ROC1/RBX1 E3 Ubiquitin Ligase Silencing Suppresses Tumor Cell Growth via Sequential Induction of G₂-M Arrest, Apoptosis, and Senescence

The SCF E3 ubiquitin ligases, consisting of Skp1, Cullins/Cdc53, F-box proteins, and the RING domain containing protein Regulator of Cullins-1 (ROC1)/Ring Box Protein-1 (RBX1; refs. 1–5), are crucial to the regulation of numerous cellular processes under both physiologic and pathologic conditions as part of the ubiquitin-proteosome system. These E3 ubiquitin ligases promote degradation of diverse substrates, including cell cycle regulatory proteins, transcription factors, and signal transducers. Importantly, SCF dysfunction can cause a variety of diseases including cancer (6, 7). For example, the oncogenic F-box protein Skp2 is overexpressed in human tumors, and promotes p27 degradation, contributing to malignant progression (8), whereas the tumor suppressor F-box protein FBW7, which degrades several proto-oncogenes (such as MYC, Notch, JUN, and Cyclin E), undergoes numerous cancer-associated mutations, and loss of FBW7 function causes chromosomal instability and tumorigenesis (9).

The core of SCF ubiquitin ligases is a complex of ROC1-cullins (7). ROC1 contains a RING-H2 finger domain (Cys42-X2-Cys45-X29-Cys75-X1-His77-X2-His80-X2-Cys83-X10-Cys94-X2-Asp97 in human), required for zinc ion binding and ubiquitin ligation (2, 10). Crystal structure studies revealed that ROC1 complexes with cullin-F-box proteins that recognize a variety of protein substrates and transfers ubiquitin from E2 to substrates for proteasome-targeted degradation (11). In yeast, ROC1 is required for ubiquitination of the cyclin-dependent kinase inhibitor Sic1 during the G₁-S cell cycle transition (5). ROC1 deletion causes yeast death, which can be rescued by human ROC1 or its family member, ROC2/SAG (2, 5, 12). In Caenorhabditis elegans, ROC1 is essential for cell cycle progression and chromosome metabolism. Depletion of ROC1 by siRNA causes pronounced defects in meiosis, mitotic chromosomal condensation and segregation, and cytokinesis (13). In Drosophila, ROC1a is required for cell proliferation and embryo development. Deletion of ROC1a results in animal death, which cannot be rescued by overexpression of ROC1b, indicating a nonredundant function between the family members (14). We recently reported that ROC1 disruption in mouse causes early embryonic lethality at E7.5 due to proliferation failure as a result of p27 accumulation, which can be partially rescued by simultaneous loss of p27 (15). Furthermore, a shRNA library–based functional genomic screen identified ROC1 as a growth essential gene in a number of human cell lines, although no further mechanistic characterization was performed (16).

In light of these studies showing the importance of ROC1 to cell growth and the known dysfunction of the SCF E3 ubiquitin ligases in a variety of cancers, we hypothesized that ROC1 overexpression is required for proliferation and survival of human cancer cells. We report here that ROC1 is indeed overexpressed in a number of solid human primary tumor tissues and many human cancer cell lines and that ROC1 siRNA silencing remarkably suppressed tumor cell growth via sequential induction of G₂-M arrest, apoptosis, and senescence, suggesting that ROC1 could serve as a potential anticancer target.

Cell culture. All the cancer cell lines used were from American Type Culture Collection and cultured in DMEM media containing 10% serum.

Immunohistochemistry staining of human tumor tissue array. Human tumor tissue arrays were provided and immunohistochemistry stained with affinity-purified ROC1-specific antibody made against COOH-terminal peptide of human ROC1 (15) by the University of Michigan Tissue Core, using the DakoCytomation EnVision+ System-HRP (DAB) detection kit.

Lentivirus-based siRNA and lentivirus infection. Lentivirus-based siRNA against ROC1 (LT-ROC1) and p53 (LT-p53) as well as LT-virus expressing scrambled control siRNA (LT-CONT) were constructed as described previously (17–19). The target sequences are as follows: LT-ROC1-01, 5′-AACTGTGCCATCTGCAGGAACCACATTTCAAGAGAATG TGGTTCCTGCAGATGGCACAGTTTTTTGT-3′; LT-ROC1-02, 5′-CTAGACAAAAAACTGTGCCATCTGCAGGAACCACATTCTCTTGAAATGTGGTTCCTGCAGATGGCACAGTT-3′; LT-p53-01, 5′-GACTCCAGTGGTAATCTACTTTCAAGAGAAGTAGATTACCACTGGAGTCTTTTTTGT-3′; LT-p53-02, 5′-CTAGACAAAAAAGACTCCAGTGGTAATCTACTTCTCTTGAAAGTAGATTACCACTGGAGTC-3′; LT-CONT-01, 5′-ATTGTATGCGATCGCAGACTTTTCAAGAGAAAGTCTGCGATCGCATACAATTTTTTGT-3′; and LT-CONT-02, 5′-CTAGACAAAAAATTGTATGCGATCGCAGACTTTCTCTTGAAAAGTCTGCGATCGCATACAAT-3′. A siRNA oliognucleotide specifically targeting ROC1 (siROC1, 5′-GACTTTCCCTGCTGTTACCTAA-3′; ref. 16), along with scrambled control siRNA (siCONT, 5′-ATTGTATGCGATCGCAGACTT-3′), were ordered from Dharmacon. A panel of human cancer cell lines was infected with LT-ROC1 or LT-CONT for 72 h, then split for assays as described below. For U87 cells, cells were infected either for 72 h and split for ATPlite cell proliferation assay and clonogenic survival assay, or for 120 h and subject to senescence-associated β-galactosidase (SA-β-gal) staining, fluorescence-activated cell sorting (FACS) analysis, and immunoblotting (IB) analysis.

