Main

E2F1 is generally dispensable for cell proliferation but is selectively activated in response to specific cues, such as DNA damage, where it drives the expression of pro-apoptotic genes. Ectopic expression of Drosophila E2F1 (dE2f1) in the developing wing causes apoptosis, giving a visible, dosage-sensitive phenotype that we have used to screen for in vivo regulators of E2F-dependent apoptosis2. Using this strategy, we found a novel interaction between dE2f1 and the Wnt signalling pathway (Fig. 1). An apoptotic, gnarled wing phenotype (Fig. 1b), caused by elevated dE2f1 in newly eclosed wing epithelial cells, was strongly suppressed by co-expression of the Drosophila β-catenin orthologue armadillo (arm; Fig. 1d), and partly suppressed by co-expression of pangolin (pan; Fig. 1e), which encodes dTCF, the transcription factor partner of arm. Moreover, expression of dominant-negative, amino-terminally truncated pan (dTCFΔN) phenocopied the dE2f1 wing phenotype (Fig. 1f), and ectopic expression of shaggy (sgg; the GSK3 orthologue and negative regulator of Arm protein stability) strongly enhanced the effects of dE2f1 (Fig. 1g, h). Using a stable and activated-mutant form of arm (arm*; S44Y mutation in sgg/GSK3 phosphorylation site3), we tested whether dE2f1 expression could modify an arm-dependent phenotype. Expression of arm* under the GMR eye specific promoter causes a rough eye phenotype (Fig. 1k) that was partly suppressed by co-expression of dE2f1/dDp (Fig. 1l). Together, these genetic interactions show a strong functional antagonism between elevated dE2F1/dDP and Arm/β-catenin signalling in vivo.

Figure 1: Functional antagonism between E2F1 and β-catenin/TCF-signalling.
figure 1

a, Normal Drosophila wing of Act88F-Gal4/+ genotype. A dE2F1-induced apoptotic wing phenotype (b) is strongly suppressed by co-expression of arm (d) or pan/dTCF (e), phenocopied by expression of dominant-negative dTCF (dTCFΔN) (f), and strongly synergizes with sgg/GSK3 co-expression (h). Expression of arm (c), sgg (g) or pan (not shown) alone does not induce a wing phenotype. i–l, GMR-mediated eye-specific expression of dE2F1/dDp strongly suppresses a rough eye phenotype induced by activated-arm (arm*; S44Y) expression. All phenotypes were compared in female progeny from F1 crosses conducted at 25 °C versus control (w1118 background). m, E2F1 expression activates a canonical pE2F4B-luciferase reporter while abrogating the S33Y-β-catenin-mediated activation of pTopFLASH. n, Inhibition of pTopFLASH activity is specific to E2F1. o, E2F1 dominantly inhibits pTopFLASH activation by p73. All data are expressed as mean ± s.d. (n = 3) of normalized relative light units (NRLU) of luciferase.

PowerPoint slide

Full size image

Wnt/β-catenin signalling is important during development and regulates diverse aspects of cell function, including proliferation, differentiation and survival4,5. The genetic interactions suggested that E2F1 might inhibit Arm/β-catenin-dependent transcription. To test this, and to ask whether the interaction was conserved in human cells, we examined the effects of E2F1 on activation of a TCF-luciferase reporter (pTopFLASH) by a stable, tumour-derived form of β-catenin (S33Y6). In human Saos2 cells—a p53- and Rb-deficient cell line that exhibits low basal Wnt activity—expression of E2F1 strongly inhibited S33Y-β-catenin transcription (Fig. 1m). Control experiments showed that the inhibition was not an indirect consequence of E2F1-induced apoptosis or cell cycle progression (Fig. 1 and Supplementary Fig. 2). Inhibition by E2F1 was comparable to the effects of dominant-negative forms of TCF1 or TCF4, enhanced by co-expression of the E2F1-dimerization partner DP1, and abrogated by mutation of DP1- or DNA-binding domains of E2F1 (Supplementary Figs 2 and 3). Remarkably, as little as 10 ng of E2F1 expression plasmid repressed pTopFLASH activity tenfold when co-transfected with DP1 (Supplementary Fig. 2c). E2F1 activated transcription from a canonical E2F-luciferase reporter (pE2F4B) in these cells (Fig. 1m, n), indicating that it has context-dependent effects. Inhibition of β-catenin was a specific property of E2F1, as other activator E2Fs had little effect (Fig. 1n). Using E2F1/3 chimaeras7, we mapped the inhibitory region to the Marked Box and adjacent domains of E2F1 (Supplementary Fig. 3), regions that allow selective interactions with other transcription factors and that determine the differences between the transcriptional signatures of E2F1 and E2F3 (ref. 8).

