Limb type-specific regulation of bric a brac contributes to morphological diversity

Morphological diversity among animal appendages facilitates their many functions in moving, feeding and sensing the environment. In insects, the antenna is the primary olfactory and auditory organ while the thoracic legs serve mainly in terrestrial locomotion. Not only are different types of appendages unique in their forms, but there also exist many modified forms of the same appendage type. One feature of appendages that varies even among the insects is the number and morphology of their distal segments (Snodgrass, 1935). In the dipteran insect Drosophila melanogaster, the distal portion of the leg, the tarsus, is divided into five segments (t1-t5), while the homologous region of the antenna consists of only two segments plus part of a third (a4 and a5 plus distal a3) (Postlethwait and Schneiderman, 1971). By studying how limb segmentation is differently regulated in homologous limbs, we will gain insights into how morphological variations among appendages are achieved and into how they evolved.

Although several genes are known to be expressed in segmentally reiterated patterns in the antenna and leg, most of these are unlikely to be regulated differently between these appendage types. For instance, Serrate, Delta and fringe, which function in the Notch pathway, are expressed and required for the formation of all joints and are not being limited to homologous subdomains of antenna and leg (Bishop et al., 1999; de Celis et al., 1998; Mishra et al., 2001; Parody and Muskavitch, 1993; Rauskolb and Irvine, 1999). Other genes, such as Bar and apterous are similarly expressed in narrow proximodistal (PD) subdomains that correspond to homologous segments of the antenna and leg (Kojima et al., 2000; Larsen et al., 1996; Pueyo et al., 2000; Rincon-Limas et al., 1999; Tsuji et al., 2000). Thus although all of these genes contribute to limb segmentation, they probably do not contribute to the morphological differences between appendages.

In contrast, bric a brac (bab), which is required specifically for distal limb segmentation, is expressed in restricted PD subdomains and exhibits distinct expression patterns between antenna and leg. bab encodes a BTB/POZ domain protein necessary for normal development of most tarsal and distal antennal joints (Godt et al., 1993). bab mutations cause fusion of tarsal segments 2 through 5 (t2-t5) of the leg and of the fourth and fifth segments (a4, a5) and arista (ar) of the antenna (see Fig. 1A-D) (Godt et al., 1993). The mature bab pattern consists of two concentric rings in the distal antenna and in four concentric rings in the distal leg (see Fig. 2C′,F′). To understand how the antenna and leg develop different numbers of distal segments, we therefore have focussed on how bab becomes differentially expressed in the antenna and the leg.

bab expression in the leg is Distal-less (Dll)-dependent (Campbell and Tomlinson, 1998), and it has been proposed that the activation of bab by Dll in the leg is mediated by Spineless (Ss) (Duncan et al., 1998). Dll encodes a homeodomain transcription factor needed for the formation of the trochanter to the distal claws of the leg, and from a2 to arista of the antenna (Campbell and Tomlinson, 1998; Cohen and Jurgens, 1989; Dong et al., 2000; Gorfinkiel et al., 1997) (reviewed by Panganiban, 2000). ss encodes a bHLH-PAS family transcription factor related to the mammalian dioxin receptor and is required for patterning of the distal part of the first tarsal (t1) and second through fourth tarsal segments (t2-t4) (Duncan et al., 1998). In addition to their proximodistal (PD) patterning roles in the antenna and leg, both Dll (Cohen and Jurgens, 1989; Dong et al., 2000; Sato, 1984; Sunkel and Whittle, 1987) and ss (Balkaschina, 1929; Burgess and Duncan, 1990; Duncan et al., 1998; Struhl, 1982) are required for antenna specification.

The observations that bab is differentially expressed between antenna and leg (Godt et al., 1993) (this work) and that its putative activators, Dll and Ss have distinct expression patterns and functions in the two limb types (Dong et al., 2000; Dong et al., 2001; Duncan et al., 1998), led us to hypothesize that the regulation of bab was likely to vary between limbs. Here, we present evidence that supports this. In particular, we find that although Dll does serve as a bab activator in the antenna and also in the wing pouch, only part of this activation in the leg and none in the wing is mediated by Ss. Interestingly, the differences between antenna and leg expression of bab are not a consequence of differences in activators, but of differences in the expression of bab repressors. We conclude that bab is differentially regulated in different limbs by tissue-specific combinations of PD patterning genes. The unique gene networks that regulate bab in each limb type thereby contribute to morphological variation by regulating the number of distal segments.

