The Drosophila bZIP transcription factor Vrille is involved in hair and cell growth

The vrille (vri) gene was previously identified as an enhancer of decapentaplegic (dpp) embryonic and adult phenotypes (George and Terracol,1997). vri RNA expression is dynamic throughout development. In the embryo, vri transcripts are maternally provided and later expressed in morphogenetically active epithelia: in foregut,pharynx, hindgut, in stripes along dorsal and lateral epidermis and in tracheal track. At third larval instar, vri is expressed in gut,brain and in imaginal discs. vri was later characterized as a clock-controlled gene expressed in circadian pacemaker cells(Blau and Young, 1999) and acting as a direct repressor of the Clock and cryptochromegenes (Glossop et al., 2003; Cyran et al., 2003).

vri encodes a bZIP transcription factor(George and Terracol, 1997)with a typical basic DNA-binding domain and a leucine zipper involved in homo-or heterodimerization (Vinson et al.,1989). Vri is closely related, in the bZIP extended domain, to proteins implicated in cell death or growth. The more closely related factors are, with 60% identity and 93% similarity with Vri, gene 9 of Xenopus(Brown et al., 1996) (which is induced by thyroid hormone during the tadpole tail resorption program) and E4BP4 (a transcription factor of the human placenta that acts as a repressor of the protein E4 of adenovirus, which is involved in apoptosis)(Cowell et al., 1992; Cowell and Hurst, 1994). This factor was later isolated as NFIL3A a transactivator of the interleukin 3(IL3) promotor in human T cells (Zhang et al., 1995). In pro-B lymphocytes, the murine homolog of NFIL3A(100% identity in the bZIP domain) is a delayed anti-apoptotic early transcription factor induced by IL3 stimulation acting through the Ras/MAPK(mitogen-activated protein kinase) and PI3K (phosphatidylinositol 3-kinase)pathways. The alteration of these pathways is likely to contribute to human B-lineage leukemia (Ikushima et al.,1997; Kuribara et al.,1999). Vri and NFIL3A are also related to the segmentation gap gene product of Drosophila Giant(Capovilla et al., 1992), to CES-2, the product of a cell death specification gene of C. elegans(Metzstein et al., 1996), and to the sub-family of proteins (PAR) (Haas et al., 1995). In mammals, the PAR bZIP proteins include DBP(Mueller et al., 1990), TEF(Drolet et al., 1991), VBP(Iyer et al., 1991) and HLF(hepatic leukemia factor) (Hunger et al.,1992). It has been postulated that members of the CES-2/PAR family are evolutionarily conserved regulators of programmed cell death(Metzstein et al., 1996; Seidel and Look, 2001). In other respects, it is noteworthy that the PAR family genes present, like vri and the chicken and mouse homologs of NFIL3A/E4BP4, circadian oscillations (Wuarin and Schibler,1990; Falvey et al.,1995; Fonjallaz et al.,1996; Blau and Young,1999; Mitsui et al.,2001; Doi et al.,2001).

We show that vri partial loss of function induces flight and other locomotory defects associated with a downward bending wing phenotype and hair defects. Furthermore, vri interacts genetically with genes encoding actin-binding proteins: bent encoding a myosin light chain kinase(Ayme-Southgate et al., 1991; Daley et al., 1998), karst encoding a βHeavy-spectrin(Thomas and Kiehart, 1994; Thomas et al., 1998; Zarnescu and Thomas, 1999),and α actinin encoding an actin crosslinking spectrin superfamily member (Fyrberg et al.,1990; Fyrberg et al.,1998; Roulier et al.,1992; Dubreuil and Wang,2000). Clonal analysis shows that vri is cell autonomous and that the absence of vri reduces cell and hair size in wing and to smaller eyes with a reduced number of ommatidia of abnormal morphology with shorter photoreceptor cell stalks. The overexpression of vri reduces salivary glands growth and nuclei size, a phenotype typical of an inhibition of endoreplication. Overexpression of vri in the embryo and in imaginal discs has an anti-proliferative effect. In the wing disc, increased apoptosis is observed and at the wing surface cells with multiple trichomes are formed, indicative of cell cycle arrest and defects in regulation of actin cytoskeleton. We propose that vri may control cell growth and proliferation via the regulation of the actin cytoskeleton.

