The Ability of E1A to Rescue ras-Induced Premature Senescence and Confer Transformation Relies on Inactivation of Both p300/CBP and Rb Family Proteins

Aberrant activation of ras oncogenes is a crucial step in tumorigenesis (1, 2). Constitutive activation of ras, either through point mutations or overexpression, is associated with a wide variety of human tumors at high frequency (3, 4). Despite the fact that activated ras genes are oncogenic, the outcome of ras activation alone is somewhat different in primary, normal cells. Instead of promoting proliferation and transformation, activated ras induces a relatively stable form of growth arrest known as premature senescence in primary fibroblasts, keratinocytes, and endothelial cells (5–7), accompanied by the accumulation of different combinations of cell cycle inhibitor proteins, such as p53, p21^WAF1, p16^INK4A, and p14/p19^ARF, decreased expression of cyclin A, and reduced kinase activity of CDK2 (5).

It is conceivable that ras-induced premature senescence serves as a defensive mechanism against oncogenic transformation. Therefore, cellular transformation by oncogenic ras requires additional, cooperating genetic alterations that bypass ras-induced senescence. A number of genetic changes have been reported that allow bypass of ras-induced senescence in rodent cells. These changes include inactivation of Rb family proteins (8, 9), loss of p53 or p19^ARF functions (10, 11), defects in transforming growth factor (TGF)-β signaling (6), and acquisition of viral oncogenes such as adenoviral E1A (5). These findings, at least in part, provide an explanation to the observation that ras needs to cooperate with other genetic alterations to transform primary rodent cells (1, 12). In human cells, however, the mechanism that allows the bypass of ras-induced premature senescence remains largely unclear despite the fact that viral oncoproteins such as the adenoviral E1A-12S protein can render primary human fibroblasts insensitive to ras-induced senescence and promote ras-mediated oncogenic transformation (5, 13, 14).

E1A is a multifunctional oncoprotein that interferes with the functions of many cellular proteins through direct interaction (15). The NH₂ terminus of E1A seems to confer its transforming activities by modulating the functions of cellular growth-regulatory proteins, such as the Rb family proteins (pRb, p107 and p130), p400, and p300/CBP transcriptional coactivators. By binding to the Rb family proteins, E1A overrides the negative growth-regulatory activities of these proteins by releasing the transcription factor E2F, which in turn induces the expression of growth-promoting genes (16, 17). Interaction between E1A and p300/CBP leads to inhibition of the intrinsic histone acetyl transferase activity of p300/CBP (18). Interaction with p400, a SWI2/SNF2-related protein, also seems to be essential for E1A-mediated transformation (19). In contrast, the COOH terminus of E1A is not only dispensable for E1A-mediated transformation (14, 20), but also negatively modulates the oncogenic activities of the NH₂ terminus and sensitizes cells to apoptosis (15, 21). The COOH-terminal region of E1A interferes with the functions of a transcriptional corepressor CtBP that regulates apoptosis (22) and the dual-specificity Yak1-related kinases (DYRK) involved in tumor metastasis (23). Deletion of the COOH terminus of E1A leads to enhanced transformation and tumor metastasis (20, 24, 25).

Despite the well-known ability of E1A to interact with multiple cellular proteins involved in growth control and transformation, the mechanism by which it blocks ras-induced senescence is unclear. In this study, we analyzed the E1A activities that were essential for rescuing ras-induced senescence in primary human fibroblasts. Our results have indicated that E1A-mediated bypass of ras-induced senescence relies on multiple E1A activities, including those that interfere with the functions of the Rb family members and p300/CBP proteins. These findings have shed light on those genetic alterations that overcome the senescence response and thus cooperate with ras to transform human cells, and have revealed a novel role of p300 and CBP in the senescence pathway.

Cell culture. BJ human foreskin fibroblasts (from Dr. J. Smith) were maintained in MEM with 10% FCS, nonessential amino acids, and glutamine. LinX-A retroviral packaging cells, WI38, and IMR90 were grown in DMEM with 10% FCS and glutamine.

Retroviral vectors and their transduction. WZLhyg- and Babepuro-HaRasV12 (5, 13), and retroviral vectors expressing full-length wild-type and mutant adenovirus 5 E1A (26), were obtained from Dr. Scott Lowe (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY). DNA fragments encoding truncated, wild-type, or mutant E1A1-143 were amplified by PCR and subcloned into Babepuro or WZLneo between BamHI and EcoRI sites. LXSN-E7 was from Dr. Galloway (27).

