Drosophila Hfp negatively regulates dmyc and stg to inhibit cell proliferation

Regulation of cell proliferation is critical for the accurate propagation of genetic material, and development of tissues and organs in a multicellular animal. When these controls fail, excessive cell proliferation can lead to the formation of tumours and developmental abnormalities. Here we show that Half pint (Hfp; pUf68 - FlyBase), the Drosophila orthologue of mammalian FBP interacting repressor (FIR), is required to inhibit cell cycle progression during Drosophila development.

FIR and its Drosophila orthologue Hfp have an evolutionarily conserved function in pre-mRNA splicing. Mammalian FIR was originally isolated as poly(U) binding splicing factor (PUF60), and together with the splicing factors p54 and U2AF, promotes RNA splicing in vitro(Page-McCaw et al., 1999). Furthermore, FIR directly interacts with U2AF65, the large subunit of U2AF(Poleev et al., 2000). FIR and U2AF65 have similar domain structures, including the multiple RNA-recognition motif (RRM) domains. The Drosophila homologue of FIR has a conserved role in regulating pre-mRNA splicing and is known as Half pint (Hfp)(Van Buskirk and Schupbach,2002), dPUF68 (Page-McCaw et al., 1999) or pUbsf(http://flybase.bio.indiana.edu/). Hfp controls RNA splicing of several Drosophila ovarian genes,including ovarian tumor (otu)(Van Buskirk and Schupbach,2002) (reviewed by Rio,2002). The hfp mutant ovary phenotype, which includes defective germline proliferation that results in reduced numbers of germline cells per egg chamber, is rescued by re-expressing an appropriately spliced otu isoform (Van Buskirk and Schupbach, 2002). Therefore Drosophila Hfp, like its mammalian counterpart FIR, has an important role in tissue-specific regulation of alternative splicing.

In addition to regulating splicing of pre-mRNA, mammalian FIR is also an important regulator of Myc gene activity(Eisenman, 2001; Liu et al., 2000). Expression of Myc is tightly regulated, at the level of transcription,translation and protein stability(Eisenman, 2001). One mechanism for control of Myc transcriptional initiation and elongation is mediated by the far upstream element (FUSE), a DNA sequence located 1500 bp upstream of the Myc promoter. The FUSE binding protein (FBP), a KH domain transcriptional activator, binds the FUSE and is absolutely required for Myc expression and cell growth in mammalian cells (Duncan et al., 1994; He et al., 2000). The FBP interacting repressor (FIR) counteracts FBP function by forming a ternary complex with FBP and the FUSE to repress Myc transcription(Liu et al., 2000). The N-terminal repression domain of FIR interacts with the basal transcription component TFIIH and interferes with promoter clearance. The in vivo importance of this mechanism is not clear; however, mutations in ERCC2 or ERCC3 (which encode TFIIH subunits corresponding to the xeroderma pigmentosum (XP) complementation groups XPD and XPB, respectively) impair regulation of Myc expression by FBP and FIR. This may contribute to cancer risk in individuals with XP mutations (Liu et al.,2001).

The proteins encoded by the myc family of proto-oncogenes are important regulators of cell growth (size and mass increase), proliferation,differentiation and apoptosis (Eisenman,2001). In response to mitogenic signalling, Myc can inhibit differentiation and either promotes cell growth and proliferation or apoptosis, depending on the context. Myc proteins form stable heterodimers with Max proteins to modulate expression of target genes by binding E box DNA sequences. Although primarily a transcriptional activator, Myc can also inhibit the expression of certain target genes. Deregulated Myc expression is potently oncogenic and is one of the most frequently observed molecular abnormalities in human cancers. Despite this, regulation of Myc expression and its role in tumourigenesis has not been clearly defined.

The Drosophila dmyc (dm - FlyBase) and dmax(Max - FlyBase) gene products also form heterodimers, bind E-box DNA sequences and activate transcription(Eisenman, 2001; Gallant et al., 1996). Gain and loss of function studies in Drosophila have revealed that the primary in vivo function of dMyc is to stimulate cell growth. dmycmutations cause cellular growth retardation; resulting in small flies with small cells (Gallant et al.,1996; Johnston and Edgar,1998; Johnston et al.,1999; Schreiber-Agus et al.,1997). Conversely, overexpression of dmyc in the wing imaginal disc promotes cell growth, leading to increased cell size(Johnston et al., 1999). Although dMyc-induced cell growth is accompanied by faster G1/S phase progression, the overall cell division rate of dmyc overexpressing cells remains normal due to an extended G2 phase(Johnston et al., 1999), which arises because the Drosophila homologue of Cdc25 phosphatase, String(Stg), is rate limiting for G2-M cell cycle progression(Edgar and O'Farrell, 1989; Edgar and O'Farrell, 1990). Stg triggers mitotic entry by dephosphorylating, and thereby activating the Cdk1/Cyclin B kinase (Edgar et al.,1994).

