Requirement of abdominal-A and Abdominal-B in the developing genitalia of Drosophila breaks the posterior downregulation rule

The Hox genes specify the morphological diversity along the anteroposterior axis of most animal species (Mann and Morata, 2000; McGinnis and Krumlauf, 1992). Hox proteins contain a DNA-binding domain, the homeodomain (Gehring et al.,1994), and act as transcription factors, controlling the expression of different downstream genes(Graba et al., 1997). The outcome of Hox gene activity is the formation of different structures. In fact, one remarkable aspect of some Hox mutations is the transformation of one structure into another one (Mann and Morata, 2000; McGinnis and Krumlauf, 1992).

These transformations are particularly striking in Drosophilaappendages. For example, the ectopic expression of the Hox gene Antennapedia in the antennal primordium converts the antenna into a leg (Frischer et al., 1986; Jorgensen and Garber, 1987; Schneuwly et al., 1987a; Schneuwly et al., 1987b). In some cases, homeotic transformations have revealed the hidden appendage structure of certain organs. Thus, the genitalia of Drosophila derive from a primordium with appendage-like characteristics(Gorfinkiel et al., 1999), and are indeed transformed into a leg or an antenna in the absence of the Hox gene Abdominal-B (Abd-B)(Estrada and Sánchez-Herrero,2001).

The genitalia and analia (collectively known as terminalia) are ectodermic structures located at the posterior region of the adult that derive from the genital disc. This disc is the only unpaired disc of Drosophila, and is formed by the fusion of three primordia corresponding to the eighth (A8),ninth (A9) and tenth (A10) abdominal segments. The different development of these primordia in males or females depends on the sex-determination signals. In females, the A8 segment forms the female genitalia and the eighth tergite,the A9 forms the parovaria and part of the uterine wall (both belonging to the internal genitalia), and the A10 forms the female analia. In males, the A8 gives rise to a tiny A8 segment, the A9 to the male genitalia and the A10 to the male analia (Keisman et al.,2001; Nöthiger et al.,1977; Schüpbach et al.,1978). The morphology of these structures, however, also depends on two other groups of genes: one group is formed by genes such as engrailed, hedgehog, decapentaplegic (dpp) and wingless, genes that are involved in signaling pathways(Casares et al., 1997; Chen and Baker, 1997; Emerald and Roy, 1998; Freeland and Kuhn, 1996; Keisman and Baker, 2001; Sánchez et al., 1997; Sánchez et al., 2001);the other group includes the Hox gene Abd-B, which is needed for the formation of the genitalia (Casanova et al., 1986; Celniker et al.,1990; Estrada and Sánchez-Herrero, 2001; Karch et al., 1985; Sánchez-Herrero et al.,1985; Tiong et al.,1985), and the Hox-like gene caudal (cad), which is required for analia development (Moreno and Morata, 1999). It is the combined activity of these three sets of elements (sex-determination genes, signaling pathways and Hox genes) that shapes the terminalia of the adult fly (reviewed by Christiansen et al., 2002; Estrada et al., 2003; Sánchez and Guerrero,2001).

Homeotic gene function in the genital primordia, however, is more complex than what we have just described. First, the abdominal-A(abd-A) Hox gene, required for the development of A1-A8 in the embryo(Sánchez-Herrero et al.,1985; Tiong et al.,1985), is expressed in the A8 of the female genital disc(Casares et al., 1997; Freeland and Kuhn, 1996). No function has been ascribed yet to such expression. Second, Abd-B is a complex gene: the use of four different promoters and the existence of specific exons give rise to several transcripts that encode two different proteins. The A (m) transcript encodes the Abd-B M (or Abd-B I) protein, and the B, C (r) and γ RNAs encode the Abd-B R (or Abd-B II) protein (Celniker et al., 1989; DeLorenzi et al.,1988; Kuziora and McGinnis,1988; Zavortink and Sakonju,1989). The Abd-B M protein has 221 amino acids more than the Abd-B R product does in its N-terminal domain but both proteins share a common C-terminal region, which includes the homeodomain(Celniker et al., 1989; Zavortink and Sakonju, 1989). In the embryonic epidermis, the Abd-B M transcript and protein are expressed in parasegments (PS) 10-13 (A5-A8 segments), whereas the Abd-B R transcript and protein are present in PS14-PS15 (A9-A10) initially, and in PS14 (A9) at late stages (Boulet et al.,1991; Celniker et al.,1989; DeLorenzi et al.,1988; DeLorenzi and Bienz,1990; Kuziora and McGinnis,1988; Sánchez-Herrero and Crosby, 1988). The γ RNA is transcribed in just a few cells of PS14 or PS15 (Kuziora and McGinnis, 1988).

The role of Abd-B M and Abd-B R products in genital development remains unclear. Abd-B m mutations transform the A5-A8 segments into the A4 segment, both in males and females; the female genitalia are lost whereas male genitalia remain intact (Casanova et al.,1986; Karch et al.,1985; Sánchez-Herrero et al., 1985; Tiong et al.,1985). By contrast, mutations in Abd-B r function eliminate genitalia and analia in both sexes(Casanova et al., 1986). Significantly, the transformations obtained in either Abd-B m or Abd-B r mutants clearly differ from those observed when all Abd-B functions are eliminated: in some of the clones mutant for Abd-B (m and r), part of the male or female genitalia are transformed into leg or antenna(Estrada and Sánchez-Herrero,2001). Therefore, the precise role of abd-A, Abd-B m and Abd-B r in genitalia development is not well defined.