The pro-apoptotic activity of E2F1 has been linked to the TP53 and TP73 tumour suppressors; intriguingly, both of these also affect β-catenin-dependent transcription from the pTopFLASH reporter9,10,11. p53, like E2F1, inhibited pTopFLASH transcription, whereas p73, a p53-related gene that is a transcriptional target of E2F1, activated pTopFLASH (Fig. 1o). In TP53-deficient Saos2 cells, E2F1 repressed β-catenin-mediated pTopFLASH activity and dominantly suppressed the stimulatory effects of p73 (Fig. 1o). Thus, E2F1 is a potent inhibitor of β-catenin/TCF-activated transcription that acts independently of p53 and is dominant over p73.

To determine the effects of E2F1 on endogenous β-catenin-regulated genes we first examined c-MYC, one of the best-studied targets of β-catenin and a key mediator of the pro-proliferative effects of deregulated β-catenin during tumorigenesis12. Levels of c-MYC messenger RNA and protein rapidly decreased after E2F1 induction in Saos2-TR-E2F1 cells (Fig. 2a, c). PPARδ and CD44, two other well-studied Wnt targets, showed similar changes (Fig. 2c). As a control, E2F1 strongly activated the well-known E2F target genes CCNE1 (Cyclin E) and TP73 (Fig. 2b). Similar effects on Wnt target genes were observed when E2F1 was expressed in colorectal cancer cells (Fig. 2f).

Figure 2: E2F1 abrogates Wnt signalling by modulating β-catenin target gene expression and inducing the GSK3-independent degradation of β-catenin.
figure 2

a, In Saos2-TR-E2F1 cells, E2F1 represses c-myc levels without affecting the levels of TCF1 or TCF4. b–d, E2F1 modulates the expression of endogenous Wnt target genes (qPCR analysis after Tet-induced E2F1 expression at 24 h). e, The kinetics of E2F1-induced AXIN2 and SIAH1 expression mirrors the E2F1 activation of CCNE1/Cyclin E and CCNA1/Cyclin A, and precedes E2F1-induced apoptosis. f, E2F1 represses the expression of Wnt targets in DLD1 colorectal cancer cells. Levels of mRNA were normalized to GAPDH and the effect of E2F1 is depicted as the ratio between samples after pCMV-empty or pCMV-E2F1 (1 µg each) expression. g, h, E2F1-induced β-catenin degradation (control expression from the same lysates shown in Supplementary Fig. 4f) is both GSK3- and caspase-independent. Saos2 cells were treated with control (DMSO), GSK3 inhibitors (20 µM SB216763 and 5 mM LiCl), or the caspase inhibitor peptide BOC-aspartyl-FMK (BAF; 100 µM) with or without Tet-induction of E2F1 (western blot). i, Co-expression of Bcl-2 (25 ng), pRB (10–25 ng), stabilized tumour-derived β-catenin mutants (10–25 ng), or the GSK3-β inhibitor SB216763 (15 μM), partially rescues E2F1-induced apoptosis. Cell death was depicted as the percentage inhibition of enhanced green fluorescent protein (EGFP) loss at 48 h after transfection with pCMV-E2F1 or pCMV-empty (100 ng each) along with EGFP expression construct. All data are expressed as mean ± s.d. (n = 3).

PowerPoint slide

Full size image

The list of known Wnt targets includes genes that control β-catenin degradation, such as AXIN1 and AXIN2. Interestingly, the expression of AXIN1 and AXIN2, as well as SIAH1, a p53-inducible, GSK3-independent promoter of β-catenin degradation13,14, were all significantly activated by E2F1 (Fig. 2d, e), consistent with previous studies suggesting that AXIN2 and SIAH1 are E2F-target genes15,16. Accordingly, the level of β-catenin protein decreased at later time points following E2F1 expression, a change that preceded apoptosis (Fig. 2g and Supplementary Fig. 4). Similarly, ectopic expression of Drosophila dE2F1/dDP reduced Arm protein levels (Supplementary Fig. 4h).

