SoxF is part of a novel negative-feedback loop in the wingless pathway that controls proliferation in the Drosophila wing disc

One of the long-standing questions in biology is how organ growth is coordinated with tissue patterning. Research during recent decades has shown that a limited set of signals and signaling pathways control this coordination. Some of these signals are mitogenic, and their production at specific sites, called signaling centers, links spatial information to cell proliferation within developing organs(Freeman and Gurdon, 2002). Normal organ growth not only needs mitogens, but also mechanisms to control their production, transport, reception and/or transduction to ensure that proliferation is limited in space and time. Alterations in these control mechanisms often lead to disease.

The Wnt/β-catenin signaling pathway promotes cell proliferation during normal development and disease (Polakis,2000). Wnts are lipid-modified glycosylated signaling molecules that can reach distant cells. Binding of Wnts to the receptor complex[composed of a Frizzled family receptor and an Arrow (LRP) co-receptor]results in the stabilization of the transcriptional co-factor β-catenin[armadillo (arm) in Drosophila]. Thereby,β-catenin/Arm accumulates in the nucleus, where it associates with Tcf/LEF DNA-binding transcription factors to regulate the expression of Wnt target genes (Gordon and Nusse,2006). Research in a number of model organisms has demonstrated that the Wnt/β-catenin pathway controls cell proliferation in a variety of tissues, including the nervous system(Chenn and Walsh, 2002; Chesnutt et al., 2004; Dickinson et al., 1994) and the progenitors of the intestine and hematopoietic systems(Pinto et al., 2003; Willert et al., 2003) in mammals, and during imaginal disc development in Drosophila(Giraldez and Cohen, 2003; Johnston and Sanders, 2003; Neumann and Cohen, 1996). It is also known that most colorectal tumors, and a number of other tumor types,are caused by aberrant Wnt/β-catenin signaling(de Lau et al., 2007; Polakis, 2000), which underlines the necessity of tight regulation of this pathway.

The range and intensity of the signaling elicited by Wnt molecules have been shown to be regulated by many different mechanisms, including negative-feedback loops. These have been particularly well studied for the main Drosophila Wnt gene, wingless (wg). wg is required in the imaginal discs for the growth and patterning of the adult body structures (Giraldez and Cohen, 2003; Johnston and Sanders, 2003). wg signaling results in the downregulation of its two receptors, Dfz-2 (fz2 - FlyBase)and fz (Cadigan et al.,1998; Muller et al.,1999) and in the upregulation of Dfz-3 (fz3 -FlyBase), a non-productive low-affinity receptor, and of the extracellular Wg inhibitor Notum (wingful)(Gerlitz and Basler, 2002; Giraldez et al., 2002; Sato et al., 1999; Sivasankaran et al., 2000). Intracellularly, high levels of wg/Wnt signaling induce the expression of two inhibitors of the pathway: naked cuticle(Rousset et al., 2001; Zeng et al., 2000) and nemo (Zeng and Verheyen,2004). All these feedback loops result in an attenuation of the signal at the sites of maximal wg production and are generally implicated in all processes in which wg is required.

The Drosophila wing disc gives rise to the wing blade, the notum(body wall) and the hinge, which joins the wing blade to the body wall and articulates its movements (see Fig. 1A-D). wg is expressed in two concentric rings in the hinge domain (Baker, 1988) and has been shown to be required for the proliferation of hinge cells(Neumann and Cohen, 1996; Zirin and Mann, 2007). Moreover, wg overexpression is sufficient to drive hinge overgrowths without causing major repatterning(Neumann and Cohen, 1996; Whitworth and Russell, 2003). Therefore, the precise regulation of the wg pathway is crucial to control the growth of the hinge. The mitogenic effect of wg on hinge cells contrasts with its effect on the neighboring wing pouch cells which,upon similar wg overexpression, are mostly driven into sensory organ differentiation (Neumann and Cohen,1996; Sanson et al.,1996). One prediction from these results is that the hinge-specific proliferative function of wg needs dedicated control mechanisms to ensure normal hinge size and shape. To identify these mechanisms, we searched genes that are differentially expressed in the hinge territory for a role in wg-mediated proliferation. SoxF(Sox15) belongs to the family of sequence-specific HMG Sox transcription factors and has been shown to be expressed in the prospective hinge of third larval stage (L3) wing discs(Cremazy et al., 2001). The functions of Sox genes have been extensively studied in mammals, in which they play essential roles during development(Kiefer, 2007). In addition,misregulation of Sox genes is often associated with cancer(Dong et al., 2004).