We have analyzed homeotic expression and requirement in terminalia development. We propose that in the embryonic genital disc, as in the larval discs, Abd-B m, Abd-B r and cad are expressed in the A8, A9 and A10, respectively. We also report that abd-A, Abd-B m and Abd-B r are needed for development of the internal female genitalia, Abd-B m for the development of female external genitalia and Abd-B r for the development of male genitalia. Strikingly, abd-A and Abd-B bear unexpected relationships in mature genital discs. In the A8 of the female genital disc, Abd-B M maintains abd-A expression. In Abd-B m mutant clones, however, another Abd-B protein maintains abd-A expression but does not prevent abd-A function, as these clones transform the A8 segment into the A4. In the male A9, Abd-B r function does not repress the Abd-B m transcript, at least in part of the primordium, and some Abd-B r mutant clones transform male genitalia into leg or antenna. These relationships between Hox genes are different from those reported in the embryonic epidermis and contravene the rule that posteriorly expressed Hox genes repress those expressed more anteriorly.

abd-A^M1 is an abd-A null mutation(Sánchez-Herrero et al.,1985); abd-A^iab3-277 is a breakpoint located at +63.0-64.5 kb [coordinates according to Karch et al.(Karch et al., 1985)], which transforms abdominal segments posterior to the A2 into the A2(Busturia et al., 1989; Karch et al., 1985). The Df109 (or DfUbx¹⁰⁹) deficiency (89D1-89E1-2)eliminates the Ubx and abd-A genes(Karch et al., 1985; Lewis, 1978; Sánchez-Herrero et al.,1985), and the DfR59 deficiency eliminates the bxd regulatory region of Ubx, the abd-A gene and the 3′ regulatory region of Abd-B(Gyurkovics et al., 1990). Dpbxd¹¹¹ is a duplication on the X chromosome that includes the abd-A and Abd-B genes(Lewis, 1981). Abd-B^D18, a small deficiency (from +148.5-150.0 to+163.5-166.5 kb) removing the Abd-B gene(Hopmann et al., 1995), and Abd-B^M1 (Casanova et al., 1986; Sánchez-Herrero et al.,1985) are Abd-B (m^–r^–) mutations. Abd-B^M5, Abd-B^M3 (Casanova et al., 1986; Sánchez-Herrero et al.,1985), Abd-B^T2N(Estrada et al., 2002) and Abd-B^D14 (Karch et al., 1985) are Abd-B m^– mutations: Abd-B^M5 is caused by a C to T base substitution that produces a stop codon at amino acid 119, and thus resulting in an Abd-B M protein of 118 amino acids instead of the normal 493 (S. Pelaz and G. Morata,personal communication); Abd-B^T2N is a derivative of a P-element insertion located at 48,957(Estrada et al., 2002)[coordinates according to Martin et al.(Martin et al., 1995)]; Abd-B^D14 is a 411 bp deletion extending from 66 bp upstream to 345 bp downstream of the transcription initiation site of the Abd-B m RNA (Zavortink and Sakonju, 1989). Abd-B^Uab1 and Abd-B^X23-1 are Abd-B r^–mutations (Casanova et al.,1986): Abd-B^Uab1 is an inversion within the bithorax complex with a breakpoint at +185 kb(Karch et al., 1985) and Abd-B^X23-1 has a breakpoint at, approximately, +189 kb(Mack et al., 1997). The abd-A-lacZ (HC7JA1)(Bender and Hudson, 2000) and hdc-lacZ (B5) (Weaver and White, 1995) lines are P-lacZ insertions that reproduce the expression of the abd-A and headcase genes,respectively. The Gal4/UAS system (Brand and Perrimon, 1993) was used to drive the expression of different genes with the following constructs: UAS-abd-A(Michelson, 1994),UAS-Abd-B m (Castelli-Gair et al., 1994), UAS-Dll(Gorfinkiel et al., 1997),UAS-GFP (Ito et al., 1997),UAS-lacZ (Brand and Perrimon,1993) and UAS-myc-EGFP^F(Allan et al., 2003). The Abd-B-Gal4^LDN line (L.d.N. and E.S.-H.,unpublished) and the cad-Gal4 line (MD509)(Calleja et al., 1996) are insertions in the Abd-B and cad genes, respectively, which reproduce the pattern of expression of the Abd-B m and cadtranscripts in the posterior embryonic abdominal segments. The dpp-Gal4 line has been previously described(Staehling-Hampton et al.,1994). The MKRS, abd-A-lacZ(Bender and Hudson, 2000),TM6B, and TM6B, abd-A-lacZ balancers were used to identify mutant larvae and embryos.

