994).

Abu-Shaar and Mann (1998) suggested that the generation of the concentric domains occurs in multiple phases. The coxopodite and telopodite regions are initially established by the expression of Dll and a homeobox gene, homothorax (hth). In coxopodite, the proximalmost region constituting future body wall and proximal segments, Dpp and Wg signaling are blocked by concerted function of Hth and nuclear localized Extradenticle (Exd), another homeodomain protein (Abu-Shaar and Mann, 1998; Gonzalez-Crespo and Morata, 1996; Wu and Cohen, 1999). In contrast, in the telopodite, Dpp and Wg signaling are active and required for proximodistal axis generation. Subsequently, a dac expression domain appears between hth and Dll domains to subdivide the telopodite region further. One more intermediate region, which expresses both dac and Dll, then arises by the early third instar.

Downstream of Dll, several genes participate in distal tarsal formation. A homeobox gene aristaless (al), expressed in the most distal tip of leg discs, is required for the pretarsus structures (Campbell et al., 1993; Schneitz et al., 1993; Campbell and Tomlinson, 1998). spineless (ss), which encodes a bHLH-PAS protein homologous to mammalian dioxin receptor, is expressed transiently in a future tarsus region and its absence results in deletion of distal tarsal segments 2-4 (Duncan et al., 1998). bric á brac (bab), encoding a nuclear factor having BTB domain, is expressed downstream of ss (Duncan et al., 1998) and is essential for the segmentation and specification of tarsal segments 2-4 (Godt et al., 1993).

Here, we show that functionally redundant homeobox genes at the Bar locus, BarH1 and BarH2 (Kojima et al., 1991; Higashijima et al., 1992a), serve as essential regulators in numerous aspects of distal tarsus development. BarH1 and BarH2 expression requires Dll. Juxtaposition of Bar-positive and Bar-negative tissues induces initiation of central folding. Antagonistic interactions between Bar and Al or other pretarsus factors establish the pretarsus/tarsal-segment-5 boundary. Bar is also required for the segmentation and specification of tarsal segments 3-5.

Fly strains used are: Canton S (wild-type), Df(1)B^263-20 (Sato et al., 1999), Df(1)BH2 (Higashijima et al., 1992b), In(1)B^M2, In(1)pdf, B^SY (Lindsley and Zimm, 1992), In(1)w^VC (Hotta and Benzer, 1976), Dll^SA1 (Gorfinkiel et al., 1997), dac³ (Mardon et al., 1994) and al¹, al² (Lindsley and Zimm, 1992). al¹ and al² are used in combination with Df(2L)al (Lindsley and Zimm, 1992). In al¹/Df(2L)al and al²/Df(2L)al flies, aristae and claws were almost completely abolished but other pretarsus structures were not significantly affected. en-lacZ (ryXho25; Hama et al., 1990), neu-lacZ (A101; Bellen et al., 1989), Bar-lacZ (P058; FlyView: http://pbio07.uni-muenster.de/) and ta5-lacZ (BM25; T. Michiue and K. S., unpublished data). B⁵⁵ was isolated by imprecise excision of P-element of P058. ptc-GAL4 (559.1; Brand and Perrimon, 1993), blk-GAL4 (40C.6; Morimura et al., 1996), ap-GAL4 (md554; Calleja et al., 1996), UAS-BarH1, UAS-BarH2 (Sato et al., 1999) and UAS-al. UAS-al was generated by inserting al cDNA (Campbell et al., 1993) into pUASV (Sato et al., 1999). For FRT/FLP mosaic analyses, hsFLP, FRT18A, FRT40A and FRT42D (Xu and Rubin, 1993) were used. All other mutations and balancer chromosomes were described in Lindsley and Zimm (1992).

Ten out of twenty eight mosaic legs from 83 flies of the genotype Df(1)B^263-20, y/In(1)w^VC, are entirely Bar–. To isolate Bar– discs, larvae having mouth hooks and denticle belts mosaic for y– were dissected. Among leg discs from eighteen such larvae, six were judged as Bar– because of the absence of anti-BarH1 antibody-staining signals.

