Pygopus, a nuclear PHD-finger protein required for Wingless signaling in Drosophila

The Drosophila protein Wingless (Wg) is a founding member of the Wnt family of secreted glycoproteins, which are present throughout the animal kingdom. Wnts have been shown to play essential roles in determining many cell fate decisions throughout development in worms, flies, amphibians and mice (Cadigan and Nusse, 1997). In addition, inappropriate activation of Wnt signaling has been linked to several forms of human cancer (Polakis, 2000).

Cells respond to Wg and many vertebrate Wnts by a highly conserved signal transduction cascade that revolves around the Armadillo (Arm; β-catenin in vertebrates) protein (Willert and Nusse, 1998). In the absence of Wg/Wnt signaling, a cytosolic pool of Arm/β-catenin is phosphorylated by a complex of Shaggy/Zeste white 3 (Sgg) (GSK3β in vertebrates), Axin and the adenomatous polyposis coli (APC) protein, then rapidly degraded via the ubiquitination/proteosome pathway (Peifer and Polakis, 2000; Polakis, 2000). Wg/Wnt signaling blocks the activity of the Zw3/GSK3β/Axin/APC complex, resulting in stabilization of Arm/β-catenin (Li et al., 1999; Salic et al., 2000). The stabilized protein then accumulates in the nucleus (Yost et al., 1996), where it forms a complex with members of the TCF/LEF1 (Pangolin – FlyBase) family of HMG group DNA-binding proteins (Molenaar et al., 1996; van de Wetering et al., 1997).

In the absence of Arm, TCF is thought to act as a transcriptional repressor of Wg target genes, through interaction with the transcriptional co-repressor Groucho (Cavallo et al., 1998). In addition, there is good evidence that the ARID domain protein Osa represses Wg target genes by acting in a chromatin remodeling complex that contains the bromodomain protein Brahma (Collins and Treisman, 2000; Treisman et al., 1997). Binding of Arm to TCF somehow blocks the action of these factors, converting TCF from a repressor to an activator (Collins and Treisman, 2000; Korswagen and Clevers, 1999).

In cultured cells, reporter genes with TCF/LEF1-binding sites are highly activated by the transient transfection of β-catenin (Molenaar et al., 1996). β-Catenin/Arm contains transcriptional activation domains both in the N- and C-terminal part of the protein (Cox et al., 1999; Hecht et al., 1999; Hsu et al., 1998; van de Wetering et al., 1997). Several factors have been found to bind to these regions of β-catenin, and to simulate its ability to activate transcription. The CBP/p300 acetyltransferases bind to the C terminus of β-catenin and synergize its transcriptional activity in cultured cells and frog embryos (Hecht et al., 2000; Miyagishi et al., 2000; Sun et al., 2000; Takemaru and Moon, 2000). The DNA helicase Pontin52 (also called TIP49a) binds to the N-terminus of β-catenin and synergizes with it in the reporter gene assay (Bauer et al., 1998). Pontin52 can also bind to the TATA box binding factor TBP, suggesting it links β-catenin/Tcf to the basal transcription complex (Bauer et al., 1998). β-catenin has also been found to bind directly to TBP (Hecht et al., 1999). The brahma ortholog Brm1 can also potentiate β-catenin activity in transient transfection assays (Barker et al., 2001). Thus, β-catenin can promote transcriptional activation from its Tcf anchor in a variety of ways.

The role of the fly homologs of these transcriptional co-activators has also been examined, sometimes with contradictory results. As predicted, pontin52 genetically acts as a positive regulator of Arm activity (Bauer et al., 2000). However, nejre (nej), the fly CBP, acts as a negative regulator, in contrast to the data summarized above (Waltzer and Bienz, 1998). With brahma, there are reports indicating a negative (Collins and Treisman, 2000) and a positive (Barker et al., 2001) role in regulating Wg targets in the wing. Some of the discrepancies may be explained by these factors acting differently in flies and vertebrates. However, the fly genetics are all based on the examination of partial loss-of-function mutants, often in sensitized backgrounds, or on expression of dominant-negative versions of the proteins. While convenient, these assays can be difficult to interpret clearly.

This report describes the identification of a new factor regulating Wg signaling, which we call Pygopus (Pygo). pygo is required for Wg signaling in at least a dozen different readouts in several tissues, suggesting that it is a core component of the pathway. Overexpression of pygo also inhibits Wg action, consistent with it acting in a complex. pygo is required downstream of Arm nuclear import and encodes a nuclear protein containing a single PHD finger, a zinc-binding motif often found in chromatin remodeling factors (Aasland et al., 1995; Capili et al., 2001; Pascual et al., 2000). Our data are consistent with a model where Pygo is necessary in order for Tcf/Arm to regulate target gene expression.

The P[GMR-Gal4] stock used (Freeman, 1996) was obtained from M. Freeman. A P[UAS-wg] line was provided by H. Krause. The P[sev-wg] transgene has been described previously (Cadigan and Nusse, 1996). These transgenes (all on the second chromosome) were recombined to create the GMR/wg and P[GMR-Gal4] P[sev-wg] stocks used for our EP and secondary screen. The Rorth collection of EP lines (Rorth et al., 1998) were obtained from the BDGP and the Bloomington stock center. The P[GMR-arm*]^F36 stock (Freeman and Bienz, 2001) was kindly provided by M. Bienz. The Gal4 driver P[Ptc-Gal4] was used and the dpp-lacZ line BS3.0 (Blackman et al., 1991) was used to mark dpp-expressing cells. The rod alleles, X1 and X2 (Scaerou et al., 1999) were obtained from R. Karess and R. Goldberg. An FRT^82B Axin mutant chromosome (Hamada et al., 1999) was provided by N. Tolwinski and E. Wieschaus. wg^CX4 is a molecular null (van den Heuvel et al., 1993).