In situ hybridization was done basically as described(Cubas et al., 1991; Wolff, 2000), with slight modifications. The DNA probe for the Abd-B m (A) transcript is a BamHI genomic probe from 50,702 to 48,864 [coordinates as in Martin et al. (Martin et al., 1995)]. RNA probes for the Abd-B m (A) or Abd-B r transcripts were obtained by the amplification of Abd-B cDNA or genomic DNA with region-specific primers followed by in vitro transcription reactions(Stoflet et al., 1988), using the Roche DIG RNA-labeling mix. The Abd-B m probe includes 671 bp of the Abd-B m-specific exon and was obtained with the following primers: 5′-CAGCAACTACAACAACAGCCGAC-3′ and 5′-ACACGCACACTGCCTAAAGAGC-3′. Two `common' probes were used to detect all Abd-B r RNAs. The first one (I) includes 220 bp of two exons common to all the Abd-B r cDNAs and 159 bp of the exon specific for the Abd-B r C cDNA, and was obtained with the following primers:5′-TGGAAAGATCAGACTTGCAGGTCACG-3′ and 5′-TGGGATGGGAACTGACGCTGGA-3′. The second one (II) includes 7 bp of an exon common to all the Abd-B cDNAS, 223 bp of two exons common to all the Abd-B r cDNAs and 66 bp of the exon specific for the Abd-B r γ cDNA, and was obtained with the following primers:5′-CGGAAGATTGTATTTGTGCGGTTG-3′ and 5′-TTGATGTCTGTGGGATGGGAAC-3′.

Double X-gal staining and in situ hybridization was performed basically as described previously (Wolff,2000). For double antibody staining and in situ hybridization in imaginal discs, in situ hybridization was carried out first(Wolff, 2000), and this was followed by four washes in PBS for 15 minutes each, incubation with the primary antibody overnight at 4°C, four washes in PBS and incubation with the appropriate biotinylated secondary antibody for two hours at room temperature; the imaginal discs where then washed and stained with the Vectastain ABC kit.

Immunohistochemistry was carried out as previously described(Sánchez-Herrero, 1991; Estrada and Sánchez-Herrero,2001), with slight modifications. The antibodies used were mouse and rabbit anti-β-galactosidase (Cappel), mouse anti-Abd-B lA2E9(Celniker et al., 1989),rabbit anti-Abd-B (DeLorenzi and Bienz,1990), rat anti-abd-A(Macías et al., 1990),guinea-pig anti-Hth (Azpiazu and Morata,2002), guinea-pig anti-Snail(Kosman et al., 1991), rabbit anti-Tsh (Wu and Cohen,2000), rabbit Anti-Dll(Panganiban et al., 1995), rat anti-Dll (Wu and Cohen, 2000)and mouse anti-Dll (Duncan et al.,1998). Secondary antibodies were coupled to Red-X, Texas Red, FITC and Cy5 fluorochromes (Jackson Immunoresearch). Staining with To-Pro-3 iodide was done as previously described(Baena-López et al.,2003).

Internal genitalia were dissected out of late pupae, pharates or adults,and the staining to detect β-galactosidase expression was carried out as described previously (Ashburner,1989).

Mitotic recombination clones were induced during the larval period by the FLP/FRT system (Xu and Rubin,1993), with or without using the Minute technique(Morata and Ripoll, 1975). Clones were identified in the adult by the yellow cuticular marker and in the imaginal discs by the loss of the lacZ or GFPmarkers. The genotypes of the larvae in which the clones were induced are the following (in females the hs-flp is carried as an heterozygous insertion).

abd-A^– clones: y hs-flp122; FRT82B abd-A^M1/FRT82B arm-lacZ.

Abd-B^– (Abd-B m^–r^–) clones: w f^36ahs-flp122; FRT82B Abd-B^D18/FRT82B arm-lacZ Dp(f⁺)M(3)w123, y hs-flp122; FRT82B Abd-B^D18/FRT82B hs-CD2 M(3)w y⁺ and y hs-flp122; FRT82B Abd-B^D18/FRT82B Ubi-GFP M(3)rpS3.

Abd-B m^– clones: y hs-flp122; FRT82B Abd-B^M5 or FRT82B Abd-B^M3/FRT82B arm-lacZ, y hs-flp122; FRT82B Abd-B^M5 or FRT82B Abd-B^M3/FRT82B hs-CD2 M(3)w y⁺, y hs-flp122; FRT82B Abd-B^M5 or FRT82B Abd-B^M3/FRT82B Ubi-GFP M(3)rpS3, w f^36a hs-flp122; FRT82B Abd-B^M5 or FRT82B Abd-B^M3/FRT82B arm-lacZ Dp(f⁺) M(3)w123,and y hs-flp hs-GFP FRT18A/Dpbxd¹¹¹ arm-lacZ FRT18A;Abd-B^D14/Abd-B^D18.

Abd-B r^– clones: y hs-flp122; FRT82B Abd-B^Uab1 or FRT82B Abd-B^X23-1/FRT82B arm-lacZ,w f^36a hs-flp122; FRT82B Abd-B^Uab1 or FRT82B Abd-B^X23-1/FRT82B arm-lacZ Dp(f⁺) M(3)w123 and y hs-flp122; FRT82B Abd-B^X23-1/FRT82B hs-CD2 M(3)w y⁺.

Abd-B^– abd-A⁺ clones: y tub-Gal4 UAS-GFP; FRT82B Gal80/FRT82B Abd-B^D18 UAS-abd-A.