Flies were raised on standard media and all crosses were performed at 25°C unless stated otherwise. vri alleles are described by George and Terracol (George and Terracol,1997) and new 25D P insertions lines at the Berkeley Drosophila Genome Project (BDGP: http://flypush.imgen.bcm.tcm.edu/pscreen). The following stocks were used: a, Actn, bt and kst alleles(FlyBase, 2003); w¹¹¹⁸, FLP/FRT (Xu and Rubin, 1993); EGUF/hid(Stowers and Schwartz, 1999); hs FLP; Sp/SM6-TM6B (Tb), GFP FRT40A, Act5C>CD2>GAL4,UASGFP [III] (Neufeld et al.,1998a); pnr-GAL4[III](Calleja et al., 1996); vg-GAL4[II] (Simmonds et al.,1995); ey-GAL4[III](Hazelett et al., 1998); en-GAL4[II] and hs-GAL4[III](Brand and Perrimon, 1993); C765-GAL4[X] (Guillén et al., 1995); MS1096-GAL4[X](Capdevila and Guerrero,1994); A9-GAL4[X](Haerry et al., 1998); F4-GAL4[II] (Weiss et al.,1998); UAS-rbf[III](Du et al., 1996); rbf¹¹, rbf¹⁴(Du and Dyson, 1999); UAS-dap[III], UAS-E2F-Dp[III] and UASCycE[III](Lane et al., 1996); UAS-string[II] (Neufeld et al.,1998a); and UAS-dPTEN(Gao et al., 2000). Balancer chromosomes and other stocks are described elsewhere(Lindsley and Zimm, 1992; FlyBase, 2003).

Standard molecular biology experiments were performed as described by Sambrook et al. (Sambrook et al.,1989). Genomic DNA preparations and PCR amplifications were performed as described by Szuplewski and Terracol(Szuplewski and Terracol,2001). Positions are from cDNA1 (3.8) start in the genomic region:primer 1 (forward) 5′-CGATTGTCTGCACGCTGG-3′ (322-339), primer 2(reverse) 5′-GTTCCTTCTCCGGCGATC-3′ (2570-2553), primer 3 (reverse)5′-CAAGGCAAGGCTGGAGAG-3′ (6083-6066) and primer 4 (reverse)5′-AGTATCTGCAGCGCACGG-3′ (7379-7352). P-lacW(Bier et al., 1989) primers:5′P (reverse) 5′-CCTCTCAACAGCAAACGTGTACTG-3′(90-67), 3′P (forward) 5′-TCTCTTGCCGACGGGACC-3′(10653-10670), lacZ (forward) 5′-GATCATCTGGTCGCTGGG-3′(1935-1952). vri⁵(Török et al.,1993) was mapped by plasmid rescue(George and Terracol, 1997)and upstream sequence from the 5′P primer is: (1218)5′-GATTCTTGCATCATTCGGCG-3′. The insertion site is identical to that determined from the P-lacW 3′ end (AG034151) by the BDGP(http://www.fruitfly.org/p_disrupt/)(Spradling et al., 1995; Spradling et al., 1999). PCR products were cloned into pGEM plasmid (Promega) and sequenced with the SP6 or T7 primers. PCR fragments and breakpoint sequences were: vri^5R7.2, 2.2 kb from primers lacZ-3; P-lacW/genomic,5′-GCAGTGCACGGCAGATACAC/TTGCCGCTTCGGTCACCCGT-3′; vri^5R1.5, 3.3 kb from primers lacZ-4; P-lacW/genomic,5′-CAACATCAAATTGTCTGCGG/CGATGATGGTGAAGTTAACG-3′; vri^5R5.24, cloned by plasmid rescue; and genomic/P-lacW,5′-TGATTTAAGCAGAGTATTTC/GCTAAATACTGGCAGGCGTT-3′. DNA was sequenced by the dideoxy-chain termination method(Sanger et al., 1977) using the US Biochemicals sequencing kit (Pharmacia) or performed on a ABI Prism^TM 377 DNA sequencer (Applied Systems). DNA sequences were compiled using the Genetics Computer Group software (GCG)(Devereux et al., 1984) and compared with the Drosophila database using the BDGP(http://www.fruitfly.org/blast/)Blast Searches program (Altschul et al.,1990).

Mitotic clones were generated using the FLP/FRT technique(Golic and Lindquist, 1989; Xu and Rubin, 1993). vri alleles were recombined onto a P[ry⁺ hs-neo FRT]40A second chromosome. To generate somatic clones in adult y w/y w; vri P[ry⁺ hs-neo-FRT]40A/CyO females were crossed to y P[ry⁺ hs FLP]1/Y; P[y+] P[ry⁺ hs-neo FRT]40A/Bcmales and yellow clones analyzed in y P[ry⁺ hs FLP]1/y w; P[y⁺] P[ry⁺ hs-neo FRT]40A/vri P[ry⁺ hs-neo FRT]40A female progeny. Clones were heat-shock induced at the third larval instar by 1 hour exposure at 38°C. To generate mitotic clones in the eye, we used the EGUF/hid method(Stowers and Schwarz, 1999). y w/y w; P[ry⁺ hs-neo FRT]40A GMR-hid l(2)CL-L1/CyO; ey-GAL4 UAS-FLP/ey-GAL4 UAS-FLP females were crossed to y w/Y; vri P[ry⁺ hs-neo FRT]40A/CyO males at 29°C. GAL4-expressing clones were induced by the FRT `flip-out' method(Struhl and Basler, 1993) by crossing P[ry⁺ hs FLP]/Y; UAS-vri/SM6-TM6B (Tb) males to Act5C>CD2>GAL4, UAS-GFP/Act5C>CD2>GAL4, UAS-GFP females. In both types of experiments, clones were heat-shock induced in the progeny 10-48 hours after egg deposition by 30 minute exposure at 37°C. Tb⁺ female larvae were dissected at mid third larval instar.