P[UAS-pygo] lines were constructed by cloning the BglII/XhoI fragment of cDNA GH25362 into the pUAST vector (Rorth, 1996). This fragment contains the entire 815 codon ORF and some 5′ and 3′ UTR sequences. The P[UAS-pygo] construct was introduced into w¹¹¹⁸ hosts by P-element-mediated transformation. Five independent lines were obtained, one of which gives phenotypes less severe than the others in the GMR/wg suppression assay (see Fig. 1). Two of the strongest lines, 1-1 and 1-2 were used for all other experiments.

Deletions of the pygo locus were created by imprecise excision of the EP(3)1076 transposon, using the Δ2-3, Sb chromosome (Robertson et al., 1988). The EP(3)1076 was outcrossed to w¹¹¹⁸ flies for six generations before isogenization. This removed at least one linked lethal. Approximately 200 white minus males were obtained from dysgenic crosses, and 10 of these had a phenotype more severe than the parental chromosome (which die as pharates). These were characterized using PCR and the relevant PCR bands were sequenced to confirm the nature of each deletion. The alleles pygo¹⁰ and pygo⁹ are used in this report.

Random clones expressing pygo were generated using the P[Actin>CD2>Gal4] transgene (Pignoni and Zipursky, 1997). P[UAS-pygo] males were crossed to yw P[Actin>CD2>Gal4]; P[UAS-GFP]; P[HS-Flp]⁹⁹/TM6B females. A 60 to 90 minute heat shock 48-72 hours after egg laying (late second and early third larval instar) recombines out the CD2 spacer, generating a clone of Actin/Gal4 cells, which are then dissected and fixed at late third larval instar. The clone is marked with GFP. Unless otherwise indicated, at least 20 clones were examined for each marker.

For clonal analysis of the pygo¹⁰ allele, it was recombined onto a FRT^82B chromosome using standard methods (Xu and Rubin, 1993). The Axin, pygo double mutant was constructed by brute force screening of the male progeny of FRT^82B Axin/FRT^82B pygo¹⁰ mothers crossed to TM3/TM6 males (the cytological locations of Axin and pygo are 99C and 100D, respectively). These males were crossed to balancer virgins and sacrificed after progeny were observable. Their genomic DNA was screened for the presence of the Axin and pygo¹⁰ alleles using specific PCR primers. Two recombinants were found out of 150 progeny examined. For the rescue experiments of pygo embryonic phenotypes described below, a P[Da-Gal4] FRT^82B pygo¹⁰ recombinant was also created.

Clones in the wing imaginal discs were created essentially as described (Xu and Rubin, 1993) with the following modifications. FRT mutant males were crossed to ywP[HS-Flp]; FRT^82B P[arm-lacZ] females. The arm-lacZ transgene was a generous gift from D. Lessing and R. Nusse. A P[Ubi-GFP] was used to mark the clones in Fig. 7G-I. Clones were induced at 48-72 hours after egg laying (second larval instar). Clones in the eye-antennal imaginal discs were induced with P[Eyeless-Flp] (Newsome et al., 2000), kindly provided by B. Dickson and marked with arm-lacZ. In both wing and eye, the discs were analyzed at late third larval instar. Unless otherwise noted, at least 20 clones were examined for each marker. pygo¹⁰ germline clones were generated using FRT^82B P[ovo^D] as described (Chou and Perrimon, 1996) with heat shocks during larval development. Embryos lacking zygotic pygo were identified by the absence of eve-lacZ.

Immunostaining was as described previously (Cadigan and Nusse, 1996). Affinity-purified rabbit anti-Wg antisera (1:60 in discs; 1:20 in embryos) was from R. Nusse, guinea pig anti-Sens (1:50) was from H. Bellen, purified rabbit anti-Dll (1:150) was from G. Panganiban and rat anti-Elav (1:100) was from G. Rubin. Rabbit anti-Eve (1:100) was from Z. Han and R. Bodmer, and rabbit anti-Arm (1:200) was from N. Tolwinski and E. Wieschaus. Mouse monoclonal anti-En supernatant (1:2) was from the University of Iowa Hybridoma Bank, mouse and rabbit anti-β-galactosidase (1:500) from Sigma and Cappel, respectively. Antibody against the human nuclear pore complex protein (NPC) MMS-120P (1:500), which crossreacts with the fly NPC, was purchased from Convance. Cy3- and Cy5-conjugated secondary antibodies were from Jackson Immunochemicals and Alexa Flour 488-conjugated secondaries were from Molecular Probes. All fluorescent pictures were obtained with a Zeiss Axiophot coupled to a Ziess LSM510 confocal apparatus. All images were processed as Abode Photoshop files.

A digoxigenin-labeled pygo antisense probe was made using T7 RNA polymerase and pygo cDNA GH25362 linearized with SacI. In situ hybridization was performed as described (Cadigan et al., 1998). Embryos were photographed with a Nikon Eclipse800 compound microscope using DIC optics. Cuticles were prepared and photographed as previously described (Bhanot et al., 1999). Adult fly eyes were frozen overnight and photographed with a Leica M10 microscope.