Flies were kept in a mixture of ethanol: glycerol (3:1) macerated in 10%KOH at 60°C for 10 minutes, dissected, washed with water, dehydrated with ethanol and mounted in Euparal for inspection under a compound microscope.

In the third instar genital disc of Drosophila, Abd-B is expressed in the A8 and A9 segments, and cad in the A10(Casares et al., 1997; Freeland and Kuhn, 1996). To study whether these expression domains are established early in development,we have analyzed Abd-B and cad transcription in the embryonic genital disc. This disc is identified by the expression of genes like snail, escargot or headcase (hdc), and we selected the hdc-lacZ B5 line, which reproduces the pattern of hdc RNA expression (Weaver and White, 1985), to mark the genital disc. At about stage 15 [stages according to Campos-Ortega and Hartenstein(Campos-Ortega and Hartenstein, 1995)], hdc is expressed in three clusters of cells, two anterior ones placed bilaterally, and a third one located in a more posterior and central position(Weaver and White, 1995)(Fig. 1A and Fig. 2A). The three clusters fuse later in development to form the genital disc. At stage 15, we counted six to seven cells at each of the two anterior groups, and two to three cells in the posterior one, making up a total of 14-17 cells (n=12). Double staining with anti-Abd-B and anti-β-galactosidase antibodies (in hdc-lacZ embryos), or with GFP and anti-β-galactosidase antibody(in cad-Gal4/UAS-GFP; hdc-lacZ/+ embryos), shows that Abd-B is expressed in the two anterior clusters and cad in the posterior one (Fig. 1A-F).

To ascertain whether the two Abd-B products (Abd-B M and Abd-B R) are present in the genital disc primordium, we compared the expression driven by an Abd-B m-Gal4 line (see Materials and methods) with the signal detected with an antibody that recognizes both Abd-B M and Abd-B R proteins(Celniker et al., 1989). In UAS-myc-EGFP^F/+; Abd-B-Gal4^LDN/hdc-lacZ embryos, we saw a GFP signal in about two cells located laterally in each of the two anterior clusters; these cells most likely express Abd-B m, and, therefore,are also labelled with the anti-Abd-B antibody(Fig. 1G,I-L). There are also 8-10 Abd-B-expressing cells not labelled with GFP, and these,probably, correspond to those expressing the Abd-B R protein(Fig. 1H,K,L). Taken together,our results suggest that the embryonic genital primordium includes three groups of cells that probably express Abd-B m, Abd-B r and cad, respectively.

We have studied the Abd-B requirement in the development of the embryonic genitalia by analyzing the disc morphology in different Abd-B mutants. In Abd-B m^– homozygous embryos, the genital primordium lacks about four hdc-expressing cells, two in each of the two anterior clusters (n=15; Fig. 2B, the wild type is shown in 2A). In Abd-B^Uab1/Abd-B^M1 (Abd-B r^–) embryos, the disc primordium is disorganized and contains about eight cells (n=15; Fig. 2C). Finally, in embryos deficient for both Abd-B functions (henceforth, we refer to embryos or cells mutant for both functions as Abd-B^–) there are a few scattered hdc-expressing cells, occasionally forming lateral groups within the A8 segment (Fig. 2D). Some of the cells from these groups ectopically express Dll (Fig. 2E-F), a marker of distal appendage development(Cohen et al., 1989; Cohen et al., 1991). We conclude that the two Abd-B functions are required to form a normal embryonic genital disc, and that Abd-B confers identity to the A8 and A9 genital disc cells already at the late embryonic stages.

While this manuscript was under review, Chen and collaborators(Chen et al., 2005) showed that whereas snail (sna) or escargot are expressed in all of the cells from the embryonic genital disc(Chen et al., 2005; Fuse et al., 1996; Hartenstein and Jan, 1992; Whiteley et al., 1992), hdc is transcribed just in a subset of these cells. We have confirmed this result by staining hdc-lacZ embryos with anti-β-galactosidase and anti-Sna antibodies: three to five cells between the two anterior hdc-expressing groups of cells, and some cells posterior to them are stained with the anti-Sna antibody (data not shown). However, whether or not we mark the disc with either hdc-lacZ or Sna, Abd-B is expressed only in the anterior cells of the disc, and the Abd-B m-Gal4 line drives signal just a subset of these cells(Fig. 1A-C,I-L; data not shown). Chen et al. (Chen et al.,2005) have also analyzed homeotic gene expression and requirement in the genital disc, and conclude, as we do, that the embryonic genital disc is formed by cells that express Abd-B m (plus abd-A), Abd-B r and cad, and which probably correspond to the A8, A9 and A10 segments, respectively.