Retroviral vectors expressing the pRb small interfering RNA (siRNA) and p16^INK4A siRNA were gifts from Drs. Scott Lowe and Greg Hannon (Cold Spring Harbor Laboratory; ref. 28). Sequences for the effective p300 siRNA (TGACACAGGCAGGCTTGAC) and CBP siRNA (TAGTAACTCTGGCCATAGC) were provided by Drs. Matt Thayer and Daniel Stauffer (Oregon Health & Science University, Portland, OR). Oligonucleotides that represent small hairpin RNAs targeting these sequences were designed and cloned into pRetroSUPER (obtained from Dr. Rene Bernards) according to a published protocol (29). To construct siRNAs for p130 and p107, oligonucleotides for small hairpin RNAs targeting several regions of each gene were designed and cloned into pRetroSUPER, transduced into BJ cells, and validated by Western blot analysis. The constructs that most efficiently knocked down the protein levels were selected for p107 (CAGCCTAGAGGGAGAAGTT) and p130 (GATAAGTCCTTCCAGAACA).

Retrovirus-mediated gene transduction was carried out as previously described using an amphotropic packaging cell line (LinX-A; ref. 30). We typically achieved 20% to 30% infection rates in BJ cells. Transduced cells were purified with 100 μg/mL hygromycin B, 400 μg/mL G418, 5 μg/mL blasticidin, and/or 1 μg/mL puromycin.

Western blot. Cells were lysed in radioimmunoprecipitation assay buffer [1% Triton X-100, 0.5% deoxycholate, 0.1% SDS, 1 mmol/L sodium vanadate, and Complete protease inhibitor cocktail (Roche) in PBS]. Protein concentrations were determined by Bradford assays. Twenty to 80 μg of proteins were separated on 6%, 8%, 10%, or 12% SDS-PAGE gel and transferred to nitrocellulose membranes. The primary antibodies used were from Santa Cruz (Ha-Ras C-20, p130 C-20, p107 C-18, and CBP C-1), Upstate Biotechnology (p300 CT, RW128), PharMingen (Rb G3-245), and Sigma (actin). The monoclonal antibody against the NH₂ terminus of E1A (M58) was a gift from Dr. Nick Dyson (Massachusetts General Hospital, Boston, MA). Reactive proteins were visualized using enhanced chemiluminescence. The chemiluminescence signals were captured by the FluorChem-8900 Imaging System (AlphaInnotech).

Immunoprecipitation-Western blot. Two 10 cm plates of BJ cells (PD28-30) transduced with hemagglutinin (HA)-tagged or untagged E1A1-143, E1A1-143-27/124, E1A1-143-RG2, or their vector control were grown to 40% to 50% confluence, washed with PBS, and lysed with 1 mL of NETN buffer [20 mmol/L Tris-HCl (pH 8.0), 1 mmol/L EDTA, 150 mmol/L NaCl, 0.5% NP40, 2 mmol/L sodium vanadate, and Complete protease inhibitor cocktail from Roche]. The lysates were sonicated on ice until the lysates were clear. After centrifugation at 14,000 × g for 10 minutes at 4°C, the supernatants were transferred to fresh tubes and their protein concentrations were determined by Bradford assay. For immunoprecipitation, 100 μL of each lysate was incubated with 80 to 100 μL of a monoclonal antibody against HA (12CA5; for HA-tagged proteins) or E1A (M58; for untagged proteins) and 20 μL (beads volume) of a 1:1 mixture of protein A-Sepharose and protein G-Sepharose beads with rocking at 4°C for overnight. The beads were washed four times with NETN. After the final washed, 100 μL of NETN buffer was left behind and 100 μL of 2× Lammeli buffer was added. The samples were heated at 95°C for 5 minutes and 30 μL of each sample was separated by SDS-PAGE for Western blot analysis.

Analysis of senescence. Cells infected with ras or control viruses were selected with proper drugs for 4 days and analyzed for senescence. The day when drug selection was completed (6 days after infection) was defined as day 0. For growth curves, cells were plated in 12-well plates at 10⁴/well in duplicates or triplicates. Every 3 to 4 days, cells were harvested and counted. At each split, 10⁴ cells were reseeded to fresh plates, and allowed to grow until next split. Population doublings (PD) were calculated with the formula PD = log(N₂ / N₁)/log₂, where N₁ is the number of cells seeded and N₂ is that of recovered (31). Cells with senescence-associated β-galactosidase activity were detected as previously described (5). At least 200 cells were counted in randomly chosen fields for each sample.