Flies were fixed in 3% glutaraldehyde, 0.1 M PBS (2 hours at room temperature and then 24 hours at 4°C), dehydrated in ethanol series and then in amyl acetate series. Electron microscopy was performed on a JEOL JSM 6100 scanning electron microscope. To observe adult retinas, flies were fixed in carnoy, embedded in paraffin wax and 7 μm sections were cut on a Leica RM 2145 microtome. Embryonic cuticles were prepared as described by Wieschaus and Nüsslein-Volhard (Wieschaus and Nüsslein-Volhard, 1986). Wings were dissected, collected in 70% ethanol and mounted in Euparal (Labosi).

Larval tissues were dissected in 1×PBS, fixed in PBT [1×PBS,0.1% Tween 20 (Sigma) plus 4% paraformaldehyde] for 25 minutes, washed three times for 5 minutes in PBT, stained for 10 minutes in PBT, 0.1 μg/ml DAPI(Sigma), washed overnight in PBT, incubated 24 hours in 80% glycerol and mounted in Citifluor (Kent Scientific Industry Project). Acridine Orange(Sigma) staining was performed as described in Gaumer et al.(Gaumer et al., 2000). Preparations were observed on a Leica DMR fluorescence microscope using a Micro Max (Princeton Instrument) camera and collected using the Metaview Imaging System software (Universal Imaging Corporation).

vri cDNAs are described by George and Terracol(George and Terracol, 1997). The 3.8 kb XbaI fragment from pBScDNA3.8 was cloned into the XbaI site of the pUAST transformation plasmid(Brand and Perrimon, 1993) and the 3.3 kb SalI-XbaI fragment from pBScDNA3.3 was cloned into the XhoI and XbaI sites of pUAST. pUAST constructs were co-injected with the pUChδ2-3 helper plasmid in the pole cell region of w¹¹¹⁸ preblastoderm embryos(Spradling and Rubin, 1982). 20 UAS-vri1 (3.8 kb) and 27 UAS-vri2 (3.3 kb) independent transformed lines were recovered. The following lines were used in this study: UAS-vri1⁶[III], UAS-vri1⁹[III], UAS-vri1¹³ [III], UAS-vri1¹⁴[X] and UAS-vri2⁸[II].

The 5′ extension of vri was characterized by the recovery of new cDNAs and P-insertion alleles. From the nine cDNAs vriidentified by the BDGP(http://www.fruitfly.org/EST/)(Rubin et al., 2000), six isolated from embryos (0-24h) had the same 5′ end as cDNA1 (3.8 kb)(Fig. 1). One cDNA from adult heads starts at +15 bp and two, from embryos, start at -6.061 kb from cDNA1 start. The complete sequence provided for one of them, RE29005, shows that the cDNA is 3,073 kb long with a 7814 bp first intron and is 380 bp shorter at the 3′ end (Fig. 1). This cDNA encodes the same 610 amino acid putative protein as cDNA2 (3.3 kb). Four new lethal and one viable P-insertion lines were mapped (BDGP, http://flypush.imgen.bcm.tcm.edu/pscreen) in the vri region (Fig. 1) that we identified as new vri alleles. BG01569-3 (vri⁹) maps at -75 bp,upstream of vri⁴, vri⁶ (-66) and vri⁸(-57), and KG03762 (vri¹⁰) maps in intron 2 near vri⁵ and vri⁷. They do not complement vrialleles. KG01220 (vri¹¹) maps at -16.7 kb and KG10174 (vri¹³) at -8.6 kb. They complement partially vri alleles. EY01490(vri¹²) is a PUAS viable line and maps at -12.6 kb.