Only two of the eight Sox family genes present in the Drosophilagenome have been studied in detail: Dichaete (D) and SoxNeuro (SoxN). They belong to the SoxB group and have prominent roles in embryonic segmentation and nervous system development(Overton et al., 2002). In addition, it has recently been shown that both genes negatively regulate the activity of the wg/Wnt pathway during cell fate specification in the embryonic epidermis (Chao et al.,2007; Overton et al.,2007).

Here, we report that SoxF, which is the sole member of this Sox group in Drosophila, is also required to restrain wgsignaling, but using a novel mechanism: the transcriptional repression of wg. In the absence of SoxF, wg transcription spreads through the hinge causing its overproliferation. SoxF is itself under the control of the canonical wg/Wnt pathway such that wg and SoxF regulate each other's transcription through a feedback loop. Moreover, the expression of rotund (rn), which is part of the proximodistal patterning mechanism of the wing disc, allows the exclusion of SoxF from a thin rim of cells, allowing them to express wg. Thereby, this rim becomes a spatially well-defined mitogen-producing center necessary to ensure normal hinge growth. This novel mode of action of a Sox gene on the Wnt pathway - the transcriptional repression of a Wnt gene - might be relevant to human disease, as loss of human SoxF genes has been implicated in colon carcinoma.

In order to determine the role played by SoxF during hinge development, we first characterized one previously isolated SoxFallele, Sox15^KG09145 [now renamed SoxF^KG09145 (Bellen et al., 2004)]. The SoxF^KG09145 allele carries an insertion of the P[SUPor-P] transposon in an intronic region of the gene,which also harbors the CG30071 transcript (see Fig. S2 in the supplementary material). Most homozygous SoxF^KG09145 flies die as pharate adults, and escapers are weak with held-out wings (see Fig. S2D in the supplementary material). This latter phenotype is indicative of hinge defects. In fact, these flies show abnormal proximal hinge structures: the sclerites, the alula and the costa are affected (see Fig. S2E,F in the supplementary material). Although the insertion does not affect SoxFcoding sequence, we observed by RT-PCR (data not shown) and in situ hybridization (see Fig. S2B,C in the supplementary material) that SoxF expression is completely lost in the wing disc of mutant L3 larvae. As this P-element carries insulator sequences, we also checked by RT-PCR that expression of CG30071 and of the 5′ neighboring gene, RpS23, was not affected by the insertion, which was indeed the case (see Fig. S2G in the supplementary material). We also generated new alleles by imprecise excision of the P transposon from the original allele. In addition to full revertants, we isolated more than ten mutant lines in which different lengths of intron sequences were deleted, without affecting the coding region, and which showed a range of phenotypic severity. These results suggest that this intronic region carries crucial elements for the regulation of SoxF expression. We also isolated some alleles that disrupt the coding sequence. Among them, SoxF²⁶ is specific to the SoxF gene and deletes the first exon and part of the first large intron, and is therefore likely to be a null allele (see Fig. S2A in the supplementary material). This allele has the same phenotype as the initial insertion. In addition, the phenotype and escaper rates of individuals carrying SoxF^KG09145 over a deficiency uncovering the SoxF locus, Df(2R)Exel7130, are the same as for homozygous SoxF^KG09145 flies. Therefore, SoxF^KG09145 behaves as a genetic null allele. Cremazy and co-workers (Cremazy et al.,2001) reported that SoxF is expressed in the embryonic peripheral nervous system (PNS). We obtained adult escapers of the molecular null allele SoxF²⁶. These animals, in addition to their abnormally folded wings, are weak and die shortly after eclosion. Other hinge mutants, such as wg spd-fg, are much healthier. Therefore, it is possible that the larval lethality and weakness of adult escapers is due to abnormal PNS development.

