RESULTS

The methodology for generating the pb exd and pb Scr exd clones: Addressing the epistatic interaction between Scr and exd required a mosaic analysis using Flip-mediated mitotic recombination (Golic 1991; Xu and Rubin 1993). The locations of the pb Scr loci on the third chromosome and the exd locus on the first chromosome are a major technical problem. This problem was overcome by moving the exd⁺ locus from the first chromosome to the right arm of the third chromosome. To move the exd⁺ locus, Cordelia Rauskolb kindly supplied a P{w⁺, exd⁺} insertion on the third chromosome (Rauskolbet al. 1993), which we mapped genetically to the distal tip of 3R (3-107.8). The position of P{w⁺, exd⁺} allowed the construction of the marked P{ry⁺, neo^r, FRT}82B M(3)95A² P{y⁺, ry⁺}96E P{w⁺, exd⁺} chromosome. This chromosome was used to set up the genotype shown in Figure 1 (see materials and methods). When a mitotic recombination event occurs between the FRT sites, the yellow clones of cells are exd¹, because P{w⁺, exd⁺} is lost, and are homozygous for pb²⁷ or pb²⁷ Scr².

We tested whether this system for generating exd¹ clones would reproduce previously observed phenotypes. The adult phenotypes of loss-of-EXD activity have been documented previously (González-Crespo and Morata 1995; Rauskolbet al. 1995). All the head phenotypes were reproduced. This includes the arista-to-tarsus transformation (Figure 2a), the antenna-to-second-leg transformation (Figure 2b), and ectopic eye formation, which is sometimes associated with a leg (Figure 2c). However, clones induced at the first/second-instar larval stage with the P{ry⁺, neo^r, FRT}82B M(3)95A² P{y⁺, ry⁺}96E P{w⁺, exd⁺} chromosome did not include the transformation of proboscis tissue into knobs of tissue with bracted bristles, as was reported in a gnandromorph analysis (Rauskolbet al. 1995). This transformation was observed when clones were induced during embryogenesis with the P{ry⁺, neo^r, FRT}82B Sb^63a P{w⁺, πM }87E P{y⁺, ry⁺}96E P{w⁺, exd⁺} chromosome (Hayden 1996). Clones generated by the loss of the ring X chromosome are generated early during embryogenesis, suggesting that, for this transformation of the proboscis, exd¹ clones must be generated during embryogenesis (Rauskolbet al. 1995).

Loss-of-EXD activity is epistatic to loss-of-SCR activity: Epistasis was tested by determining whether pb exd and pb Scr exd proboscis cells adopted aristal or tarsal identity. Clones of exd¹ proboscis cells induced during the first/second larval stage were wild type (Figure 3b). Both pb²⁷ and pb²⁷ Scr² clones of ectodermal proboscis cells adopted aristal identity (Figure 3, c and e; Table 2). Both pb²⁷ exd¹ and pb²⁷ Scr² exd¹ clones of ectodermal proboscis cells adopted tarsal identity (Figure 3, d and f; Table 2). Hence, loss-of-EXD activity is epistatic to loss-of-PB and loss-of-PB SCR activity. We have observed the same result with the P{ry⁺, neo^r, FRT}82B Sb^63a P{w⁺, πM }87E P{y⁺, ry⁺}96E P{w⁺, exd⁺} chromosome inducing clones during embryogenesis (Hayden 1996).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Generation of clones for the analysis of epistasis. In the example shown, pb²⁷ Scr² exd¹ clones are being generated. The genotype of the first, second, and right arm of the third chromosomes are indicated in A. The right arm of the third chromosome is shown at the G2 phase of the cell cycle when site-specific recombination generates detectable mitotic recombination events. The results of these events after cytokinesis are shown at the bottom in B and C with only the genetic constitution of the first and third chromosomes indicated.

In a pb²⁷ Scr² mosaic analysis, the proboscis is transformed into a complete antenna that includes the third antennal segment and arista (Percival-Smithet al. 1997). However, clones of pb²⁷ Scr² exd¹ cells did not adopt third-antennal-segment identity (Table 2). This indicates that pb²⁷ Scr² exd¹ clones of ectodermal proboscis cells were transformed to second-leg identity; however, we were unable to assign the exact identity of pb²⁷ Scr² exd¹ clones in the proboscis because an unambiguous apical bristle, which is a specific marker of second-leg identity, was not observed.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Mosaic analysis with FLP-mediated mitotic recombination reproduces previously observed phenotypes. (a–c) Clones generated with the P{ry⁺, neo^r, FRT} M(3)95A² P{y⁺, ry⁺} 96E P{w⁺, exd⁺} chromosome and induced at 36–60 hr after egg laying. a shows an arista-to-tarsus transformation, b shows an antenna-to-second-leg transformation, and c shows a rostrum-to-ectopic eye, with associated leg, transformation. (d) A clone induced during embryogenesis and generated with the P{ry⁺, neo^r, FRT} Sb^63b P{w⁺, πM}87E P{y⁺, ry⁺} 96E P{w⁺, exd⁺} chromosome. The clone of exd¹ proboscis cells has bristles with bracts. The clones in a and d are enclosed with dashed lines, tarsal claws are shown by arrowheads, and the apical bristle by an arrow.