Bar–, Dll– and dac– clones, respectively, were generated in larvae whose genotypes are Df(1)B^263-20, y FRT18A/FRT18A; hsFLP, Sb/+, y w hsFLP; FRT42D Dll^SA1/FRT42D лM45F and y w hsFLP; dac³FRT40A/лM36F FRT40A. To examine the ability of B^SY to rescue Df(1)B^263-20 mosaic phenotypes, clones were generated in larvae of the genotype Df(1)B^263-20, y FRT18A/FRT18A/B^SY; hsFLP, Sb/+. In all cases, clones were induced by a 90 minute heat shock at 37°C during late first-second instar.

Several independent UAS-Bar and UAS-al lines were used to misexpress Bar and al. UAS-BarH1^M6 and UAS-BarH2^F9 driven by ptc-GAL4 gave leg phenotypes similar to those of blk-GAL4-driven UAS-BarH1^M12, UAS-BarH1^M13, UAS-BarH2^F11 or UAS-BarH2^M9. In combination with ap-GAL4, UAS-BarH1^M13, UAS-BarH2^M9 and UAS-BarH2^F11 gave leg phenotypes similar to one another. Five independent lines of UAS-al showed little leg defects when driven by ptc-GAL4 or blk-GAL4.

Antibody staining was carried out according to Sato et al. (1999). For confocal microscopy, Cy3-conjugated secondary antibody (Jackson Immune Research) or biotinilated secondary antibody (Vector) followed by avidin-FITC (Promega) were used. Antibodies used were: rabbit anti-BarH1, rabbit anti-BarH2 (Higashijima et al., 1992b), rat anti-Ap (Lundgren et al., 1995), mouse anti-Fas II (Lin et al., 1994), mouse anti-Dac (Mardon et al., 1994), rat anti-Al (Campbell et al., 1993), mouse anti-Dll (Diaz-Benjumea et al., 1994), rabbit anti-β-gal (Cappell), mouse anti-β-gal (Promega) and Rhodamine-phalloidin (Molecular Probe).

During development of Drosophila legs, the disc epithelium folds concentrically; genes expressed circularly in the leg disc just before the onset of folding may thus be essential for leg morphogenesis. BarH1 and BarH2, may belong to such a class of genes. Staining for BarH1, BarH2 and Bar-lacZ (Fig. 1E,F) indicated that BarH1 and BarH2 are coexpressed circularly in all three types of third-instar leg discs. Since BarH1 and BarH2 are functionally redundant to each other (see below), they are hereafter referred to as Bar collectively.

Indications of circular Bar expression (Bar ring) were first evident at 76 hours AEL (after egg laying) at 25°C and became more evident at 80 hours AEL (Fig. 1A,B). The formation of the central fold, from which distalmost leg segments are generated (see below), began at 84 hours AEL as a circular indentation along the periphery of the Bar ring, starting dorsally and completing ventrally by 88 hours AEL (Fig. 1C,D,G-G′′). As the indentation became deeper in mid third instar, graded Bar expression gradually became apparent along the proximodistal axis (Fig. 1H-H′′). In the central fold of a 112 hours AEL leg disc, future tarsal segment 3, lacking Bar expression, became clearly identifiable (Fig. 1I-I′′). At 5 hours APF (after puparium formation), Bar expression was closely related to lines of demarcation of tarsal segments (Fig. 1J-L,O). Bar expression was strongest in tarsal segment 5, clearly evident in tarsal segment 4 and not detected in tarsal segment 3 and the pretarsus except for future claw regions.

ta5-lacZ is a lacZ reporter driven by a tarsal-segment-5-specific Bar enhancer (Fig. 1J-M), while ap is a LIM-homeobox gene expressed in tarsal segment 4 (Fig. 1M-P; Cohen et al., 1992). Fig. 1Q shows that these markers begin to be expressed as adjacent circles within the Bar ring during the central fold formation, indicating that tarsal segments 4 and 5 are derivatives of the early Bar compartment. Because of the absence of a suitable molecular marker, we could not determine whether future tarsal segment 3 is included in the early Bar ring. However, since genetic analysis (see below) shows that Bar function in tarsal segments 3-5 is essential for segmentation of distal tarsus, we tentatively conclude that tarsal segments 3-5 are derivatives of the early Bar ring and distal tarsus separation starts just prior to the completion of the central fold formation at early third instar.

