Soma-dependent modulations contribute to divergence of rhomboidexpression during evolution of Drosophila eggshell morphology

Evolutionary changes in gene regulation are a major cause of the diversification of morphological traits in animals(Carroll et al., 2001; Davidson, 2001; Simpson, 2002). Although differences in the expression patterns of genes with instructive roles in morphogenesis have been demonstrated in various species(Averof and Patel, 1997; Gompel and Carroll, 2003; Shapiro et al., 2004; Simpson et al., 1999; Stern, 1998), few studies have addressed the molecular mechanisms underlying the diversification of these gene expression patterns (Belting et al.,1998; Gompel et al.,2005; Wittkopp et al.,2002). The dorsal appendages (DAs) are specialized respiratory structures on the drosophilid eggshell(Hinton, 1969; Spradling, 1993a). The mechanisms of their morphogenesis have been extensively studied in Drosophila melanogaster, whose eggshell has two DAs(Fig. 1B)(Berg, 2005). There is remarkable diversity in the DA morphology among species, including their size,shape and number, although these characteristics are stereotyped within each species (Kambysellis and Craddock,1997; Patterson and Stone,1952; Throckmorton,1962). Thus, comparative studies of gene expression patterns during DA formation are useful for understanding the mechanisms underlying the evolutionary diversification of gene expression(Barkai and Shilo, 2002; Derheimer et al., 2004; James and Berg, 2003; Nakamura and Matsuno, 2003; Peri et al., 1999; Perrimon and Duffy, 1998).

The egg chamber of Drosophila consists of a single oocyte and 15 nurse cells surrounded by a layer of somatic follicle cells(Margolis and Spradling, 1995; Spradling, 1993a). In D. melanogaster, the DA primordia arise from a subset of follicle cells that are specified by epidermal growth factor receptor (EGFR) signaling(Nilson and Schupbach, 1999; Wasserman and Freeman, 1998). Gurken (Grk) is a transforming growth factor-α-like protein and an oocyte-specific ligand for EGFR that localizes to the dorsal-anterior end of the oocyte and is presented to the overlying follicle cells(Neuman-Silberberg and Schupbach,1993; Nilson and Schupbach,1999; Wasserman and Freeman,1998). Grk induces the expression of rhomboid(rho), which encodes a serine protease, in a single population of dorsal anterior follicle cells (Lee et al., 2001; Ruohola-Baker et al., 1993; Urban et al.,2001), and the Decapentaplegic signaling pathway helps limit the expression of rho in these cells(Peri and Roth, 2000). In these follicle cells, Rho processes Spitz, a transmembrane ligand for EGFR, to a secreted and active form, which in turn amplifies EGFR signaling in the follicle cells (Sapir et al.,1998; Schweitzer et al.,1995; Wasserman and Freeman,1998). The high EGFR signaling activity triggers the expression of argos at the dorsal anterior midline and subsequently establishes a negative feedback loop, resolving the single peak of EGFR signaling into twin peaks (Wasserman and Freeman,1998). Consequently, a single DA primordium is formed at each region of peak EGFR signaling activity(Wasserman and Freeman,1998).

D. virilis, a species that diverged from D. melanogaster40-60 million years ago, has four DAs (Fig. 1A,C) (Powell,1997). We previously showed that rho is expressed differently between D. melanogaster and D. virilis(Fig. 1D,E). At stage 10A, rho is expressed in D. melanogaster in a dorsal anterior saddle-shaped zone that includes the midline(Fig. 1D), but in D. virilis there are two dorsal-lateral domains(Fig. 1E). At stage 10B, rho expression is refined into two L-shaped stripes in D. melanogaster (Fig. 1D);the pattern in D. virilis becomes a V-shaped stripe with its apex missing (Fig. 1E). Finally, the rho expression is maintained in two regions of follicle cells where the two DAs will form in D. melanogaster at stage 12(Fig. 1D), whereas in D. virilis, rho expression is restricted to four domains corresponding to the positions of the future DAs (Fig. 1E). Thus, changes in the regulation of rho expression could be responsible for the divergence in DA number(Nakamura and Matsuno, 2003). In this study, we investigated the mechanisms by which rho expression diverged during the evolution of D. melanogaster and D. virilis. Our results suggest that divergence of the trans-acting landscape regulating the rho expression probably had a crucial role in the evolution of different DA numbers in these two species.