The Wingless-signalling pathway regulates both dmyc and stg expression during Drosophila wing development. During third instar larval development, the dorsoventral compartment boundary of the wing imaginal disc forms a zone of cells arrested in G1 or G2, termed the ZNC. A band of Wingless (Wg) expression controls cell cycle arrest within the ZNC(Johnston and Edgar, 1998). While dmyc is expressed in proliferating zones within the wing,expression is normally low in the ZNC, and ectopic expression of dmycin the ZNC prevents cell cycle exit(Johnston et al., 1999). Furthermore, inhibition of Wg signalling in the ZNC, via expression of dominant negative TCF, results in ectopic dmyc expression in the ZNC(Johnston et al., 1999). These studies show that Wg signalling represses dmyc expression within the ZNC; however, whether the repression of dmyc transcription by TCF is direct or indirect is unknown. While the posterior region of the ZNC is comprised solely of G1-arrested cells, the anterior compartment of the ZNC contains a band of G1-arrested cells at the dorsoventral boundary that is sandwiched between anterior-dorsal and anterior-ventral G2-arrested domains. Wg signalling is required for the downregulation of stg and associated G2-arrest in the anterior of the ZNC. This occurs indirectly, via Wg upregulating achaete and scute, which in turn downregulate stg, resulting in the G2-arrested cells in the ZNC(Johnston and Edgar, 1998). Whether Achaete and Scute act directly on the stg promoter is unknown.

Here we describe an alternative role for Hfp, as a negative regulator of cell cycle progression in Drosophila imaginal tissues. We show that within the ZNC of hypomorphic hfp mutant wing discs, cells undergo ectopic S phases, suggesting that, like FIR, Hfp might control cell proliferation by regulating dmyc expression. Indeed, elevated dmyc expression was detected in hfp mutant clones, and reducing the dosage of hfp rescued the dmyc mutant ovary phenotype. Unlike dmyc overexpression, hfp mutants did not affect cell growth, although cell proliferation was increased. This can be explained via an affect of Hfp on the G2-M phase transition, since hfp mutants can rescue the cycle 14 G2-arrest phenotype of an stg mutant. Furthermore, Hfp protein was elevated in response to Wg pathway signalling. Taken together, these results are consistent with Hfp playing an important role in cell cycle arrest downstream of Wg signalling. These findings suggest that Hfp links patterning signals to cell growth and proliferation in the Drosophila wing.

Since the hfp mutant strain EP3058 contained additional lethal mutations (Van Buskirk and Schupbach, 2002), recombination was used to isolate the hfp^EP allele. The purified hfp^EPfailed to complement deficiency Df(3L)Ar14-8, which has breakpoints 61C4-62A8 covering hfp (Van Buskirk and Schupbach, 2002). To generate the UAS-hfpconstruct, full-length hfp cDNA was subcloned into pUAST and transgenic flies were generated as previously described(Richardson et al., 1995). UAS-hfp transgenes on the second and third chromosomes were used for all experiments. Recombinants of GMRGAL4 and UAS-hfp on the second chromosome were used to test for genetic interactions at 25°C. Recombinants of hfp^EP and stg^AR2 were generated and balanced over TM6B,abdA-lacZ, and double-mutant embryos were selected based on the absence of AbdA-LacZ staining. All general fly stocks were obtained from the Bloomington Stock Centre, except the UAS-TCF^DN and en-GAL4,UAS-GFP (from Laura Johnston), axin, FRT82B (from Jessica Treisman), C96-GAL4 (from Bruce Edgar) and GMR-p21(from Iswar Hariharan). For analysis of clones hs-FLP; FRT80B,Tb-GFP females were crossed to FRT80B,hfp^EP/TM6B males, clones were generated by heat shocking (at 37°C for 1 hour) second instar larvae and wandering third instar larvae were dissected and analysed. Similarly, axin clones were generated by crossing hs-FLP; FRT82B, Ub-GFP females to FRT82B, axin/TM6B males.

mRNA in-situ hybridization was carried out as described in previous methods(Dorstyn et al., 1999) except the signal was detected using fast-red substrate (Roche). Following in-situ hybridization, clones were distinguished using a rabbit anti-GFP polyclonal antibody (Molecular Probes), detected using an anti-rabbit-biotin conjugated secondary antibody, followed by streptavidin-Alexa488 (Molecular Probes). After in-situ hybridization of ovaries, DNA staining was carried out with Oligreen (Molecular Probes) to assist with staging.

Immunohistochemistry, including TUNEL and BrdU labelling of Drosophila larval tissues and embryos, was carried out as previously described unless otherwise indicated(Quinn et al., 2000; Quinn et al., 2001). The monoclonal Hfp antibody (Trudi Schupbach) was detected using an anti-mouse-biotin conjugated secondary antibody followed by streptavidin-lissamine rhodamine (Jackson). TUNEL staining was carried out using the in-situ cell death detection kit TRred (Roche). Other antibodies used were anti-BrdU monoclonal antibody (Becton Dickinson) and rabbit anti-phosphohistone H3 (Santa Cruz), rabbit anti-GFP (Molecular Probes),rabbit anti-Cyclin B (David Glover), rat anti-Geminin(Quinn et al., 2001), rabbit anti-βgal (Rockland) and rabbit anti-Stg (Bruce Edgar). Ovaries were stained with phalloidin-rhodamine, 0.1% in PBT for 1 hour (Sigma), prior to staining with Oligreen (Molecular Probes). All fluorescently labelled samples were analysed by confocal microscopy (Biorad MRC1000). Scanning electron micrographs of adult eyes were generated as previously described using a Field Emission Scanning Electron Microscope(Secombe et al., 1998).