From late third instar stages onwards, Bar was also expressed in putative claw cells (insets of Fig. 1E,F). Staining for BarH1 and en-lacZ (Fig. 1R; Hama et al., 1990) showed the anterior edge of the posterior claw region coincides with that of the posterior compartment, suggesting that the center of the Bar ring is situated in the anterior compartment, between the paired claw regions. During late third instar, Bar-negative patches in future tarsal segment 5 became detectable (Fig.1E,F,S,T). Staining for BarH1 and neu-lacZ (Bellen et al., 1989; Huang et al., 1991) showed that these correspond to sensory organ precursors and/or their derivatives (Fig. 1S,T).

As a first step to elucidate Bar functions during leg development, we carried out gynandromorphic mosaic analysis (Hotta and Benzer, 1976) using a larval-lethal deletion Df(1)B^263-20, which uncovers BarH1 and BarH2 along with forked (f; Hoover et al., 1993; Ishimaru and Saigo, 1993), Fimbrin (Fim; Ishimaru et al., unpublished data), and an unknown gene, X2 (see Fig. 2N). In legs totally composed of cells hemizygous for Df(1)B^263-20, tarsal segments 2-5 were fused together into a small bulb-like, non-segmental structure having neither claws nor pulvilli, whereas other leg segments were normal (Fig. 2A,B). The first appreciable morphological event in distal leg development is central folding along the periphery of the Bar ring (see Fig. 1). As shown in Fig. 2K, no central folding occurred in Df(1)B^263-20 leg discs; more proximal folds (mf and pf) were normally formed (for the wild-type, see Fig. 1D). Together, these results may indicate that at least one of the five genes uncovered by Df(1)B^263-20 is essential for distal leg morphogenesis.

To determine which cells require Df(1)B^263-20 gene activity for normal distal leg development, small Df(1)B^263-20 clones were generated using the FRT/FLP method (Xu and Rubin, 1993). Partial fusion of tarsal segments 2-5 were frequently observed (see Fig. 3P,Q). Central fold formation was also prevented in Df(1)B^263-20 clones generated along the periphery of the early Bar ring (data not shown), consistent with the notion that both central folding and segmentation among tarsal segments 2-5 require the Df(1)B^263-20 gene(s). In mosaic legs normal in appearance (n=132), all mutant clones, except for a single case, in which a mutant bristle was situated at the proximal tip of tarsal segment 3, were observed outside tarsal segments 3-5 (Table 1), indicating that Df(1)B^263-20 gene activity in future tarsal segments 3-5, which are the presumed derivatives of the early Bar ring, is essential for normal leg development. Among five Df(1)B^263-20 genes, f, Fim, BarH2 and X2 appeared dispensable for distal leg development other than bristle morphogenesis at least in the presence of BarH1. Indeed, no defect in gross morphology of distal legs was observed when Df(1)B^263-20 homozygous clones were generated in flies carrying BSY, a Y chromosome with a Bar locus fragment including BarH1 but not other Df(1)B^263-20 genes (Fig. 2N, unpublished data). Consistently, no defect in distal leg gross morphology was observed in Df(1)BH2, uncovering f, Fim and BarH2 (Fig. 2N, unpublished data; Higashijima et al., 1992b).

BarH1 may be functionally redundant to BarH2 as has been shown in other developmental contexts (Hayashi et al., 1998; Sato et al., 1999). B^M2 and pdf are inversion mutants with a distal breakpoint between BarH1-and BarH2-coding sequences (Fig. 2N; Tsubota et al., 1989). In these mutants, early BarH1 expression was almost completely missing but early BarH2 expression along with central folding occurred normally (Fig. 2E,H). We conclude that the genes essential for distal leg morphogenesis are BarH1 and BarH2, which are functionally redundant to each other.