For the GFP-Pygo chimeric protein, the Pygo ORF was PCR amplified and cloned into the pEGFP vector (Clontech). A fragment containing GFP fused in frame to the N terminus of Pygo was cloned into pAc5.1/V5-His A (Invitrogen), which contains the Actin 5C promoter. Sequencing of the chimeric gene revealed a single conservative amino acid substitution (G>V) at amino acid number 732. Drosophila S2 cells were obtained from R. Nusse and J. Dixon and cultured in Schneider’s Drosophila Media plus 10% fetal calf serum (Gibco). These cells were transfected with 1 μg of plasmid/six-well plate using Cellfectin as described by the manufacturer (Gibco) and fixed and immunostained with anti-NPC antibody 40 hours later, as described previously (Bhanot et al., 1996).

Drosophila have typical compound insect eyes (Fig. 1A). Misexpression of wg using the eye-specific GMR-Gal4 driver (Freeman, 1996) in combination with UAS-wg results in a dramatically reduced eye size (Fig. 1B). This phenotype can be used as the starting point in a misexpression screen for Wg signaling antagonists. If a Wg antagonist such as Axin is co-expressed with wg in the eye, the small eye phenotype is greatly suppressed (Willert et al., 1999). We used ‘EP’ P elements, which contain a Gal4-dependent promoter (Rorth, 1996), to randomly co-express genes with wg in the eye. A collection of 2300 EP lines were crossed to a P[GMR-Gal4]/P[UAS-wg] (GMR/wg) stock and the progeny were scored for suppression of the small eye phenotype.

The initial positives (36) from the screen were crossed to a P[GMR-Gal4], P[sev-wg] line. P[sev-wg] eyes lack interommatidial bristles but are otherwise morphologically normal (Cadigan and Nusse, 1996). Two of the positives suppressed the ability of Wg to inhibit bristles (data not shown), suggesting that they may be bona fide Wg signaling antagonists. One of these lines is inserted adjacent to a known negative regulator of the pathway, zw3 (EP(X)1576). Overexpression of zw3 is known to suppress Wg signaling (Steitz et al., 1998). The second line (EP(3)1076), significantly suppresses the GMR/wg phenotype (Fig. 1C) and corresponds to a novel gene.

The position of EP(3)1076 has been mapped by the Berkeley Drosophila Genome Project (BDGP; http://www.fruitfly.org/index.html) using inverse PCR, which was confirmed by PCR with genomic and EP element-specific primers (data not shown). It is located in a small intron in the 5′ UTR of a gene (BEST:LD21971) we refer to as pygopus (pygo; Fig. 2A). Northern blot analysis reveals the major isoform of the gene to be approximately 5 kb in length (data not shown). Sequencing of a 4 kb cDNA (GH25362; obtained from the BDGP) confirmed the predicted splice sites (GenBank Accession Number, AY075095). More recent searches have revealed an EST that extends the 5′ end of the transcript (LD18280, Fig. 2A), indicating the pygo transcript overlaps the transcript of rough deal (rod) (Scaerou et al., 1999) by at least 14 bases. This start site gives a transcript length of approximately 4.5 kb, which roughly agrees with the data from northern blots when polyadenylation is taken into consideration.

The EP(3)1076 transposon is inserted in the proper orientation to misexpress full length pygo, and this was confirmed experimentally. First, RT-PCR shows that expression of pygo, but not of rod, increases dramatically when a heat shock promoter is used to drive Gal4 expression (data not shown). Second, random clones of cells expressing Gal4 under the control of an Actin promoter (Pignoni and Zipursky, 1997) in a EP(3)1076 background cause a dramatic increase in pygo transcript levels (Fig. 2H). Third, and most directly, a P[UAS-pygo] transgene strongly suppresses the GMR/wg phenotype (Fig. 1D). These data argue that pygo is responsible for the antagonistic effects of EP1076 on Wg signaling.

The predicted Pygo protein contains 815 amino acids and possesses two recognized motifs, a predicted NLS at the N terminus (residues 39-44) and a PHD domain at the C terminus (residues 747-811; Fig. 2A). The function of PHD domains is unclear. They are zinc finger-binding motifs (Capili et al., 2001; Pascual et al., 2000), and are found in many transcription factors and chromatin remodeling proteins (Aasland et al., 1995).

Database searches reveal that there are potential vertebrate Pygo homologs. An embryonic mouse cDNA (Accession Number, AK011208) and human cDNA (XM_034083) have C-terminal PHD domains similar to that of Pygo (42-47% identity; 63-67% similarity). No other PHD domains in the human or fly genomes are more closely related. There is only one other region of sequence similarity outside this domain, encompassing the predicted NLS of the three genes. Further experiments will be required to determine if these vertebrate proteins function in Wnt signaling.