The Drosophila stock EP(3)3058(hfp^EP) harbours a recessive lethal P element insertion in the 5′ UTR of hfp, 94 bp upstream of the initiating methionine codon (Van Buskirk and Schupbach, 2002). Homozygous hfp^EP larvae were of similar size to age-matched wild type third instar larvae. However,the pupariation of hfp^EP larvae was consistently delayed by approximately 2 days, and continued growth during this period resulted in wandering larvae and pupae ∼20% larger than wild-type third instar larvae(Fig. 1A,C). The duration of the pupal stage was normal for hfp^EP mutant animals;however, they failed to eclose and died as pharate adults that were larger than wild type (Fig. 1B). The hfp^EP/hfp^EP terminal phenotype included duplication of superior scutellar macrochaete, and malformation of legs, wings and sex combs (data not shown).

The pleiotropic phenotype of hfp mutant animals indicated that Hfp might be involved in several stages of development. In Drosophila,maternal transcripts are transferred during oogenesis and serve to sustain early embryonic development until stage 5, after which zygotic transcription commences. Northern analysis revealed that hfp mRNA was maternally deposited in the early embryo; however, zygotic hfp expression was low during late embryonic and early larval stages(Fig. 1D). hfptranscripts were also detected in third instar larvae, pupae and adults. We observed a marked decrease in hfp mRNA in hfp^EP/hfp^EP and hfp^EP/Df(3L)Ar14-8 larvae compared with age-matched wild-type third instar larvae(Fig. 1E). However, hfp transcript was still detectable, consistent with the notion that hfp^EP is not a null allele(Van Buskirk and Schupbach,2002). In wild-type animals, expression of hfp during third instar (Fig. 1D)coincides with the onset of differentiation in imaginal discs. We examined Hfp protein expression in wing discs using an antibody recognizing Hfp(Van Buskirk and Schupbach,2002) and used an antibody to Geminin, which is abundant in late S phase and G2 but absent in G1 cells (Quinn et al., 2001), to visualize the ZNC(Fig. 1F; see Introduction). Hfp protein was detected in the nucleus of most wing disc cells, with higher staining in cells in the ZNC (Fig. 1G). Consistent with northern analysis, Hfp protein level was significantly reduced in wing discs from hfp^EP/hfp^EP larvae(Fig. 1H).

In order to investigate whether Hfp regulates cell proliferation during Drosophila development, we measured BrdU incorporation in wing discs from wandering hfp^EP/hfp^EP larvae. In wild-type wing discs the ZNC is clearly marked by the absence of BrdU labelling (Fig. 1I). The number of S-phase cells was markedly increased in hfp^EP mutant wing discs, BrdU incorporation was uniform across the disc and cell cycle arrest was not evident in the ZNC region(Fig. 1L). Strikingly,anti-phosphohistone H3 antibody staining of mitotic cells(Hans and Dimitrov, 2001), was also elevated, indicating an overall increase in cell proliferation in hfp wing discs (Fig. 1M; 127±7 mitotic cells per disc) compared with wild type(Fig. 1J; 75±8 mitotic cells; n=5 discs, P<0.01).

The developing eye is a sensitive system for analysis of cell proliferation. During wild-type eye development, a wave of differentiation moves from posterior to anterior across the third instar eye imaginal disc. Within the morphogenetic furrow (MF) cells are arrested in G1 and posterior to the MF a subset of cells enter a synchronous S phase(Fig. 1O) while other cells begin differentiation to form ommatidial pre-clusters, followed by a band of mitotic cells known as the second mitotic wave(Fig. 1P). Analysis of the band of S phases posterior of the MF (Fig. 1R) and the second mitotic wave(Fig. 1S) in hfpmutant eye discs revealed that the S-phase band is generally broader than for wild type, but the second mitotic wave does not occur prematurely, suggesting that Hfp might normally be required for the pre-cluster cells to cease division.

Despite increased proliferation in hfp^EP mutant discs,they were not overgrown compared with wild type (data not shown). We did not observe an obvious difference in cell size between hfp^EPmutant wing disc cells by either cross section(Fig. 1V compared with wild type, Fig. 1U) or by transverse section (data not shown), suggesting that increased cell death may accompany increased proliferation in this tissue to account for the fact that the discs are similar in size to wild type. Indeed, TUNEL staining revealed an increase in the number of apoptotic cells in the wing imaginal discs of hfpmutants (Fig. 1X; 143±17 apoptotic cells per disc) compared with wild-type larvae(Fig. 1W; 26±9 apoptotic cells, n=5 discs, P<0.005).

Therefore, although increased cell cycles were observed in hfpmutant wing discs, the overall disc size was similar to wild type, as ectopic proliferation was apparently balanced by increased apoptosis. The elevated cell death observed in hfp mutant wing discs is likely to be a secondary consequence of deregulated cell proliferation. In Drosophila, compensatory cell death in the face of hyperproliferation appears to be a general mechanism for maintaining normal compartment size and is also observed in imaginal discs upon ectopic expression of dmyc(Johnston et al., 1999), the cell cycle transcription factor E2F (Asano et al., 1996) or both the G1-S phase regulator Cyclin E and the G2-M phase regulator Cdc25/Stg (Neufeld et al., 1998).