The Drosophilastrains used as wild type were: D. melanogaster,Canton-S; D. virilis, stocks #15010-1051.0 and#15010-1051.87 (Tucson Drosophila Stock Center, http://stockcenter.arl.arizona.edu). The D. melanogaster yw^7-5 and D. virilis w (stock#15010-1051.53, white⁵⁰¹¹²) mutant strains were used as the D. melanogaster and D. virilis transformation hosts,respectively. The D. melanogaster EGFP-Vasa strain(Sano et al., 2002) was used as a pole cell transplantation host.

Immunofluorescent staining of ovaries was performed as described(James and Berg, 2003), except that Block Ace (Dainippon Pharmaceutical) was used as a blocking reagent. The following primary antibodies were used: rabbit anti-green fluorescent protein(GFP) (1:1000, MBL), mouse anti-β-galactosidase (Gal) (1:500, Promega)and rabbit anti-Broad-Complex (BRC) core (1:2000)(Dequier et al., 2001). Alexa Fluor-488-conjugated antirabbit immunoglobulin G (IgG) (1:400, Molecular Probes) and Cy3-conjugated anti-mouse IgG (1:400, Rockland) were used as secondary antibodies. Immunofluorescent staining of embryos was performed as described (Hayashi et al.,2004). Embryos were stained with rabbit anti-Vasa (1:500) and mouse anti-GFP (1:500, Wako Pure Chemicals) antibodies, and then treated with Alexa Fluor-568-conjugated anti-rabbit IgG (1:500, Molecular Probes) and Alexa Fluor-488-conjugated anti-mouse IgG (1:500, Molecular Probes).

The RNA probes were labeled with digoxigenin (Roche), and in situ hybridization was performed as described(Wasserman and Freeman,1998).

Partial DNA fragments of mirr for in situ hybridization were amplified using genomic DNAs from Canton-S and D. virilis (#15010-1051.87) as template DNAs by PCR. The primers were:Mirr_FW, 5′-GATATGATGACCGACC-3′ and Mirr_RV,5′-CCTATAAGCTCTGATTGC-3′.

The DNA probe of the D. virilis rho cDNA was labeled with digoxigenin (DIG DNA labeling Mix, Roche) by random-primed labeling(Nakamura and Matsuno, 2003). A D. virilis phage genomic library (gift of T. C. Kaufman, Indiana University, Bloomington, IN) was screened with this probe using standard conditions (Sambrook and Russel,2001). Twenty-one overlapping phage clones that hybridized with the D. virilis rho probe were isolated. The three largest genomic fragments isolated from these phage clones were digested with SalI and subcloned into the SalI site of pBluescript KS(-). These subclones were sequenced, aligned and ligated. Finally, a genomic DNA fragment 12-kb upstream of the start site of the D. virilis rho cDNA was obtained (GenBank accession number AB278158).

A genomic fragment covering the 2.2-kb upstream region of the D. melanogaster rho gene was generated by PCR using the D. melanogaster genomic P1 clone DS02734 (GenBank accession number AC004343)as a template (Kimmerly et al.,1996), and subcloned into the BamHI site of pCaSpeR-NLSlacZ (gift of C. Thummel, Howard Hughes Medical Institute,University of Utah, Salt Lake City, UT). The primers were: rho2.2-FW,5′-CGGGATCCCGCAAGCTTTTCCTCTGCTC-3′ and rho2.2-RV, 5′-CGGGATCCCGTTCTCTGCTTGCACCCAC-3.(Restriction enzyme cloning sites are underlined.)