Antibody stainings and immunohistochemistry were carried out as described previously (Halder et al., 1998). Antibodies used were: rabbit anti-Hth (Pai et al., 1998), rabbit anti-Dll (Panganiban et al., 1995), mouse anti-Dll (Vachon et al., 1992), rat anti-Bab (a gift from F. Laski), mouse anti-Myc (a gift from S. Blair), rabbit anti-β-gal (a gift from S. Carroll), rabbit anti-Sal (a gift from R. Schuh), mouse anti-Dac (University of Iowa Developmental Studies Hybridoma Bank) and rabbit anti-BarH1 (a gift from T. Kojima). Secondary antibodies coupled to Cy2, Cy3 and Cy5 were obtained from Jackson ImmunoResearch. Imaging was carried out on BioRad MRC1024 confocal and Zeiss Axiophot microscopes.

The following fly strains were employed: (1) dpp-GAL4 (A.3)/TM6B (Morimura et al., 1996); (2) act>CD2>GAL4 (=actin promoter-FRT-CD2-FRT-GAL4) (Pignoni and Zipursky, 1997); (3) w; UAS-GFP-hth8/TM6B, Tb, Hu (Casares and Mann, 1998); (4) w; FRT82B hth^P2 (Pai et al., 1998); (5) w; dac⁴ FRT40A, (Graeme Mardon); (6) w; FRT43D Dll^SA1 (Dong et al., 2000); (7) y hs-FLPase; FRT43D 2piM (Dong et al., 2000); (8) Dll¹/CyO, wg-lacZ; (9) Dll³/CyO, wg-lacZ; (10) bab^ARO7/TM6B, Tb Hu e ca (Frank Laski); (11) Df (3L) bab^AR07 th st cp FRT80B/TM3 (Artyom Kopp); (12) dac-lacZ (Graeme Mardon); (13) w; UAS-Dll/In (2LR) Gla, Gla Bc Elp (Konrad Basler); (14) Df(3R)ss^D114.4/TM6 (Ian Duncan); (15) ss^D114.7/TM6 (Ian Duncan); and (16) y hs-FLPase; FRT82B M piM (S. Blair). Stocks constructed by us for these experiments were: (1) y hs-FLPase; FRT82B 2piM; (2) y hs-FLPase; UAS-GFP-hth8/TM6B, Tb Hu; (3) y hs-FLPase; UAS-dac; (4) Dll^GAL4/CyO, wg-lacZ; (5) w; Df(2L)sal⁵ FRT40A; (6) y hs-FLPase; 2piM FRT40A; and (7) y hs-FLPase; UAS-sal.

Dll hypomorphic larval imaginal discs were generated by crossing heterozygous Dll mutant animals in which each Dll mutant chromosome was balanced over CyO, wg-lacZ. Mutant animals were identified by the absence of X-gal staining in the larval tails. Ectopic expression of Dll, homothorax (hth) and dachshund (dac) was induced using the GAL4/UAS binary system (Brand and Perrimon, 1993). dpp-GAL4 was used to activate UAS-GFP-hth along the anterior-posterior compartment boundary of the developing imaginal discs. Clones of cells ectopically expressing Hth, Dac or Sal were generated using a modified GAL4/UAS system (Pignoni and Zipursky, 1997) in which y hs-FLPase; UAS-GFP-hth/TM6B, Tb Hu (or y hs-FLPase; UAS-dac or y hs-FLPase; UAS-sal) flies were crossed to act>CD2>GAL4 flies. The resulting larvae were heatshocked at 37°C for 10 minutes at 72-96 hours after egg laying (AEL) to induce site-specific recombination between the FRT sites, which in turn results in constitutive GAL4 expression in the clones. Dll, hth, dac and sal null clones were generated using the FLP/FRT system (Xu and Rubin, 1993). Animals of the genotypes: (1) y hs-FLPase; FRT43D 2piM/FRT43D Dll^SA1; (2) y hs-FLPase; FRT82B piM/FRT82B hth^P2; (3) y hs-FLPase; 2piM FRT40A/dac⁴ FRT40A; and (4) y hs-FLPase; 2piM FRT40A/Df(2L)sal⁵ FRT40A were heatshocked at 37°C for 1 hour at 48-72 hours AEL and examined in mid- to late-third instar. To make large hth null clones, the hth⁺ FRT chromosome carried a Minute (M) mutation. Animals of the genotype: y hs-FLPase; FRT82B M piM/FRT82B hth^P2 were heatshocked at 37°C for 1 hour at 120-144 hours AEL and examined in mid- to late-third instar (because the heterozygous Minute animals develop slowly, this was ∼240-288 hours AEL). The ss null genotype examined was Df(3R)ss^D114.4/ss^D114.9. The bab null genotype examined was Df(3L)bab^ARO7/Df(3L)bab^AR07 th st cp FRT80B/TM3.