The expression profile of pygo in embryos was examined using in situ hybridization. pygo is expressed at relatively high levels in pre-blastoderm embryos (Fig. 2C) and this staining is absent in germline clones of pygo (Fig. 2D), indicating that it is maternal in origin. pygo expression drops rapidly after this early high level, and low levels of signal are observed throughout the embryos for the rest of embryogenesis. For example, at full germband extension, pygo expression is at such low levels that visualization of the message requires overstaining, as judged by significant signal in embryos lacking maternal and zygotic pygo (compare Fig. 2E,F). We believe that the allele used (pygo¹⁰) is a molecular null (Fig. 2B), although we cannot rule out that some small amount of aberrant mRNA is produced. In either case, the data indicate a high degree of maternally provided message, followed by a low level of ubiquitous zygotic expression. This continues into larval development, where pygo appears to be expressed at low levels in no distinctive pattern (e.g. the wing imaginal disc in Fig. 2G).

The suppression of the GMR/wg and P[sev-wg] phenotypes by pygo overexpression is consistent with the notion that high levels of Pygo block Wg signaling. However, the data can also be explained by pygo interacting with the targets of Wg in the eye or the promoters driving wg expression. To address this, we examined the effect of pygo overexpression on Wg readouts in other tissues.

In the third instar wing imaginal discs, wg is expressed in a stripe of cells along the dorsoventral border (Couso et al., 1994; Phillips and Whittle, 1993). Wg secreted from these cells is thought to act as a morphogen, regulating both short- and long-range targets (Neumann and Cohen, 1997; Zecca et al., 1996). In addition, Wg signaling refines the distribution of Wg protein by negative autoregulation of wg expression (Rulifson et al., 1996) and downregulation of the Wg receptor Frizzled2 (Fz2) (Cadigan et al., 1998). Thus, the wing imaginal disc offers several readouts to monitor Wg signaling.

Random clones of cells expressing high levels of pygo were generated as described (Pignoni and Zipursky, 1997), except that GFP was used to mark the clones (Cadigan et al., 1998). If a clone is positioned in the endogenous wg stripe, it blocks Wg expression (data not shown). This is not seen when Wg signaling is inhibited in these cells (Rulifson et al., 1996), indicating that pygo overexpression has consequences not related to Wg signaling. However, if the clone is adjacent to the wg stripe, then Wg protein is upregulated (Fig. 3A-C), consistent with a block in Wg signaling. The extent of Wg expansion is similar to that observed in clones mutant for dishevelled (a positive regulator of Wg signaling), which is due to derepression of Wg synthesis (Rulifson et al., 1996) and increased Wg stability produced by high levels of Fz2 (Cadigan et al., 1998).

To examine targets that are positively regulated by Wg in the wing, we chose the zinc-finger protein Senseless (Sens) and the homeodomain protein Distal-less (Dll). Sens is expressed in the proneural clusters on either side of the dorsoventral border, immediately adjacent to the Wg expression domain (Nolo et al., 2000). Inhibition of Wg signaling with a dominant-negative TCF blocks Sens expression (data not shown), demonstrating that it is a short-range target of Wg action. pygo-expressing cells outside the Wg expression domain completely lack Sens expression (Fig. 3D,E,F). The long-range target Dll (Neumann and Cohen, 1997; Zecca et al., 1996) is also always lost in clones overexpressing pygo (Fig. 3G-I). For reasons that are not clear, occasionally some expression persists just inside the clonal border (Fig. 3G-I, see arrows).

Finally, we show that pygo overexpression causes derepression of decapentaplegic (dpp) in leg imaginal discs. In the developing leg, wg and dpp are expressed in wedge-like domains just anterior to the posterior compartment, with wg highly enriched in the ventral half and dpp in the dorsal part. If Wg signaling is blocked, dpp expression becomes derepressed (Brook and Cohen, 1996; Jiang and Struhl, 1996; Theisen et al., 1996). If pygo is misexpressed using the patched-Gal4 driver, which is active in both the dpp and wg expression domains, then dpp expression (as judged by dpp-lacZ) is extended into the ventral compartment (compare arrows in Fig. 3J,K). This derepression of dpp expression is seen in the vast majority of leg discs examined and is again consistent with pygo overexpression antagonizing Wg signaling.

Overexpression of pygo clearly results in phenotypes consistent with a block in Wg signaling. However, a more physiologically relevant test of the importance of pygo for Wg function is the analysis of pygo mutants. Deletions of the pygo locus were created via imprecise excision of the EP(3)1076 transposon. Several deficiencies were generated, the most useful of which is pygo¹⁰ (Fig. 2B). This deletion removes the splice acceptor of the first intron and the first 295 residues of the Pygo ORF. Therefore, we believe it is null for pygo activity. It fully complements a null allele of rod (rod^X2) (Scaerou et al., 1999), in contrast to pygo⁹, which removes the transcription start site of rod (Fig. 2B). Mutants in γ-cop, the gene on the other side of pygo, do not exist, so we can not directly test whether pygo¹⁰ compromises its activity. However, as the 3′UTR of γ-cop and the intergenic region between it and pygo are unaffected in pygo¹⁰, we consider this unlikely. Thus, pygo¹⁰ is a deletion specific for pygo.

pygo¹⁰ homozygotes (zygotic mutants) have an early pupal lethal phenotype, as do pygo¹⁰/pygo⁹ transheterozygotes. However, embryos lacking maternal pygo failed to hatch, even when zygotic pygo is provided from wild-type males. pygo mutant embryos were subjected to cuticle analysis. Wild-type embryos have a distinctive patterning of denticles on their ventral cuticle, with each denticle belt arranged in a trapezoidal pattern with intermittent naked cuticle (Fig. 4A). A wg mutant does not form naked cuticle and has a characteristic denticle lawn phenotype (Fig. 4B) (Nusslein-Volhard and Wieschaus, 1980). When mothers producing pygo¹⁰ mutant eggs are crossed with pygo¹⁰ heterozygotes, two classes of mutant phenotype are observed. Approximately half the cuticles exhibit a denticle lawn extremely similar to wg mutants (compare Fig. 4C with 4B). The other half have a reduction in the number of denticle belts, with some denticle fusions (Fig. 4D; Table 1). This phenotypic class was also observed when the fathers were wild type for pygo (Table 1), indicating that they are pygo maternal mutants. Thus, embryos lacking both maternal and zygotic pygo have a cuticle phenotype indicating a loss of Wg signaling.