Apoptotic assays. The rate of apoptosis was measured 6 days postinfection (after drug selection was completed) by two assays (32). In the first, cells with <1N (sub-G₁) DNA contents were identified by flow cytometry following propidium iodide staining. BJ cells were seeded at a density of 8 × 10⁴ cells/10 cm dish. After 48 hours, cells were harvested by trypsinization, fixed in 70% ethanol, permeabilized with 2 N HCl/0.5% Triton X-100, neutralized with 0.1 mol/L Na₂B₄O₇, stained with 5 μg/mL propidium iodide, and analyzed by flow cytometry. In the second, apoptotic cells with condensed chromatins and broken nuclei were identified by staining with Hoechst 33342 and propidium iodide. Cells were seeded at a density of 2 × 10⁴/well in a six-well plate, and after 48 hours were stained with 5 μg/mL Hoechst 33342 and 2 μg/mL propidium iodide for 2 hours. Percentage of apoptotic cells, identified by condensed, bright blue chromatins or pink chromatins (indicating increased propidium iodide uptake due to breakdown of nuclear membranes in late-stage apoptotic cells), was determined by counting >200 cells under an UV microscope in triplicates.

p53-dependent luciferase reporter assays. To create retroviral p53-dependent reporter PG-Luc or its non–p53-binding control MG-Luc, a fragment containing 13 copies of a wild-type or mutant p53-binding site and a TATA box was excised from PG13-Luc or MG13-Luc (from Dr. Bert Vogelstein; ref. 33) and ligated together with a luciferse gene from pGL3 (Promega) into pBabeBlast-SIN vector in which the promoter region in 3′ long terminal repeat (LTR) had been deleted. BJ cells were transduced with PG-Luc or MG-Luc at PD30; selected with blasticidin; and cotransduced with E1A1-143, E1A1-143-RG2, p300 siRNA, CBP siRNA, or their vector controls (pWZLneo for E1A and pRetroSUPERneo for siRNAs) and Ha-RasV12 or its vector control (WZLhyg) at PD32. After selection with G418 and hygromycin, cells were split into 12-well plates on day 7 postinfection and lysed on day 8. Luciferase activity was determined and normalized to protein concentrations as measured by Bradford assay. Fold induction by ras was calculated by dividing the activity from ras cells by that from vector control cells.

Transformation assays. BJ cells were transduced with p16 siRNA. CBP siRNA, p300 siRNA, E1A1-143-47/124 or their vector control (pRetroSUPERneo or WZLneo) at PD27, with HygMaRXII-MDM2 (30) at PD31, and with E1A1-143WT, E1A1-143-RG2, E1A1-143-47/124 or their vector control (pBabepuro), and Ha-RasV12 or its vector control MSCVblast at PD33. Infected cells were subjected to transformation assays after selection with appropriate drugs.

For anchorage-independent growth assays, 10⁴ cells at PD35 were resuspended in 3 mL of growth medium containing 0.3% low-melting temperature agarose (SeaPlaque GTG, FMC, Rockland, ME) and plated onto 4 mL of solidified bottom layer medium containing 0.5% low-melting agarose in 6 cm plates in duplicates. Cultures were fed once weekly. Colonies were stained with 0.02% Giemsa and counted 3 weeks after seeding.

For tumor formation assays, 2 × 10⁶ cells at PD38 were resuspended in 100 μL of serum-free RPMI 1640 and injected s.c. into 6-week-old female HSD:athymic nude mice (Rodent Breeding Colony, The Scripps Research Institute). Tumor growth was monitored once weekly for 7 weeks.

Senescence bypassing activity resides in the NH₂ terminus of E1A. To understand the mechanism by which ras-induced premature senescence is bypassed during cellular transformation, we analyzed functional domains and activities of E1A involved in the senescence bypass in primary BJ human fibroblasts. Expression of full-length E1A alone in BJ cells greatly inhibited cell growth (Fig. 1A). In fact, we failed to establish a cell line stably expressing E1A. The growth inhibition by E1A was at least partly due to apoptosis. When compared with the control cells in a cell cycle analysis done 6 days after transduction, the E1A-expressing population contained a significantly higher percentage of cells with sub-G₁ DNA contents (Fig. 1B,, top table). E1A expression also increased percentage of apoptotic cells characterized by condensed chromatins and increased propidium iodide uptake due to breakdown of nuclear membranes in late-stage apoptosis (Fig. 1B,, bottom). Cells transduced with Ha-RasV12 showed growth inhibition (Fig. 1A) and a G₁ arrest (Fig. 1B,, top table) and displayed senescence-associated β-galactosidase activity (Fig. 3C) consistent with premature senescence. E1A-induced growth inhibition and apoptosis were essentially prevented when E1A was cotransduced into BJ cells with Ha-RasV12 (Fig. 1A and B), confirming a previous finding in rodent cells that oncogenic ras rescued E1A-induced apoptosis (34). In addition, premature senescence induced by Ha-RasV12 was also blocked by E1A (Fig. 1A and B). As a result, cells transduced with both E1A and ras proliferated rapidly, despite the fact that expression of E1A or ras alone led to growth inhibition.