bab is required for the proper segmentation of the distal leg and antenna. Segmental fusion is seen in a bab null leg and antenna (Fig. 1A-D) (Godt et al., 1993). The two distalmost tarsal joints are most sensitive to loss of bab, but fusion of all tarsal segments is seen in bab null animals. This also results in shortening the overall length of the tarsus (Godt et al., 1993). In the antenna, because of loss of the a4/a5 and a5/arista joints, the arista is fused to a4. Consistent with its function, bab is expressed at the distal end of each tarsal segment during pupal leg development and is expressed in distal a3 and in the a4 and a5 segments of the antenna (Fig. 1E,F).

In order to gain insight into possible positive and negative regulators of bab expression during appendage development, we first examined the dynamics of wild-type bab expression. In late second/early third instar antenna and leg discs, bab is expressed throughout most of the Dll domain, with the highest levels of expression in the centers of both antennal and leg discs (Fig. 2A,A′,D,D′). By mid-third instar, bab expression is lost from the distalmost cells, and a ring of Bab is apparent in the antennal and leg discs (Fig. 2B,B′,E,E′). Still later in the third instar, the bab expression pattern resolves into a set of concentric rings, two in the antenna and four in the leg (Fig. 2C,C′,F,F′). Based on its wild-type expression, we postulated that bab is activated initially throughout the Dll domain, and that the mature bab expression patterns are achieved in part by subsequent repression distally and proximally. In addition, bab either is partially repressed or its expression is not maintained between concentric rings.

To examine the relationship between bab and Dll in the antenna, leg and wing, we investigated bab expression in Dll hypomorphic discs and in clones of cells lacking Dll. Bab is not detected in either Dll hypomorphic antenna (Fig. 3A) or leg (Campbell and Tomlinson, 1998) (not shown) discs, or in Dll null clones in either the antenna or leg (Fig. 3B,C,C′). Interestingly, although bab expression is lost cell-autonomously in Myc-marked Dll null clones in both antenna and leg (arrows in Fig. 3B,C,C′ and not shown), bab expression is also lost non cell-autonomously under certain conditions. In the antenna, for instance, bab expression is lost in cells surrounding the Dll null clones when those clones are near the boundary of the normal dac and bab expression domains (arrowheads in Fig. 3B). Non cell-autonomous loss of the modulation between bab rings is also seen occasionally near these Dll null clones. These phenomena likely are due to the non cell-autonomous activation of dachshund (dac) around the Dll null clones (Dong et al., 2001), since Dac is a bab repressor (see below and Discussion). In the leg, we also observe non cell-autonomous loss of bab expression around Dll null clones (e.g. arrowhead in Fig. 3C,C′). However, in this case Dll expression is also lost (J. C. and G. P., unpublished data). Thus the non cell-autonomous loss of bab expression in the leg is likely due to the loss of the bab activator Dll, and mechanistically different from the non cell-autonomous loss of bab expression in the antenna.

bab expression partially overlaps with Dll in the wing pouch. Although bab expression is not visibly reduced in Dll hypomorphic wings (not shown), bab expression is lost in some Dll null clones in the wing pouch (arrow in Fig. 3D,D′). There is no detectable adult phenotype in bab null wings (not shown). We conclude that bab activation requires Dll in the antenna, in the leg and in part of the wing pouch.