The mechanism of E2F1-dependent β-catenin downregulation is probably distinct from the changes observed during epithelial–mesenchymal transition, or following the disruption of adherens junctions or focal-adhesions, as markers for these processes were unperturbed by E2F1 expression (Supplementary Fig. 4i). Instead, E2F1 induced the post-translational degradation of β-catenin in a GSK3- and caspase-independent fashion (Fig. 2h and Supplementary Fig. 4g). E2F1-mediated degradation of β-catenin is functionally significant because re-expression of stable, tumour-derived mutants of β-catenin, or treatment with GSK3-inhibitors, partly abrogated E2F1-dependent apoptosis (Fig. 2i). Taken together, these results show that E2F1 inhibits β-catenin activity via transcriptional antagonism and β-catenin degradation, and that this inhibition contributes to E2F1-induced apoptosis.

β-Catenin-dependent transcription is crucially important for cell proliferation in colorectal cancer cells. Mutations in APC or CTNNB1 (β-catenin) occur early in colorectal tumorigenesis, leading to pre-malignant polyps. Additional mutations contribute to the transition to malignant adenocarcinoma4,5. An unusual feature of colorectal cancer cells is that they rarely (if ever) acquire mutations in the RB1 tumour suppressor gene. Paradoxically, RB1 copy gains are frequently found in colorectal cancer cells, often resulting in protein overexpression17. Conditional inactivation of murine Rb by Villin-Cre leads to aggressive tumours in various tissues, but rarely in the gastrointestinal tract18,19, and the knockdown of pRB reduces cell proliferation and anchorage-independent growth of human colon cancer cell lines20. Accordingly, we found elevated levels of pRB and β-catenin co-localized within the epithelium of ApcMin colonic tumours in mice (Fig. 3a, b). We hypothesized that colorectal tumour cells might select for mechanisms that limit the activity of E2F1, and that in this context the pRB tumour suppressor might act to sustain high levels of β-catenin/TCF-dependent transcription.

Figure 3: pRB inactivation abrogates β-catenin/TCF-dependent transcription.
figure 3

a, b, High levels of pRB and β-catenin co-localize within the tumour epithelium of an ApcMin colonic tumour. c, Endogenous pRB was depleted for 6 days in U2OS-shRb cells (containing DOX-inducible short-hairpin-Rb and GFP transgenes). d, pRB depletion activates E2F and represses TCF activity. Basal and activated E2F and TCF activity was determined by transfection of their respective luciferase reporter plasmids (for 24 h) in the absence (basal) or presence (activated) of E2F1 or S33Y-β-catenin expression constructs. e, In SW480 colorectal cancer cells, E2F1 expression is sufficient to activate E2F activity and repress endogenous β-catenin activity in a dose-dependent manner. f, Expression of shRb (400 ng) is sufficient to repress pTopFLASH activity in three different colorectal cancer cell lines (SW480, DLD1 and HCT116). g, pRB inactivation reduces clonogenic survival of SW480 cells and is partly rescued by S33Y-β-catenin/TCF1E, but not by Bcl-2 expression. SW480 cells were Amaxa nucleoporated with either control LLP-GFP or -Rb silencing constructs (1 µg each) with or without S33Y-β-catenin (250 ng), TCF1E (250 ng) or Bcl-2 (500 ng) expression constructs, and survival was determined at day 5 by MTT assay. All data are expressed as mean ± s.d. (n = 3; *P < 0.015 by t-test).

PowerPoint slide

Full size image

To test this hypothesis, we used a stable line of U2OS osteosarcoma cells containing a doxycycline (Dox)-inducible short-hairpin RNA targeting Rb (U2OS-shRb)21. Depletion of pRB increased transcription from an E2F reporter and inhibited transcription from the pTopFLASH reporter (Fig. 3c, d). In SW480 colorectal cancer cells that contain mutant APC and have deregulated β-catenin22, the expression of E2F1 sufficed to activate the E2F-luciferase reporter and to inactivate basal pTopFLASH transcription (Fig. 3e). Moreover, a short hairpin RNA (shRNA) vector that targets pRB inhibited the activity of endogenous β-catenin/TCF (Fig. 3f) and strongly inhibited cell proliferation (Fig. 3g), an effect partly rescued by co-expression of S33Y-β-catenin/TCF, but not Bcl-2 (Fig. 3g). Together, these observations provide a molecular explanation why colorectal tumour cells maintain the expression of pRB.