UAS-SoxF was generated by cloning the full-length SoxFcoding region from the cDNA clone IP09065 as an EcoRI/XhoI fragment into the pUASt plasmid. Transgenic flies were generated by standard methods. To analyze, comparatively, the effects of gene overexpression using the GAL4/UAS system (Brand and Perrimon, 1993), the genotypes were synthesized to contain the same number of UAS sequences by including a `neutral'UAS-GFP if needed.

wg-GAL4, Bx^MS1096-GAL4 (FlyBase); zfh2^MS209-GAL4(Whitworth and Russell,2003).

UAS GFP-Wg (Pfeiffer et al.,2002), UAS-Arm^S10 (UAS-Arm^*), UAS-dTCF^DeltaN (TCF^DN) and UAS-GFP (FlyBase); UAS-rn(St Pierre et al., 2002) was a gift from J. P. Couso; UAS-dsSoxF (number 45482, DrosophilaGenetic Resource Center, Kyoto, Japan).

puckered-lacZ (puc^E69), wg-lacZ(FlyBase), SpFlag-lacZ (Neumann and Cohen, 1996), rn-lacZ(St Pierre et al., 2002).

wg^spd-fg (Neumann and Cohen, 1996), wgCX3(Klein and Arias, 1998)w¹¹¹⁸; Df(2R)Exel7130, P+Pbac[XP5.WH5] Exel7130/CyO (FlyBase).

The allele SoxF^KG09145 was recombined onto an FRT42D chromosome using standard genetic procedures. Mitotic recombination SoxF^KG09145 clones were generated by the FRT/FLP method (Xu and Rubin,1993) in L1-2 larvae from the cross between FRT42D SoxF^KG09145/CyO males to hsFLP122;;FRT42D ubi-GFPfemales. To induce the clones, 24-72 hours after egg laying (AEL) larvae were heat shocked at 37°C for 30 minutes. Mutant tissue was detected by the absence of the GFP marker.

SoxF and dTCF^DeltaN overexpression clones were obtained by incubating yw hs-FLP122; act>y+>Gal4, UAS-GFP/+UAS-SoxF/+ or UAS-dTCF^DeltaN/+ larvae for 10 minutes at 35.5°C at two developmental times (48-72 hours and 48-96 hours AEL). wg, rn and arm overexpression clones were obtained by crossing males of their respective UAS lines to yw122,act>hsCD2>GAL4 females (Basler and Struhl, 1994). Larvae from the crosses were heat shocked for 10-20 minutes at 35.5°C between 48 and 96 hours AEL. To mark the clones,CD2 was induced by subjecting late L3 (wandering) larvae to a 30-minute heat shock at 37°C, followed by a 30 minute recovery period at 25°C just prior to dissection.

The antibodies used for immunostaining were: mouse anti-β-galactosidase (Sigma, 1/1000) and rabbit anti-β-galactosidase(Cappell, 1/1000), mouse anti-GFP (Invitrogen, 1/1000), mouse anti-CD2(Serotec, 1/400), mouse anti-Nub (Ng et al., 1995), rabbit anti-Tsh(Wu and Cohen, 2000), mouse anti-Wg (4D4, Developmental Studies Hybridoma Bank, Iowa University, 1/100),guinea pig anti-Hth (Casares and Mann,1998), mouse anti-Arm (N27A1, Developmental Studies Hybridoma Bank, 1/50), rabbit anti-cleaved Caspase 3 (Cell Signaling, 1/500). Appropriate secondary antibodies were conjugated to Alexa 488, 568 or 647(Invitrogen, 1/800). After dissection and fixation, larvae were incubated with primary antibodies overnight at 4°C or for 2 hours at room temperature.

Rhodamine-phalloidin staining (Invitrogen, 1/400) was performed during secondary antibody incubation or was added directly to the mounting medium.