A secondary P-induced mutagenesis was previously performed to generate vri deletion alleles by imprecise excision of the vri⁵ P-lacW transposon mapping in the first intron(George and Terracol, 1997)(Materials and Methods). Five independent white lethal lines (among 17) showing new patterns by Southern analysis and nine independent w⁺ lethal lines (among 30) were selected for amplification of genomic DNA by PCR. The nine w⁺lines had red eyes, while vri⁵ eyes were orange and amplification of their DNAs from primers 1-5′P and 3′P-2 gave the same bands as vri⁵. Therefore, they probably result from the insertion of a secondary transposon on the chromosome. vri^5R1.5, vri^5R5.7, vri^5R7.2and vri^5R8.4 DNA amplified from primers 1-5′P and vri^5R5.7, vri^5R5.24 and vri^5R8.4 DNA from primers 3′P-2 giving the same bands as vri⁵. Thus, vri^5R5.7 and vri^5R8.4 were generated by internal rearrangements in the P element, vri^5R5.24 by imperfect excision of the 5′end and vri^5R1.5 and vri^5R7.2 by imperfect excision of the 3′end (Fig. 1). vri^5R7.2 deletion eliminates 7627 bp in the transposon 3′ end and 3735 bp (1218-4952) in vri with a breakpoint in the 3′ non-coding end of cDNA1. The deletion eliminates most of the vri gene downstream from the P-lacW insertion,including the bZIP domain. The remaining 5′ end of cDNA1 encodes a putative 66 amino acid protein (the first 27 vri amino acids plus 19 amino acid from the first intron 5′ end). vri^5R7.2 might represent a null allele. vri^5R1.5 deletion eliminates 5779 bp in the 3′ end of P-lacW and 5771 bp (1218-6988) in genomic DNA. The 5′ end is identical to the vri^5R7.2 end,but the deletion extends downstream from vri and breaks within the GadFly gene CG14024 (BDGP, http://www.fruitfly.org/annot/)(Adams et al., 2000)corresponding to the BDGP cDNA GH11985. vri^5R5.24deletion eliminates 19623 bp in the vri 5′ end, and 2223 bp in the transposon 5′ end. The breakpoint maps 1665 bp upstream from KG01220 insertion site and downstream from the putative GadFly gene CG14030 corresponding to the BDGP cDNA LD22858. The short vri cDNA(3.3 kb) is potentially preserved. The exact sizes of the genomic deletions were deduced from our own sequences and the GadFly sequence (SEG, AE003609; http://www.fruitfly.org/annot/)(Adams et al., 2000).

vri¹¹ and vri¹³over vri lethal alleles and Df(2L)tkv^Sz2(25D2-4; 25D6-E1) lead to flies with pleiotropic phenotypes (100% penetrance). The wings are shorter and downward (or more rarely upward) bending, and posterior scutellar macrochaete are shorter and upturned(Fig. 2A,B). They are poorly viable and present locomotory and flight defects. vri¹¹/vri¹³ flies present the same phenotypes. Wings are notably smaller and regions of atrophic or missing hairs are present at the margin and on the surface of the wing(Fig. 2C,D). The downward bending phenotype has been described for other mutants. This is the case for mutants in the arc gene, which encodes an adherens junction-associated PDZ domain protein(Liu and Lengyel, 2000) and in the karst gene encoding β-heavy-spectrin involved in cell integrity, polarity and adhesion (Thomas and Kiehart, 1994; Thomas et al., 1998; Zarnescu and Thomas, 1999). This phenotype is also observed for mutations in bent encoding a myosin light chain kinase expressed in indirect flight muscle and tubular muscle(Ayme-Southgate et al., 1991; Daley et al., 1998). Viable mutants in bent are flightless, this is also the case for mutants in muscle actin, and in actin-binding proteins such as flightless Iencoding a gelsolin family homolog (Davy et al., 2001) and α actinin encoding an actin crosslinking protein of the spectrin superfamily(Fyrberg et al., 1990; Fyrberg et al., 1998; Roulier et al., 1992; Dubreuil and Wang, 2000). We have tested for interactions in double heterozygotes between vri and arc, karst, bent and α actn alleles. No significant interaction was observed with arc, but only the hypomorphic a¹ allele was tested. With the three other genes, reduced viability and wing hair phenotypes similar to those of vri are observed (Fig. 2E,F). With bent (b¹) and Actn⁸ (null), Actn¹⁴ and Actn^G0077, flies also present progressive locomotory defects affecting rear legs.