To confirm that loss of pygo activity is responsible for the phenotypes described above, pygo mutant phenotypes were examined in the presence of P[UAS-pygo] and P[Daughterless-Gal4] (P[Da-Gal4]), which is ubiquitously active during embryogenesis (Wodarz et al., 1995). When pygo maternal mutants contain P[Da-Gal4] but not P[UAS-pygo], 99% of the embryos have a reduction of abdominal denticle belts (Table 1). The cuticle shown in Fig. 4D has six denticle belts; no embryos had all eight. When pygo maternal mutants contain P[Da-Gal4] and P[UAS-pygo], there is a considerable shift of the phenotypic range to the right (Table 1). The cuticle shown in Fig. 4E has seven abdominal belts, and 4% of the progeny had all eight. Keeping in mind that only half of progeny contain P[Da-Gal4] (see Table 1 footnote), the data suggest that ubiquitous expression of pygo can significantly rescue the reduction of denticle belts in maternal pygo mutants.

P[Da-Gal4]/P[UAS-pygo] can also significantly rescue pygo maternal/zygotic mutants. In control crosses where P[UAS-pygo] is omitted, approximately half (52%) of the embryos have a denticle lawn. When the pygo transgene is included, only a quarter (25%) of the progeny have a lawn of denticles, which is the predicted result as only half the embryos have zygotic P[Da-Gal4]. In fact, the ratio of full to lesser lawns is even better than expected, which could be due to some maternal expression of P[Da-Gal4]. There is also an increase in the number of progeny with seven or eight denticle belts (22% versus 2%), which are presumably rescued pygo maternal mutants. These results indicate that pygo is the gene responsible for the embryonic phenotypes we observe in pygo¹⁰ mutant embryos.

The pygo embryonic phenotype was further characterized using molecular markers. The Engrailed (En) protein is normally expressed in epidermal stripes of single segment periodicity (Fig. 4F). In wg mutants, the En stripes are initiated normally but fade from the epidermis by full germband extension (van den Heuvel et al., 1993) (Fig. 4G). In embryos lacking maternal and zygotic pygo, the En pattern begins normally but alternating stripes become slightly irregular during germband extension (data not shown). By full germband extension, the En stripes are largely absent but some expression does remain (see arrows Fig. 4H). Wg signaling positively regulates its own striped expression at the same stage, indirectly through maintenance of Hedgehog expression and by a more direct Wg autoregulatory loop (Hooper, 1994). The Wg stripes are normal at early stages (data not shown) but fade at full germband extension in a pygo mutant (Fig. 4K). In addition, the dorsal derepression of Wg expression seen in wg mutants (van den Heuvel et al., 1993) is also observed in the pygo mutants (arrow in Fig. 4K). Both the En and Wg expression patterns in pygo maternal/zygotic mutants are consistent with Wg signaling being severely compromised in the absence of pygo.

In the mesoderm, Wg is needed for expression of Even skipped (Eve) in a subset of pericardial cells (Wu et al., 1995) (Fig. 4L,M, arrows). In pygo null mutants, the pericardial Eve expression is completely absent (Fig. 4N). Eve is still expressed in the CNS in both wg and pygo mutants, though the pattern is more severely disrupted in pygo (compare arrowheads in Fig. 4M and 4N). The Eve-positive RP2 neurons, which are absent in wg mutants (Chu-LaGraff and Doe, 1993), are also missing in pygo mutants (data not shown), once again consistent with pygo being required for Wg signaling.

As stated above, pygo¹⁰ homozygotes and pygo¹⁰/pygo⁹ transheterozygotes die around mid-pupation, indicating that maternal pygo expression can provide enough activity for viability until this stage. However, the imaginal discs of third instar pygo¹⁰ homozygotes are severely reduced in size and display abnormal morphology. Therefore, we examined the role of pygo in these tissues using mosaic analysis.

Fig. 5A-E show the effects of inducing pygo mutant clones in the wing. In the wing imaginal disc, Wg signaling at the dorsoventral boundary of the presumptive wing blade is required for formation of an adult structure known as the wing margin (Couso et al., 1994; Phillips and Whittle, 1993). Wings from flies containing clones of pygo¹⁰ frequently contain notches, like the one in Fig. 5A, caused by a loss of wing margin. To confirm that these notches are due to a loss of Wg signaling, molecular markers were examined at third larval instar as previously determined in pygo-overexpressing cells (Fig. 3A-I). Loss of pygo causes derepression of Wg adjacent to the stripe and does not affect Wg expression in the stripe (Fig. 5B). pygo clones in the wing disc also show a cell autonomous loss of expression of the Wg targets Sens and Dll (Fig. 5C-E). The Sens result was observed with 100% penetrance (n>20). In the case of Dll, all clones had reduced expression, with Dll completely absent 28% of the time (Fig. 5D), a large reduction with 57% frequency (not shown) and a modest reduction in 15% of the clones (Fig. 6E; n=92). Thus, as in the embryo, loss of pygo results in a dramatic reduction in several Wg-dependent readouts, though the results with En and Dll suggest that Wg signaling may still occur at a modest level without pygo.