To study homeotic gene function in the genital disc after embryonic stages,we analyzed abd-A and Abd-B expression and function in larvae and pupae. The mature female genital disc expresses abd-A(Casares et al., 1997; Freeland and Kuhn, 1996)(Fig. 3A) in a domain within the A8 segment that, according to fate maps, corresponds to the presumptive internal female genitalia (Epper,1983; Littlefield and Bryant,1979). We have confirmed this correspondence by studying abd-A expression throughout late larval and pupal development with an abd-A-lacZ line that reproduces abd-A expression in the third instar female genital disc (Bender and Hudson, 2000) (Fig. 3B, compare with 3A). In abd-A-lacZ pupae, the internal female genitalia are intensively stained, including the uterus, seminal receptacle, oviducts and spermatheca, but not the parovaria(Fig. 3C,D). To explore whether abd-A is required for female internal genitalia development, we used the abd-A^iab3-277 allele (see Material and methods). In abd-A^iab3-277/abd-A^– female genital discs, abd-A expression is substantially reduced(Fig. 3E), and in abd-A^iab3-277/abd-A^– females most of the internal genitalia disappear or are highly abnormal(Fig. 3F, compare with wild-type internal genitalia in 3C). As previously reported(Karch et al., 1985) these mutant females lack ovaries.

We wanted to study the expression and requirement of the different Abd-B transcripts and proteins in the genital disc. A previous report showed that an Abd-B m enhancer trap labels the A8 of female and male genital discs, whereas an Abd-B r transcript is expressed in the A9 of the male disc (Casares et al.,1997). To confirm these results, we hybridized genital discs with probes specific for Abd-B m and Abd-B r transcripts(Fig. 4A). Our results show,within our limits of detection, that Abd-B m expression is strong in the A8 of male and female genital discs, and weak in the A9 of the male disc(Fig. 4B-E). This A9 signal is also observed in the embryo when similar probes are used(Sánchez-Herrero and Crosby,1988; Kuziora and McGinnis,1988; Boulet et al.,1991), and may represent genuine Abd-B m transcription or may be due to hybridization to Abd-B r unprocessed RNA. In support of the latter conclusion, we note that the Abd-B-Gal4^LDN insertion (although it drives expression only where Abd-B m levels are high) is co-expressed with two markers of the A8 segment, teashirt and high levels of homothorax (data not shown)(Estrada and Sánchez-Herrero,2001; Gorfinkiel et al.,2003). Abd-B r is expressed in the A9 of male and female genital discs, with some expression in a few cells in the lateral regions of the female A8 (Fig. 4B,C,F,G).

As Abd-B M is the main or only product expressed in the A8, most Abd-B m mutant clones in this primordium should behave as clones that eliminate both Abd-B functions, that is, they should transform part of the female genitalia into leg or antenna(Estrada and Sánchez-Herrero,2001). However, Abd-B m mutant clones, similar to Abd-B m escapers (Casanova et al., 1986; Karch et al.,1985; Sánchez-Herrero et al., 1985; Tiong et al.,1985), transform the dorsal eighth tergite into an anterior tergite, and the genitalia into structures sometimes resembling a sternite(Fig. 5A,B). In males, these clones also transform the tiny A8 into an anterior abdominal segment(Fig. 5C).

The difference between Abd-B^– and Abd-B m^– clones is also observed in the A8 of the female genital disc. Although Abd-B^– clones present smooth borders and sometimes activate Dll(Estrada and Sánchez-Herrero,2001), Abd-B m^– clones are indented and do not express Dll (Fig. 5D-F). We note that most of the Abd-B^–clones that induce Dll transcription are located in the region where abd-A is expressed. Because abd-A represses Dll in the embryo (Cohen et al.,1991; Simcox et al.,1991; Vachon et al.,1992), we decided to examine in more detail the relationship between abd-A, Abd-B and Dll. abd-A mutant clones do not activate Dll (Fig. 5G-I) and do not change Abd-B expression(Fig. 5J-L). Reciprocally, Abd-B m^– clones do not alter abd-Atranscription (Fig. 5M-O).

Surprisingly, abd-A disappears or is strongly downregulated in Abd-B^– (Abd-B^D18) clones(Fig. 5P-R). We note that abd-A also seems to be absent in wild-type cells around some of the large clones, but this is probably due to the bulging of the epithelium in and around mutant cells, as double staining of the clones with a nuclear ubiquitous marker, or with To-Pro-3 iodide, suggests that abd-Arepression is cell autonomous (not shown). The absence of abd-A in Abd-B^– clones indicates that Abd-B is required to maintain abd-A expression in the A8 of the female genital disc (perhaps through iab regulatory regions), which is the opposite of its role in the embryonic ectoderm, where Abd-B represses abd-A (Karch et al.,1990; Macías et al.,1990; Sánchez-Herrero,1991). In Abd-B^– clones, Dll is ectopically expressed (Fig. 5P,R), and this is independent of abd-A, as in clones that simultaneously lose Abd-B and gain abd-A expression(Abd-B^– abd-A⁺ clones) Dllsignal is still present (Fig. 5S-W). Moreover, ectopic abd-A expression in Abd-B^– abd-A⁺ clones does not repress Dll in the male primordium of the genital disc or in the leg discs(Fig. 5X and not shown). By contrast, ectopic Abd-B m protein represses Dll in the leg disc (data not shown). Strikingly, ectopic Dll expression represses abd-A(Fig. 5Y). Two conclusions can be drawn from these results: first, that although the Abd-B mtranscript is the main or only Abd-B transcript expressed in the female A8, a very different outcome is obtained if Abd-B m or both Abd-B m and Abd-B r functions are eliminated in this primordium; second, that the relationships between Abd-B and abd-A, and between abd-A and Dll, are the opposite of those observed in the embryo.