To analyze further the effect of the absence or reduction of vriin the adult cuticle, we generated somatic clones with different vrilethal alleles. vri⁵, a P-insertion allele, vri¹ and vri², both with a stop codon at position 924 upstream from the bZIP domain yielding 202 and 83 amino acid putative proteins(George and Terracol, 1997),were used. We tested also vri^5R7.2, a bZIP deletion allele, and vri^5R5.24 a 5′-deletion allele (described here). In clones (marked with yellow) observed on the whole body, all types of bristles are thinner and reduced in size with all the alleles tested or are missing, especially with vri¹, vri² and vri^5R7.2 alleles(Fig. 3A,B). Wing clones show pigmented regions (Fig. 3C)that could result from necrosis. The wing surface contains small holes and is wrinkled. These phenotypes could be due to cytoskeletal defects. Clones were particularly studied in the stout bristle cells of the anterior margin triple row region. In addition to the reduction in size of the bristles, the margin is often concave suggesting defects in adhesion molecules(Fig. 3D). The mutation is cell autonomous, the size of vri bristles is the same at the center of the clone as at the margin. Reduction in bristle size is generally an indication of a reduction in the cell size, as each bristle corresponds to a single cell. The average distance between bristles within vri clones was scored showing a significant reduction compared with vri⁺ clones(Fig. 3E). The vricells are smaller and are about 70% of the wild-type size with vri¹, vri², vri⁵ and vri^5R7.2. However, vri^5R7.2 bristles are more atrophic or are missing leaving many holes in the margin. This phenotype was also observed with viable combinations. vri^5R524 cells are 85%of wild-type size, suggesting that the allele is weaker.

We induced vri clones in the eye using the EGUF/hidmethod (Stowers and Schwarz,1999), which generates eyes that are composed exclusively of clones because all other genotypes are eliminated. This is done by using the dominant photoreceptor cell lethal GMR-hid and a recessive cell death mutation. The vri eyes are smaller with a rough aspect(Fig. 4A,B), and ommatidia are disorganized with abnormal morphology and duplicated or missing bristles(Fig. 4C,D). The total number of ommatidia is significantly reduced by 10% in vri⁵, 17% in vri¹and 19% in vri^5R7.2 eye clones compared with control eyes (Fig. 4E). The mean size of ommatidia, measured from at least 60 ommatidia from several independent heads, was significantly reduced (Student's t-test, P<0.001) in vri^5R7.2 eyes only and was 80±9% of wild-type size. In vri^5R7.2retina, depth is markedly reduced and the photoreceptor stalks are distorted and their length is shortened by half (Fig. 4F,G). The analysis of wing and eye clones shows that vri^5R7.2 is the stronger allele and that, as expected, it might be a null allele.

We used the GAL4-UAS system (Brand and Perrimon, 1993) to assess the effect of overexpressing vri in different types of cells. Two types of UAS transgenes containing the two major cDNA forms were constructed (Materials and Methods). UAS-vri1 (3.8 kb) and UAS-vri2 (3.3 kb) encoding,respectively, 729 and 610 amino acid putative proteins(George and Terracol, 1997)(Fig. 1). The results are similar with the two types of transgenes. Overexpression of vri1 and vri2 is lethal with most of the GAL4 drivers or leads to severe alterations in the targeted tissues.

Overexpression of vri using the driver pannier, the expression of which is restricted to dorsal tissues throughout development(Calleja et al., 1996; Heitzler et al., 1996), leads to embryonic lethality due to defects in the dorsal epidermis and dorsal closure (Fig. 5B). Use of the vg-GAL4 driver (Simmonds et al.,1995), which directs GAL4 expression in the wing pouch, leads to notching at the wing margin (Fig. 5E). Overexpressing vri with the leaky hsp70-GAL4 (Brand and Perrimon,1993) driver results in atrophic or missing bristles on the adult cuticle (Fig. 5H). Similar phenotypes are observed by overexpressing with the same drivers the inhibitor of proliferation Rbf (Datar et al.,2000) (Fig. 5C,F,I)or Rho1, a small GTPase involved in actin cytoskeleton regulation(Hariharan et al., 1995). Overexpression of vri driven with ey-GAL4(Hazelett et al., 1998), the expression of which is specific to the eye disc and starts in the embryonic eye disc primordia, results in atrophy of the eye(Fig. 6). We observed similar phenotypes (not shown) by overexpressing Rbf(Datar et al., 2000) or Rho1(Hariharan et al., 1995). According to the strength of the UAS-vri transgene and the temperature, a progressive reduction in the size of the eye and a rough aspect are observed (Fig. 6A-D). The ommatidia are disorganized and bristles are either duplicated or missing. In some flies, the eyes are totally absent(Fig. 6E) and dissection of third instar larvae revealed no eye disc and atrophic brain (not shown). The stronger phenotype observed at 29°C leads to an absence of head(Fig. 6F). These phenotypes are compatible with an inhibition of proliferation.

