Since genes encoding highly conserved activites often function across species boundaries, we attempted to evaluate mouse Cul1 (mCul1) function in Drosophila. To this end, we used a GAL4 transgene to drive overexpression of mCul1 in the wing disc. MS1096 is an X-linked transgene that expresses GAL4 strongly on the dorsal surface and more weakly on the ventral surface of the wing primordium in third instar larval wing imaginal discs. Because of dosage compensation, GAL4 levels in males (MS1096/Y) are twice those in heterozygous females (MS1096/+). In males (MS1096/Y;UAS-mCul1/+), mCul1 overexpression led to a strong reduction in the size of the wing without any gross patterning defects (fig. 1A,1B). Females (MS1096/+;UAS-mCul1/+) displayed a weaker phenotype that was sometimes indistinguishable from wild-type (see fig. 2A). In males, further increasing expression by using two copies of the UAS-mCul1 transgene reduces the wing to a rudimentary stump (not shown). The reduced wings also occasionally showed a multiple wing hair phenotype in which several trichomes instead of just one were produced by individual intervein cells, particularly in the region near the anterior cross-vein. Introduction of a single mutant copy of the Cul1^k01207 allele 22 or a deficiency that uncovers Cul1 (Df(2R)NCX8) dominantly enhanced the wing reduction (fig. 1C). In contrast, decreasing the level of endogenous Cul3/guftagu (another Cullin family member, see http://flybase.bio.indiana.edu/) or expressing a GFP marker had no effect. Coexpression of Drosophila Cul1 (dCul1) suppressed the phenotype induced by expression of mCul1 (fig. 1D). These data show that overexpression of mouse Cul1 has a dominant negative effect on Cul1 function, or that dCul1 suppresses a neomorphic activity of mCul1.

To test if overexpression of mCul1 is sensitive to the dosage of other SCF components, we asked whether changing the copy number of both dRoc1 and dSkpA can modify the mCul1-induced phenotype. dRoc1 and dSkpA are located at the tip of the X chromosome in region 1B (Flybase, http://flybase.bio.indiana.edu/). The small wing phenotype seen in flies overexpressing mCul1 was enhanced when flies also had a deficiency uncovering this region (fig. 2A,2B), while coexpression of dSkpA rescued the wing phenotypes (fig. 2C,2D). These results show that SCF function limits the mCul1 overexpression phenotype.

Cul1 and SCF function are required for proteasome-dependent degradation of a number of proteins in other organisms. We thus tested whether reduction in proteasome function would modify the phenotype caused by mCul1 overexpression. The DTS5 and DTS7 alleles are mutations in the β6 and β2 subunits of the proteasome 2324. Both mutations, which have antimorphic activity, acted dominantly to strongly enhance the small wing phenotype generated by mCul1 overexpression (fig. 2E) supporting the idea that the phenotype is due to reduced degradation of an SCF substrate.

To directly assess Cul1 function, we sought a target protein whose destruction might be limited by the activity of the Cul1/proteasome pathway. While SCF triggers ubiquitination and proteasome action degrades many substrates to completion, a few targets are instead cleaved and partially degraded. This is true of the transcription factor Cubitus interruptus (Ci): it is cleaved and its C-terminal portion is degraded 252627. Cleavage and degradation require phosphorylation, SCF^Slmb, and proteasomal activity. While complete inactivation of SCF^Slmb causes cell death, it was shown that a cell-viable hypomorphic slimb allele resulted in accumulation of unprocessed Ci. Based on this, and findings that Ci accumulation is especially sensitive to pharmacological inhibition of the proteasome (data not shown), it appears that Ci levels are particularly sensitive to partial reduction in the activity of the SCF/proteasome pathway of protein processing/degradation. Inhibition of Ci processing can be scored by the accumulation of a C-terminal epitope 26. This assay showed that reduction in the function of Slimb or Roc1, two activities that contribute to SCF function, is associated with an increase in Ci staining 2627. Similarly, reduction of Cul1 function by mutation in Cul1 (data not shown and 28) or by overexpressing mCul1 (fig. 2F and 2G) led to Ci accumulation.

Together these results indicate that mCul1 overexpression leads to a Cul1 dominant negative action that reduces SCF function. Based on the strength of the phenotype and its enhancement by removal of an endogenous copy of Cul1, we infer that the mCul1 expression only partially suppresses Cul1 function to create a hypomorphic phenotype. We infer that the wing defect results from a reduction in the rate of degradation of one or more targets of SCF action.