The results presented above prompted us to investigate Abd-Bexpression in Abd-B^– and Abd-B m^– clones. In Abd-B^D18(Abd-B^–) clones there is, as expected, no Abd-B expression (not shown). Surprisingly, the Abd-B signal does not change in Abd-B^M5, Abd-B^M3 or Abd-B^D14 mutant clones(Fig. 6C-E and data not shown). This result was observed with two different antibodies that recognize both Abd-B proteins (Celniker et al.,1989; DeLorenzi and Bienz,1990) and cannot be attributed to the presence of an abnormal Abd-B M protein because the alleles used do not make Abd-B M product, and the Abd-B^D14 mutation does not even make Abd-B m RNA(Boulet et al., 1991; Casares and Sánchez-Herrero,1995; Celniker et al.,1990: DeLorenzi and Bienz,1990; Zavortink and Sakonju,1989) (Fig. 6A-B). One possible interpretation of this result is that Abd-B r is derepressed in Abd-B m^– clones induced in the female A8. To confirm this, we looked at Abd-B r expression (using the`common' probe I; see Materials and methods) in Abd-B^M5mutant clones. However, as shown in Fig. 6F-H, such derepression was not observed.

Our phenotypic analysis of Abd-B m function has been confined so far to the external genitalia and the eighth tergite. However, Abd-B m transcripts are also present, although at low levels, in the presumptive internal genitalia of third instar genital discs(Fig. 4D). To follow Abd-B m transcription at later stages, we examined lacZ expression in the internal female genitalia of Abd-BGal4^LDN/UAS-lacZ late pupae and pharate adults. As shown in Fig. 6I,J,and similar to the expression observed in abd-A-lacZ females(Fig. 3C), X-gal staining is detected in the spermathecae, the uterus and, weakly, in the seminal receptacle, but not in the parovaria; however, and in contrast to the results obtained with the abd-A insertion, oviducts were not stained(Fig. 6I). In females mutant for Abd-B m, the internal female genitalia do not develop properly and, in some cases, parovaria are the only structures remaining(Fig. 6K). In males, Abd-B m mutations do not affect the internal genitalia, although the testes are reduced in size and are not connected properly with the internal genitalia(data not shown).

Abd-B r transcripts are expressed mainly or only in the A9 segment of male and female genital discs (Fig. 4F,G), but Abd-B r mutations eliminate genitalia and analia in both sexes (Casanova et al.,1986; Celniker and Lewis,1987; Karch et al.,1985; Kuhn et al.,1981). To define the requirement for the Abd-B R product during larval development, we induced Abd-B r^– mitotic recombination clones during larval stages. These clones are normal in the analia of both sexes and in the external female structures. In the male genitalia, we observe two types of clones: some of them, probably induced in the internal genitalia, transform genital structures into distal leg or, less frequently, antenna (Fig. 7A-D), thus resembling the transformations observed in Abd-B^– clones(Estrada and Sánchez-Herrero,2001). By contrast, clones in the lateral plates or clasper teeth are wild type or cause only slight pattern alterations. In accordance with the transformations observed in the adult, some Abd-B r^–clones induced in the male A9 eliminate Abd-B expression(Fig. 7E-G), and a few of these activate Dll (Fig. 7H-J); these clones have smooth borders and are located mostly in the penis apparatus presumptive region(Bryant and Hsieh, 1977; Bryant, 1978). Clones in other regions are, in general, indented, do not activate Dll and show normal or slightly reduced Abd-B expression (see summary in Fig. 7K). Some clones mutant for the Abd-B^x23-1 hypomorph allele also show a reduction in or elimination of Abd-B protein expression in the male A9 and transformation into leg tissue in the adult.

In Abd-B r mutant flies, the internal genitalia of males is absent(not shown), while that of females is abnormal. A common feature of these females is the absence of parovaria, and in about half of them we observe three and even four spermathecae instead of the normal two(Fig. 7L). Supernumerary spermathecae are also detected in females bearing Abd-B r^– clones (data not shown). In a few cases, duplications of part of the internal genitalia are observed in Abd-B r mutant females.

In this work, we describe the expression and function of abd-A and Abd-B in Drosophila genitalia development. As in the embryonic cuticle, abd-A and Abd-B m are needed in the A8 whereas Abd-B r is required in the A9. The relationship between these homeotic products in the mature genital discs, however, clearly differs from what is observed in the embryonic epidermis. In what follows, we discuss several aspects of abd-A and Abd-B expression and function in genital development.

We report that the embryonic genital disc has three distinct cell populations at stages 15/16: some anterior-lateral cells transcribe Abd-B m, anterior-central and middle cells express Abd-B r and posterior cells transcribe cad, although the expression of these products may overlap. Because the genital disc is formed by the fusion of cells coming from the A8, A9 and A10 segments(Nöthiger et al., 1977; Schüpbach et al., 1978),and by analogy to the expression of these genes in the mature genital discs(Casares et al., 1997; Freeland and Kuhn, 1996) (this report), we conclude that Abd-B m, Abd-B r and cad are probably expressed in the A8, A9 and A10 segments, respectively, of the embryonic genital disc.