Because E2F1 is a potent inhibitor of β-catenin, we reasoned that tumour cells might select for additional ways to limit its activity. We generated transgenic flies that allowed us to knockdown dE2F1 in a tissue-specific manner (dE2f1RNAi ; Fig. 4 and Supplementary Fig. 5). This approach reduced dE2F1 activity in vivo and generated phenotypes that were used to screen for factors that were rate-limiting for dE2F1-dependent proliferation in vivo (see Methods). We identified dCdk8c01804 , a hypomorphic mutant allele of Drosophila Cdk8, as a strong and specific suppressor of dE2f1RNAi phenotypes in both the eye and wing (Fig. 4a–c and Supplementary Fig. 5). CDK8, Cyclin C, MED12 and MED13 form a sub-module of the Mediator complex, a large multi-subunit regulator of transcription23,24. We observed increased expression of dE2F1-regulated genes in dCdk8- or dCycC-mutant larvae (Fig. 4d). RNA interference (RNAi)-mediated depletion of dCDK8 or dCycC in Drosophila SL2 cells caused similar changes (Supplementary Fig. 6a). In addition, RNAi-mediated depletion of dCDK8, or other components of the CDK8 sub-module, partly suppressed the cell proliferation defects caused by depletion of dE2F1 (Supplementary Fig. 6b, c).

Figure 4: CDK8 antagonizes E2F1 activity.
figure 4

a–c, The dCdk8c01804 mutant dominantly suppresses a rough eye phenotype caused by the eye-specific expression of a dE2F1RNAi transgene. d, The expression of the dE2F1 target genes PCNA and MCM5 is upregulated in dCdk8 (null or hypomorphic (hypo) alleles) or dCycC null mutant Drosophila larvae, whereas the expression of the dE2F2 target gene vasa is unaffected. e, dCDK8 physically interacts with dE2F1 by GST-pulldown assay. f, Co-immunoprecipitation (IP) of human E2F1 and CDK8 from Saos2 whole-cell extracts. As control, E2F1 does not associate with the small-Mediator subunit CRSP70. g, CDK8 binds to and specifically phosphorylates E2F1. Kinase-assay was performed after CDK8 or CRSP70 (as control) immunoprecipitation from Saos2 cells. E2F1, E2F4 or DP1 were re-immunoprecipitated and resolved in 12% SDS–polyacrylamide gel electrophoresis (SDS–PAGE). h, The expression of wild-type CDK8, but not kinase-dead, abrogates the inhibitory effects of E2F1 on β-catenin/TCF-dependent transcription. pTopFLASH assays were determined at 48 h in co-transfection experiments with 200 ng CDK8 or CDK8KD in Saos2 cells. i, Expression of shRb (400 ng) and CDK8KD (300 ng) cooperatively repress pTopFLASH activity in SW480 colorectal cancer cells. All data are expressed as mean ± s.d. (n = 3).

PowerPoint slide

Full size image

Glutathione S-transferase (GST)-pulldown assays demonstrated a strong physical interaction between dCDK8 and dE2F1 (Fig. 4e) that mapped to the dE2F1 transactivation domain (Supplementary Fig. 7). The physical interaction between E2F1 and CDK8 is conserved between species: human E2F1 co-immunoprecipitated with CDK8 from Saos2 cell extracts (Fig. 4f). Moreover, E2F1 was specifically phosphorylated by CDK8 when complexes were incubated in kinase buffer (Fig. 4g). Thus, CDK8 physically interacts with E2F1 and is a conserved negative regulator of E2F1-dependent transcription.

The interaction between CDK8 and E2F1 is particularly notable, as a concurrent study has found significant CDK8 copy number gains in colorectal cancers: approximately 40% of tumours have copy number gains in both CDK8 and RB1 (ref. 1). Chromatin immunoprecipitation (ChIP) experiments on SW480 colorectal cancer cells confirmed that CDK8 and E2F1 are both present at E2F-regulated promoters as well as the c-MYC promoter (Supplementary Fig. 8), suggesting an interplay between E2F1, CDK8 and β-catenin/TCF. In Saos2 cells, which contain low basal Wnt activity, expression of CDK8 enhanced β-catenin activation from the pTopFLASH reporter, whereas expression of kinase-dead CDK8 (CDK8KD) had no effect (Fig. 4h). Moreover, expression of CDK8, but not CDK8KD, suppressed the inhibitory effect of E2F1 on the pTopFLASH reporter (Fig. 4h). Similar results were observed for CDK8 expression in HCT116 colorectal cancer cells (Supplementary Fig. 8). Hence, CDK8 and E2F1 have antagonistic effects on β-catenin-mediated transcription, and increasing the levels of CDK8 protects β-catenin from inhibition by E2F1. In agreement with this, the expression of CDK8KD reduced pTopFLASH activity in APC-deficient SW480 colorectal cancer cells, and enhanced the inhibition caused by short-hairpin targeting of pRB (Fig. 4i).