BrdU incorporation followed standard protocols(Sullivan et al., 2000). Discs were incubated for 30 minutes in a 10 mM BrdU (Roche) solution, and BrdU was detected with a mouse anti-BrdU antibody (Roche, 1/400).

Fluorescent in situ hybridization was performed as described(Vanzo and Ephrussi, 2002)with minor modifications. SoxF antisense RNA probes were synthesized from a plasmid that contains the coding sequence of SoxF or the probe described by Cremazy et al. was used(Cremazy et al., 2001), with incubation at 65°C or 55°C, respectively. Probes were labeled with digoxigenin (Dig), and detected with an alkaline phosphatase-conjugated anti-Dig antibody (1/1000), both from Roche. Signal was developed using Fast Red tablets (Roche) followed by standard immunostaining(Vanzo and Ephrussi, 2002). Confocal image acquisition was performed on a SP2-AOBS confocal system(Leica). Stacks of (x,y) sections were recorded along the z-axis every 1 μm. Single z-sections (`cross-sections')were recorded as (x,z) confocal sections, with a z-step of 1μm. In some cases, maximum or average projections of the z-series were produced in order to visualize the total signals in the samples. Confocal data processing was performed using Adobe Photoshop.

RNA extraction was performed using the RNeasy Kit (Qiagen). For each genotype, eight larvae were collected in lysis buffer (RLT, RNeasy, Qiagen)and ground with a pestle in an Eppendorf tube. The lysate was passed through a QIAshredder column (Qiagen) to optimize extraction and DNA digestion was performed during the process of extraction.

For the RT-PCR reactions, 5-7 μg of RNA was used for the first-strand cDNA synthesis (SuperScript First-Strand Synthesis Kit, Invitrogen). PCR was performed using 2 μl of the first-strand synthesis reaction with GoTaq polymerase (Promega). PCR conditions were: 30 cycles of 30 seconds at 95°C, 30 seconds at 55°C, 30 seconds at 42°C. Primers were:L1-SoxF (5′-TGCAACTGCAACAACATCAA-3′) and R1-SoxF(5′-GTCAGATAGCCACCGTGCTC-3′), which amplify a fragment specific to the SoxF transcript; L1RpS23 (5′-AGATCTTGGGCGTTCCTTCT-3′)and R1Rps23 (5′-TTGCAATCCAAATCACAGGA-3′) for the RpS23gene; L1CG30071 (5′-AGAAGCTGGAGCAGAAGCTG-3′) and R1CG30071(5′-GCTGCTGAATTCTTGGAAGG-3′) for the CG30071 gene;L1-8394 (5′-GCGATGGCGAGTATAGGAAC-3′) and R1-8394(5′-CAGCGATACGATGAACATGC-3′) for the CG8394 gene. For the SoxF, CG8394 and RpS23 genes, amplification was specific for the corresponding messenger RNAs, as the primers were designed against coding sequences that are separated by introns in the pre-mRNAs.

During the three larval stages (L1-3), the wing disc is progressively subdivided in three concentric domains: the prospective body wall, the hinge and the wing blade (Fig. 1). In L3 discs, these different domains are clearly demarcated by folds in the epithelium (Fig. 1A). The hinge is formed by two concentric bands of tissue(Fig. 1A,B) that will give rise to the distal hinge, which is contiguous with the wing blade, and the proximal hinge, which forms the axillary sclerites of the wing articulation(Bryant et al., 1978; Casares and Mann, 2000; del Alamo Rodriguez et al.,2002; Neumann and Cohen,1996).