In the developing eye, misexpression of wg at low levels with the sevenless promoter (P[sev-wg]) results in a morphologically normal eye, except that the interommatidial bristles are absent (Cadigan and Nusse, 1996). Expression at higher levels with GMR-wg represses the bristles and causes a severe reduction in eye size (Fig. 1B). Clones of pygo¹⁰ in either misexpression background completely suppress the effects of Wg (data not shown).

Finally, reduction of pygo in the leg disc gives phenotypes consistent with loss of Wg signaling. As outlined previously, Wg signaling is required for ventral leg identity, at least in part by repressing the dorsally expressed gene dpp (Brook and Cohen, 1996; Jiang and Struhl, 1996; Theisen et al., 1996). To examine the role of pygo in the leg, we used a hypomorphic allele. Besides the Gal4-dependent phenotypes observed with EP(3)1076, we also found Gal4-independent recessive phenotypes. To avoid confusion, we refer to the EP(3)1076 allele as pygo^EP in this context. pygo^EP homozygotes are late pupal lethal, and exhibit several defects in their exoskeleton, including malformed legs (Fig. 5I). The sex comb, a stout row of bristles seen on the ventral side of the first leg in Drosophila males (Fig. 5F, arrow) is missing in the pygo^EP legs (Fig. 5I, arrow). At the molecular level, the dpp-lacZ reporter, which is normally expressed primarily dorsally (Fig. 5G), becomes derepressed ventrally in pygo^EP legs (Fig. 5H). Once again, loss of pygo results in a phenotype consistent with a loss of Wg signaling. The fact that every Wg readout we examined is pygo-dependent suggests that it is a core component of the pathway.

If pygo is a core component of Wg signaling in the fly, where does it act in the pathway? We approached this question with epistasis analysis. Initially, this was achieved via overexpression. In the absence of Wnt signaling, β-catenin (and by extension Arm) is believed to be phosphorylated at serine and threonine residues at its N terminus via the GSK3β/Axin/APC complex (Peifer and Polakis, 2000; Polakis, 2000). If these residues are deleted or substituted, β-catenin becomes resistant to degradation (Yost et al., 1996). In flies, these mutant forms of Arm (Arm*) activate Wg signaling independently of Wg (Pai et al., 1997). When placed under the control of the GMR promoter, Arm* causes a small eye phenotype similar to that of GMR-wg (Fig. 1E) (Freeman and Bienz, 2001). Co-expression of pygo severely suppresses this phenotype (Fig. 1F). This strongly suggests that pygo overexpression blocks Wg signaling downstream of Wg-induced Arm stabilization.

To examine the position of pygo in the pathway using loss-of-function genetics, we created Axin, pygo double mutants. In Axin mutants, the signaling pathway is constitutively activated because of stabilization of Arm (Hamada et al., 1999; Tolwinski and Wieschaus, 2001; Willert et al., 1999). As found in vertebrate systems, Axin functions in a complex with Sgg to phosphorylate Arm (Willert et al., 1999; Yanagawa et al., 2000). The Wg target gene Sens was used as a readout in wing imaginal discs. As shown before (Fig. 5C), in pygo clones, Sens expression adjacent to the dorsal/ventral Wg stripe is lost (Fig. 6A). In Axin clones, Sens is activated (Fig. 6B), no matter where in the presumptive wing blade the clones are located (data not shown), as loss of Axin constitutively activates Wg signaling (Hamada et al., 1999). In Axin, pygo double mutant clones, Sens expression was always lost (Fig. 6C). Thus, pygo acts downstream of Axin in this assay.

Epitasis analysis was also performed in the eye. At the beginning of the third larval instar, a wave of apical constriction of the columnar epithelial cells, called the morphogenetic furrow (MF) sweeps across the presumptive eye from the posterior to the anterior (Wolff and Ready, 1993). Behind the MF, ordered clusters of photoreceptors develop (red stain in Fig. 6D). When Wg signaling is activated in the primordial eye, such as in Axin mutant clones, no photoreceptors are specified (Lee and Treisman, 2001). Thus, the eye offers another test of whether pygo is epistatic to Axin.

Photoreceptor development, as judged by Elav staining, appears normal in pygo mutant cells (Fig. 6D; clones are marked by a lack of green signal). Even at higher magnification, no detectable difference was observed in the photoreceptor clusters between pygo-positive and pygo mutant cells (data not shown). As previously reported (Lee and Treisman, 2001) clones that lacked Axin lack any evidence of photoreceptor development (Fig. 6E). This dramatic phenotype is completely rescued in Axin, pygo double mutant clones (Fig. 6F), clearly demonstrating that pygo is epistatic to Axin. This is consistent with the overexpression studies that suggest pygo acts downstream of Arm stabilization.

We attempted to extend the epitasis analysis by examining the cuticles of Axin, pygo germline clones, but despite the fact that many Axin or pygo germline eggs could be obtained, only a few malformed eggs that were not fertilized were obtained with Axin, pygo double mutants.