The reduction in wing size in mCul1 overexpressing flies could result from changes in cell size or cell number. Each cell of the wing blade produces a hair and hair density indicates cell density and hence cell size. Wing hair density was unchanged by mCul1 expression, even when there was a dramatic effect on wing size (62.6 ± 3.0 cells/7950 μm² for MS1096/Y; 60.8 ± 2.3 cells/7950 μm² for MS1096/Y;UAS-mCul1/+). This result suggests that mCul1 overexpression causes a reduction in cell number either by a coordinate reduction in cell growth and proliferation or by increasing cell death. To distinguish between these possibilities, we dissected wing imaginal discs from late third instar larvae and stained them for mitotic or apoptotic cells. Labelling with an anti-phospho-histone 3 antibody did not reveal any significant changes in the number of mitotic cells (not shown). However, acridine orange staining showed a large increase in cell death throughout the wing pouch when mCul1 was overexpressed (fig. 3A,3B). Coexpression of baculovirus anti-apoptotic protein p35 rescued the reduction in wing size (fig. 3C,3D) indicating that mCul1 overexpression induces elevated apoptosis. mCul1-induced apoptosis was insensitive to the coexpression of dominant negative forms of p53 (not shown). p35 expression also suppressed the multiple wing hair phenotype indicating that this phenotype too requires apoptosis. As dead cells do not contribute to adult structures, this phenotype must be a secondary consequence of cell death. We conclude that limiting SCF function induces apoptosis. Note that this conclusion does not argue against a cell cycle role for SCF, as defects in cell cycle progression can trigger apoptosis.

Having determined that the mCul1 overexpression phenotype results from a dominant negative effect, we were in a position to test for genetic interactions between mCul1 overexpression and mutations in potential cell cycle targets of Cul1. Because SCF complexes have been implicated in cell cycle control in other organisms, we tested whether changing the dosage of cell cycle regulators dominantly modified the mCul1 overexpression phenotype. Reductions in the levels of a variety of cell cycle genes – cyclin E, dacapo, roughex, DP, RBF, string, cyclin A, rca1, wee1 and cdc2 – had no effect on wing size (not shown). In contrast, reducing the level of E2F by 50% completely restored wild-type wing size (fig. 4A,4B,4C), an effect that correlated with a strong reduction in cell death in the disc (fig. 4E,4F). The presence of a single hs-E2F transgene gave a weak and incompletely penetrant enhancement in the absence of heat shock, and this enhancement was stronger and completely penetrant when the transgenic strain was exposed to a heat shock treatment every 24 h during third instar and early pupal development (fig. 4D). Since the presence of the hs-E2F transgene and its induction by our heat shock protocol (low level overexpression compared to Asano et al., 1996) had no phenotype when in a wild-type background, we conclude that mCul1 expression sensitizes the cells to E2F overexpression. The reciprocal suppression by E2F mutation and enhancement by E2F overexpression show that the apoptosis triggered by mCul1 is mediated by or requires E2F function. Because a high level of ectopic expression of E2F is known to induce apoptosis 1213, these results suggest that mCul1 overexpression might interfere with normal E2F downregulation.

Immunostaining for E2F in the furrow of the eye disc showed an abrupt drop in protein levels at the position where most cells exit G1 and initiate S phase 12. Using a double label approach, we were able to test directly for overlap between E2F presence and S phase as marked by 5-bromodeoxyuridine (BrdU) incorporation in dividing cells of the wing disc despite the asynchrony of division. No E2F staining was detected in BrdU labelled cells (green) of wild-type wing discs, whereas non-S phase cells showed strong E2F (red) staining (fig. 5A, 5B, 5C) (no E2F staining was observed in over a thousand BrdU positive cells from 8 different discs). Thus, E2F is downregulated in advance of significant BrdU incorporation in wild-type wing disc cells.

When mCul1 is overexpressed, a small fraction of S phase cells (3% of 800 BrdU positive cells from 6 independent wing discs) contains detectable amounts of E2F (fig. 5D, 5E, 5F, arrowheads). Since E2F is normally degraded prior to S phase, the overlap in staining suggests that the partial suppression of Cul1 caused by mCul1 expression slowed E2F destruction to the point that it persisted briefly in S phase. We suggest, but have not demonstrated, that the low percentage of cells showing the overlap in the staining is due to a remaining capacity of the residual SCF activity to promote slower but still effective elimination of E2F during S phase. In any case, our finding shows that when dCul1 function is reduced, the normal pattern of E2F degradation in S phase is disrupted, leading to inappropriate persistence of E2F. While it could also be suggested that this effect might be secondary to other disruptions induced by dominant negative Cul1, the genetic suppression of phenotypes by reduction of E2F dose (above) suggests otherwise, as do the analyses of the role of the F-box protein, Slimb (below).

The experiments described above demonstrate that overexpression of mCul1 produces a phenotype that can be modified by second site mutations and therefore could be used for a more general genome-wide modifier-screen. To determine whether other suppressors or enhancers might be isolated, we tested if deficiencies on the third chromosome uncover other potential interacting genes.