During L2, prospective distal hinge cells express the POU gene nubbin (nub) (Ng et al.,1995; Zirin and Mann,2007), while proximal hinge cells express the zinc-finger gene teashirt (tsh) (Azpiazu and Morata, 2000; Fasano et al., 1991; Soanes and Bell,2001) (summarized in Fig. 1I). At the beginning of L3, a ring of wg expression appears in the prospective distal hinge, the so-called wg inner ring(IR). The wg IR expression is included within the nub domain in cells that also express the Kruppel-like transcription factor rn(del Alamo Rodriguez et al.,2002), and is driven by a specific regulatory element, the spade-flag (spd-fg) enhancer(Neumann and Cohen, 1996). Starting in early L3, wg IR drives intercalary proliferation between the nub and tsh domains generating a region that expresses neither of the two, the so-called gap domain(Zirin and Mann, 2007). By late L3, a second ring of wg, called the wg outer ring (OR),appears in the prospective proximal hinge, and abuts the distal limit of tsh expression (see Fig. 1I). In this paper, we focus on the regulation of the expression and function of the wg IR domain, as it has a major role in controlling hinge proliferation.

To identify genes that are differentially expressed in the hinge, we genetically marked hinge cells by driving GFP with the hinge-specific driver zfh-2^MS209-GAL4 (Fig. 1D). This driver reproduces the pattern of the zfh-2(zfh2) gene, which is expressed in most hinge cells (Terriente et al., 2008; Whitworth and Russell,2003). GFP⁺ (hinge) and GFP^- (body wall plus wing blade) cells were FACS sorted and their transcriptome profiles compared(a full account of this analysis will be published elsewhere). This experiment identified Drosophila Sox15 (CG8404) as the transcript most over-represented in hinge cells. Recently, Bowles and co-workers reassigned this gene to the SoxF group of the Sox family, making it the sole Drosophila member of the group, which in mammals includes Sox7,Sox17 and Sox18 (Bowles et al., 2000). We adopt their nomenclature and hereafter refer to this gene as SoxF.

SoxF had been reported to be transcribed in the hinge of late L3 wing discs (Cremazy et al.,2001). We further mapped the SoxF domain relative to wg and rn reporters by in situ hybridization. The SoxF domain abuts rn and the wg IR on its distal border and extends into the proximal hinge overlapping the late wg OR(Fig. 1E-H; see Fig. S1 in the supplementary material; data not shown). Therefore, the realm of SoxFexpression straddles the gap domain (Zirin and Mann, 2007). This adjacent, non-overlapping expression between SoxF and the domains of rn and wg is also observed at earlier stages (see Fig. S1 in the supplementary material).

In order to determine SoxF function, we analyzed wing imaginal discs from larvae homozygous for the null allele SoxF^KG09145 (see Materials and methods). SoxFmutant wing discs showed hinge overgrowths(Fig. 2) that caused misfolding of both its dorsal and ventral regions. However, the wing pouch and body wall regions seemed unaffected. To determine the origin of these overgrowths within the hinge, we mapped them relative to the expression of nub, tsh and the intervening gap domain (Fig. 2A,B). In SoxF mutant discs, both the prospective distal hinge, which expresses nub, and the gap domain were significantly enlarged (Fig. 2A,B). In addition, the overgrown hinge still expressed high levels of homothorax (hth) (not shown), which is a hinge marker(Azpiazu and Morata, 2000; Casares and Mann, 2000). Therefore, the tissue overgrowth observed in the SoxF mutants correlates with the SoxF expression domain, suggesting that SoxF has an autonomous effect on the control of hinge proliferation. In addition, the overgrowth cannot be explained by changes in cell fate because we still detected normal expression of hinge-specific markers.

In SoxF mutant discs, we detected elevated levels of incorporation of the S-phase marker BrdU specifically in the hinge, indicating that the overgrowths were in fact caused by increased cell proliferation(Fig. 2C,D). In addition, we noted an increase in apoptotic cell death, as detected by activated Caspase 3(Decay - FlyBase) staining (see Fig. 4D). This apoptosis is associated with activation of the Jnk pathway, as indicated by the upregulation of a transcriptional reporter of the Jnk target puckered (not shown).