The 2.2-kb upstream fragment of the D. melanogaster rho gene and fragments containing the 4.2- and 12-kb upstream regions of the D. virilis rho gene were subcloned into the BamHI and KpnI/BamHI sites of pSL[hsp27mp-NLS-EGFP] (T.N.,unpublished), respectively. The resulting constructs were digested with AscI, and each insert was cloned into the AscI site of pBac[3xP3-DsRedaf].

Germline transformation of the P-element CaSpeR vectors was performed according to a standard protocol(Spradling, 1993b). For transformation with the piggyBac vector, piggyBac constructs were co-injected into the eggs of D. melanogaster yw and D. virilis w with phsp-pBac, and flies were screened for 3xP3-DsRed expression under a fluorescence stereoscopic microscope (gift of E. A. Wimmer,Georg August University, Göttingen, Germany). For each transgene, at least three independent insertions were isolated and characterized.

Pole cell transplantation was performed as described(Kobayashi et al., 1996), with the following modifications. As a marker for the D. melanogastergermline, we used an EGFP-vasa construct that expresses GFP specifically and continuously in the germline throughout the life cycle(Sano et al., 2002). We transplanted pole cells from D. virilis (#15010-1051.0) embryos[200-250 minutes after egg laying (AEL) (25°C)] into D. melanogaster embryos [100-150 minutes AEL (25°C)] carrying the EGFP-vasa. The donor embryos were at the cellular blastodermal stage.

In D. melanogaster, a region of the rho gene 2.2-kb upstream of its transcriptional start site contains a cis-regulatory element that is responsible for its expression during oogenesis(Dorman et al., 2004; Ip et al., 1992; Sapir et al., 1998; Ward and Berg, 2005). This rho enhancer, termed Dmel rho^2.2, is sufficient to drive the expression of a reporter gene in the developing follicle cells(Dorman et al., 2004; Sapir et al., 1998; Ward and Berg, 2005). At stage 10B, Dmel rho^2.2 is activated in L-shaped stripes on either side of the midline (Dorman et al.,2004; Sapir et al.,1998; Ward and Berg,2005). This expression pattern resembles that of endogenous rho, which is essential for DA formation, at the same stage(Fig. 1D and Fig. 3C)(Dorman et al., 2004; Sapir et al., 1998).

We speculated that a similar cis-regulatory element should exist in the D. virilis rho gene. Therefore, we isolated a 12-kb genomic region upstream of the D. virilis rho cDNA start site (GenBank accession number AB278158). To test whether the 12-kb fragment or a 4.2-kb upstream fragment derived from it contained a regulatory sequence that governs rho expression in the follicle cells, we analyzed the activity of these fragments in D. virilis using piggyBac- mediated transgenesis (Handler, 2002). When placed upstream of a GFP reporter gene, the 12- and 4.2-kb upstream fragments drove reporter expression in the follicle cells of D. virilis during DA formation in essentially the same manner(Fig. 2 and data not shown). Therefore, we used the 4.2-kb fragment in the following studies, because a shorter fragment is preferable for generating transgenic lines efficiently in D. virilis and D. melanogaster. At stage 10B, Dvir rho^4.2 drove reporter expression in a pattern identical to the distribution of endogenous D. virilis rho mRNA(Fig. 2A,C). At stage 12, Dvir rho^4.2 also drove GFP expression in the four domains that subsequently form the four DAs (Fig. 2D); this pattern was also similar to that of endogenous rho at the same stage (Fig. 2B). We concluded that Dvir rho^4.2 consists of an enhancer orthologous to Dmel rho^2.2.