vri overexpression driven in the wing with dpp^blk-GAL4, ap-GAL4 or en-GAL4 is lethal therefore we used the ubiquitous drivers, A9-GAL4(Haerry et al., 1998), C765-GAL4 (Guillén et al.,1995) and MS1096-GAL4 (Capdevilla and Guerrero, 1994)expressed predominantly in the dorsal surface. With weaker UAS-vritransgenes, results are similar with the three drivers(Fig. 7; C765-GAL4 not shown). The L5 vein is often shortened and ectopic bristles are found with the A9-GAL4 driver (Fig. 7C). The cells are larger than normal, as shown by the more widely spaced wing bristles, and possess multiple, abnormally oriented trichomes(Fig. 7C,D,G,H). A stronger viable phenotype is observed in flies with two copies of UAS-vri2⁸. The wings are severely reduced in size and possess cells with multiple trichomes of abnormal morphology(Fig. 7I,J). We observed the same phenotype when overexpressing rbf or the cyclin dependant kinase inhibitor, dacapo (De Nooij et al., 1996) (not shown).

We tested the rescue of vri MS1096-GAL4/+; UAS-vri2⁸/+phenotype by co-overexpressing the cell cycle activators, dE2F(Duronio et al., 1995; Duronio and O'Farrell, 1995)and two of its targets, Cyclin E, which promotes S phase initiation, and the string-encoded phosphatase that regulates the G2/M phase transition(Neufeld et al., 1998a). No rescue was observed with any of the transgenes neither when co-expressing the growth inhibitor, dPTEN of the PI3K pathway(Goberdhan et al., 1999; Huang et al., 1999; Gao et al., 2000). We also performed tests in double heterozygotes with vri and the null alleles rbf¹¹ and rbf¹⁴(Du and Dyson, 1999) and with available members of the Ras/MAPK and PI3K pathways regulating cell growth but we did not recover any genetic interaction.

To observe the effect of vri overexpression at earlier stages,larval tissues were studied. With the strong UAS-vri transgenes such as UAS-vri1⁹, their combination with MS1096-GAL4is pupal lethal and leads to increased apoptosis in the undifferentiated proliferating imaginal wing disc (Fig. 8A,B). Overexpression of vri was driven in the salivary gland using the F4-GAL4 transgene whose expression starts in late stage 13 of embryogenesis, once cell proliferation in the salivary primordium is complete, between the first and second rounds of endoreplication. The driver stays active throughout larval stages. F4-GAL4/+;UAS-vri1⁹/+ glands grow to about one half the size of the control gland (Fig. 8C,D). Nuclei are smaller indicating an inhibition of endoreplication and often present a degenerative aspect characterized by a condensation of the chromatin(pyknosis), suggestive of high levels of apoptosis. Overexpression in the fat body, another polytenic tissue, using the `flip out' technique, results in small clones with smaller cells, indicating an inhibition of endoreplication(Fig. 8E,F).

Overexpression phenotypes are often difficult to interpret, for example, it has been shown that ectopic expression of other transcription factors is able to generate headless flies by inhibition of cell proliferation(Jiao et al., 2001). However,we have observed consistent phenotypes with different drivers and even if overexpression has not a consequence opposite to loss-of-function mutations,which is not rare, both phenotypes are compatible.

The functional analysis of vri performed by mutant clone induction observed in adult indicates that vri is cell autonomous and involved in hair and cell growth. The fact that vri acts in a strict cell-autonomous manner suggests that it does not regulate the expression of a diffusive molecule such as a growth factor or a hormone. Smaller cells are recovered on the whole cuticle with shorter thinner or atrophic bristles. In the wing, clones have an abnormal shape and degenerative tissues are observed. These defects could be due to cytoskeleton defects. Clones induced in the eye precursor cells result in smaller eyes with a significantly reduced number of ommatidia with an atrophic morphology and a reduced size with the stronger allele. The photoreceptor cell stalks are shorter and atrophic. These results suggest that vri cells grow more slowly and are less viable than vri⁺ cells, even when they are not surrounded by wild-type cells

Smaller cells and bristles can result from metabolic defects, for example,mutations in tRNAs, ribosomal proteins (Minute) or rRNAs. This is the case for mutations in the S6 kinase gene, a regulator of ribosomal protein production (Montagne et al.,1999) and in Drosophila myc, diminutive, which is involved in growth and metabolism(Johnston et al., 1999). They are also found with genes regulating the cell cycle, such as rbf(Datar et al., 2000), and with genes of the Ras/MAPK and Insulin/PI3K pathways controlling growth(Stocker and Hafen, 2000; Oldham et al., 2000; Prober and Edgar, 2001). Hypomorphic mutations in the rbf gene, which encodes a homologue of the retinoblastoma protein, Rb, an important regulator of cell proliferation and differentiation leads to atrophic bristles and rough eyes(Du, 2000). Null-clone cells in wing discs are smaller than wild-type cells and exhibit many pyknotic nuclei, suggesting elevated levels of apoptosis. rbf-null clones induced in the eye exhibit slight to moderate hypoplasia, missing or duplicated bristles and fused ommatidia. However, RBF inhibits cell cycle progression, rather than cellular growth directly(Datar et al., 2000). Cells homozygous for partial loss-of-function mutations in components of the Ras/MAPK pathway such as Ras1 grow slowly and remain small. Ras1 controls growth, survival and differentiation in the eye(Wassarman and Therrien,1997; Halfar et al.,2001) and coordinates cellular growth and cell cycle progression in the wing (Diaz-Benjumea and Hafen,1994; Karim and Rubin,1998; Prober and Edgar,2000). Cells devoid of the lipid kinase PI3K(Leevers et al., 1996; Weinkove et al., 1999) have a reduced growth and size, while loss of its antagonist dPTEN results in bigger cells than wild type (Goberdhan et al.,1999; Huang et al.,1999; Gao et al.,2000).