Cancer cells acquire multiple mutations during tumorigenesis, and it is a major challenge to explain how each change contributes to malignancy. However, the absence of mutations can also give new insights. It has long been known that colorectal tumour cells fail to mutate RB1 and typically express elevated levels of this tumour suppressor. The discovery that E2F1 is a potent inhibitor of β-catenin-dependent transcription provides an unexpected and simple explanation to this conundrum. This interaction may also explain why colorectal tumours frequently overexpress c-myc-induced microRNAs that target E2F1 (refs 25 and 26). The discovery by Firestein and colleagues1 showing significant RB1 and CDK8 copy number gains in colorectal cancers is especially intriguing given the evidence that CDK8 is an important modulator of both β-catenin and E2F1. Whereas CDK8 enhances the activity of β-catenin, it represses the activity of E2F1. Consequently, the amplification of CDK8 may act as a switch, allowing increased β-catenin-dependent transcription that is also resistant to E2F1 inhibition (see model in Supplementary Fig. 1). Reversing this process, such as the inhibition of CDK8 combined with the activation of E2F1, may be useful as a two-pronged strategy to target cancer cells that are driven by deregulated β-catenin activity.

Methods Summary

Unless otherwise noted, all fly crosses were conducted at 25 °C and phenotypes are depicted in female progeny. Transgenic dE2f1-dsRNA (dE2f1RNAi ) flies were created using a system developed previously27. SW480 (A. Burgess), DLD1 (W. Hahn), U2OS-shRb (S. Lowe), and Saos2-TR-E2F1 cells2 were cultured in DMEM supplemented with 10% fetal bovine serum. Drosophila Schneider line 2 (SL2) cells were maintained as previously described2. Transient transfections were performed using Fugene-6 reagent (Roche), and in some cases with CellFectin (Invitrogen), according to the manufacturer’s instructions. For SW480 survival and DLD1 quantitative PCR (qPCR) experiments, high-efficiency (approximately 70%) gene transfer was accomplished by using Amaxa nucleofection according to the manufacturer’s protocol (Program T-020; Kit-T and Kit-L, respectively). All Drosophila RNAi in SL2 cells was performed as described2, using 50-µg double-stranded RNA (dsRNA) synthesized with T7 RiboMax (Promega) with all conditions normalized using luciferase-dsRNA. Luciferase reporter assays were performed as previously described for SL2 cells2 and mammalian cells28. MTT viability assay was performed as previously described2. Western blot and immunohistochemical analysis was performed using standard techniques. Dissected third-instar larval disc immunohistochemistry was performed using anti-dE2F1 antibody (T. Orr-Weaver). Immunohistochemical staining of ApcMin mouse tumours was performed as described19 using anti-Rb (Santa Cruz, sc-50) and β-catenin antibodies (BD Biosciences, 610054). ChIP and data analysis were performed as previously described29. Gel shift assays were performed as described30. Detailed information on antibodies, fly stocks and plasmids, as well as qPCR, GST-pulldown assay, co-immunoprecipitation and IP-kinase assay conditions, is given in Methods.

Online Methods

Fly stocks, transgenes and genetic crosses

The following stocks were used for these studies: GMR-Gal4, ptc-Gal4, UAS-arm (S2; 4783); UAS-dTCF/pan (4837); UAS-dTCFΔN/pan (4784); UAS-GSK3/sgg (5361) (Bloomington Stock Center); GMR-arm* (Y55; M. Bienz); dCdk8K185 (null) and dCycCY5 (null) (H.-M. Bourbon); GMR-wIR (R. Carthew); and PCNA-GFP (B. Duronio). The dE2f1RNAi primer sequence (sub-cloned inverted into the pWIZ vector) was 5′-TTATTTCAAACGCCCTACCG-3′ and 5′-GAATTGCATCTGCAGTGAGC-3′ (verified by sequencing). All transgenic fly embryo injections were performed by the MGH Cutaneous Biology Research Center Transgenic Fly Core. Approximately 30 different transgenic lines carrying one or multiple transgenes were balanced and recombined with different Gal4 lines using standard Drosophila genetic methods. Because the dE2f1RNAi phenotypes in both the Drosophila eye (GMR-Gal4,UAS-dE2f1RNAi #10) and wing (ptc-Gal4,UAS-dE2f1RNAi #3) could be modified by known factors that interact with the RB-E2F pathway (J.-Y.J. and N.J.D., unpublished observations), we designed and performed a dominant modifier genetic screen. We screened approximately 6,500 piggyBac transposon insertion lines of the Exelixis mutant collection (J.-Y.J., A.H. and N.J.D., unpublished results) that was maintained at the Harvard Drosophila Collection (S. Artavanis-Tsakonas). A detailed description of the genetic screen will be presented elsewhere.