The observation that loss of Hfp promotes cell cycle entry prompted us to examine whether overexpression of Hfp could block cell proliferation. We generated transgenic flies containing a UAS-hfp transgene in order to ectopically express Hfp using various GAL4 drivers (Brand and Perrimon, 1993). Ubiquitous expression of Hfp using armadillo-GAL4(arm-GAL4) partially rescued the pupal lethality of hfp^EP/hfp^EP animals, verifying transgene function (data not shown). We specifically overexpressed hfp in cells posterior to the MF in the eye disc using the GMR-GAL4 driver. Expression of two copies of UAS-hfp under control of GMR-GAL4 (GMR-GAL4,UAS-hfp/+; UAS-hfp/+)resulted in flies with disorganized adult eyes that were slightly smaller than wild type (Fig. 2B). Third instar eye discs from GMR-GAL4,UAS-hfp/+; UAS-hfp/+ larvae showed reduced BrdU incorporation in the S-phase band posterior to the MF(Fig. 2D) compared with wild type (Fig. 2C). In addition,reduced numbers of cells staining with anti-phosphohistone H3 were observed posterior to the S-phase band of GMR-GAL4,UAS-hfp/+; UAS-hfp/+ eye discs (Fig. 2F) compared with wild type(Fig. 2E). Thus overexpression of hfp can inhibit S phase entry and mitoses posterior to the MF,consistent with a role for Hfp in negatively regulating cell cycle progression.

We then examined the effect of overexpressing hfp in the wing disc by using engrailed-GAL4 (en-GAL4), which drives transgene expression in the posterior compartment of the wing disc(Kornberg et al., 1985). Defects were observed in the posterior wing compartment in en-GAL4;UAS-hfp/+ adults. The phenotype varied in severity from slight wing vein abnormalities and decreased wing size (Fig. 2H and 2I) to disrupted, small and blistered wings (Fig. 2J). To analyse wing discs, the posterior compartment of third instar larval wing discs was marked by co-expression of a UAS-GFPtransgene with the en-GAL4 driver. The posterior wing compartment overexpressing hfp was small compared with the wild type posterior wing compartment (compare GFP in Fig. 2L with wild type in Fig. 2K). As the phenotype resulting from overexpression of the hfp transgene with the en-GAL4 driver was slightly variable,presumably as a consequence of subtle variations in the level of transgene expression, we evaluated the reduction in compartment size by comparing the area of the posterior compartment of en-GAL4,UAS-GFP/+; UAS-hfp/+wing discs with the same from control en-GAL4,UAS-GFP. As the area corresponding to the posterior compartment of the wing is marked by GFP in each case, we used this to determine that Hfp overexpressing tissue was reduced by 31.7% (mean of the average number of pixels=66,931±19247; n=5) compared with the control (mean of the average number of pixels=96,616±7566; n=5). The region surrounding the ZNC of the wing disc is normally highly proliferative(Fig. 2M); however,overexpression of Hfp in the posterior compartment of the wing resulted in fewer BrdU-labelling cells (Fig. 2N). Thus overexpression of hfp in either eye or wing imaginal discs results in cell cycle inhibition and is associated with reduced overall size of these tissues. Taken together with the loss-of-function studies (above) these findings suggest that Hfp normally functions to inhibit cell cycle progression.

Overexpression of human p21 (an inhibitor of G1-S cyclin-dependent kinases)posterior to the MF, under the control of the GMR promoter, inhibits S-phase entry posterior to the MF and results in a rough eye phenotype in adults (de Nooij and Hariharan,1995). The GMR-p21 rough eye phenotype can be modified by reducing the dose of cell cycle regulators(Secombe et al., 1998),providing a sensitive system to investigate the role of putative cell cycle regulators. When the dosage of hfp was reduced in a GMR-p21background, the rough adult eye phenotype was dominantly suppressed; GMR-p21/+, hfp^EP/+ eyes were larger and contained fewer fused ommatidia (Fig. 3C)than GMR-p21/+, +/+ eyes (Fig. 3B). Similarly, we found that the mild rough eye phenotype caused by overexpression of the Drosophila p21/p27 homologue dacapo(dap) was dominantly suppressed by mutation in hfp (data not shown).

In third instar larval eye discs, GMR-p21 abolishes the band of S phases posterior to the MF and the second mitotic wave(Fig. 3E,H,K). Compared with GMR-p21/+, eye discs from GMR-p21/+, hfp^EP/+ contained more S-phase cells posterior to the MF(Fig. 3F compared with 3E), and an accompanying increase in mitotic cells (Fig. 3I compared with 3H). Thus, the dominant suppression of the GMR-p21 rough eye phenotype by hfp^EP can be explained by this partial rescue of cell proliferation in the eye imaginal disc. These results, together with the analysis of hfp^EP/hfp^EP wing discs,suggest that Hfp inhibits cell cycle entry in larval imaginal discs.