Broad-Complex (BR-C) is expressed in the presumptive DA-forming cells(Deng and Bownes, 1997; Tzolovsky et al., 1999). The juxtaposition of rho-expressing and BR-C-expressing cells is required for DA formation in D. melanogaster(Ward and Berg, 2005). In cells expressing rho, Spitz is converted into an active and secreted ligand for EGFR (Lee et al.,2001; Urban et al.,2001), so that EGFR signaling is activated and triggers the expression of BRC in adjacent cells(Schweitzer et al., 1995; Ward and Berg, 2005). In D. virilis and D. melanogaster, the rho and BR-C expression domains are mutually exclusive at stage 10B(Nakamura and Matsuno, 2003),indicating that the genetic cascade involving rho and BR-C is conserved between these species. To test whether Dvir rho^4.2carries sufficient information to restrict its activation anterior to the BR-C-expressing domain, we analyzed the expression of Dvir rho^4.2-GFP and BR-C in D. virilis. Dvir rho^4.2was activated in a single row of cells anterior and adjacent to BR-C-expressing cells (Fig. 2E-G). These results suggested that Dvir rho^4.2contains sufficient information for comparative studies with Dmel rho^2.2 in these two species.

Phylogenetic analyses indicated that four-DA eggs are ancestral compared with two-DA eggs (Fig. 1A)(Powell, 1997; Throckmorton, 1962). This suggests that changes occurred in the trans-acting landscape regulating the cis-regulatory elements of rho and/or in these cis-regulatory elements themselves during the evolution from the four- to two-DA eggs. We assumed that if divergence of the trans-regulatory landscape was wholly responsible for the species-specific expression patterns of rho, the exogenous rho enhancers would adopt the same expression pattern as the endogenous rho expression (i.e. in D. virilis, Dmel rho^2.2 should be activated in a similar pattern to Dvir rho^4.2). By contrast, if evolutionary modifications of the cis-regulatory elements were responsible for the divergence in rhoexpression, the activation pattern of Dvir rho^4.2 and Dmel rho^2.2 in the heterologous species should match the expression pattern of the endogenous rho in their species of origin (i.e. their homologous species).

To address this issue, we introduced the Dmel rho^2.2-GFP reporter construct into D. virilis. As reported previously, Dmel rho^2.2 was activated in L-shaped domains at stages 10B and 12 in D. melanogaster (Dorman et al.,2004). Interestingly, Dmel rho^2.2 was activated in the V-shape with its apex missing in D. virilis at stage 10B(Fig. 3A). This pattern was very similar to the expression patterns of Dvir rho^4.2-GFP and the endogenous rho in D. virilis at this stage. Furthermore, at stage 12, Dmel rho^2.2 drove expression in a pattern resembling that of Dvir rho^4.2-GFP and endogenous rho in D. virilis, which was significantly different from the activation pattern of the Dmel rho^2.2 in its homologous species, D. melanogaster (Fig. 3B). These results suggested that the Dmel rho^2.2enhancer could respond to the heterologous trans-regulatory landscape in a manner similar to the endogenous rho enhancer in D. virilisduring DA formation. Given these results, we speculated that diversification in the trans-regulatory landscape, rather than modification of the cis-acting elements, was mostly responsible for the differences in rhoexpression patterns during DA formation in these two species.

We then introduced Dvir rho^4.2-GFP into D. melanogasterto examine whether our hypothesis held true in the reciprocal situation. In D. melanogaster, both Dmel rho^2.2(Fig. 3C) and Dvir rho^4.2 (Fig. 3E)drove GFP expression in the L-shaped pattern at stage 10B. This result was consistent with our idea that the trans-regulatory landscape diverged significantly between D. virilis and D. melanogaster, and this divergence was responsible for the evolutionary changes in the rho expression patterns in these species.

Our results suggested that the trans-regulatory landscapes that direct rho expression are evolutionarily divergent between D. virilis and D. melanogaster. However, we also noted that the expression domain of the Dvir rho^4.2-GFP was slightly expanded posteriorly compared with that of the Dmel rho^2.2-GFP in D. melanogaster (compare Fig. 3E with Fig. 3C). This difference became more obvious at stage 12 (compare Fig. 3F with Fig. 3D). At this stage, the Dvir rho^4.2-GFP was expressed in a square on either side of the midline, rather than in the L-shaped pattern(Fig. 3F).