One modifying deficiency, Df(3R)e-R1, removes region 93B6;D3-4 and acts as an enhancer of the small wing phenotype. One gene in this region, slimb, is an interesting candidate because this gene encodes an F-box protein that is part of an SCF complex 2629. Several independent alleles of slimb were tested and all enhanced the dominant-negative mCul1 overexpression phenotype (fig. 6, compare A and B). Four deficiencies uncovering other predicted F-box containing genes (CG4911, CG3428, CG2010 and CG9003) or mutations in the F-box protein genes ppa and ago 3031 had no effect on the phenotype (not shown) showing that the effect is specific. Interestingly, co-reduction of E2F and slimb function reversed the slimb enhanced mCul1 overexpression phenotype (fig. 6A and 6C) and loss of a copy of slimb attenuated the suppression of the mCul1 phenotype resulting from loss of one copy of E2F (compare fig. 4C and 6C). This mutual co-suppression suggests that slimb and E2F exert opposing activities on the same process and perhaps that Slimb acts on E2F or on a critical regulator of E2F activity.

We thus wished to test whether loss of Slimb had an effect on E2F protein in imaginal discs. As slimb null mutants die as first instar larvae, direct analysis of the mutant discs was not possible. The severity of the slimb mutant phenotype suggested that the mutation reduced crucial SCF functions much more than did expression of mCul1. We used a slimb mutant stock carrying a slimb transgene under the control of the HSP 70 promoter to modify the time of onset of the slimb mutant phenotype. In the absence of heat shock, slimb homozygotes exhibited the early lethality of slimb mutants, but a single 30 minute heat shock at 37°C during first instar was enough to allow some larvae to reach the wandering third instar stage 4 days later. These larvae have small imaginal discs. Labelling of these discs with the anti-E2F antibody revealed that nearly all cells expressed E2F (fig. 7A). Many of these cells also incorporated low levels of BrdU (fig. 7B and 7C), usually in regions of the nucleus with lower E2F staining. A few cells also had high levels of BrdU incorporation and these also expressed E2F, although at lower levels (asterisks on fig. 7A and 7B). An hour and a half after a heat shock that restored Slimb expression, BrdU-labelled cells no longer contained detectable levels of E2F (fig. 7D,7E,7F). No accumulation of E2F mRNA can be seen in slimb mutant discs (fig. 8) suggesting that any effect of Slimb on E2F occurs posttranscriptionnally. These results indicate that elimination of E2F during DNA replication requires Slimb.

The above results raise the possibility that an SCF^Slmb complex can target E2F for degradation. If this is the case, it is expected that Slimb would recruit E2F to the complex. To test this hypothesis, we asked whether Slimb and E2F physically interact. To this end, ³⁵S-labelled E2F was produced in vitro in a wheat germ lysate and incubated with the WD repeat region of Slimb fused to GST and bound to beads. Analysis of bead associated proteins (fig. 9) showed association of radiolabeled E2F with the fusion containing the WD-repeat region of Slimb (lane 3) but not with GST (lane 2). The interaction between Slimb and E2F depends on a phosphorylation event as it is prevented by the presence of active alkaline phosphatase (lane 4) but not when the phosphatase is inhibited by sodium vanadate (lane 5). Similar results were obtained with E2F2, the other Drosophila E2F family member (not shown).

We identified several lead ESTs in the Berkeley Drosophila Genome Project database. Sequencing of one of the corresponding cDNAs (LD20253, obtained from Genome Systems Inc., St. Louis, MO) revealed that it encodes a protein with 61% identity to the mouse and human Cul1 proteins. Expression of this cDNA in yeast is able to rescue the cdc53-1 mutation (not shown) indicating that the protein can perform at least some of the functions of the yeast Cul1. We used the same strategy to identify a Skp1 cDNA that also rescues skp1 mutations in yeast (not shown). Because the Drosophila genome contains several Skp1-like genes, we call the fly Skp1 homologue dSkpA.