Previous studies have shown that the NH₂ terminus of E1A is sufficient to transform primary rodent and human cells in cooperation with ras (14, 20). Therefore, we investigated whether the COOH-terminal region of E1A was dispensable for the bypass of ras-induced senescence. A series of COOH-terminal deletion mutants of E1A were transduced into primary BJ fibroblasts together with oncogenic ras (Ha-rasV12) and analyzed for their ability to rescue premature senescence. Consistent with the reports that binding of E1A to CtBP sensitized cells to apoptosis and inhibited transformation by E1A (25, 35), deletion of the COOH-terminal 19 residues containing the CtBP-binding site essentially eliminated the growth inhibitory and proapoptotic effect of E1A (Fig. 1A,, bottom). Furthermore, deletion of the CtBP binding site did not affect the ability of E1A to rescue ras-induced senescence (Fig. 1A,, bottom). Thus, CtBP binding seems to be dispensable for bypassing ras-induced senescence. Further deletion analysis revealed that an E1A deletion mutant containing only residues 1 to 143 also failed to induce apoptosis (Fig. 1A and B), but was fully capable of rescuing ras-induced senescence. E1A1-143 rescued ras-induced growth inhibition (Fig. 1A,, top) and G₁ arrest (Fig. 1B,, top table) and prevented the induction of senescence-associated β-galactosidase by ras (Fig. 3C). These results indicate that the senescence bypassing activity resided in the NH₂-terminal 143-amino-acid residues of E1A and does not involve cellular proteins bound to the COOH-terminal half of E1A. The NH₂-terminal 143 residues of E1A are sufficient to bypass ras-induced premature senescence without inducing apoptosis.

Inactivation of Rb family proteins is not sufficient to rescue senescence. One of the major activities of the NH₂ terminus of E1A is to bind to and interfere with the functions of Rb family proteins, pRb, p107, and p130. To investigate the role of Rb inactivation by E1A, we examined whether inactivation of Rb proteins by other means was sufficient to confer resistance to ras-induced senescence. Like E1A, the human papillomavirus type-16 E7 protein binds to Rb family proteins pRb, p107, and p130, resulting in the release of E2F transcription factors and abrogation of the growth regulatory functions of Rb proteins (36). In addition to its ability to sequester Rb from E2F, E7 also enhances the degradation of the Rb proteins, thus further down-regulating their expression levels (36, 37). When transduced into BJ or WI38 fibroblasts, E7 failed to rescue ras-induced premature senescence despite its ability to substantially reduce the protein levels of all three Rb family members (Fig. 2B). BJ-E7 and WI38-E7 cells became growth arrested (Fig. 2A and Supplementary Fig. S2A) and displayed senescence-associated β-galactosidase activity (Supplementary Fig. S2B and data not shown) upon transduction of Ha-RasV12. This finding indicates that inactivation of the Rb family proteins is not sufficient to rescue oncogenic ras-induced senescence. Therefore, E1A-mediated senescence bypass must rely on additional activities of E1A.

Senescence bypass by E1A requires both Rb- and p300/CBP-binding but not p400-binding activities. To determine whether the NH₂-terminal activities of E1A are necessary for E1A to rescue premature senescence, we analyzed a series of E1A mutants that had selectively lost these activities. In addition to Rb family proteins, the NH₂-terminal 143 residues of E1A (E1A1-143) are sufficient to bind to other cellular proteins, such as p300/CBP and p400, that are critical for oncogenic transformation (15). Conserved region 2 (CR2, residues 120-140) and part of the conserved region 1 (CR1, residues 40-60) of E1A are required for interference with the Rb family proteins (38–40). A point mutant 47/127, containing Tyr⁴⁷γHis in CR1 and Cys¹²⁴γGly in CR2, has lost binding to all Rb proteins but retained p300/CBP binding (41). Binding of E1A to p300/CBP relies on the residues at the very NH₂ terminus (residues 1-25) and part of CR1 (residues 61-85) that seem to be dispensable for Rb binding (41, 42). Deletion of residues 2 to 11 (Δ2-11) and an Arg²γGly point mutation (RG2) in the NH₂ terminus, as well as Δ61 to 85 deletion in CR1, abolished or greatly reduced p300/CBP binding while having little effect on the binding to Rb, p107, and p130 (41). Binding of p400 relies on residues 26 to 35 and does not overlap with p300/CBP- or Rb-binding activities (19).