To study further the differential activation patterns between Dvir rho^4.2 and Dmel rho^2.2 in D. melanogaster, we introduced Dvir rho^4.2-GFP and Dmel rho^2.2-β-Gal simultaneously into D. melanogaster, and observed the expression patterns driven by Dvir rho^4.2 and Dmel rho^2.2 with anti-GFP and anti-β-Gal antibody staining, respectively. Both the GFP andβ-Gal products of the reporter constructs carry a nuclear localization signal. Thus, the expression of these reporters was detected as nuclear staining at the single-cell level. The expression of Dvir rho^4.2-GFP and Dmel rho^2.2-β-Gal differed temporally and spatially (Fig. 4). When Dmel rho^2.2-β-Gal expression started,early in stage 10B, the Dvir rho^4.2-GFP expression was not yet detectable (Fig. 4A-C). This difference was not owing to features of the GFP and β-Gal reporters (for example, their relative translational efficiencies), because when the GFP andβ-Gal reporters were expressed under the control of the same enhancer,Dmel rho^2.2, their expression patterns were essentially identical,temporally and spatially (see Fig. S1 in the supplementary material). Late in stage 10B, both Dvir rho^4.2-GFP and Dmel rho^2.2-β-Gal were detectable(Fig. 4D-F). The anterior border of the expression domain of Dvir rho^4.2-GFP and Dmel rho^2.2-β-Gal was the same(Fig. 4F). However, whereas Dmel rho^2.2-β-Gal was expressed in a single row of cells(Fig. 4D), Dvir rho^4.2-GFP was expressed in one or two extra rows posterior to the single row where both reporters were expressed(Fig. 4E). At stage 12,expression of the Dmel rho^2.2-β-Gal was still restricted to the single row of cells (Fig. 4G), whereas the expression domain of the Dvir rho^4.2-GFP had expanded posteriorly(Fig. 4H). These results suggest that the functions of the rho enhancers also evolved between these two species, although both enhancers maintained the ability to respond to the positional information that defines the anterior borders of their activation. However, we also noted that a few of the anterior-most cells expressed Dvir rho^4.2-GFP but not Dmel rho^2.2-β-Gal at stage 12 (Fig. 4I),suggesting that some other functions might have also changed between the two enhancers.

During DA patterning, the germline oocyte provides positional information to the somatic follicle cells through the Grk signal(Nilson and Schupbach, 1999). A mathematical model predicts that the distribution and amounts of Grk could play a key role in determining the number of DAs(Shvartsman et al., 2002). Considering that the Grk signal serves as the first cue for the positional information that defines the expression patterns of rho in the follicle cells, we proposed the following two hypotheses: first, that some divergence in the pivotal mechanism for determining the number of DAs occurred in the Grk signal itself, and second, that the trans-regulatory landscape that modifies the Grk signal in somatic follicle cells changed.

To distinguish between these possibilities, we generated a chimeric egg chamber that consisted of the D. virilis oocyte and D. melanogaster follicle cells (Fig. 5A). If the Grk signal diverged among Drosophila species,such a chimeric egg chamber would result in D. virilis-like appendages. The germline progenitors, or pole cells, were transplanted from D. virilis donor embryos into D. melanogaster host embryos carrying the EGFP-vasa construct, which expresses GFP in the germline(Sano et al., 2002). In the D. melanogaster hosts, the D. virilis germline was recognized by the absence of GFP expression.

First, we examined whether the D. virilis pole cells could enter the gonad of the D. melanogaster embryo and whether the D. virilis germline cells could form the chimeric egg chamber in concert with D. melanogaster follicle cells in D. melanogasterovaries (Fig. 5B-B″,C). The transplanted D. virilis pole cells, which expressed only endogenous Vasa (Fig. 5B″, purple), were incorporated into the gonad of D. melanogaster together with the D. melanogaster pole cells(Fig. 5B″). Furthermore,these germline cells derived from D. virilis formed a chimeric egg chamber together with the D. melanogaster follicle cells (white arrowhead in Fig. 5C), despite their phylogenic divergence. In contrast to the prediction from the mathematical model, we found that all the chimeric eggs (n=16)observed in four host females had two DAs(Fig. 5E), identical to the pattern of D. melanogaster (Fig. 5D).