In pdf and B^M2/Df(1)B^263-20 flies, tarsal segments 3-5 were partially fused with each other (Fig. 2C). B⁵⁵ is a newly isolated

Bar deletion mutant uncovering BarH1, X2 and a late Bar enhancer (ta5) but not BarH2 (Fig. 2N). In B⁵⁵ mutants, BarH1 expression is missing throughout development (Fig. 2F,G), while BarH2 expression was normal in early third instar larval stages but extensively reduced afterwards (Fig. 2I,J). As with pdf flies, tarsal segment 3 and more distal segments were fused together (Fig. 2D), while tarsal segment 2 and central fold were normally formed (Fig. 2D,F,I). The pdf and B⁵⁵ phenotype was eliminated by B^SY (Lindsley and Zimm, 1992; unpublished data). Based on these results, we conclude that late Bar function is involved in segmentation among tarsal segments 3-5.

Juxtaposition of Bar-positive and Bar-negative tissues may result in folding. To test this idea, BarH1 or BarH2 was misexpressed along the AP border using the UAS/GAL4 system (Brand and Perrimon, 1993). blk-GAL4 or ptc-GAL4 was used as a GAL4 driver. As anticipated, no appreciable difference in morphological changes were detected between BarH1 and BarH2 misexpression. Fig. 2L shows that ectopic folding was formed along outer circumference of Bar-misexpressing region proximal to the Bar ring, while the formation of the authentic central fold was prevented in the regions where endogenous Bar-expressing cells and cells solely expressing exogenous Bar were juxtaposed to each other. It would thus follow that central folding along the early Bar ring is an outcome of the juxtaposition of Bar-positive and Bar-negative tissues. It should, however, be emphasized that no ectopic folding ever occurred either prior to the onset of central folding (Fig. 2M), or within the Bar ring and future pretarsus (Fig. 2L), suggesting the involvement of temporal and regional factors other than Bar.

Using mosaic analysis and misexpression, we examined whether Bar activity is required for the expression of markers for tarsal segments 4 (ap) and 5 (ta5-lacZ). Neither Ap nor ta5-lacZ expression was detected in Bar– clones (Fig. 3A-F). Bar misexpression along the AP border caused ectopic expression of either ta5-lacZ or Ap or both in Bar-misexpressing regions proximal to the Bar ring (Fig. 3G-I). Furthermore, Bar– clones within tarsal segments 4 and 5 were occasionally associated with campaniform sensilla (arrowhead in Fig. 3P), normally situated only at the dorsodistal tip of tarsal segment 3 (Fig. 3O). Based on these observations, we conclude that Bar is essential for ap and ta5-lacZ (Bar) expression in distal tarsi and prevents cells expressing Bar at later stages from adopting the tarsal segment 3 fate.

ap-GAL4 is a GAL4 enhancer trap of the ap locus (Calleja et al., 1996). In ap-GAL4/UAS-Bar leg discs, not only did ta5-lacZ misexpression occur weakly in presumptive tarsal segment 4 (Fig. 3M,N), but also tarsal segment 4 occasionally showed partial transformation into tarsal segment 5. The ventralmost bristle situated at the distal tip of tarsal segment 4, normally thick and black (arrow in Fig. 3R), was frequently transformed into a thin and less-pigmented bristle (Fig. 3T) characteristic of tarsal segment 5 (arrowheads in Fig. 3R,T). Specific expression of ap or ap-GAL4 in future tarsal segment 4 reflects that subdivision of the early Bar ring has already occurred. Thus, the above findings may indicate that Bar misexpression in future tarsal segment 4 at later stages is still capable of causing future tarsal segment 4 cells to adopt tarsal segment 5 fate at least partly, and hence, suggests that late strong Bar expression in wild-type tarsal segment 5 primordia is important for them to acquire and maintain the tarsal segment 5 identity. It should, however, be noted that ap expression is not eliminated upon Bar misexpression (Fig.3I,L,M), suggesting that a factor other than Bar is involved in repression of ap in tarsal segment 5. Together, these findings indicate that graded Bar expression at later stages may be essential for specification of tarsal segments 3-5.