Smaller cells and bristles can also result from cytoskeletal defects. For example, the miniature and dusky genes encoding ZP proteins required for cytoskeletal reorganization are involved in growth and morphogenesis of cells and hairs in the wing(Roch et al., 2003). Mutant wings are smaller, phenotype also observed in vri viable combinations. Drosophila bristles are single cells with very long extensions supported by actin bundles of crosslinked actin filaments(Tilney et al., 2000a; Tilney et al., 2000b). The assembly of actin filaments is under the control of the Rho GTPases, including Rho1, Drac1 and DCdc42. Rho GTPases act as molecular switches involved in many processes such as morphogenesis, chemotaxis, axonal guidance and cell cycle progression (Hall,1998; Van Aelst and Symons,2002). They regulate actin cytoskeleton and are involved in planar polarization, hair morphology and outgrowth processes(Eaton et al., 1995; Eaton et al., 1996; Guichard et al., 1997; Hall, 1998; Adler et al., 2000), and photoreceptor morphogenesis. The dominant-negative form of Drac1, N17Drac1,leads to reduced and disordered rhabdomeres(Chang and Ready, 2000; Colley, 2000), while overexpression of Rho1 induces rough eyes and causes atrophy of the retina (Hariharan et al.,1995). Regulation of actin cytoskeleton is coupled with other pathways, including MAPK and PI3K pathways(Hall, 1998). For example Drac1 acts upstream of JNK cascade and DCdc42 downstream of Dpp pathway in dorsal and thorax closure processes (Ricos et al., 1999). lilli acts in cytoskeleton regulation,control of cell identity and cell growth, in parallel with the Ras/MAPK and PI3K/PKB pathways. It is noteworthy that retinal cells and wing margin bristles lacking lilly are significantly smaller than wild type(Wittwer et al., 2001; Tang et al., 2001). The flightless I-mediated cytoskeletal regulation involves PI3K, and the small GTPases Ras, RhoA and Cdc42 (Davy et al.,2001). Viable mutations in flightless I cause also ultrastructural defects in the indirect flight muscles.

Overexpression of vri in the embryo and imaginal discs induces an atrophy of the resulting tissues. With strong transgenes, the discs do not proliferate and the adult structures are totally absent. Similar phenotypes are observed when the inhibitor of proliferation Rbf or the regulator of actin cytoskeleton Rho1 are overexpressed. In wing discs, strong transgenes lead to lethality and increased apoptosis is observed. On the wing surface larger cells with multiple trichomes are observed. With strong viable transgenes, the wing is reduced and trichomes are disoriented with abnormal morphology. This phenotype has been attributed to cell cycle arrest in G1 phase. Inactivation of Cdc2 kinase which regulates entry into mitosis or of the myb gene required for G2/M transition and maintenance of diploidy leads to a similar phenotype (Weigmann et al.,1997; Katzen et al.,1998). This phenotype is observed when the human cyclin-dependant kinase inhibitor P21CIP/WAF1 is overexpressed with UASP21 driven by the dpp-GAL4 transgene (Karim and Rubin, 1998). We observed the same phenotype when overexpressing dacapo, the Drosophila P21 homolog(De Nooij et al., 1996), under the control of MS1096-GAL4. DAP binds to Cyclin E/Cdk2 complexes inducing cell cycle arrest in G1 phase and epidermal cell proliferation inhibition (Lane et al.,1996). The same phenotype is also observed when disrupting cytoskeleton regulation by overexpressing the activated form of Drac1,DRac1N17 (Eaton et al., 1995)or a dominant-negative form of DRacGAP,DRacGAP^δEIE. DRacGAP is a negative regulator of the Rho-family GTPases, Drac1 and DCdc42, regulating actin cytoskeleton via EGF/Ras signaling pathway in the developing wing. The P21 overexpression phenotype is suppressed by UAS-Ras1^V12, an activated form of Ras1, or UAS-DRacGAP (Sotillos and Campuzano, 2000).