This result suggests that the germline Grk signal did not change significantly during the evolution from four- to two-DA eggs. However, this interspecific transplantation could have introduced a non-specific effect that allowed the chimeric egg to develop two DAs, even if the Grk signal had diversified between the two species. For example, an incompatibility between the D. virilis Grk and the D. melanogaster EGFR could compensate for a difference in Grk signals. To address this problem, we took advantage of the fact that the Grk signal has two roles during oogenesis(Nilson and Schupbach, 1999). In addition to its role in DA formation, the Grk signal has an essential role in the dorsal-ventral patterning of the egg(Neuman-Silberberg and Schupbach,1993; Nilson and Schupbach,1999). When the dorsal-ventral patterning is aberrant in D. melanogaster, the DAs are shifted with respect to the dorsal midline,dorsally or ventrally, respectively, with the loss or gain of Grk signaling(Neuman-Silberberg and Schupbach,1993; Nilson and Schupbach,1999; Schupbach,1987). Therefore, we measured the distance between the roots of the two DAs in wild-type and chimeric eggshells, both of which were laid by a single transplanted host female to avoid any other effects, such as nutritional status. The distances between the two DAs in the wild-type(48.2±4.6 μm, n=57) and chimeric (49.2±3.6 μm, n=12) eggshells showed no statistically significant difference(Fig. 5F,G,0.5>P>0.2, Student's t-test), suggesting that the dorsal-ventral patterning occurred normally in the chimeric eggs. It is unlikely that the two different patterning events, DA formation and dorsal-ventral patterning, both of which are triggered by Grk signal-dependent positional information, were both properly compensated for by chance. Thus, we speculate that the role of Grk in DA patterning is conserved between D. virilis and D. melanogaster, suggesting that subsequent events,downstream of the Grk signal, diverged to elicit the different expression of rho between these two species.

In D. melanogaster, mirr, which encodes a homeobox transcription factor, is expressed in dorsal anterior follicle cells(Jordan et al., 2000; Zhao et al., 2000). This expression pattern depends on the Grk signal, and mirr subsequently activates the expression of rho during oogenesis(Jordan et al., 2000; Jordan et al., 2005; Zhao et al., 2000). Therefore, mirr might introduce the evolutionary changes into the landscape of trans-regulatory factors that are responsible for the divergence of rho expression in D. virilis and D. melanogaster. To test this hypothesis, we analyzed the mirr expression patterns in these two species.

In both species, mirr expression began similarly, at the dorsalanterior region of the follicle cells, at stage 9, prior to the onset of rho expression (Fig. 6A,E). After this stage, the mirr expression followed species-specific patterns, which resembled the rho expression pattern in each species. In D. melanogaster, the expression of mirrwas repressed in the dorsal midline at stage 10A (white arrowhead in Fig. 6B) and was elevated in the follicle cells at the anterior boundary (white arrow in Fig. 6B), which corresponded to the anterior row of rho-expressing cells in the L-shaped pattern at stage 10B (Fig. 1D). This anterior expression continued until stage 10B (white arrow in Fig. 6C). By contrast, early in stage 10A in D. virilis, the mirr expression was repressed only at the dorsal-anterior region, not the entire dorsal midline (white arrowhead in Fig. 6F). Late in stage 10A, the region lacking mirr expression expanded and became a triangle (Fig. 6G), which consequently divided the mirrexpressing region into two dorsal-lateral domains that were reminiscent of the rho-expressing domains of D. virilis at stage 10A(Fig. 1E). After stage 10B, the expression of mirr was barely detectable in either species(Fig. 6D,H). The expression of D. virilis mirr was not restricted to the domains expressing rho (Fig. 1E). Therefore, the expression of mirr does not always cause the expression of rho. Nevertheless, our results suggest evolutionary changes in the mirr expression patterns as a candidate mechanism for the divergence of rho expression at stage 10A in D. virilisand D. melanogaster.