Since expression of wg at the IR is necessary for, and sufficient to induce, the proliferation of hinge cells(Neumann and Cohen, 1996; Zirin and Mann, 2007), we asked whether its expression was altered in SoxF mutants. We compared the expression of a wg-lacZ transcriptional reporter in wild-type and SoxF^KG09145 mutant wing discs. In wild-type discs, wg-lacZ is expressed in two distinct rings in the hinge, IR and OR,separated by a non-expressing region (Fig. 3A). However, in SoxF mutant discs, wgtranscription spread throughout the hinge and no wg-negative territory remained (Fig. 3B). When we examined the effect of removing SoxF function in clones, we observed effects on wg expression only in clones spanning the hinge. In these clones, wg expression filled the domain between the IR and OR rings, which thus became connected (Fig. 3C). Sox mutant clones in the wing pouch or prospective notum had no effect on wg (not shown). These results indicate that SoxF is required cell-autonomously to repress wgtranscription in the domain that separates the IR and OR.

The expression of wg in the IR is controlled by the spd-fg enhancer which, when linked to lacZ, drives reporter gene expression in the IR and wing margin expression domains of wg(Neumann and Cohen, 1996). In wild-type discs, spd-fg-lacZ expression was seen as a narrow stripe centered in the prospective distal hinge fold(Fig. 3D,D′). However, in discs in which the expression of SoxF had been knocked down by RNAi(zfh-2^MS209-GAL4; UAS-dsSoxF), the expression of this reporter was considerably wider (Fig. 3E,E′), now filling the distal hinge fold and abutting the proximal hinge fold (Fig. 3E′). Therefore, the repression of wg by SoxF is likely to occur through the wg spd-fg enhancer.

As wg expression was derepressed in SoxF mutant conditions, we tested whether SoxF was sufficient to block wg transcription. We observed that SoxF-expressing clones were able to repress wg expression in the IR at both the protein(Fig. 3G,G′) and transcriptional (Fig. 3F,F′) level. According to our previous observations, the expression of SoxF also blocks the expression of the spd-fg-lacZenhancer (Fig. 3G) in a cell-autonomous manner, reinforcing the idea that the regulation of wg IR by SoxF works through the spd-fg enhancer. Clones overlapping the OR showed no effects on wg expression, in agreement with the co-expression of SoxF and wg OR found in normal discs (data not shown).

The correlation between wg derepression and hinge overgrowth,together with the known role of wg as an essential mitogen in the hinge, led us to test whether wg was itself required for the overgrowths. We recombined the SoxF^KG09145 allele into a wg^spd-fg background. wg^spd-fg is a regulatory mutation that deletes the enhancer that drives wgexpression in the IR (Couso et al.,1994; Neumann and Cohen,1996; Tiong and Nash,1990). Accordingly, in spd-fg mutant discs, the prospective distal hinge underproliferates and spd-fg mutant adults lack distal hinge structures (Neumann and Cohen, 1996). We verified that in spd-fg mutant discs there is no increase in apoptosis, as monitored by anti-activated Caspase 3 staining (not shown), which confirms that reduced hinge proliferation is the major cause of the spd-fg adult phenotype. In wg^spd-fg, SoxF^KG09145 double mutants, the distal hinge was not overgrown (Fig. 4C). This result indicates that wg is required for the overproliferation observed in SoxF mutant discs. In fact, the reduction of the distal hinge was even stronger in the double mutant discs than in wg^spd-fg discs, as indicated by the width of the distal hinge fold and the almost complete loss of the wg IR(Fig. 4B,C). This stronger phenotype can be accounted for by the apoptosis that we still detect in the hinge region of wg^spd-fg, SoxF^KG09145 discs(Fig. 4E), which is similar to that observed for SoxF single mutants and which does not occur in wg^spd-fg discs.