Given that mammalian FIR protein negatively regulates the cell cycle via Myc, and the above data showing that Hfp might normally inhibit cell cycle progression through p21/Dacapo, we tested Drosophila Myc (dMyc)for genetic interactions with p21/Dacapo. The dMyc mutant enhances the GMR-p21 phenotype (Fig. 3N females of genotype dmyc^P0/+;GMR-p21/+ and Fig. 3O males dmyc^P0/Y; GMR-p21/+ compared with the control female GMR-p21/+, in Fig. 3M). Conversely, the GMR-GAL4, UAS-dacapo reduced/rough eye phenotype is suppressed by co-expression of a UAS-dmyc transgene(Fig. 3R compared with Fig. 3P). The finding that the inhibitory affect of p21/Dacapo on the G1 to S transition can be suppressed by either reducing the dose of hfp or by overexpressing dmyc,suggests that Hfp and dMyc may have antagonistic effects on the G1 to S transition in the eye imaginal disc.

Given that mammalian FIR protein is a negative regulator of Myc,and the above data showing that Hfp inhibits cell cycle progression, we investigated the possibility that Hfp regulates dmyc in Drosophila. If the role of Hfp as a negative regulator of dmyc has been conserved, we hypothesized that reducing the dose of hfp might suppress the dmyc mutant phenotype. The three characterized hypomorphic dmyc alleles, diminutive¹ (dmyc^dm1)(Gallant et al., 1996), dmyc^P0, and dmyc^P1, are all recessive female sterile (unpublished data). Analysis of ovaries from dmyc^P0 and dmyc^P1 females revealed that early stage (stage 2-9) egg chambers were of normal appearance but then arrested between stages 10-11 of oogenesis with smaller ovarioles(Fig. 4B,D).

Progression beyond stage 10 of oogenesis requires dumping of the nurse cell cytoplasm into the oocyte, which is followed by nurse cell apoptosis(Buszczak and Cooley, 2000). An initial step of dumping is formation of dense bundles of actin filaments in the nurse cell cytoplasm, essential for structural support of nurse cell nuclei (Gutzeit, 1986). Although actin filaments were present in stage 10 dmyc^P0/dmyc^P0 egg chambers(Fig. 4B compared with Fig. 4A), cytoplasmic actin bundles failed to develop around stage 11 nurse cell nuclei(Fig. 4D compared with Fig. 4C). Thus dmyc^P0/dmyc^P0 ovaries fail to undergo nurse cell death, which is required for progression to stage 12 of oogenesis(Fig. 4G compared with Fig. 4F, and measured by TUNEL,data not shown). Strikingly, the hfp^EP mutation dominantly suppressed these defects in dmyc^P0/dmyc^P0 ovaries(Fig. 4E,H). The actin network appeared normal in nurse cells from stage 10 dmyc^P0/dmyc^P0; hfp^EP/+ ovaries(Fig. 4E), and TUNEL positive nuclei were obvious at stage 12 (data not shown). Indeed dmyc^P0/dmyc^P0; hfp^EP/+ and dmyc^P1/dmyc^P1; hfp^EP/+ females yielded mature oocytes(Fig. 4H) that gave rise to viable embryos (data not shown).

DNA endoreplication in nurse cells and follicle cells also occurs during stage 10 of oogenesis. Follicle cells undergo genomic endoreplication until stage 10A and switch to amplification of specific loci, including the chorion genes at stage 10B (Calvi et al., 1998; Edgar and Orr-Weaver, 2001). Reduced chorion gene amplification was observed in dmyc^P1/dmyc^P1 follicle cells (Fig. 4J,M) compared with the wild-type control (Fig. 4I,L). Consistent with results above, reducing the dosage of hfp restored chorion gene amplification to normal levels in dmyc^P1/dmyc^P1 ovaries(Fig. 4K,N).

To test whether increased dmyc mRNA was associated with reduced hfp gene dosage in ovaries, in-situ hybridization analysis was performed. In wild-type egg chambers, abundant dmyc expression was observed in nurse cells and follicle cells(Fig. 4O,R), consistent with previous findings (Gallant et al.,1996). As expected, dmyc mRNA abundance was reduced in dmyc^P0/dmyc^P0 nurse cells compared with wild type, and was almost absent in follicle cells and the oocyte(Fig. 4P,S). Increased dmyc expression was observed in follicle cells from dmyc^P0/dmyc^P0; hfp^EP/+ egg chambers compared with those from dmyc^P0/dmyc^P0 flies(Fig. 4Q,T). The relative increase in dmyc mRNA in nurse cells of dmyc^P0/dmyc^P0; hfp^EP/+ ovaries is less striking and is likely to be a consequence of cytoplasmic dumping, which is impaired in dmyc^P0/dmyc^P0 but occurs in dmyc^P0/dmyc^P0; hfp^EP/+ ovaries (see above). These data suggest that, like mammalian FIR, Hfp functions as a negative regulator of dmyc.