Because E1A1-143 is sufficient to confer resistance to ras-induced senescence without inducing apoptosis in BJ cells, we constructed E1A mutations that selectively disrupt binding to p400, Rb proteins, or p300/CBP in the context of E1A1-143. These mutants were transduced into BJ cells and tested for their ability to rescue ras-induced senescence. We initially focused on a small deletion mutant Δ26 to 35 that was defective in p400 binding (19) and two point mutants, 47/124 and RG2, that specifically disrupted binding to Rb family proteins and p300/CBP, respectively (41), because point mutations and small deletions were less likely to alter protein conformation or affect other functions of E1A. These mutant E1A proteins were expressed at comparable levels as the wild-type E1A1-143 when transduced into BJ cells (Fig. 3E) and did not induce cell death, as indicated by the lack of an increased population of cells with sub-G₁ DNA contents (Supplementary Table S1). Deletion of the p400-bindig site (Δ26-35) had no effect on the ability of E1A to bypass ras-induced senescence (Fig. 3A), suggesting the p400-binding activity was dispensable for senescence bypass. In contrast, E1A mutants carrying the 47/124 or RG2 mutation failed to rescue premature senescence in BJ cells (Fig. 3B and C). Upon ras transduction, cells expressing E1A-27/124 or E1A-RG2 underwent growth arrest (Fig. 3B) and displayed senescence-associated β-galactosidase activity (Fig. 3C). Both RG2 and 47/124 mutants also failed to rescue ras-induced senescence in two other primary human fibroblast cell lines, WI38 and IMR90 (Supplementary Fig. S1). Therefore, both Rb-binding and p300/CBP-binding activities are required for E1A to mediate bypass of ras-induced senescence. The Rb-binding defective 47/124 mutant and p300/CBP-binding defective RG2 mutant complemented each other in bypassing senescence when cotransduced into BJ (Fig. 3D) or WI38 cells (Supplementary Fig. S2), confirming that the inability of these mutants to confer senescence rescue was indeed due to loss of two separate activities and that the Rb-binding and p300/CBP-binding activities could cooperate in trans to rescue ras-induced senescence.

To verify that the 47/124 and RG2 mutants had selectively lost binding to Rb family proteins and p300/CBP, we examined the ability of these mutants to bind to the cellular proteins in vivo. The E1A complexes were immunoprecipitated from BJ cells with an anti-HA antibody that recognized the HA-tag fused to E1A proteins, and the presence of the cellular proteins were detected with their respective antibodies in a Western blot analysis (Fig. 4). Wild-type E1A1-143 bound to p300, CBP, and all three Rb family proteins (pRb, p107, and p130). However, the 47/127 mutant had lost binding to the Rb family proteins but not to p300 and CBP, whereas the RG2 mutant was unable to bind p300 and CBP, but retained the ability to bind to all Rb proteins. The same results were obtained with untagged E1A proteins isolated from cells using an E1A-specific antibody (M58; data not shown). These results confirmed that the 47/124 and RG2 mutants have lost Rb-binding and p300/CBP-binding, respectively. Thus, the inability of these to mutants to rescue premature senescence indicates that both activities are essential.

The requirement for the Rb- and p300/CBP-binding activities of E1A in senescence bypass was further investigated by analyzing additional E1A mutants that have been previously shown to selectively lose these activities (Supplementary Table S2). Again, all these mutants were constructed in the context of E1A1-143 that failed induce apoptosis. Most of these mutant E1A proteins were expressed at comparable levels as the wild-type E1A1-143 when transduced into BJ cells (Fig. 3E; data not shown). Our results showed that any mutant that was unable to bind to Rb proteins or to p300/CBP also failed to rescue ras-induced senescence. Among the mutants that were defective only in Rb binding, E1A1-120 lacked the CR2 domain that directly contact Rb proteins, and Δ48 to 60 and Δ37 to 68 lacked part of that CR1 domain essential for Rb binding (38–41). Mutants that were defective only in p300/CBP binding included Δ2 to 11 and Δ2 to 24, lacking the NH₂-terminal p300/CBP-binding region, and Δ61 to 85 lacking part of the CR1 that was essential for p300/CBP binding but not for Rb binding (41, 42). In addition, as anticipated, E1A mutants defective in binding to both p300/CBP and p400 (Δ2-36), to both Rb and p400 (Δ30-49), or to all these three proteins (Δ30-85) failed to mediate bypass of ras-induced senescence. Taken together, our data showed that the rescue of ras-induced senescence in primary human fibroblasts relies on at least two of the E1A activities, Rb binding and p300/CBP binding.