When Wg signaling is activated, Arm is stabilized and translocates to the nucleus. In Drosophila, it has proved very difficult to detect nuclear Arm, even in cells receiving high levels of endogenous Wg (Tolwinski and Wieschaus, 2001). However, it has recently been shown that Axin maternal and zygotic mutant embryos display high levels of nuclear Arm (Tolwinski and Wieschaus, 2001). Because attempts to make Axin, pygo germline clones were unsuccessful, clones in the wing disc were generated to investigate Arm levels and localization. In clones of cells lacking pygo, Arm is present at low levels at the cell periphery, consistent with its role in adherence junctions (Fig. 7A,D,G) (Willert and Nusse, 1998). In Axin clones, Arm protein levels are greatly increased in both the nucleus (Fig. 7B,E,H) and cytoplasm (data not shown). Axin, pygo double mutant clones also have high levels of cytosolic and nuclear Arm (Fig. 7C,F,I), though the nuclear levels of Arm appear slightly less than in Axin clones (compare Fig. 7B with 7C). We interpret these data to mean that Arm is still stabilized in the absence of pygo (as would be expected if pygo acts downstream of Axin) and that, for the most part, pygo is not required for Arm nuclear import.

Consistent with these data, a GFP-Pygo fusion protein localizes to the nucleus in a Drosophila cell line. Compare Fig. 7J,K,L, which show GFP-Pygo localization in relation to the nuclear membrane and indicate that, under these conditions, the vast majority of Pygo is in the nucleus. Thus pygo acts genetically downstream of Arm stability and nuclear import, consistent with the nuclear localization of the Pygo fusion protein.

In this study, a total of twelve distinct readouts of Wg signaling from embryos and leg, wing and eye imaginal discs were found to be significantly (two readouts) or completely (ten readouts) blocked in cells lacking pygo (Figs 4,Fig. 5-6). The effects of pygo loss in clones were completely cell autonomous for Wg and Sens expression in the wing (Fig. 5B,C). In addition, pygo transcripts are ubiquitously expressed at low levels throughout embryonic and larval tissues (Fig. 2C,E,G). It is formally possible that pygo acts to produce a factor that is required for Wg signaling or acts in parallel to the pathway. However, the fact that pygo is required for Wg action in so many contexts favors a model where Pygo acts directly in the signal transduction cascade of cells that receive Wg.

In the case of the embryonic En stripes (Fig. 4H) and Dll expression in the wing blade primordia (Fig. 5E), the loss of pygo activity resulted in a less severe effect than that observed in wg or Wg signaling component mutants (Neumann and Cohen, 1997; van den Heuvel et al., 1993; Zecca et al., 1996). The current model of Wg action in the wing, where Wg is thought to act as a morphogen, postulates that the Dll promoter requires a low level of Wg signaling for its activation (Neumann and Cohen, 1997; Zecca et al., 1996). Perhaps loss of pygo does not completely abolish Wg signaling, so that there is still some activation of Dll and en. Alternatively, there could be a redundant factor that can partially replace pygo, or specific promoters are less sensitive to loss of pygo than others. However, the ability of pygo mutants to block the high levels of signaling induced by loss of Axin (Fig. 6A-F), argues that many targets absolutely require pygo even when Wg signaling is greatly elevated.

Where does Pygo act in the Wg signaling pathway? Our experiments indicate that pygo acts downstream of Axin (Fig. 6), an activated form of Arm (Fig. 1F) and Arm nuclear import (Fig. 7A-I). Consistent with this, a tagged form of Pygo is nuclear (Fig. 7J-L). Taken together these data strongly suggests that Pygo acts in the nucleus, probably at the transcriptional level.

How pygo influences transcription of Wg target genes in the nucleus could occur in several ways. Simple explanations include that pygo could simply be required for the interaction of Arm with TCF, or for TCF to bind to DNA. However, the fact that Arm still accumulates to high levels in the nuclei of Axin, pygo mutant cells (Fig. 7C,F,I) may indicate that the Arm/TCF/DNA complex still forms in the absence of pygo. It has been shown that expression of a dominant-negative version of TCF (which lacks the Arm-binding domain but retains its ability to bind DNA) prevents Arm nuclear accumulation (Tolwinski and Weischaus, 2001). This supports the idea that TCF acts as a nuclear tether for stabilized Arm. Using this line of reasoning, Arm is still found in the nuclei of Axin, pygo mutant cells because it is still bound by TCF, which is still localized properly on the DNA. It should be noted that we do see a subtle reduction in nuclear Arm accumulation in Axin, pygo versus Axin mutant cells (compare Fig. 7C with 7B). However, the small difference suggests that this effect by be indirect.

Another line of evidence suggesting that pygo is not required for TCF to bind to DNA comes from a detailed analysis of the pericardial enhancer of the eve gene. Eve expression in the pericardial cells is absent in wg and pygo mutants (Fig. 4M,N). Mutation of a single high-affinity site in the pericardial enhancer significantly reduces expression (Halfon et al., 2000). However, mutation of all the sites bound by TCF in vitro revealed a depression of the enhancer throughout the dorsal mesoderm, suggesting that in the absence of Wg signaling, Tcf represses the eve pericardial enhancer (Knirr and Frasch, 2001). As no such derepression of Eve expression was observed in pygo mutants (Fig. 4N), this suggests that TCF can bind to the eve enhancer and repress transcription in the absence of pygo.