UAS-mCul1, UAS-dCul1 and UAS-dSkpA were generated by subcloning the coding region of the corresponding cDNA in pUAST 38. The resulting vectors were then used to transform y¹w⁶⁷ embryos. The MS1096 stock is as described 39. The Cul1 allele k01207 22 will be described elsewhere. The following mutations were tested for genetic interaction with mCul1 overexpression: E2F^rm729, E2F⁹¹, E2F⁷¹⁷², cycA^C8LR1 (a gift from Christian Lehner), cycA^neo114, cycE^AR95, stg^7B, stg^AR2, cdc2^B7, cdc2^E1–9, DP^vr10–13, Df(2R)vg56, Dwee1^ES1 (a gift from Shelagh Campbell), Df(2L)WO5, gft¹, gft⁶⁴³⁰ (gifts from Hemlata Mistry), rca1⁰³³⁰⁰, rca1², dap^3X1, DTS5 and DTS7 (gifts from John Belote). The Bloomington Stock Center (Bloomington, IN; http://flystocks.bio.indiana.edu/) provided the l(2)k01207 and T(1;Y)B106 stocks and the deficiency kit for the third chromosome. The Umeå Stock Center (Umeå, Sweden) provided the Df(2R)NCX8 stock. UAS-p35 and UAS-PI3KD954A were given by Bruce Edgar and Michael Waterfield, respectively. Slimb alleles were provided by Tian Xu (slmb^e4-1) and Bernadette Limbourg-Bouchon (slmb^P245, slmb⁸ and slmb⁸, hs-slmb).

For modification of the mCul1 overexpression phenotype, MS1096; UAS-mCul1/TM3,Sb virgin females were crossed to males bearing the tested mutation or transgene over a marked chromosome (L or If for the second chromosome, TM3,Sb or TM6B for the third chromosome) and males from the progeny were inspected for modification of the phenotype. All crosses were kept at 25°C. The wings of all MS1096; UAS-mCul1/+ flies reach the tip of the abdomen and all modifiers are highly penetrant. A mutation was considered an enhancer if the wings were significantly shorter and a suppressor if they were clearly longer. Mutations on the X chromosome were first recombined with MS1096 and interactions were tested in females by crossing MS1096/Y;UAS-mCul1/TM3 males to virgin females carrying the mutation and MS1096 over a balancer. As a control, MS1096;UAS-PI3KD954A/TM3 virgin females were crossed to males bearing modifiers of the mCul1 phenotype. Modifiers that had the same effect on both the mCul1 and the dominant negative PI3K phenotypes were discarded as they presumably affected the GAL4 driver. Wings were dissected and mounted in glycerol/ethanol (1:1) and pictures were taken under brightfield illumination on a Leica microscope equipped with a CCD camera. Heat shock treatments consisted of two 10 minutes incubations at 37°C separated by a 20 minutes interval at 23°C. Cell size was assessed by counting trichomes in a 7950 μm² area on wings from 10 different flies.

For acridine orange staining, discs were dissected in Schneider's medium, incubated for 5 minutes in medium containing 1.6 × 10^-7 M acridine orange, washed briefly and put on a slide in a drop of medium. Pictures were taken immediately. For E2F/BrdU double staining, discs were first incubated 30 minutes in a drop of Schneider's medium containing 10% fetal calf serum and 1 mg/ml BrdU then fixed 20 minutes with 2% formaldehyde in PLP (10 mM NaIO₄, 75 mM Lysine, 37 mM NaH₂PO₄, pH 7.2). Fixed discs were then permeabilized in PBS containing 0.1% saponin and 1% goat serum and incubated overnight with an anti-E2F serum 12 at 1/250. Two rounds of amplification (reaction time: 10 minutes) with the tyramide biotin system (NEN) were used to detect E2F. Following this, the discs were processed for BrdU detection as previously described 40. Images were taken with a Princeton Instruments CCD camera on an Olympus inverted microscope using the DeltaVision microscopy system software. Cubitus interruptus was detected using the rat monoclonal antibody 2A1 41. To compare protein levels, control and mCul1-expressing discs were processed together. E2F mRNA was detected by in situ hybridisation using an antisense probe as described 42. A sense probe used as control gave no signal. To compare levels, discs were processed together.

Plasmid pGEX4T1-Slmb(176–510) allowing expression of fragment 176–510 of the Slimb protein fused to GST was kindly provided by Anne Plessis and Matthieu Sanial. GST fusion proteins were purified from Escherichia coli BL21 as described 43, except that bacteria were grown at 30°C. In vitro translated (IVT) proteins were produced and radiolabelled with ³⁵S-methionine using the TNT coupled wheat germ extract system from Promega. IVT proteins were incubated with the GST fusion proteins on beads for 1 hour at 4°C in 500 μL of 40 mM HEPES pH 7.4, 100 mM KCl, 5 mM MgCl₂, 1 mM EDTA, 1% Triton X100, 1 mM ATP, 1 μM microcystin, 14 mM b-mercaptoethanol. For phosphatase treatment, IVT proteins were first incubated in 130 μl of binding buffer containing 20 U of alkaline phosphatase in the presence or absence of 10 mM Na₃VO₄ for 1 h at 37°C. After incubation, beads were then washed three times with incubation buffer containing 130 mM KCl. Bound proteins were then analysed by SDS-PAGE and autoradiography.