E7 and a p16^INK4A small interfering RNA complement the Rb-binding defective mutant of E1A to rescue ras-induced senescence. The requirement of the Rb-binding activity of E1A suggested that inactivation of Rb proteins might contribute to the rescue of ras-induced senescence in cooperation with other E1A activities, although it alone was not sufficient to bypass the senescence response. To test this notion, we examined whether inactivation of Rb with E7 or a siRNA for p16^INK4A could rescue ras-induced senescence in cooperation with the E1A-47/124 mutant that was defective in Rb binding but competent in p300/CBP binding.

Consistent with previous reports (36), expression of E7 in BJ cells led to reduction in the protein levels of pRb, p130, and p107 (Fig. 2B). Although E7 alone failed to block ras-induced senescence, it did so together with E1A-47/124 that was capable of p300/CBP-binding in both BJ (Fig. 5A and C; Supplementary Table S1) and WI38 (Supplementary Fig. S2) primary fibroblasts. p16^INK4A is up-regulated during ras-induced senescence and plays a critical role in mediating the growth arrest and senescence phenotypes (5, 43). Like E7, a p16^INK4A siRNA (28) that substantially reduced ras-induced p16^INK4A expression (Fig. 5B) also complemented the E1A-47/124 mutant in rescuing premature senescence when stably transduced into BJ cells (Fig. 5A and C). p16^INK4A mediates growth arrest by inhibiting the cyclin D–dependent protein kinases that phosphorylate the Rb family proteins and inactivate their growth-suppressing functions. It has been shown that growth inhibition by p16^INK4A requires not only pRb but also other Rb family proteins (44). Therefore, knocking down p16^INK4A expression by siRNA likely interferes with the functions of multiple Rb family members. These results have thus indicated that interference with the function of Rb family proteins by E7 or the p16 siRNA, together with additional activities provided by the E1A-27/124 mutant, leads to the rescue of ras-induced senescence.

Inactivation of individual Rb family members is not sufficient to complement the Rb-binding defective mutant of E1A in rescuing ras-induced senescence. The observation that inactivation of Rb family proteins by E7 or the p16^INK4A siRNA contributes to senescence bypass prompted us to investigate the role of each individual Rb family member in ras-induced senescence. To this end, we constructed siRNA for pRb (28), p130, and p107. When stably transduced into BJ cells via retroviruses, each one of these siRNAs specifically and effectively knocked down the level of each individual Rb family protein without affecting the levels of the others (Fig. 5B). However, none of these Rb protein siRNAs, either alone or in combination with E1A-27/124, rescued ras-induced premature senescence (Fig. 5A and C). This finding suggests that the rescue of ras-induced senescence relies on the ability of E1A to inactivate multiple Rb family proteins, or, alternatively, an Rb protein that has yet to be identified. Our finding is consistent with previous findings that in mouse embryonic fibroblasts, bypass of ras-induced senescence required inactivation of pRb and p107 or all three Rb proteins (8, 9).

A CBP small interfering RNA and a p300 small interfering RNA complement the p300/CBP-binding defective mutant of E1A to rescue ras-induced senescence. Mutagenesis analysis of E1A revealed that the p300/CBP-binding activity of E1A was essential for bypassing ras-induced premature senescence. To determine whether the requirement for the p300/CBP binding in E1A-mediated senescence bypass is due to the functional interference with p300/CBP proteins, we examined the consequence of p300/CBP inactivation on senescence. siRNAs for p300 and CBP were constructed that effectively knocked down the p300 and CBP levels, respectively, when stably transduced into BJ cells via retroviruses (Fig. 6C). Although the p300 siRNA or CBP siRNA alone had no effect, they both complemented the E1A-RG2 mutant to rescue ras-induced premature senescence in both BJ (Fig. 6A and B; Supplementary Table S1) and WI38 (Supplementary Fig. S2) fibroblasts. Therefore, inactivation of p300 or CBP contributes to the bypass of ras-induced senescence in cooperation with other activities present in the E1A-RG2 mutant.

p300 and CBP are essential for the induction of p53 activity by ras. p300 and CBP were initially identified as transcriptional coactivators for cyclic AMP–responsive element binding protein (18, 45) and were later found to be involved in the activities of a variety of transcription factors, including p53 (18, 46). It is possible that p53 activity is essential for the senescence induction in BJ cells and that inactivation of p300/CBP blocks p53 activity by depriving of its coactivators. To address this possibility, we analyzed the transcriptional activity of p53 using a p53-dependent luciferase reporter gene stably transduced into BJ cells via retroviral infection. A luciferase reporter cassette containing p53-binding sites was transduced into BJ cells with a self-inactivating retrovirus to create a stable p53-reporter cell line. The self-inactivating virus has a promoterless LTR, which prevents p53-independent transcription of luciferase from the LTR. When transduced into the p53-reporter cell line, Ha-RasV12 induced transcriptional activity of p53, and this induction was blocked by E1A1-143, the p300 siRNA, and the CBP siRNA (PG-Luc; Fig. 7). As a control, Ha-RasV12, E1A, or p300/CBP siRNAs had no effect on the transcription from a promoter containing mutant p53-binding sites (MG-Luc; Fig. 7) in the same assay. These results showed that both p300 and CBP were required for the induction of p53 activity by ras, suggesting that p300 and CBP might be transcriptional coactivators for p53 in the senescence pathway, and that inactivation of p300/CBP by either E1A or siRNA might contribute to senescence bypass by preventing p53 activation.