Ectopic incomplete segmental joints were frequently formed in the vicinity of the boundary between Bar– clones and Bar-expressing tarsal segments 4-5 tissues (Fig. 3P). Tarsal segments 4 and 5 were partially fused with each other in ap-GAL4/UAS-Bar legs (Fig. 3S). Thus, the juxtaposition of tissues expressing different levels of Bar may be essential for proper segmentation in distal tarsi.

In Drosophila, antennae possess segmental structures homologous to those in legs and similarly differentiate through circular folding. Arista and basal cylinder, probably corresponding to pretarsus and distal tarsus (Postlethwait and Schneiderman, 1971), are derivatives of the central knob of antennal discs. After the onset of third instar, Bar expression occurred initially in a central region of the antennal disc and gradually became a ring similar to that observed in the leg disc (Fig. 4A,B). As with legs, central folding occurred just outside the Bar ring (Fig. 4B). In antennal discs lacking Bar activity, no central fold was formed (Fig. 4C); Bar– antennae frequently lost arista and basal cylinder (Fig. 4D,E). Thus, Bar is concluded to be essential not only for leg but also for antennal development.

Establishment of the tarsus/pretarsus boundaryal is a homeobox gene expressed at the center of leg and antennal discs from early third instar onwards (Campbell et al., 1993; Schneitz et al., 1993). Initially, the Al expression domain and early Bar ring overlapped slightly in leg discs (Fig. 5A). Al/Bar overlapping could be more clearly seen in early antennal discs (Fig. 5C). Up to 90 hours AEL, the central part of the leg disc was strictly divided into two regions, Bar-positive/Al-negative and Bar-negative/Al-positive circular domains (Fig. 5B). In antennal discs, such discrimination in Bar/Al expression may be incomplete (Fig. 5D).

Regionally exclusive expression of Bar and Al may be due to mutually antagonistic interactions between Bar and Al. When Al expression was examined in mid to late third instar larval leg discs having Bar– clones, Al expression invaded a Bar– presumptive tarsus region (Fig. 5F), while Al expression was considerably attenuated by Bar misexpression along the anteroposterior compartment border in mid third instar discs (Fig. 5I,J). Ectopic patches of Bar expression were frequently observed in the presumptive pretarsus of hypomorphic al leg discs in late third instar (Fig. 5G). Bar derepression due to reduction in Al activity is more clearly observed in antennal discs; on a hypomorphic al mutant background, Bar was expressed in the centralmost region even at late third instar larval stages (Fig. 5H). That no appreciable repression of Bar was detected when al was misexpressed by ptc-GAL4/UAS-al (Fig. 5K,L) may indicate the involvement of factors other than Al in Bar repression in presumptive pretarsus. In addition, the lack of Al invasion into Bar– clones in leg discs at early third instar larval stages (Fig. 5E) may indicate that Bar is dispensable for repressing al when their expression is initiated.

At late third instar, cells in the distalmost region of leg discs are densely packed at the apical surface and distinguishable from surrounding loosely packed cells (Condic et al., 1991; Fig. 5M). Double-staining with rhodamine-phalloidin and anti-BarH1 antibody revealed that the former corresponds to Bar-negative pretarsus cells, and the latter, Bar-positive tarsus cells (Fig. 5M,N). Proximalmost pretarsus cells (border cells) were frequently rectangular in apical shape (Fig. 5M). Staining for Fasciclin II (Fas II; Grenningloh et al., 1991) showed that border cells prominently express Fas II at late third instar (Fig. 5O). Fas II expression was interrupted by Bar– clones (Fig. 5Q). Fas II misexpression was induced along Bar-misexpressing presumptive pretarsus, while endogenous Fas II expression was repressed (Fig. 5R). Thus, Bar would upregulate and downregulate Fas II expression in Bar-negative border cells and Bar-positive non-border cells, respectively. We, thus, conclude that Bar is essential for the establishment of the boundary between tarsal segment 5 and pretarsus.