To further investigate regulation of dmyc expression by Hfp, we generated clones of homozygous hfp mutant tissue in wing imaginal tissues using FLP/FRT-induced mitotic recombination of the hfp^EP allele (Xu and Rubin, 1993). Analysis of hfp mutant clones revealed reduced levels of staining with the anti-Hfp antibody in third instar eye discs, compared with surrounding non-clonal, GFP-positive tissue (data not shown). Previous mRNA analysis has shown that dmyc is expressed in proliferating regions in the wing disc, with lower expression in the non-proliferating ZNC (Johnston et al.,1999). Analysis of mosaic wing discs revealed elevated dmyc mRNA expression specifically in hfp^EP mutant clones, including those spanning the ZNC, compared with surrounding hfp^EP/+ cells and wild type clones(Fig. 4V-Z). Increased levels of dmyc transcript were also observed in hfp mutant clones in the eye disc (data not shown); therefore, Hfp acts to repress dmyctranscript accumulation in Drosophila imaginal tissues.

The evidence above suggests that Hfp negatively regulates accumulation of dmyc transcript; however, the finding that reducing the dose of dmyc does not rescue the hfp hypomorphic phenotype (data not shown), suggests that the pupal lethality associated with the hfpmutant is not simply a consequence of increased levels of dmyc. Therefore, if the hfp mutant lethality is not exclusively due to increased dmyc expression, Hfp may regulate other essential genes.

Examination of genetic interactions between dMyc and Hfp in the eye also suggested a second role for Hfp. dmyc overexpression in wing discs results in larger cells due to increased growth, an accelerated G1 phase and a compensatory extension of G2 phase due to the fact that Cdc25c/Stg, the rate limiting factor for G2-M progression, is not upregulated by dMyc(Johnston et al., 1999). Similarly, overexpression of dmyc using the eye driver GMR-GAL4 results in larger cells posterior to the MF in third instar larvae (Fig. 5G,H compared with wild type, Fig. 5A,B), larger adult ommatidia (Fig. 5Icompared with wild type, Fig. 5C) and an oversized adult eye(Fig. 5J compared with wild type, 5D). Reducing the level of hfp in this genetic background results in a further increase in the overall size of the dmyc overexpressing adult eye(Fig. 5P compared with Fig. 5J) with more disorganized, slightly larger ommatidia(Fig. 5M,N,O compared with Fig. 5G,H,I).

To determine whether cell cycle progression was also affected, we analysed S phase and mitosis in third instar larval eye discs. Assuming that overexpression of dmyc affects cell cycle progression in the eye in a similar manner to that in the wing, it would be expected that eye cells overexpressing dmyc would spend more time in G2 and relatively less time in S phase; thus, a thinner S-phase band would result. Indeed, ectopic dmyc expression in the posterior part of the eye via GMR-GAL4 resulted in an S-phase band posterior to the MF that was slightly thinner than wild type (Fig. 5K compared with Fig. 5E). BrdU labelling represents a snap-shot of S phases, and therefore a thinner BrdU band suggests that fewer cells are in S phase at a particular time compared with wild type, consistent with G1-S progression and S phase being accelerated and an extended G2 phase.

Indeed, as expected in the event of a G2 delay, the band of mitotic cells was reduced in GMR-GAL4, UAS-dmyc/+ eye discs(Fig. 5L compared with wild type, Fig. 5F). Reducing the dose of hfp increased both the number of S-phase cells(Fig. 5Q) and restored M-phase entry (Fig. 5R). The increased mitotic cells observed upon reducing the dose of hfp suggests that more of the dmyc overexpressing G2-delayed cells progress into mitosis. This cannot be explained by the effect of increased dmyclevels when hfp is reduced and suggests that Hfp may normally negatively regulate a cell cycle component that is required for promotion of G2-M progression.

The increased number of S-phase cells observed upon halving the dose of hfp may be a consequence of passage of G2-delayed cells through mitosis into another S phase. To examine the possibility that Hfp might regulate G2-M progression via an inhibitory affect on Stg (the rate limiting regulator of G2-M), we generated hfp^EP, stg^AR2 double mutants and analysed mitoses in mutant embryos using anti-phosphohistone H3 (PH3) staining(Fig. 6A-H). Analysis of hfp^EP mutant embryos revealed an apparently normal pattern of PH3 staining in cycle 16 mitotic domains when compared with wild type(Fig. 6C compared with 6A). Closer inspection of the mitotic figures from hfp^EP mutant embryos revealed abnormal chromosome morphology; including many lagging chromosomes that are often mis-segregated due to closure of the contractile ring prior to sister chromatid separation (Fig. 6D). Maternal Stg enables mitoses prior to embryonic cycle 14; however, after interphase 14 zygotic transcription of stg is required for G2-M progression, and as a consequence stg mutants arrest in G2 of cycle 14 (Edgar and O'Farrell,1990). As expected, cycle 14 stg^AR2 mutant embryos lacked PH3 staining (Fig. 6E), and were comprised solely of large G2 cells(Fig. 6F). Strikingly, mitotic entry was restored in hfp^EP, stg^AR2double mutant embryos (Fig. 6G), and consequently cell size was restored to the wild type range (Fig. 6H). Furthermore,in contrast to the complete embryonic lethality of stg mutant embryos, hfp^EP, stg^AR2 double mutants survive embryogenesis and die between first and second instar. Thus, in addition to negatively regulating dmyc and G1-S progression, these results suggests that Hfp normally acts to negatively regulate mitotic entry via negative regulation of stg.