Inactivation of p300/CBP contributes to E1A and Ras-mediated transformation of primary human fibroblasts. The requirement of both Rb-binding and p300/CBP-binding activities in bypassing ras-induced senescence prompted us to investigate whether these activities contribute to ras-mediated cellular transformation. Primary BJ fibroblasts can be fully transformed by the combination of E1A1-143, Ha-RasV12, and MDM2 (14). We thus analyzed the behavior of the E1A mutants in this cellular transformation model. Whereas wild-type E1A1-143 in combination with MDM2 and Ha-RasV12 supported anchorage-independent growth and promoted tumor formation in athymic nude mice, the p300/CBP-binding defective RG2 mutant or Rb-binding defective 47/124 mutant failed to efficiently transform BJ cells (Fig. 8). When coexpressed in BJ cells, the RG2 and 47/124 mutants complemented each other in transformation (Fig. 8). These results suggest that both Rb-binding and p300/CBP-binding functions are essential, and that these activities can be provided in trans to induce transformation. Furthermore, inactivation of CBP or p300 by siRNA also complemented the RG2 mutant in transforming BJ cells in the presence of MDM2 and Ha-RasV12 (Fig. 8). The efficiency of transformation mediated by p300/CBP siRNA and E1A1-143RG2 was less robust compared with that induced by wild-type E1A1-143 or by E1A1-143-47/124 and E1A-RG2. Cells transduced with p300/CBP siRNA, E1A-RG2, MDM2, and Ha-RasV12 formed less soft agar colonies and formed tumors in only 50% of the injected nude mice (Fig. 8). This was most likely due to the incomplete inhibition of p300/CBP by siRNA although it cannot be ruled out that the E1A-47/124 mutant contained additional transforming activities besides inactivation of p300/CBP. Despite the lower transformation efficiency, our results clearly indicated that p300/CBP siRNA could at lease partially replace the p300/CBP-binding activity of E1A in transformation. Therefore, the p300/CBP-mediated senescence response is likely to be a key barrier in this transformation model, and inhibition of p300/CBP facilitates cellular transformation.

Interestingly, although the p16 siRNA complemented the 47/124 mutant in rescuing ras-induced senescence (Fig. 6A and B), it did not support cellular transformation together with the 47/124 mutant, MDM2, and HaRas-V12 (Fig. 8), indicating that the 47/124 mutant might have lost an additional p16/Rb-independent activity that was required for transformation but not for senescence rescue. Alternatively, the residual p16 level after siRNA knockdown may be sufficient to block transformation, but not to induce senescence.

Ras-induced premature senescence is generally considered as a tumor-suppressing defense mechanism in normal cells. As a result, additional genetic alterations are required to bypass the senescence response for oncogenic ras to cause transformation. A recent study indicated that ras did not induce senescence when expressed from its endogenous locus in mouse embryonic fibroblasts (47). Although a different ras gene, K-rasG12D, was used in their report, it raised a possibility that all oncogenic ras genes, including Ha-rasV12 used in our study, would confer transformation without inducing senescence in vivo. However, in mouse skin tumor models, although chemical carcinogens initially induced mutations in the endogenous Ha-ras gene, tumorigenesis did not occur without secondary mutations in genes that mediated ras-induced senescence, including p53, p21^WAF1, p16^INK4A, and p19^ARF (48, 49). This indicates that, at least in the cases of skin carcinogenesis, bypass of ras-induced senescence is essential for oncogenic transformation in vivo. It is likely that an attenuated form of senescence response may operate in vivo during tumorigenesis. Alternatively, oncogenic ras may induce senescence only in certain cell types or in combination with other stress signals present in the cellular microenvironment. Analysis of genetic events that regulate ras-induced senescence will help clarify the role of senescence in tumor suppression and provide insights into the mechanism of tumorigenesis.