ss null mutations result in loss of part of t1 and of the t2, t3 and t4 segments of the leg, including their joints. Thus in ss null legs, there remains only a single distal joint positioned between t1 and t5 (Duncan et al., 1998). Consistent with the adult cuticular phenotype, in late third instar ss null animals, we observe a single ring of bab expression in the leg (Fig. 4A,A′). Double labeling of the ss mutant leg discs with both Bar antibodies that label the t4 and t5 segments (Kojima et al., 2000) and Bab antibodies (not shown), indicates that the remaining Bab represents the distal-most ring. Thus while the proximal tarsal rings of bab depend on Ss, the distal ring expression of bab is Ss-independent. Similarly, in ss null antennal discs, there is a distal ring of bab expression (Fig. 4B,B′). This is consistent with the presence of a single distal joint, between the transformed a3 and the transformed arista of ss null animals (Duncan et al., 1998). However, because the ss null antenna exhibit transformations toward leg, it is difficult to evaluate the requirements for ss in normal antennal joint formation. Thus, while we can conclude that Ss mediates only part of the Dll activation of bab in the leg, the relationship between ss and bab during normal antennal development remains unknown.

Although bab is expressed throughout most of the Dll domain at early third instar, by late third instar, bab becomes restricted to a portion of the Dll domain. We therefore hypothesized that there are bab repressors that are differentially expressed in the complementary part of the Dll domain. One candidate bab repressor in the leg is Dac. Dac is a novel nuclear protein required for patterning of the trochanter, tibia and proximal tarsal segments (Mardon et al., 1994). The distal boundary of dac expression is dynamic, encompassing increasingly more distal segments as the leg disc grows (Lecuit and Cohen, 1997). When dac is first activated in the leg at late second instar, its expression overlaps only slightly with that of Dll. However, by late third instar, dac expression also includes the tibia and the first and second tarsal segments (t1 and t2) of the leg disc, where it overlaps with Dll (Lecuit and Cohen, 1997) and with the t1 and t2 bab rings (Fig. 5A,A′). To test whether Dac is a repressor of bab in the leg, we carried out both loss- and gain-of-function experiments. We observe strong activation of bab in dac null clones (asterisk in Fig. 5B,B′) and repression of bab by ectopic Dac (Fig. 5D,D′). Therefore, Dac is a repressor of bab in the leg. We note that bab can be non cell-autonomously activated near dac null clones (arrowhead in Fig. 5B,B′), although we do not yet know the mechanism by which this occurs. Dll also is activated in the dac null clones that exhibit strong bab expression and Dll is repressed by ectopic Dac (Abu-Shaar and Mann, 1998; Dong et al., 2001). Thus Dac repression of bab in the leg could be mediated in part by reducing Dll activity. Nonetheless, the fact that Dll, dac and bab are coexpressed in t1 and t2 by late third instar, indicates that Dac is insufficient to repress Dll or bab.

In addition to setting the proximal boundary of the bab domain in the leg, Dac also plays a role in modulating bab expression between rings. In strong dac hypomorphic leg discs, Bab loses its inter-ring modulation and is uniformly expressed at high levels in t1-t3 (Fig. 5E,E′). Interestingly, the uniform Bab expression in dac null legs is correlated with phenotypes similar to bab loss of function, namely the t1-t3 tarsal segments and joints are lost (P. D. S. D. and G. P., unpublished results). Thus, alternating high and low levels of Bab appear to be critical for tarsal joint formation.

It has been proposed that mutually repressive interactions in the leg, e.g between Bar and aristaless (al) (Kojima et al., 2000) and between Dll and dac (Dong et al., 2001), play critical roles in the subdivision of the PD axis. We therefore wished to test whether the antagonism of Dac for Bab is reciprocal. In bab null antenna and leg discs, Bab protein cannot be detected, but dac expression is normal (Fig. 5C and F). Consistent with this, dac is expressed in a medial ring instead of a circle that encompasses the bab domain prior to bab activation (not shown). Thus Dac represses bab, but Bab does not repress dac.