Changes in gene expression patterns during evolution can be attributed to two distinct mechanisms. First, alterations in the cisregulatory sequence of a gene can be responsible for the divergence of its expression pattern. Second,changes in trans-regulatory factors can cause gene expression patterns to diverge, even if the cisregulatory elements of these genes are conserved during evolution. Diversification in enhancer elements is known to contribute predominantly to the evolution of animal morphology(Belting et al., 1998; Gompel et al., 2005; Wittkopp et al., 2002). The gain and loss of cis-acting elements have played central roles in the divergence of the expression patterns of genes that play crucial roles in the generation of specific characteristics in different species(Gompel et al., 2005). In this study, we investigated the contribution of these two processes to the evolutionary diversification of DA numbers in D. virilis and D. melanogaster. In addition to the importance of cisregulatory elements demonstrated previously, our findings suggest that the landscape of trans-regulatory factors could also change and affect morphological divergence during evolution.

In D. melanogaster, rho expression has an instructive role in defining the pattern of DA precursor cell formation(Sapir et al., 1998; Ward and Berg, 2005). In addition, we previously demonstrated that the expression patterns of rho diverged and were correlated with the position and number of DAs in D. virilis and D. melanogaster(Nakamura and Matsuno, 2003). Therefore, in this study, we mostly focused on the enhancers of rhoin these species. To distinguish whether divergence in the trans-regulatory landscape or the cis-regulatory elements is important for the evolutionary change in rho expression patterns between D. melanogasterand D. virilis, we introduced reporter constructs of Dvir rho^4.2 and Dmel rho^2.2 into these two species. Phylogenic analyses of Drosophila species suggest that the four DAs are an ancestral characteristic, and that the flies with two DAs evolved from four-DA ancestors (Fig. 1). Thus, the characteristics of Dmel rho^2.2 were probably derived from the ancestral Dvir rho^4.2 enhancer. We found that Dvir rho^4.2 and Dmel rho^2.2 adopted the expression pattern of the endogenous rho of the heterologous species. These results suggest that Dvir rho^4.2 and Dmel rho^2.2 did not diverge in terms of their ability to respond to the trans-acting factors in follicle cells. Therefore, we speculated that changes in the cis-regulatory elements from Dvir rho^4.2 to Dmel rho^2.2 were not the main cause for divergence in the activation patterns of these enhancers in their homologous species.

Although the DNA sequences of Dvir rho^4.2 and Dmel rho^2.2 diverged drastically, several putative binding sites for transcription factors, such as ETS, Su(H) and BR-C, were common to both (data not shown), which could explain the conserved function of the two enhancers. Recently, it was reported that BR-C represses the activity of Dmel rho^2.2 in a cell-autonomous manner during DA patterning(Ward et al., 2006), and this repression allows the enhancer to be activated in the L-shaped region. These two rho enhancers share five overlapping binding sites for BR-C (data not shown). Thus, these BR-C-binding sites might serve as cis-regulatory elements to transmit the conserved functions of these two enhancers. Notch (N)signaling also regulates Dmel rho^2.2(Ward et al., 2006), and we found that one binding site for Su(H) is conserved among all six Drosophila species examined (data not shown). Conservation of the binding sites for these various transcription factors and their possible involvement in the evolution of DA patterning suggest that rhoexpression is controlled by complex responses to multiple transcription factors, instead of by a simple EGFR-signal feedback system, which is consistent with the model proposed by Peri and Roth(Peri and Roth, 2000).