Abd-B is not only expressed, but also required in the embryonic genital primordium. In the absence of Abd-B m, the number of hdc-expressing cells in the disc is reduced, most likely because these cells adopt now a more anterior fate, as occurs in the cuticle(Sánchez-Herrero et al.,1985; Tiong et al.,1985). When Abd-B r is absent, the genital primordium lacks some cells and is disorganized, and when both Abd-B products are absent,the primordium is reduced to a few, dispersed cells, some of which express Dll ectopically, suggesting a transformation into a leg primordium.

The A8, A9 and A10 primordia of the mature genital discs bear anterior and posterior compartments, with expression of en and wg in each of these three primordia (Casares et al.,1997; Chen and Baker,1997; Freeland and Kuhn,1996; Sánchez et al.,1997). Curiously, although we can define three primordia in the embryonic disc, based on the expression of Abd-B m, Abd-B r and cad, neither en nor wg is expressed in the three separate domains at this stage (Casares et al., 1997). This may suggest, as was also recently proposed(Chen et al., 2005), that new bands of en and wg expression may be formed later in development, in precise concordance with the three primordia defined previously by the Abd-B m, Abd-B r and cad genes (see Casares et al., 1997). We note that late en expression is also characteristic of the antennal primordium of the eye-antennal disc(Morata and Lawrence,1978).

We have shown that abd-A is expressed in the whole internal female genitalia except for the parovaria, and this is consistent with experiments indicating that parovaria derive from the female A9 segment(Keisman et al., 2001). abd-A has been shown to be required for gonad development(Boyle and Dinardo, 1995; Brookman et al., 1992; Cumberledge et al., 1992; Greig and Akam, 1995; Karch et al., 1985; Karch et al., 1990; Moore et al., 1998; Warrior, 1994), and in the abd-A^iab-3/Df mutant combinations ovaries are also absent. However, the defects we observe in the female internal genitalia are not simply due to an indirect effect of the lack of gonads, as iab-4mutations prevent the formation of the ovaries but do not alter internal genitalia formation (Cumberledge et al.,1992).

Our results indicate that Abd-B m is required for the development of female external and internal genitalia, both derived from the female A8(Nöthiger et al., 1977; Schüpbach et al., 1978). The internal genitalia of Abd-B-Gal4^LDN/UAS-lacZfemales were stained with X-gal except in two structures, the oviducts and parovaria. The absence of oviduct staining in Abd-B-Gal 4^LDN/UAS-lacZ females is probably due to the particular expression driven by this reporter, and does not imply an absence of Abd-B m transcription in these organs, for two reasons: first, Abd-B m transcripts are present in the whole A8 segment of the female genital disc; and second, oviduct development is affected in Abd-B mmutant females. Parovaria, by contrast, are not stained in Abd-B-Gal 4^LDN/UAS-lacZ or abd-A-lacZ females, and this agrees with their A9 provenance(Keisman et al., 2001). This is supported by the observation that in some Abd-B m mutant females parovaria are the only structures that remain in the internal female genitalia.

Abd-B M seems to be the main or only Abd-B product present in the female A8, so it was expected that elimination in this segment of just Abd-B M or of all Abd-B proteins would give similar results. This is not so. Some Abd-B^– clones transform part of the female genitalia into leg or antenna (Estrada and Sánchez-Herrero, 2001), whereas Abd-B m mutant clones convert the eighth tergite, and probably the female genitalia, into an anterior abdominal segment. The differences between Abd-B m^– and AbdB^– clones in the A8 of the female genital disc reveal the existence of unsuspected regulatory interactions between the abd-A and Abd-B genes: whereas Abd-B m^– clones do not affect abd-A, in AbdB^– clones abd-A expression is eliminated. This is a surprising result, because it is contrary to what is observed in the embryo, where Abd-B represses abd-A(Karch et al., 1990; Macías et al., 1990; Sánchez-Herrero,1991).

Abd-B m^– clones induced in the female A8 do not alter abd-A expression but do not change Abd-B expression levels either. This is observed with mutations that do not make Abd-B M protein, so the Abd-B protein detected is not the Abd-B M product. Surprisingly, although we detect some Abd-B r expression in the female A8, we do not see uniform Abd-B r expression throughout this primordium and Abd-B r transcripts seem not to be derepressed in Abd-B^M5 mutant clones. We have no explanation for this conundrum. Perhaps the probe used, although it includes sequences complementary to all of the Abd-B r cDNA sequences that have been published (Celniker et al.,1989; DeLorenzi et al.,1988; Kuziora and McGinnis,1988; Zavortink and Sakonju,1989), does not efficiently detect all of the non-Abd-B mtranscripts.

The differences in regulatory and functional interactions among gene products in the embryo and the genital discs are not limited to those of Abd-B and abd-A that have been discussed above. First, there may be changes in phenotypic suppression: the transformation of the eighth tergite to the fourth one in Abd-B m^– clones is due to abd-A (E.S.-H., unpublished). Because in these clones Abd-B protein is still present, this suggests that abd-A may phenotypically suppress Abd-B, differently from what is generally observed in the embryo (Castelli-Gair et al.,1994; Sánchez-Herrero et al., 1994). Second, Abd-B r represses Abd-B min the embryo (Casanova et al.,1986; Boulet et al.,1991), but some Abd-B r^– clones do not activate Abd-B m in the male disc (see below). Finally, abd-A represses Dll in the embryo(Cohen et al., 1991; Simcox et al., 1991; Vachon et al., 1992), but not in the female genital disc, and ectopic Dll can repress abd-A instead. abd-A does not repress Dll in the leg discs either, and this resembles Ubx function, which represses Dll only early in development(Castelli-Gair and Akam,1995). By contrast, we have shown that Abd-B represses Dll in the embryo (this report), in the larval genital disc(Estrada and Sánchez-Herrero,2001), and in the leg disc when ectopically expressed (this report).