Although viral oncogenes, such as E1A, have been shown to rescue ras-induced premature senescence, little is known about the cellular mutations that allow the senescence bypass. In this study, by dissecting the senescence bypassing activities of E1A, we have shown that genetic alterations leading to the disruption of the p16^INK4A/Rb pathways and those interfering with the function of p300/CBP proteins both contribute to the bypass of ras-induced senescence. These two types of genetic alterations (as achieved by E7 or the p16^INK4A siRNA and the p300 or CBP siRNA in our study), rescued ras-induced senescence in cooperation with the Rb-binding defective and the p300/CBP-binding defective mutants of E1A, respectively. However, neither one of these alterations alone rescued ras-induced senescence. These findings suggest that oncogenic ras triggers at least two parallel pathways that lead to premature senescence, with one engaging the p16^INK4A/Rb functions and the other requiring p300 and CBP. The ability of E1A to inactivate both of these two pathways allows it to bypass ras-induced senescence, leading to oncogenic transformation. Because these observations were made in multiple primary human fibroblast cell lines, inactivation of Rb proteins and p300/CBP may be part of a common mechanism for bypassing ras-induced senescence in human fibroblasts.

We have shown a novel role of p300/CBP proteins in the senescence response triggered by oncogenic ras. Inactivation of these proteins, together with additional activities present in the E1A-RG2 mutant, can override this response. Therefore, p300 and CBP play a critical role in the pathway that limits the oncogenic potential of ras. Previous studies in cancer patients and mouse models have suggested a tumor suppressing function of p300 and CBP (18, 50). Results from the present study suggest that loss of p300/CBP, in conjunction with other genetic alterations such as those inactivating the Rb pathway, can contribute to tumorigenesis by disrupting the antioncogenic defense response against ras and other oncogenes. Consistent with this notion, inhibition of p300 or CBP by siRNA cooperated with the E1A-RG2 mutant, MDM2 and Ha-RasV12 to transform primary BJ fibroblasts. These results have suggested that the p300/CBP-mediated senescence response may be a major barrier for cellular transformation, and, thus, have provided a mechanistic basis for the tumor suppressing function of p300 and CBP.

Although the precise role of p300/CBP in senescence awaits further investigation, our results show that p300 and CBP are essential for the induction of p53 activity by ras. This finding suggests a potential mechanism underlying the essential role of p300/CBP in senescence. p300/CBP may act as transcriptional coactivators for p53 in the senescence pathway, and inhibition of these proteins renders p53 inactive, thus leading to bypass of senescence.

Our study has revealed that the ability of E1A to bypass ras-induced senescence also relies on its binding to the Rb family proteins. Furthermore, although inactivation of the Rb pathway by E7 or a p16^INK4A siRNA alone is not sufficient to allow the senescence bypass, it did so in cooperation with the Rb-binding defective mutant of E1A. In contrast to E7 and the p16^INK4A siRNA that interfere with all the Rb family proteins, inhibition of individual Rb family members (pRb, p130, or p107) by siRNA failed to complement the Rb-binding defective E1A mutant to rescue ras-induced senescence. Thus, bypass of ras-induced senescence requires inactivation of multiple, but not a single, Rb protein. This finding mirrors two previous observations in mouse embryonic fibroblasts, which showed that simultaneous deletion of both Rb and p107 or all three Rb members rescued ras-induced senescence, whereas deletion of a single Rb protein was not sufficient to do so (8, 9). Therefore, in both human and mouse fibroblasts, different Rb family members seem to have redundant functions in the senescence pathway, and bypass of ras-induced senescence thus requires inactivation of multiple Rb proteins. On the other hand, our study also revealed a fundamental difference in the mechanism of ras-induced senescence between human fibroblasts and mouse embryonic fibroblasts. Whereas inactivation of multiple Rb members is sufficient to rescue ras-induced senescence in mouse embryonic fibroblasts, senescence bypass in human cells requires inactivation of Rb proteins as well as additional alterations such as loss of p300/CBP functions.

Grant support: NIH grants R01 CA91922 and CA106768, Department of Defense grant DAMD17-01-1-0390, and a scholarship from The Ellison Medical Foundation (AG-NS-0081-00; P. Sun), and NIH grant R01CA78343 (D. Tedesco) awarded to Dr. Steven I. Reed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Jim Smith for BJ primary human fibroblasts, Dr. Scott Lowe for full-length wild-type and mutant E1A vectors, Drs. Scott Lowe and Greg Hannon for pRb and p16 siRNA vectors, Dr. Nick Dyson for the antibody against E1A (M58), Dr. Bert Vogelstein for PG13-Luc and MG13-Luc reporter plasmids, Dr. Rene Bernards for pSUPERretro, Drs. Matt Thayer and Daniel Stauffer for sharing the sequences of effective p300 siRNA and CBP siRNA, and Ellen Fiss for administrative assistance.