Cell culture and gene transfer

Plasmids used included pCMV-E2F1, pCMV-DP1, pCMV-E2F2, pCMV-E2F3, pCMV-E2F4, pCMV-E2F1ΔX, pCMV-E2F1-121, pCMV-E2F1-143, pCMV-E2F1-144, pGST–E2F1, pGST–E2F4, pGST–DP1 (K. Helin); pSFFV-p53 (S. Korsmeyer); pCMV-neoBam, pCMV-β-gal4, pCMV-pRB, pE2F4B (F. Dick); pCMV-DNDP1 (Δ103–126) (M. Classon); 8×SuperTop, 8×SuperFop, pCMV-CDK8, pCMV-CDK8KD, pLKO-shCDK8-1, pLKO-shGFP (W. Hahn); pCMV-E2F1/3 chimaeras (J. Nevins); pcDNA3-DNTCF1, pcDNA3-DNTCF4 (H. Clevers); pEVR-TCF1E (M. Waterman); pGL3OT, pGL3OF (E. Hay); pEGFP-N1 (BD Biosciences); pLPC-empty, pLPC-TAp73, pClneo-Δ45-β-catenin, pClneo-S33Y-β-catenin, pClneo-β-catenin, LLP-shGFP and LLP-shRbCD (targeting sequence GGTTGTGTCGAAATTGGATCA) (J. Rocco). Inhibitors and chemical reagents include LiCl GSK3 inhibitor (Fisher, 121-500), SB216763 GSK3-β inhibitor (Sigma, S3442), BOC-Aspartyl-FMK caspase inhibitor (Enzyme Systems Products, FK-011), MG101 (Sigma, A6185) and MG132 (Sigma, C2211).

Luciferase reporter and EGFP viability assays

For pTopFLASH assays, Saos2 cells were transfected in six-well plates with 100–200 ng pTopFLASH reporter plus 100 ng S33Y-β-catenin expression construct, along with 100 ng pCMV-β-gal as normalization control after β-gal assay. Unless otherwise specified, luciferase assays were performed 48 h after transfection (data are expressed as mean ± s.d., n = 3). The S33Y-β-catenin construct was omitted for pTopFLASH assays in colorectal cancer cells. For EGFP co-transfection viability assay, 100 ng of pEGFP-N1 was co-transfected with the indicated plasmids. Whole-cell lysates were prepared with GFP homogenizing buffer and assayed fluorometrically 48 h after transfection (excitation/emission λ = 488/511 nm). Tet-induced survival experiments were done under low serum (0.5%) conditions.

Antibodies for western and immunohistochemical analyses

Other antibodies used included dE2F1 (polyclonal anti-rabbit, C. Seum), E2F1 (Santa Cruz, sc-193), E2F4 (Santa Cruz, sc-1082), DP1 (Santa Cruz, sc-610), CRSP70 (Santa Cruz, sc-9426), c-myc (Santa Cruz, sc-40), TCF1 (Santa Cruz, sc-8589), TCF4 (Santa Cruz, sc-8632; Upstate, 05-511 for ChIP), Cyclin E (Santa Cruz, sc-247), Cyclin A (Santa Cruz, sc-596), pan-MAPK (BD Biosciences, 612641), β-catenin (Cell Signaling, 9562), CDK8 (Santa Cruz, sc-1521), pRB (Santa Cruz, sc-50), Arm (N27A1; Developmental Studies Hybridoma Bank), β1-integrin (BD Biosciences, 610467), E-cadherin (BD Biosciences, 610181), N-cadherin (BD Biosciences, 610920), α-catenin (BD Biosciences 610193), FAK (BD Biosciences 610087), phospho-FAK Tyr397 (BD Biosciences, 611806), GSK3 (Cell Signaling, 9315), phospho-GSK3 (Upstate, 05-413), Src (Cell Signaling, 2108), phospho-Src Tyr416 (Cell Signaling, 2101), STAT3 (Cell Signaling, 9132), phospho-STAT3 Tyr705 (Cell Signaling, 9131), phospho-STAT3 Ser727 (Cell Signaling, 9134), STAT5 (Cell Signaling, 9310), phospho-STAT5 Tyr694 (Cell Signaling, 9356), SHC (BD Biosciences, 610081), anti-HA-epitope (Sigma, H6908), and GFP (Sigma, G1544).