We have also investigated bab regulation in the developing wing. In the wing pouch, bab expression resembles that of Dll, except that bab is not expressed in the middle of the wing pouch along the anterior-posterior compartment boundary. This absence of detectable bab expression coincides with the domain of spalt (sal) expression (Fig. 6A). sal encodes a zinc finger transcription factor (Kuhnlein et al., 1994) required to position the L2 wing vein (de Celis et al., 1996; Sturtevant et al., 1997). To test whether Sal represses bab, we made Df(2L)sal⁵ clones. We found that bab is derepressed in these sal null clones (arrowheads in Fig. 6B,B′,B′′). Therefore, bab expression is restricted by Sal in the wing. bab activation in the sal null clones is unlikely to be mediated via Dll, since Dll is not expressed or activated in many of these clones (arrowheads in Fig. 6B′′′). bab is activated similarly in sal null clones in the haltere (Fig. 6C,D) and this is not mediated by Dll either (not shown). In fact, the wild-type expression of bab does not overlap with Dll in the haltere (not shown). Together, our results indicate that some, but not all, bab activation in the wing is mediated by Dll and that Sal is a potent repressor of bab in both wing and haltere.

As demonstrated above, Dac represses bab in the leg, and Sal represses bab in the wing pouch. The expression of both Dac and Sal are more or less complementary to that of bab in the antenna, as is that of Hth (Fig. 7A,E,H). Thus all three genes are potential bab repressors in the antenna. Hth is a TALE homeodomain transcription factor required for the nuclear localization of Extradenticle (Exd), a Pbx-related homeodomain protein (Abu-Shaar et al., 1999; Kurant et al., 1998; Pai et al., 1998; Rieckhof et al., 1997). Both Hth and Exd are required for patterning of the coxa and trochanter of the leg and of the first and second antennal segments (Abu-Shaar and Mann, 1998; Aspland and White, 1997; Gonzalez-Crespo and Morata, 1995; Gonzalez-Crespo and Morata, 1996; Pai et al., 1998; Rauskolb et al., 1993; Rauskolb et al., 1995; Rieckhof et al., 1997). In addition, Hth and Exd are required cell-autonomously throughout the antenna to specify antennal identity (Casares and Mann, 2000; Gonzalez-Crespo and Morata, 1995; Pai et al., 1998). The antennal expression of both dac and sal is dependent on Hth and Exd (Dong et al., 2000; Dong et al., 2001). Dac is coexpressed with Hth, nuclear Exd (n-Exd) and Sal in a3 (Fig. 7A and not shown), while Sal is coexpressed strongly with Hth and n-Exd in a2 (Fig. 7E and not shown). The expression of Sal in a3 is strong at early third instar and weak at late third instar. Thus the combination of Hth, n-Exd, Dac and Sal marks the proximal limit of bab expression. To test whether these genes repress bab to set its proximal boundary, we carried out both loss- and gain-of-function experiments.

bab is derepressed in distal a3 in dac null clones (Fig. 7B), and the bab domain expands proximally in strong dac hypomorphs (Fig. 7C). dac null clones and dac null antennal discs do not exhibit derepression of bab in a1 or a2 (Fig. 7C and not shown). Further, bab is repressed in flipout clones ectopically expressing Dac (Fig. 7D). Based on these results, we conclude that Dac is a bab repressor in distal a3. As in the leg, loss of inter-ring modulation of bab is observed in the antenna in strong dac hypomorphs (Fig. 7C) and this loss is correlated with loss of the a4/a5 and a5/arista joints (P. D. S. D., J. S. Dicks and G. P., unpublished data). Thus modulation of bab expression appears to be necessary for distal antennal as well as distal leg joint formation.

bab is derepressed weakly in sal null clones in a2 (Fig. 7F) and repressed in flipout clones ectopically expressing Sal (Fig. 7G). sal null clones do not exhibit derepression of bab outside of a2 (Fig. 7F and not shown). We conclude that Sal is a bab repressor in a2. bab often is derepressed in hth null clones in both a2 and a3 (Fig. 7I). We have shown previously that while sal is lost in hth null clones (Dong et al., 2000), dac is derepressed in hth null clones in both a2 and a3 (Dong et al., 2001). Thus hth null clones in a2 and a3 often coexpress bab and its repressor dac. This supports the leg data described above indicating that Dac is insufficient for bab repression. We conclude that Dac probably requires Hth (or the product of a gene activated by Hth) as a corepressor in the antenna.