The stg mutant used in the above experiment is a null, which suggests that Hfp affects accumulation or stability of the maternally supplied stg transcript or Stg protein, which are both normally actively degraded prior to cycle 14 of embryogenesis. In-situ hybridization to hfp mutant wing clones, using a DIG labelled stg probe(Fig. 6I-K), revealed no difference between levels of stg mRNA in clonal tissue. This suggests that the affect on stg is not via Hfp stabilizing stg mRNA. However, using an Stg antibody, we found increased levels of Stg protein in hfp mutant clones (Fig. 6L-N), suggesting that Hfp might normally regulate factors required for Stg translation or protein degradation.

The cell cycle arrest and repression of dmyc normally observed in the wing disc ZNC requires Wg expression and a functional Wg pathway(Johnston and Edgar, 1998). Hence, expression of a dominant negative form of TCF (TCF^DN) in cells of the ZNC causes ectopic induction of dmyc and cell cycle entry (Johnston et al., 1999). As both dmyc expression and cell proliferation in the wing disc appear to be inhibited by Hfp, we hypothesized that Hfp expression may be under the control of the Wg pathway. To test this, we activated the Wg pathway in the posterior compartment of the wing disc by expressing a dominant negative form of Shaggy, Sgg^DN, using the en-GAL4 driver. Shaggy is the Drosophila orthologue of vertebrate glycogen synthase kinase 3 (GSK3), an inhibitory component of the Wg signalling pathway(Siegfried et al., 1992). Therefore, expression of the dominant negative transgene (Sgg^DN)results in ectopic activation of the Wg signalling pathway. As expected,control en-GAL4,UAS-GFP larval wing discs showed GFP expression restricted to the posterior region of the disc and ubiquitous staining for Hfp(Fig. 7A-C). Increased staining with the anti-Hfp antibody was observed in the posterior region of en-GAL4,UAS-hfp wing discs as expected(Fig. 7D-F). Significantly,activation of the Wg pathway by Sgg^DN resulted in similar high levels of ectopic Hfp expression in the posterior region of the wing disc(Fig. 7G-I). To further confirm that Hfp is upregulated by Wg signalling, we analysed axin mutant clones, in which the Wg pathway is constitutively active, since Axin normally downregulates Armadillo (Hamada et al.,1999). Indeed, increased Hfp protein was observed in d-axin mutant clones, marked by the absence of GFP(Fig. 7J-L). Conversely, when Wg signalling was blocked by expressing a dominant negative form of TCF(TCF^DN) in the ZNC using the C96-GAL4 driver(Johnston et al., 1999) Hfp protein was reduced in all TCF^DN expressing cells, which are marked by coexpression of GFP (Fig. 7N,O, compared with the normal high level of Hfp in ZNC cells from wild type, Fig. 7M). Reduced numbers of ZNC cells (i.e. fewer cells stained for GFP) are observed as a consequence of TCF^DN overexpression, since cells die by apoptosis when Wg signalling is blocked (Johnston and Sanders, 2003). Therefore, ectopic activation of the Wg pathway is associated with increased levels of Hfp in the wing disc, and blocking Wg signalling reduces Hfp expression. Taken together, these results show that ectopic activation of the Wg pathway increases the level of Hfp in third instar wing discs, consistent with the notion that Wg may normally act by inducing hfp to inhibit dmyc expression in the ZNC.

In this study we have shown that Hfp is a negative regulator of cell cycle entry in Drosophila as evidenced by: (1) ectopic S phases in the ZNC of hfp mutant wing discs and increased S phase in the second mitotic wave in the eye disc; (2) inhibition of S phases in larval imaginal tissues by overexpression of the UAS-hfp transgene; and (3) dominant suppression of the GMR- driven human p21 or dacapo rough eye phenotypes and rescue of the posterior band of S phases in GMR-p21eye discs by reducing the level of hfp. These data suggest that Hfp normally has a role in preventing S-phase entry in cells destined to differentiate in the eye and wing imaginal discs. Furthermore, we show that this negative regulation of the cell cycle by Hfp is partly a consequence of inhibitory affects on dmyc, since: (1) an increased level of dmyc mRNA transcript occurred in hfp-/- clones; and (2)reduced levels of Hfp could rescue the dmyc mutant ovary phenotype,by restoring levels of dmyc mRNA to more wild-type levels. Indeed,upregulation of dmyc expression in Hfp mutants may explain the rescue of S phases in eye discs overexpressing p21 or Dacapo, consistent with the observation that dmyc mutants dominantly enhance the GMR-p21and GMR-driven dacapo rough eye phenotypes(Fig. 3). Mammalian Myc stimulates cyclin E expression, activation of Cdks(Bouchard et al., 1999),antagonizes the action of Cdk inhibitors, including p27(Vlach et al., 1996; O'Hagan et al., 2000), and can downregulate p21 transcription(Claassen and Hann, 2000; Gartel et al., 2001) and p21 activity via direct c-Myc-p21 protein-protein interaction(Kitaura et al., 2000). In Drosophila, dMyc has been shown to lead to an increase in Cyclin E protein levels by a post-transcriptional mechanism(Prober and Edgar, 2002),which by itself could explain the suppression of the GMR-p21 eye phenotype by reducing the dose of hfp. Whether dMyc can also inhibit p21 or Dacapo activity in Drosophila is unknown.