The wg pathway activates several inducible antagonists that modulate its signaling activity, some of which, such as Notum and nmo, act in the wing (Gerlitz and Basler, 2002; Zeng and Verheyen, 2004). The fact that the domain of SoxFexpression in the hinge coincides with the region under wgproliferative control prompted us to ask whether SoxF itself could be induced by the wg pathway. Indeed, clones of Arm^*-expressing cells, in which the pathway is constitutively active, caused cell-autonomous activation of SoxF expression in regions of the notum close to the hinge(Fig. 5A). Although we have not performed a detailed study of where Arm^* clones activate SoxF using markers for distinct domains within the notum, we noted that SoxF induction does not occur in the dorsal-most notal region,suggesting that factors such as pannier and/or the iro-Cgenes, which are expressed in this region(Calleja et al., 2000; Diez del Corral et al., 1999),might be limiting the competence to activate SoxF in response to Wnt signaling. In agreement with this restricted competence, SoxF is not found associated with two other domains of wg expression in the disc(not shown) that map to the prospective tegula of the wing(Casares and Mann, 2000) and the medial notum (Calleja et al.,1996). Similar to Arm^*, clones expressing Wg also led to ectopic expression of SoxF, although this time the induction was, in part,non-autonomous, reflecting the diffusible nature of the Wg ligand(Fig. 5B). Conversely, clones in which the wg pathway is blocked by the expression of a dominant-negative form of TCF (Pangolin - FlyBase), dTCF^DeltaN, resulted in autonomous repression of SoxF (Fig. 5C). Similar results were observed when the expression of dTCF^DeltaN was driven in the dpp domain that intersects SoxF expression (Fig. 5D). Therefore, these results reveal a cross-regulatory loop between wg and SoxF: wg induces SoxF,which in turn acts as a wg antagonist by blocking the spreading of its transcription throughout the hinge and, potentially, by attenuating its pathway. We noted that SoxF is expressed in late L2 wing discs, well before the rings of wg expression in the hinge are established (see Fig. S1 in the supplementary material). In order to determine whether wg is required for SoxF early in development, and not only in the hinge, we examined SoxF transcription in late L2 discs of wg^CX3 mutants. wg^CX3 is a regulatory mutant that lacks the earliest wg expression in the wing disc, and,as a consequence, wg^CX3 discs fail to establish the wing field (Klein and Arias, 1998). These mutant discs lack SoxF expression (see Fig. S1A,B in the supplementary material), indicating that wg is required throughout development for SoxF transcription in the wing disc.

Although wg activates SoxF expression, only the cells adjacent and proximal to the wg IR, but not the wg-expressing cells themselves, express SoxF. This indicates that an additional regulatory mechanism operates to limit, in space, the transcription of SoxF.

The activation of wg expression in the hinge is coupled to the mechanisms that pattern the wing disc along its proximodistal (PD) axis(Azpiazu and Morata, 2000; Casares and Mann, 2000; del Alamo Rodriguez et al.,2002; Terriente Felix et al.,2007; Whitworth and Russell,2003; Wu and Cohen,2002). One of the genes required for wg IR expression is rn, a transcription factor that is expressed in the prospective distal hinge and wing pouch. The wg IR, which abuts the SoxFexpression, appears at the edge of the rn domain(del Alamo Rodriguez et al.,2002). Therefore, the rn and SoxF expression domains are mutually exclusive (Fig. 1E,G; see Fig. S1 in the supplementary material). We checked whether rn could be repressing SoxF and thus polarizing its activation along the PD axis in the hinge. Ectopic clones of rnrepressed SoxF expression autonomously(Fig. 5E,E′), suggesting that this is indeed the case. The reciprocal repression, of SoxF on rn, did not seem to take place, as the domain of the rn-lacZreporter did not change in SoxF mutant discs (not shown). Therefore, SoxF is linked to the mechanism of PD axis formation of the disc in a way that ensures its directional activation by wg specifically straddling the gap domain of the hinge, the cell population whose proliferation is controlled by IR wg. Here, SoxF performs a key role in restricting the activation of the wg pathway and, by doing so, controls hinge growth.

Here we describe a novel negative-feedback mechanism in the wgpathway that is required to restrain the expression of wg itself, and which is essential to control organ growth.