Real-time quantitative PCR

Total RNA was prepared using RNeasy Extraction Kit (Qiagen). Reverse transcription PCR (RT–PCR) was performed using Taq Man Reverse Transcription (PE Applied Biosystems) according to the manufacturer’s specification. RT–PCR was performed using an ABI prism 7900 HD Sequence Detection system. Relative mRNA levels were determined using the SYBR Green I detection chemistry system (Applied Biosystems). Quantification was performed using the comparative CT method as described in the manufacturer’s manual, and GAPDH or 18S rRNA was used as normalization control. All primers were designed with Primer Express 1.0 software (Applied Biosystems) following the manufacturer’s suggested conditions.

Forward and reverse primer sequences

These included: E2F1-f (CATCCCTCACCACAGATCCC), E2F1-r (AACAGCGGTTCTTGCTCCAG); c-MYC-f (CGGATTCTCTGCTCTCCTCG), c-MYC-r (CCACAGAAACAACATCGATTTCTT); AXIN1-f (ACAGCATCGTTGTGGCGTACT), AXIN1-r (CACAGTCAAACTCGTCGCTCAC); AXIN2-f (ATTCGGCCACTGTTCAGACG), AXIN2-r (GACAACCAACTCACTGGCCTG); SIAH1-f (ATGGTCATAGGCGACGATTGAC), SIAH1-r (AAGCTGTGCAATGCTGGTGTC); CD44-f (CTGCAAGGCTTTCAATAGCACC), CD44-r (CGTGCCCTTCTATGAACCCAT); PPARδ-f (CTTCCACTACGGTGTTCATGCA), PPARδ-r (GCACTTGTTGCGGTTCTTCTTC); CCNE1-f (TGCAGAGCTGTTGGATCTCTGTG), CCNE1-r (GGCCGAAGCAGCAAGTATACC); CDC25-f (TGAATACGAGGGAGGCCACATCAA), CDC25-r (ACACGCTTGCCATCAGTAGGTACA); TP73-f (CGTACTCCCCGCTCTTGAAG), TP73-r (TCCGCTTTCTTGTAAACAGGC); GAPDH-f (CATGTTCGTCATGGGTGTGAACCA), GAPDH-r (GTGATGGCATGGACTGTGGTCAT).

Chromatin immunoprecipitation

Chromatin extracts were incubated at 4 °C overnight with antibodies specific for E2F1, CDK8 and anti-HA antibody as control. Immunocomplexes were recovered with protein A and G Sepharose beads. DNA was recovered and dissolved in 150 µl of water. RT–qPCR was performed as described above.

Forward and reverse primer sequences

These included: P107-f (GGCCAAGGACAGGTCTTTCAG), P107-r (AAAGACGCCCAGAGATGCAGC); CCNA1-f (CTGCTCAGTTTCCTTTGGTTTAC), CCNA1-r (AAAGACGCCCAGAGATGCAGC); ACYLCOA-f (CCTTCATTGGGATCACCACG), ACYLCOA-r (GGAGATGAGTACCAGCAGGTTG); MYC-EBE-f (TCCGCCTGCGATGATTTATAC), MYC-EBE-r (CAGAGTAAGAGAGCCGCATGAA); MYC-TBE1-f (TCTCCCGTCTAGCACCTTTGA), MYC-TBE1-r (CACGGAGTTCCCAATTTCTCAG); MYC-TBE2-f (GCTCTCCAAGTATACGTGGCAA), MYC-TBE2-r (TCAGAGCGTGGGATGTTAGTGTA).