SCR activity required for sex comb determination does not require EXD activity: Clones of pb²⁷ cells in the proboscis form sex combs. We suggest that this is because SCR activity is released from an interaction with PB (Percival-Smithet al. 1997). SCR activity is required cell autonomously for sex comb determination (Struhl 1982). Clones of pb²⁷ exd¹ cells in the proboscis were transformed to tarsal identity (Figure 3d), and in addition these clones formed sex combs (Figure 4; Table 2). This indicated that SCR activity does not require EXD activity in determination of sex combs, and pb²⁷ exd¹ clones in the proboscis were transformed into tarsi with first-leg identity. The transformation of the proboscis to first-leg identity requires SCR activity, as no sex comb formation was observed in pb²⁷ Scr² exd¹ clones (Table 2).

Intracellular localization of EXD protein: Our genetic analysis suggests that EXD activity is turned OFF in the ectoderm as a result of SCR expression in the mesoderm and that SCR activity does not require EXD activity for sex comb determination. We investigated whether the ON/OFF state of EXD activity correlated with a change in accumulation or localization of EXD protein. We performed double immunolocalization of SCR and EXD in late third-instar larval leg and antennal imaginal discs (Figure 5, a–h). In all leg imaginal discs, EXD protein was cytoplasmically localized in the central (distal) region of the discs (González-Crespo and Morata 1996; Aspland and White 1997). SCR was expressed and localized in the nuclei of the ectodermal cells of the T1 leg imaginal disc (Glickman and Brower 1988). In the region that includes the primordium of the sex combs, SCR protein is highly expressed, but EXD protein is cytoplasmically localized (Figure 5, a–c; Glickman and Brower 1988; Pattatucci and Kaufman 1991). Since EXD activity is proposed to be inactive when cytoplasmically localized (Rieckhofet al. 1997), EXD may be inactive in the sex comb primordia of the ectoderm, which may explain why SCR activity required for sex comb determination does not require EXD activity. SCR activity is required at late third-instar larval stage for sex comb determination (Pattatucciet al. 1991).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Analysis of epistasis of pb²⁷ and pb²⁷ Scr² with exd¹. (a–f) Clones in the proboscis representing a distinct genotype. a and b are pb⁺ Scr⁺, c and d are pb²⁷ Scr⁺, and e and f are pb²⁷ Scr². a, c, and e are exd⁺, and b, d, and f are exd¹. Clones of yellow ectodermal tissue are enclosed within dashed lines in a, b, d, and f. The asterisks show aristal structures, arrows show tarsal claws and the arrowhead shows third antennal segment identity.

In all late third-instar larval leg imaginal discs, SCR was expressed in the adepithelial cells and EXD was cytoplasmically localized in the distal region of the discs (Figure 5, d–f; Glickman and Brower 1988; Aspland and White 1997). This suggests that EXD activity is OFF in the determination of the tarsi, and, as SCR is not expressed in the adepithelial cells of antennal discs of late third-instar larvae (Glickman and Brower 1988; Figure 5, g and h), it may have been expected that EXD protein would be nuclearly localized. SCR was expressed in the disc stack of the antennal disc, but not the adepithelial cells (Figure 5, g and h); however, EXD was not nuclearly localized throughout the antennal disc, but rather was cytoplasmically localized in the distal region that is the primordium of the arista (Figure 5, g and h; González-Crespo and Morata 1995; Rieckhofet al. 1997). This suggests that EXD activity is OFF in the arista primordia of late third-instar larval discs; yet EXD activity is required for aristal determination (González-Crespo and Morata 1995; Rauskolbet al. 1995; Figure 2a). This paradox is resolved by the observation that SCR activity is required at the late second/early third-instar larval stage for tarsus determination (Percival-Smithet al. 1997). When we assayed EXD expression and localization in early third-instar larval imaginal discs, EXD protein was nuclearly localized both proximally and distally in the ectoderm of antennal discs (Figure 5, i and j), but was cytoplasmically localized distally in leg imaginal discs (Figure 5, k and l).

To confirm that the localization of EXD correlates with determination of tarsus vs. arista identity, EXD localization was assayed in genetically manipulated situations. Ectopic expression of PB^Q50Ka at early third-instar larval stage was induced by administering a 20-min, 36.5° heat-shock to P{hsp/pb^Q50Ka, ry⁺}A larvae; the larvae were allowed to recover for 3 hr. EXD was nuclearly localized in the leg imaginal discs of these larvae (Figure 5m), which correlates with the tarsus-to-arista transformation observed with ectopic expression of PB^Q50Ka (Percival-Smithet al. 1997). We also assayed EXD localization in labial imaginal discs of early third-instar larvae derived from pb⁴/pb²⁰ and pb²⁷/pb²⁰ mutants. We found that EXD was nuclearly localized in pb⁴/pb²⁰ labial discs (Figure 5n) and cytoplasmically localized in the distal cells of pb²⁷/pb²⁰ labial discs (Figure 5o), correlating with the aristal and tarsal identity adopted by these transformed labial imaginal discs, respectively. The cytoplasmic localization of EXD in pb²⁷/pb²⁰ labial imaginal discs was restricted to the most distal cells and was not as extensive as observed in normal leg imaginal discs (compare Figure 5, o with k). This difference in localization of EXD may explain the difference between the structure of a normal T1 leg and the transformed proboscis of pb null mutants. The leg transformation of pb null mutants is a near perfect tarsus but the tibia and femur are short and fused (Kaufman 1978). The clear cytoplasmic location of EXD in the center of the pb²⁷/pb²⁰ labial disc explains the strong tarsus transformation of the proboscis, and the mixed and nuclear localization of EXD in the region around the center of the pb²⁷/pb²⁰ labial disc may explain the poor formation of the tibia and femur (Figure 5o).