During Drosophila development, the wg pathway often leads to the activation of genes that attenuate its signaling pathway. This is the case, for example, for Notum and Dfz-3, which are expressed in the wing disc in response to peak levels of signaling to reduce ligand availability for the Wg receptors(Sivasankaran et al., 2000),and for nemo, which acts intracellularly to block the signal transduction pathway (Zeng and Verheyen,2004). In all cases described, these negative-feedback components act in all domains of wg expression and none regulates wgexpression at the transcriptional level. However, in the case investigated here, the putative transcription factor SoxF is activated non-autonomously by wg in a hinge-specific manner. SoxF in turn represses wg transcription driven by the wg spd-fgenhancer, thus restricting the production of wg to the thin IR domain. Interestingly, the SoxF phenotype is similar to those of dominant Dichaete (D) mutations. D is a SoxB gene not normally expressed in the wing disc(Mukherjee et al., 2000). However, flies carrying dominant D mutations show reduced hinge structures. This phenotype is caused by ectopic D expression in the prospective hinge region of the disc(Russell, 2000). One of the salient features of D discs is the repression of the wg IR(Russell, 2000), which is reminiscent of the wg repression by SoxF we have described. Therefore, and taking into account the similarity between Sox proteins in their HMG DNA-binding domain, the ectopic D might be mimicking the repression of wg that is normally exerted by SoxF.

The tight regulation of the growth of the hinge depends critically on the wg-induced activation of SoxF in the growing territory. Nevertheless, this activation is `polarized' along the PD axis, taking place only in cells adjacent and proximal to the IR. We propose that this directionality in SoxF activation results from the mechanisms that pattern the wing disc along its PD axis. It has been suggested that wg is activated non-autonomously by a signal produced by the vg-expressing wing pouch cells, but excluded from them(del Alamo Rodriguez et al.,2002). This would generate a circular domain of wgexpression surrounding the wing pouch. However, in the absence of SoxF, the domain of wg is abnormally broad and causes hinge overgrowth. This ectopic wg expression does not seem to result from a misregulation of hinge-specific genes: the expression of nub, tsh,hth and rn and their relative positioning in the hinge are unaffected in SoxF mutant discs (Figs 2 and 4; data not shown). Therefore,it seems that in the absence of SoxF, hinge cells cannot respond to the wg activating signals with enough precision to give rise to a thin ring of wg expression. Our results show that this precision is achieved through a double repression mechanism. First, wg activates its own transcriptional repressor, SoxF. This would lead to the extinction of wg expression if it were not for rn, which acts as a repressor of SoxF. Second, rn, by repressing SoxF, permits wg transcription. The result is that wg expression becomes restricted to a narrow circular stripe at the edge of the rn domain that provides a highly localized source of Wg. This signal activates, simultaneously and in the same cells, proliferation and the upregulation of SoxF, which restricts the production of the signal (Fig. 6). Therefore, SoxF joins SoxN and SoxD (Sox102F -FlyBase) (Chao et al., 2007; Overton et al., 2007) as the third Drosophila Sox known to antagonize the wg pathway. The vertebrate Sox proteins Sox9 (Mori-Akiyama et al., 2007), XSox3 (Zorn et al., 1999) and XSox17 (Sinner et al., 2004) have also been shown to downregulate the Wnt/β-catenin pathway. Therefore, this antagonism seems evolutionarily conserved.

The relationship between SoxF genes, the wg/Wnt pathway and the control of tissue proliferation seems to extend to disease. The SoxF Sox17 is normally expressed in the gut epithelium where it downregulates Wnt signaling via degradation of β-catenin and TCF. In colon carcinomas, the expression of the SoxB gene Sox17 is often reduced, and this is associated with tissue overproliferation(Sinner et al., 2007). Moreover, inactivation of the SoxE gene Sox9 leads to increased cell proliferation and hyperplasia in the mouse intestine(Bastide et al., 2007). The authors concluded that Sox9 is essential for the fine-tuning of the transcriptional activity of the Wnt pathway(Bastide et al., 2007). Interestingly, the expression of Sox9 is regulated by the Wnt pathway itself (Blache et al., 2004). Our results in Drosophila point to the possibility that the transcriptional regulation of Wnt expression by Sox genes might be a common feature of this proliferation-associated feedback loop.