Interestingly, in large hth null clones that encompass much of the distal antenna, bab often (40/66 discs=61%) is derepressed uniformly throughout a2 and a3 (Fig. 8A), while at other times (26/66 discs=39%) the derepressed bab is modulated and appears in rings. Among the hth mutant antennal discs in which the derepressed bab is modulated, we frequently (11/26 discs=42%) see four rings (Fig. 8C). Thus large clones with four rings constitute 17% (11/66) of the total large hth null clones examined. This frequency resembles that (16%; 10/61) at which five distinct tarsal segments separated by four joints can be detected in adult antennae harboring large hth null clones (Fig. 8D). In these experiments, the remaining 51/61 antennae, while visibly transformed toward leg, possessed three or fewer distinct tarsal-like joints and four or fewer tarsal segments (Fig. 8B). Thus inter-ring modulation in hth mutant antennae can be correlated with joint formation, and bab modulation probably is essential for normal distal joint formation in both the antenna and the leg.

Ectopically expressed Hth can repress bab in the antenna (Fig. 7J), but expression of dac (Dong et al., 2001) and sal (Dong et al., 2000) is also activated in these cells. We therefore cannot distinguish whether ectopic Hth is sufficient for bab repression or whether Hth cooperates with Dac and Sal to repress bab. Nonetheless, the loss- and gain-of-function data presented here indicate that all three factors function in limiting bab expression in the antenna. Sal and Hth represses bab in a2, while Dac and Hth repress bab in a3.

The precision of pattern formation in development is the result of fine control and balance of dynamic webs of gene regulation. Here we have investigated a portion of the patterning processes in the appendages of Drosophila. The differences in how bab is regulated in different limb types and within a single limb type along the PD axis indicate that there are multiple ways to alter bab expression by modulating activator and/or repressor activity. Because bab is required for formation of distal joints, these possibilities are likely to facilitate variations in where joints form and how many segments a limb possesses, thereby contributing to the morphological diversity of legs and antennae.

Both ss and bab expression depend on Dll in the antenna and leg (Campbell and Tomlinson, 1998; Duncan et al., 1998) and this work). We report here that only a subset of the Dll-dependent bab expression depends on Ss. In particular, the tarsal rings of bab expression in distal t1, distal t2 and distal t3 are both Dll and Ss dependent, whereas the ring of bab in distal t4 depends on Dll but not on Ss. Thus for the t1, t2 and t3 rings, Ss either mediates the Dll activation of bab or cooperates with Dll to activate bab, while the t4 ring is activated independent of Ss, possibly directly by Dll (Fig. 9). bab expression in the t4 ring may be regulated via a different enhancer than bab expression in the t1, t2 and t3 rings, but whether or not there are distinct bab enhancers for different rings is unclear. We think it likely that bab expression in each of the four tarsal rings represents differential activation in response graded Dpp and Wg signals in conjunction with the Dll and/or Ss selector(s).

In the antenna, the effects of Ss on bab are difficult to interpret because ss null antennae exhibit transformations toward leg. In these transformed discs, the bab expression pattern resembles that of ss null leg discs, namely only a single distal ring is observed. This may represent the distal leg ring. However, since bab enhancer elements have not yet been identified, we cannot determine whether the distal ring in both antenna and leg represents activation via a shared enhancer or utilizes distinct enhancers in each limb type.

We find that only part of the bab expression depends on Dll in the wing, and that in a homologous structure, the haltere, bab is not Dll-dependent. It has been suggested that the insect wing may derive from a proximodorsal part of a branched limb of their aquatic ancestors (Averof and Cohen, 1997; Kukalova-Peck, 1978). There is expression of bab proximal to the Dll domain in both antenna and leg discs. Thus if the wing has derived from the proximal leg, it could be that the regulation of bab expression in the wing shares features with that of the proximal leg.

Here we have demonstrated that repression by Dac establishes the proximal border of bab expression in the leg, that Sal restricts bab expression in the wing, and that Dac, Sal and Hth establish the proximal limit of bab expression in the antenna (Fig. 9). We have shown previously that coexpression of Dll and Hth/Exd determines antennal identity, and activates sal (Dong et al., 2000) and dac (Dong et al., 2001). Here we demonstrate that differential expression of these genes leads to significant differences in bab expression. In particular, bab expression is restricted to a small domain containing two rings in the antenna in contrast to a larger domain of four rings in the leg. As bab is necessary for joint formation, we propose that these differences in bab expression contribute to the differences in the number of distal segments found in the antenna and the leg.