We identified Mirr as a candidate for the difference in the landscape of trans-regulatory factors between D. melanogaster and D. virilis. The distribution of the mirr transcript was significantly different between these species. mirr induces rho expression, and regulates N signaling by repressing fringe, probably thereby regulating rho(Blair, 2000; Bruckner et al., 2000; Moloney et al., 2000). Although whether or not Mirr function is also involved in the regulation of rho transcription in D. virilis remains to be tested, it is conceivable that changes in the expression patterns of mirr may account, at least in part, for the divergence in the activation patterns of Dvir rho^4.2 and Dmel rho^2.2 in D. melanogasterand D. virilis.

In D. melanogaster, rho is expressed in a saddle-shaped pattern at stage 10A. We analyzed the genomic region within 26.2-kb upstream and 11.8-kb downstream of the transcription initiation site of rho, but failed to identify an enhancer element responsible for this early expression pattern(data not shown). The function of this early rho expression in DA formation has not yet been studied. Therefore, we could not exclude the possibility that an enhancer that regulates the early expression of rho is involved in the diversification of the rho expression pattern. However, we speculate that this early expression of rho does not play a significant role in determining the number of DAs, because D. pseudoobscura and D. melanica have eggs with two DAs, but the saddle-shaped pattern of rho expression was not detected in these species (T.K., Y.N. and K.M., unpublished). Therefore, the subsequent expression of rho is probably what plays a crucial role in determining the DA number.

Our present analysis revealed that the functions of Dvir rho^4.2and Dmel rho^2.2 are largely conserved. However, we also found that Dmel rho^2.2 had evolved a novel trait during its diversification from Dvir rho^4.2. In D. melanogaster, both Dvir rho^4.2 and Dmel rho^2.2 were activated in the L-shaped pattern at stage 10B. However, Dvir rho^4.2 was activated in one or two extra rows of cells posterior to the single row of cells where Dmel rho^2.2 was active at this stage. At stage 12, Dvir rho^4.2 was activated much more posteriorly, although Dmel rho^2.2 was still active only in the single row of cells. Given that Dvir rho^4.2 is ancestral to Dmel rho^2.2, we speculate that Dmel rho^2.2 lost a cis-acting element capable of being activated in this posterior region, or gained a cis-acting element that suppresses its activity in this region at stage 12. Indeed, it is likely that this posterior activation of rho is an ancestral characteristic,because the endogenous expression of rho in this region is found in D. virilis but not D. melanogaster.

For the formation of DAs, the patterning of EGFR signaling activity in the follicle cells plays crucial roles in D. melanogaster(Wasserman and Freeman, 1998). Two major events are involved in the regulation of EGFR signaling activity in these cells. First, Grk specifically localizes to the dorsal anterior part of the oocyte and activates EGFR in the overlying follicle cells(Wasserman and Freeman, 1998). Second, in the follicle cells, positive and negative feedback loops elaborate the pattern of EGFR signaling activity that ultimately determines the number of DAs (Wasserman and Freeman,1998). Thus, the first and second events are germ- and soma-derived events, respectively.

As predicted from the above model, the intensity of Grk expression and the width of its expression domain in the oocyte are thought to define the number of DAs (Shvartsman et al.,2002). A mathematical study predicted that changes in the amount and distribution of Grk protein in the oocyte can account for the evolution of eggshells with zero to four DAs in Drosophila species(Shvartsman et al., 2002). However, our experiments involving a chimeric egg chamber suggest that changes in the follicle cells, but not in the oocyte, have an instructive role in determining the number of DAs. These results suggest that the change in Grk signaling did not contribute to the evolution of DA numbers in these species. This is consistent with a previous finding that the distribution and amount of grk mRNA do not show a significant difference between D. melanogaster and D. virilis(Peri et al., 1999). However,our results do not exclude the possibility that changes in Grk signaling play major roles in the diversification of DA numbers during the evolution of other Drosophila species.

We thank J.-A. Lepesant (Institut Jacques Monod, Paris, France) for the anti-BR-C sera, E. A. Wimmer for the piggyBac vectors, C. Thummel for the pCaSpeR-NLSlacZ, and the Bloomington and Tucson Stock Centers for fly stocks. We are grateful to the members of the Matsuno Laboratory for support and assistance.