Increased levels of dmyc transcript were observed in hfpmutant clones, consistent with Hfp acting to repress dmyc transcript accumulation in Drosophila imaginal tissues. The upregulation of dmyc mRNA in hfp mutant tissue could occur through alterations in dmyc transcription (initiation or elongation),pre-mRNA splicing, mRNA message stability or a combination of these processes. Mammalian FIR was first shown to regulate pre-mRNA splicing by binding to RNA polypyrimidine tracts and cooperating with the essential splicing factor U2AF(Page-McCaw et al., 1999). Consistent with this, recent studies in Drosophila show that the FIR orthologue Hfp is required for correct splicing of several genes in the developing ovary (Van Buskirk and Schupbach, 2002). Mammalian FIR has been shown to have a second role as transcriptional repressor of Myc, through first forming a complex with the Myc activator FBP and interfering with the basal transcription apparatus by then binding TFIIH, thereby disrupting helicase function (Liu et al., 2000). The data described here suggest that the cell cycle inhibitory function of Hfp is partly a consequence of negatively regulating dmyc expression. Therefore, the dual roles of transcription regulation and mRNA splicing appear to have been evolutionarily conserved between Drosophila Hfp and mammalian FIR. It remains to be determined whether Hfp inhibits dmycexpression by a mechanism analogous to the mammalian FIR/FBP/FUSE interaction. A FUSE element has not been identified upstream of the dmyc promoter,and although the Drosophila splicing factor PSI is a highly conserved orthologue of FBP (Labourier et al.,2002), it has not been reported whether PSI can activate dmyc expression.

Our finding that hfp mutants do not phenocopy dmycoverexpression suggested that inhibition of dmyc expression is not the only role of Hfp. Although increased S phases are observed in hfpmutant wing discs, this is not associated with increased cell size, as occurs with dmyc overexpression in the wing disc. Rather, in hfpmutant wing discs the ZNC, which normally contains domains of G1- and G2-arrested cells (Johnston and Edgar,1998), has ectopic S-phase and M-phase cells. Since cells in hfp mutant wing discs are of normal size and ectopically enter S phase, it is possible that progression through G2 may also be accelerated. Indeed, the increased number of mitotic cells observed in eye imaginal discs when the level of Hfp is reduced in a dmyc overexpression background,suggests that Hfp normally negatively regulates G2-M phase progression. Furthermore, the abnormal mitotic figures observed in hfp^EP mutant embryos are consistent with accelerated cell cycle progression (Quinn et al.,2001). Most importantly, the hfp mutant rescued the cycle 14 G2-arrest that normally occurs in stg mutant embryos, and hfp mutant clones have increased levels of Stg protein, suggesting that Hfp normally exerts an inhibitory affect on G2-M progression via negatively regulating Stg translation or protein stability. Thus, Hfp may be required for negatively regulating both the G1-S phase transition by downregulating dmyc and the G2-M transition by negatively regulating stg.

The Wg pathway is required to downregulate both dmyc and stg expression in order to limit cell proliferation in the ZNC during wing development (Johnston and Edgar,1998; Johnston et al.,1999). Activation of the Wg pathway, using either dominant negative Shaggy or by generation of axin clones, resulted in a strong and specific increase in Hfp protein, demonstrating that Wg pathway activation is sufficient to cause Hfp induction. Our findings supported a model in which Wg signalling causes induction of Hfp in the wing disc ZNC, which in turn inhibits dmyc expression (to elicit the posterior, G1 arrest) and stg expression or activity (to provide the anterior, G2-arrested domains) (Fig. 8). The involvement of Achaete and Scute in this process, which have been previously shown to play a role in the negative regulation of stg(Johnston and Edgar, 1998)remains to be elucidated. Previous studies have shown that Ras signalling through Raf/MAPK upregulates dmyc post-transcriptionally in wing disc cells and is required to maintain normal dMyc protein levels in the wing disc(Prober and Edgar, 2000; Prober and Edgar, 2002). In contrast, since hfp clones have increased dmyc mRNA, Hfp must normally inhibit dmyc mRNA accumulation. Furthermore,overexpression of Hfp inhibits cell proliferation in all wing and eye imaginal discs, suggesting that Hfp may normally override mitogenic signals and lead to cell cycle arrest during particular stages of development.

In summary, our results suggested that Hfp negatively regulates cell proliferation by inhibiting dmyc transcription and Stg protein accumulation. Hfp is required for the developmentally regulated cell cycle arrest within the ZNC and is responsive to the Wg signalling pathway that regulates this arrest, suggesting that Hfp links patterning signals to cell proliferation during Drosophila development.

We thank Trudi Schupbach for anti-Hfp antibody, Laura Johnston for the UAS-TCF^DN and en-GAL4,UAS-GFP strains and for critical reading of the manuscript, Iswar Hariharan for the GMR-p21strain, Jessica Treisman for the axin, FRT82B stock and Bruce Edgar for the anti-Stg antibody, stg cDNA and C96-GAL4 fly strain. We thank Grant McArthur, David Levens, and Izhak Haviv for critical reading of the manuscript, and members of the Bowtell and Richardson labs for helpful discussions. G.R.H., D.D.L.B. and H.R. are supported by grants from the Australian National Health and Medical Research Council. H.R. is a Wellcome Senior Research Fellow in Medical Science.