Circular Dac expression appeared in second-instar leg discs before Bar ring appearance (Fig. 6A). This early Dac-ring was associated interiorly with Bar-positive Keilin’s organ cells, which are situated along the interior circumference of or within the early Bar ring (see Figs 1A, 6A,B). Although they were separated from each other by a Bar-negative, Dac-negative region just before the onset of central fold formation (Fig. 6C), Dac and Bar rings were immediate neighbours at earlier stages (Fig. 6B). Dac expression was derepressed in Bar– clones observed in early third instar (Fig. 6F), while repressed by Bar misexpression (Fig. 6G,H), indicating that Bar is essential for distal restriction of Dac expression. Since early Bar expression normally occurred in dac– clones (Fig. 6D), dac appears dispensable for proximal restriction of the early Bar ring. Interestingly, Bar misexpression occurred in future trochanter in dac– mutants (Fig. 6E), indicating that Dac represses Bar in future trochanter.

Dll is expressed from the beginning of leg development (Abu-Shaar and Mann, 1998). The Dll domain includes all Bar-expressing cells (see Fig. 6I,J). Bar expression was abolished in Dll– clones (Fig. 6I), while Dll continued to be expressed in Bar– clones (Fig. 6J). Bar expression thus appears to require Dll activity but Dll does not require Bar.

Distal leg segmentation is a multiple-step process involving various aspects of development. Our results are summarized in Fig. 7. Most events of distal leg segmentation occur during third instar larval stages.

By early third instar, the leg disc has been divided into four domains through hth, dac and Dll expression (Abu-Shaar and Mann, 1998). At early third instar, circular expression of Bar and al begins within the Dll domain; Dll is required for the expression of Bar (Fig. 6I and unpublished data) and al (Campbell and Tomlinson, 1998). Future tarsal segment 2 may be generated in the distalmost region of the Dac ring possibly through repression of dac expression by an unknown factor, X (Fig. 7), since Dac is expressed in the region immediately proximal to the early Bar ring (future tarsal segments 3-5) at early stages (Fig. 6B) but not in tarsal segment 2 at later stages (Abu-Shaar and Mann, 1998; Lecuit and Cohen, 1997).

Expression of molecular markers for tarsal segments 5 (ta5-lacZ) and 4 (ap) becomes apparent within the Bar ring just after the onset of central folding but before distal tarsus segmentation (Fig. 1Q). Cells in the proximalmost region of the early Bar ring may also be committed to become tarsal segment 3 at this stage. Bar expression within the early Bar ring is nearly homogeneous (Fig. 1B) and thus the initial subdivision of this ring into future tarsal segments 3-5 may require factor(s) other than Bar. ss is expressed transiently in the future tarsus region in late second to early third instar and ss mutant legs lack tarsal segments 2-4 but not 5 (Duncan et al., 1998). ss may thus be responsible for differential gene expression between future tarsal segments 4 and 5. Repression of Bar expression in future tarsal segment 3 may be due to a putative Bar repressor, Y (Fig. 7).

At later stages, Bar expression is strong in tarsal segment 5 (Fig. 7, deep green), moderate in tarsal segment 4 (light green) and absent from tarsal segment 3. Genetic and morphological analyses (Table 1; Fig. 3) strongly suggest that Bar upregulates ap and/or its own expression in tarsal segments 4 and 5. Since (1) only partial transformation of tarsal segment 4 into 5 occurs upon Bar overexpression in tarsal segment 4 (Fig. 3T) and (2) Bar misexpression fails to repress Ap expression (Fig. 3G-I), an unknown factor (Z) is likely involved in ap repression in future tarsal segment 5 (Fig. 7).

Duncan et al. (1998) suggest that tarsal development takes place in two steps: establishment of a uniform tarsal region followed by subdivision of this ring into segments. Our results indicate that there are several more intermediate steps in this process. Abu-Shaar and Mann (1998) propose three phases of leg-disc subdomain formation during early development probably prior to the onset of Bar expression. Thus, leg segmentation requires repeated subdivision of leg epithelium along the proximodistal axis with the result that smaller region-specific transcription factor domains are generated from larger ones in all instances of subdivision.