GST-pulldown assay

GST–fusion proteins were expressed in Escherichia coli BL21 cells and purified from the lysates by glutathione sepharose beads (Pharmacia) with extensive washing. The amount of bead-bound GST protein was determined by SDS–PAGE and Coomassie blue staining and normalized. The dE2F1 fragments, shown in Supplementary Fig. 7, were subcloned into pGEX-2TKN vector. To make Myc-tagged dE2F1 and HA-tagged dCDK8, full-length cDNAs were subcloned into pcDNA4TO or pcDNA3HA, respectively. Each construct was expressed in 293T cells and extracted in binding buffer (20 mM HEPES pH 7.6, 140 mM NaCl, 0.1 mM EDTA, 10% glycerol, 1 mM dithiothreitol (DTT), 1 mM benzamidine, 0.2 mM phenylmethylsulphonyl fluoride (PMSF), and 1 µg ml-1 aprotinin) with 0.5% NP-40. The whole-cell extracts were diluted fourfold with binding buffer without detergent. S2 cell extracts were prepared similarly. Whole-cell extract was applied to 50 µl of GST–fusion protein beads and incubated at 4 °C for 3 h. Beads were washed seven times with 1 ml wash buffer containing 50 mM Tris-HCl at pH 8.0, 250 mM KCl, 0.1 mM EDTA, 10% glycerol, 0.1% NP-40, 1 mM DTT, 1 mM benzamidine, 0.2 mM PMSF and 1 µg ml-1 aprotinin. The interacting proteins were then eluted with two bead volumes of binding buffer plus 0.3% sarkosyl at 4 °C for 1 h.

Co-immunoprecipitation

Saos2-TR-E2F1 cells were collected and rinsed with PBS after culture in the presence of 0.2 µg ml-1 of tetracycline for 12 h. Cell pellets were resuspended in immunoprecipitation buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.1 mM EDTA, 10% glycerol, 0.1% NP-40, 1 mM DTT, 0.2 mM PMSF, 1 mM benzamidine and 1 µg ml-1 aprotinin). Whole-cell extract was prepared by passing through 20(1/2)G needle for five times followed by incubation on ice for 15 min and centrifugation at 12,000g for 20 min. For immunoprecipitation, whole-cell extracts were incubated with 30 µl of protein-G beads pre-coated with antibodies against CDK8 (sc-1521), E2F1 (sc-193), control normal rabbit serum (IgG) or CRSP70 (goat anti-human CRSP70, Santa Cruz sc-9426). After rotating at 4 °C for 2 h, the beads were washed five times with 1 ml immunoprecipitation buffer. The antigen-associated proteins were eluted with 0.3% sarkosyl in immunoprecipitation buffer, resolved on 10% SDS–PAGE, and visualized using standard western analysis.

Immunoprecipitation-kinase assay

CDK8 or control (CRSP70) proteins were immunoprecipitated from whole-cell extracts of Saos2-TR-E2F1 cells as described in the co-immunoprecipitation section. The beads were washed twice with 1 ml immunoprecipitation buffer followed by two washes with 1 ml kinase assay buffer (50 mM Tris-HCl pH 7.4, 50 mM NaCl, 10 mM MgCl2, 1 mM MnCl2, 10% glycerol, 0.2 mM PMSF, 5 mM DTT and 1 µg ml-1 aprotinin). For kinase assays, the resulting beads were incubated in 50 µl of kinase assay buffer containing 20 µCi of [γ-32P]ATP at room temperature for 30 min. The reaction was terminated and the kinase-associated proteins were eluted with 0.3% sarkosyl in immunoprecipitation buffer. After dilution ten times with immunoprecipitation buffer, an equal amount of the supernatant was applied to protein-A beads pre-coated with antibodies against IgG (normal rabbit serum), E2F1 (sc-193), E2F4 (sc-1082) or DP1 (sc-610). The mixture was rocked at 4 °C for 2 h and the beads were washed four times with 1 ml immunoprecipitation buffer. The samples were boiled, resolved in 12% SDS–PAGE and then transferred to a PVDF membrane. The in vitro phosphorylated proteins were visualized by autoradiography.

Electrophoretic mobility gel shift assay

Briefly, 2 μg of nuclear extracts were incubated with 32P-labelled double-stranded oligonucleotides containing E2F or TCF sites site (less than 10 pg) for 30 min at 4 °C in binding buffer. For supershift experiments, antibodies were pre-incubated for 30 min. Samples were then loaded on a 4% polyacrylamide gel and visualized by autoradiography.

Double-strand oligonucleotide probe sequences

These included: MYC-wtE2F (5′-GCTTCTCAGAGGCTTGGCGGGAAAAAGAA); MYC-mtE2F (5′-GCTTCTCAGAGGCTTGTAGGGAAAAAGAA); MYC-TBE1 (5′-GTCTAGCACCTTTGATTTCTCCCAAACCC); MYC-TBE2 (5′-CTGGGACTCTTGATCAAAGCGCGGCCCTT); TOP (5′-GCCCCCTTTGATCTTACCCCCTTTGATCT); Canonical wtE2F (5′-ATTTAAGTTTCGCGCCCTTTCTCAAATTT); Canonical mtE2F (5′-ATTTAAGTTTCGATCCCTTTCTCAAATTT).