In salivary glands, vri overexpression leads to glands whose size is reduced by about one half with numerous pyknotic nuclei, suggesting inhibition of endoreplication and elevated levels of apoptosis. A reduction in the size of the gland and an inhibition of the level of endoreplication is observed when the genes encoding cell cycle regulators cyclin E,dacapo (Follette et al.,1998; Weiss et al.,1998) and rbf (Datar et al., 2000) are overexpressed. Pulses of Cyclin E are required to drive endocycle S phase, and continuous expression inhibits endoreplication cycles. In the fat body (polytenic tissue), vri-overexpressing clones are small with smaller cells often presenting a degenerative aspect.

vri was previously identified as an enhancer of dppphenotypes both in embryo and adult(George and Terracol, 1997). dpp encodes a TGFβ homolog closely related to BMP4 (bone morphogenetic protein 4), which acts as a morphogen at different stages of development (reviewed by Podos and Ferguson, 1999). Dpp plays a proliferative role in all imaginal discs at larval stages but induces a cell cycle arrest in G1 phase in the eye-antennal disc during third larval instar(Horsfield et al., 1998). Dpp might promote cell growth and/or proliferation directly, in a cell-autonomous manner (Burke and Basler, 1996; Martin-Castellanos and Edgar,2002). Other genes have been identified or described as enhancers of dpp phenotypes. Some of them are integral members of the dpp pathway, which is the case for tkv(Terracol and Lengyel, 1994), Mad and Med (Raftery et al., 1995), but others [such as lilliputian(Su et al., 2001) and cyclope (Szuplewski and Terracol, 2001)] have been shown to act in different pathways. Interestingly, lilli has been also identified in other screens as a dominant suppressor of activated MAP kinase pathway phenotypes(Dickson et al., 1996; Neufeld et al., 1998b; Rebay et al., 2000). lilli acts in parallel with the Ras/MAPK and PI3K/PKB pathways in the control of cell identity, cell growth and/or cytoskeletal arrangement(Wittwer et al., 2001; Tang et al., 2001). vri mutations do not alter dpp pathway target gene expression and probably act in a parallel pathway.

vri overexpression phenotypes suggest a role in cell cycle and proliferation. However, these phenotypes are not rescued by simultaneous overexpression of the genes encoding activators of proliferation, Drosophila E2F, cyclin E or string. Therefore, it is unlikely that Vri is either a direct repressor of genes that activate proliferation or an activator of those acting as inhibitors of proliferation like rbf or dacapo. It could act upstream in the Ras/MAPK or PI3K pathways regulating growth and involved in the regulation of the mammalian homolog of Vri, NFIL3A, acting mostly as a repressor. Genetic interactions have been tested in double-heterozygotes with available members of these pathways and vri, but no reduction in viability nor any strong phenotypes was recovered. This could result from genes with non-limiting products and/or be due to the functional redundancy of vri. Alternatively, vri may control cell size independently of growth signals.

vri loss-of-function and overexpression phenotypes, more probably,could result from primary defects in cytoskeletal actin network. Although cytoskeletal integrity and adhesion are altered in mutants of regulators of cell growth and proliferation, these effects are indirect. Wing hair atrophy phenotypes were observed in interaction with the karst gene involved in cytoskeleton arrangement. The downward-bending wing and the reduction of the photoreceptor stalk size (Pellikka et al., 2002) are two other phenotypes observed in vri and kst. We observed new vri phenotypes affecting wing shape flight and locomotion. Locomotory defects could result from neurological or muscular alteration (or both). We found interaction with the α actn and bent genes involved in muscle actin function, which suggests that the effect is rather at the muscular level, although we have not observed gross defect in indirect flight muscle. However these defects appear degenerative and must be studied in more detail. We also observe hair atrophic phenotypes in interaction with these two genes, suggesting an effect in other cell types. Although the locomotory and hair defects are not necessarily related, it is notable that the genes interacting with vri affect different types of actin, muscle and non-muscle actin. It will be interesting to search for the direct targets of Vri to understand its implication in locomotion and cytoskeletal integrity.

We thank A. Kropfinger for helpful comments on the manuscript, D. Montero for performing scanning electron microscopy, C. Dubucs for performing most of the sequences, and M. Bienz, C. Lehner, P. Leopold, M. B. O'Connor, J. Silber and the Bloomington Stock Center for providing Drosophila stocks. This work was supported by the Centre National de la Recherche Scientifique,by a grant from the Association pour la Recherche sur le Cancer (ARC)N°9894 to R.T., and by grants from the Ministère de l'Enseignement,de la Recherche et de la Technologie (MERT), the ARC and the Fondation pour la Recherche Médicale to S.S.