In both antenna and leg discs, bab expression is derepressed in dac null clones that are proximal to normal bab expression domain. However, the mechanisms of bab derepression probably differ between limb types. In the leg, derepression of bab is associated with derepression of Dll. Therefore, Dac likely restricts bab expression via bab activators. In the antenna, Dll and dac expression normally overlap, and Dll continues to be expressed in dac null clones. This suggests a more direct repression mechanism may exist in the antenna. There is substantial overlap between dac and bab expression in the leg at late third instar, indicating that the presence of Dac protein is not sufficient for bab repression and that the presence or absence other factors is required in order for Dac to repress bab. This is perhaps not surprising since Dac does not contain a known DNA binding motif and forms complexes with Eyes absent and Sine oculis to regulate transcription during eye development (Chen et al., 1997; Pignoni et al., 1997). Based on existing genetic information, candidate Dac corepressors are Sal, Hth and n-Exd in the antenna, and Antennapedia (Antp) in the leg. Alternatively, because Dac overlaps with Dll in both the antenna and leg where Dac serves as a bab repressor, it also is possible that Dac functions by precluding the ability of Dll and/or Ss to activate bab.

At early third instar, bab is expressed throughout the Dll domain in the leg, with uniformly strong expression in the presumptive tarsal segments and weaker expression in presumptive femur and tibia. By late third instar, bab expression is modulated in the presumptive tarsus and is strongest at the distal part of tarsal segments 1-4, immediately proximal to each of the tarsal joints. Weaker expression of bab is observed between these concentric rings. This modulation persists at least through early pupal stages. Various mechanisms by which this modulation is achieved can be imagined, including both active repression and/or the failure to maintain bab in the inter-rings.

We have indirect evidence suggesting that active repression plays a role. Namely, in dac null antenna and leg discs, there is loss of bab inter-ring modulation as well as proximal derepression. This lack of inter-ring modulation correlates with joint loss and segment fusion in the dac null animals (Dong et al., 2001) (this work). In the leg, bab modulation by Dac occurs cell-autonomously. However in the antenna, there is loss of bab modulation even in cells that do not normally express detectable levels of Dac. Therefore in addition to repressing bab cell-autonomously, Dac also may non cell-autonomously regulate the production of bab inter-ring repressor(s) (Fig. 9). Since Dac is a nuclear protein, it could be that Dac directs the expression of a secreted molecule that in turn mediates inter-ring modulation.

We have presented evidence that bab is differentially regulated in different Drosophila appendage types and that this regulation contributes to variations in their morphology. It is necessary to determine whether bab is required for limb segmentation in other animals, and, if so, whether variations in bab regulation contribute to the interspecific variations of homologous appendages. Investigation of the genetic hierarchies that pattern the appendages of other arthropods and of closely related phyla are needed to address these issues and to provide further insights into the generation of morphological diversity.

We are grateful to John Fallon and Artyom Kopp for comments on the manuscript and to Gerard Campbell for sharing unpublished results. The insightful comments of an anonymous reviewer led us to investigate certain aspects of bab regulation in more detail and to improvements in the manuscript. We thank Sean Carroll for access to his confocal microscope, Carol Ditack for assistance with figures, Richard Mann for UAS-GFP-hth Drosophila lines, Konrad Basler for UAS-Dll Drosophila lines, Henry Sun for the FRT-hth Drosophila line, Adi Salzberg for the rabbit Hth antibody, Frank Laski for bab mutants and antibody, Graeme Mardon for dac mutants, Dianne and Ian Duncan for ss mutants, the University of Iowa Developmental Studies Hybridoma Bank for the Dac antibody, and the Bloomington Drosophila Stock Center for many of the Drosophila lines used in these experiments. J. C. and P. D. S. D. were supported by NIH Predoctoral Training Grant T32-HD07477. G. P. is the recipient of a Young Investigator Award in Molecular Studies of Evolution from the Sloan Foundation and the National Science Foundation. This work was supported in part by NIH grant GM59871-01A1 to G. P.