Abd-B r expression is restricted to the A9 segment in male genital discs, but shows expression in the A9 and in some cells of the A8 in female genital discs. In spite of this, Abd-B r^– clones in the external female genitalia (A8) are phenotypically wild type. In the male A9, some Abd-B r mutant clones eliminate Abd-B, activate Dll and transform part of the genitalia into distal leg or antenna. This is similar to the result obtained in some Abd-B^– clones, and it implies that Abd-B mis not derepressed in these mutant clones. However, Abd-B m is perhaps derepressed in those Abd-B r mutant clones where Abd-B signal (detected by the common antibody) remains.

Although Abd-B r^– clones affect, almost exclusively, male genitalia development, Abd-B r hemizygous or trans-heterozygous flies lack genitalia and analia in both sexes(Casanova et al., 1986). This probably reflects the absence of proper interactions between the different primordia needed for the growth of the genital disc(Gorfinkiel et al., 2003). In Abd-B r mutant females, the internal genitalia are abnormal, and in some of these females we observe an absence of parovaria and the presence of three or four spermathecae. This phenotype is consistent with a segment-autonomous transformation of A9 derivatives (parovaria) into A8 structures (spermathecae), similar to the embryonic cuticular transformation of A9 into A8 observed in Abd-B r mutations(Casanova et al., 1986). A transformation of parovaria into spermathecae has been previously described in Polycomblike mutants (Duncan,1982), and may also indicate a transformation of A9 to A8.

Our results illustrate that there are quite different Hox cross-regulatory interactions in the embryo and in the genital disc(Fig. 8A). The effects in the genital discs contradict the general rule that genes transcribed more posteriorly suppress or downregulate the expression of more anterior ones(Hafen et al., 1984; Struhl and White, 1985). This rule has, nevertheless, some exceptions in genes of the Antennapedia complex(Miller et al., 2001). Further, differences in Hox cross-regulation between the embryo and imaginal discs are not unprecedented: the proboscipedia (pb) Hox gene is positively regulated by Sex combs reduced in the embryo(Rusch and Kaufman, 2000), but pb activates Sex combs reduced in the labial imaginal disc(Abzhanov et al., 2001).

It has been proposed that the primordia of female and male genitalia could be subdivided into an `appendage-like' and a `trunk-like' region(Estrada and Sánchez-Herrero,2001). We can now define these two regions of the female A8 more precisely. The `appendage-like' region would be that expressing abd-Aand low levels of Abd-B, and corresponds approximately to the presumptive internal female genitalia (region I in Fig. 8B). This domain is roughly coincident with the region of expression of a reporter insertion in buttonhead, the gene that defines ventral appendage development (C. Estella, PhD Thesis, Universidad Autónoma de Madrid, 2003)(Estella et al., 2003), and this is also, approximately, the domain where Abd-B<sup>–</sup>clones may activate Dll (Estrada and Sánchez-Herrero, 2001). If this subdivision is correct,the `appendage' specification defined by buttonhead would be repressed in the wild type by Abd-B, which both limits Dllexpression to a few cells of the A8 primordium and prevents Dllfunction (Estrada and Sánchez-Herrero, 2001). Abd-B^–clones in this region eliminate abd-A expression and promote leg or antenna development. This subdivision may also apply to the male disc, the penis apparatus presumptive region being the main `appendage' domain. Similar to what we have described, the labial disc possesses a large `appendage'region that is revealed by Dll derepression in pb mutations(Abzhanov et al., 2001). This characteristic, and the changes in Hox gene cross-regulation between the embryo and the imaginal disc, are two features shared by pb/labial disc and Abd-B/genital disc.

We thank G. Morata for support and discussions; N. Azpiazu and D. L. Garaulet for help with the experiments; A. Busturia, I. Guerrero, G. Morata and M. Suzanne for comments on the manuscript; S. Pelaz and G. Morata for communicating unpublished results; and R. González for technical assistance. We also thank M. Akam, W. Bender, S. Celniker, I. Duncan, I. Guerrero, F. Karch, A. Michelson, I. Miguel-Aliaga, R. White and the Bloomington Stock Center for stocks; and N. Azpiazu, M. Bienz, J. Casanova, F. Casares, S. Celniker, S. Cohen, A. Michelson, G. Panganiban and R. Reuter for antibodies. This work has been supported by grants from the Dirección General de Investigación Científica y Técnica (number BMC 2002-00300), the Comunidad Autónoma de Madrid (number 08.1/0031/2001.1)and an Institutional Grant from the Fundación Ramón Areces. D.F. and L.d.N. are supported by fellowships from the Spanish Ministerio de Educación y Ciencia.