Bar and Al expression begins essentially at the same time at early third instar. Initially, Al expression partially overlaps Bar expression (Figs 5A,C, 7, yellow), and no invasion of Al into Bar– clones was found when mosaic clones were observed at this stage (Fig. 5E). However, at slightly later stages, Al and Bar expression became mutually exclusive (Fig. 5B,D) and Al invaded into Bar– clones (Fig. 5F). It may thus follow that al expression is initiated Bar independently and, after a while, Bar protein accumulated to some extent begins to repress al expression. al may also regulate Bar expression in a similar fashion. Indeed, Bar misexpression occasionally occurred in leg and antennal discs of al hypomorphic mutants (Fig. 5G,H), although Bar expression was not repressed by al misexpression (Fig. 5K,L). The failure of Bar repression by al may suggest the involvement of other pretarsus gene(s) that function cooperatively with al. We actually identified two new genes that function in the pretarsus and show mutant phenotypes similar to al (T. Tsuji, T. K. and K. S., unpublished data).

Bar misexpression experiments (Fig. 3G-I) showed pretarsus to be the region where ta5-lacZ expression is very difficult. Al and/or other pretarsus factors may thus possibly compete with Bar for the late Bar enhancer, ta5, to repress pretarsus Bar expression (see Fig. 7).

Fig. 5O shows that Fas II, a protein mediating homophilic adhesion (Grenningloh et al., 1990), is concentrated in border cells demarcating the proximal pretarsus border. Our results (Fig. 5Q,R) also shows that Bar is capable of inducing Fas II expression in cells distally adjacent to Bar-expressing cells. Thus, Bar may establish the boundary between the pretarsus (Bar-negative) and tarsal segment 5 (Bar-positive) by regulating the expression of cell adhesion molecules such as Fas II. Interestingly, BarX2, a mouse gene encoding a Bar-related homeodomain protein, has been reported to regulate the expression of Fas II-like cell adhesion molecules (Jones et al., 1997).

At early third instar, proximal neighbors of the Bar ring initiate folding in a Bar-dependent manner (Fig. 1G-G′′). Similarly, Bar concentration differences in future tarsal segments might be essential for the normal development of tarsal segments 3-5 (see Fig. 3P,S). Since folding and/or segmentation are likely to be caused by change in local cell adhesiveness, Bar may also regulate some cell adhesion molecule(s) responsible for central folding and/or distal-tarsus segmentation.

Mechanisms similar to antagonistic interactions between Bar and al may also be involved in vertebrate limb development. Hoxa11 and Hoxa13 are homeobox genes expressing in a region-specific manner in vertebrate limb buds. At early developmental stages, both genes are expressed in the distalmost region of the limb, although Hoxa11 expression expands more proximally than that of Hoxa13 (Yokouchi et al., 1991). At later stages, Hoxa11 and Hoxa13 expression domains are separated from each other through Hoxa11 repression in the Hoxa13 expression domain. As in the case of Bar expression in Drosophila pupal legs, the boundary between Hoxa11 and Hoxa13 expression domains appears intimately related to cartilage segmentation (Yokouchi et al., 1991).

As with Bar and Dll in Drosophila (this work; Campbell and Tomlinson, 1998; Wu and Cohen, 1999), Hoxa13 has been reported to regulate local cell adhesiveness (Yokouchi et al., 1995). Thus, the control of genes encoding cell adhesion proteins by region-specific transcription factors may be one of fundamental mechanisms involved in both insect and vertebrate development.

We thank C. S. Goodman, J. B. Thomas, S. M. Cohen, and G. Mardon for antibodies and G. Campbell for antibody and Cdna clone. We are also grateful to Y. Hotta, G. Mardon, I. Guerrero, R. Ueda, FlyView and Bloomington Drosophila stock center for fly strains. This work was supported in part by grants from the Ministry of Education, Science Culture and Sports of Japan to T. K. and K. S.