If Pygo does not promote DNA binding of TCF or formation of the Arm/TCF/DNA complex, what might it be doing? Pygo could help positive factors like Pontin52 (Bauer et al., 2000) to complex with Arm/TCF, or it could normally prevent negative factors like Groucho (Cavallo et al., 1998) or Osa (Collins and Treisman, 2000) from localizing to Wg target genes. In addition, there are a multitude of additional negative regulators of TCF activity identified in vertebrates (see Hecht and Kemler, 2000 for a noncomprehensive list). Pygo could negatively regulate any of these factors.

While the above possibilities must be addressed, the presence of the PHD domain in the Pygo protein suggests another model. PHD domains are often found in chromatin remodeling factors (Aasland et al., 1995). These complexes are thought to alter chromatin structure to allow activation or repression of specific genes (Mahmoudi and Verrijzer, 2001; Urnov and Wolffe, 2001). The PHD domain is a zinc-binding domain that does not bind DNA, and is thought to be involved in protein-protein interactions (Capili et al., 2001; Linder et al., 2000; O’Connell et al., 2001). Therefore, it is possible that pygo is a member of such a chromatin remodeling complex.

The finding that overexpression of full-length Pygo inhibits Wg signaling is consistent with Pygo acting in a multisubunit complex. For example, a heterotrimeric complex consisting of A/Pygo/B could be disrupted by an abundance of Pygo, shifting the equilibrium to A/Pygo and B/Pygo heterodimers. While this is speculation, examples of similar situations are known for histone octomers (Meeks-Wagner and Hartwell, 1986), Apterous/Chip tetramers (Fernandez-Funez et al., 1998) and cytoskeletal complexes (Stokes et al., 2000).

The genetic case for the importance of Pygo in Wg signaling is so convincing because we could directly remove pygo in many tissues and see phenotypes specifically related to Wg signaling. This is in contrast to other Wg transcriptional regulators like Groucho, Osa and Nejire, where complete loss of activity results in pleiotropic phenotypes (Akimaru et al., 1997; Cavallo et al., 1998; Paroush et al., 1994; Treisman et al., 1997; Waltzer and Bienz, 1998). For these factors, and for Pontin52 and Reptin52 (Bauer et al., 2000), the genetic evidence was limited to partial reduction of gene activity, often in a background where Wg signaling was already attenuated. Because of this, it is impossible to explore fully the importance of these genes in the regulation of Wg targets. Even in the case of TCF, its inconvenient location on the fourth chromosome has prevented a detailed genetic analysis (i.e. germline and somatic clonal analysis), though very specific embryonic phenotypes are obtained with zygotic mutants (Brunner et al., 1997; van de Wetering et al., 1997). In this regard, pygo is most similar to arm, whose loss-of-function phenotype has been carefully analyzed in many contexts (Cadigan and Nusse, 1996; Neumann and Cohen, 1997; Peifer et al., 1991; van den Heuvel et al., 1993).

Does pygo have any functions other than regulation of Wg signaling? The phenotypes obtained in clonal analysis in the wing and eye suggests that pygo is highly specific for Wg signaling in these tissues. The fact that clones of pygo in the eye have no detectable defects in morphogenetic furrow progression and photoreceptor recruitment (Fig. 6D,F) or in the morphology of the adult eye (data not shown) is especially impressive. Eye development requires a cadre of transcription factors, including Eyeless, Sine oculis and Eyes absent, that act in concert with Hedgehog, Dpp, Notch and Ras signaling act to specify eye identity in the growing eye-antennal disc (Kumar and Moses, 2001). However, while the data in the embryo suggests that pygo primarily affects Wg signaling (Fig. 4), it is also required for non Wg-dependent processes. For example, when pygo germline clones are zygotically rescued, they have cuticles that appear, at least on a superficial level, to be of the pair-rule class (Fig. 4D). Such phenotypes are not seen in wg, dsh or arm mutant embryos (Peifer et al., 1991; van den Heuvel et al., 1993). Pygo is not a pair-rule mutant in the classical sense, as even in the complete absence of pygo En and Wg stripes are normal at cellular blastoderm. Rather, the decrease in alternative En stripes begins during germband extension (data not shown). pygo maternal/zygotic mutants also have morphological abnormalities not seen in wg mutants, such as incomplete germ-band extension (Fig. 4H,K) and less organized epithelium (data not shown), another indication of a wider role for pygo. Thus, while pygo is highly dedicated to Wg signaling, it clearly has other roles as well.

The authors thank everyone mentioned in Materials and Methods for providing fly stocks and antisera. Special thanks to T. Laverty from the BDGP for coordinating the shipment of EP lines. Thanks also to R. Cohen and A. Singal for help with maintaining the EP lines and performing the screen. S. Klinedinst participated in the initial characterization of EP(3)1076, and R. Angeles and A. Singal contributed to the cell transfection experiments. H. Lin confirmed that Sens expression was inhibited by dominant-negative Tcf. Thanks to W. Lockwood and R. Bodmer for helpful discussions and to R. Nusse for encouragement and inspiration. D. S. P. and J. J. were supported by the Genetics Training Grant of the University of Michigan. This work was supported by a Rackham Grant from the University of Michigan and NIH grant RO1 GM59846 to K. M. C.