ATPlite cell proliferation assay and clonogenic survival assay. Cells, infected with LT-ROC1 or LT-CONT, or transfected with siROC1 or siCONT, were split and seeded into 96-well plates with 3,000 cells per well in quadruplicate for ATPlite cell proliferation assay at various time points, or seeded into 6-well plates with 100 cells per well in triplicate, followed by incubation for 9 d. The colonies formed were fixed, stained, and counted (19).

Soft agar assay. Ten thousand cells after infection with LT-CONT or LT-ROC1 were seeded in 0.33% agar containing 1 × cell culture medium and 10% fetal bovine serum in 60-mm Petri dish, and grown at 37°C for 14 d. The cells were stained with p-iodonitrotetrazolium (1 mg/mL; Sigma) overnight and the colonies were counted (17).

FACS analysis. Cells were harvested and fixed in 70% ethanol at −20°C for 4 h, stained with propidium iodide (18 μg/mL) containing 400 μg/mL RNaseA (Roche) with shaking for 1 h, and analyzed by flow cytometry for apoptosis and cell cycle profile (19). Apoptosis was measured by the percentage of cells in sub-G₁ population.

SA-β-gal staining. The expression of SA-β-gal in cells was determined by SA-β-gal staining (20).

IB analysis. Cell lysates were prepared and subjected to IB analysis using antibodies against ROC1 (15), Bax, Bad, caspase 3, caspase 7, caspase 8, cyclin B1, cIAP1, cIAP2 (Cell signaling), p16, cdc25c, Cdt 1, cyclin D1, cyclin E1, pRB, Puma, PARP, Mcl-1, 14-3-3σ, Cdc2, p53, IκB, c-Jun (Santa Cruz), XIAP, Bcl-XL, p21, p27, Cdc25b (BD Transduction Laboratories), caspase 9 (Novocastra), β-actin (Sigma), Bak (Upstate), Bim (Imgenex), Noxa (Oncogene Science), Apaf-1 (Trevigen), Bcl-2 (Dako), survivin (Novus biologicals), and phosphor-γH2AX (Ser 139; Millipore).

Phorspho-γH2AX immunofluorescent staining. U87 cells were infected with LT-ROC1 or LT-CONT for 120 h with DMSO or 25 μmol/L etoposide added at last 12 h. Cells were fixed with 10% formalin and blocked with 3% horse serum and incubated with anti–phorspho-γH2AX monoclonal antibody at 1:100, followed by incubation with Rhodamine Red-labeled anti-mouse IgG at 1:250. Cellular nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI). The stained cells were observed under fluorescent microscope.

Statistical analysis. The statistical significance of differences between groups was assessed using GraphPad Prism4 software (version 4.03). The unpaired, two-tailed t test was used for the comparison of parameters between groups. The level of significance was set at a P value of <0.05.

ROC1 was overexpressed in diverse primary human tumor tissues and cell lines. ROC1 expression in human tissues was determined by immunohistochemistry staining of human tumor tissue arrays. As shown in Fig. 1A (left), ROC1 was expressed weakly in normal tissues, including lung, liver, and breast, but was overexpressed in carcinomas of lung, liver, breast (Fig. 1A,, right), as well as carcinomas of colon and ovary (data not shown). Overexpressed ROC1 was detected only in tumor mass, not in adjacent stroma tissues. To determine the frequency of cancer tissues with ROC1 overexpression, immunohistochemistry staining of a human lung cancer tissue array, consisting of 17 normal tissues and 38 tumor tissues of adenocarcinoma (n = 19) and squamous carcinoma (n = 19) was performed. Based on staining intensity, we classified the samples into five groups with increasing staining intensity from the weakest (±) to the strongest (++++; Fig. 1B). As summarized in Fig. 1C, ROC1 staining in normal tissues was very weak with 77% samples in group 1 and the remaining 23% in group 2. In contrast, ROC1 staining was very high in lung tumor tissues. For adenocarcinoma, only 5% samples were in group 1, 21% in group 2, and the remaining 74% in groups 3 to 5. More strikingly, only 5% of squamous carcinomas were in groups 1 and 2 and the remaining 95% in groups 3 to 5. Thus, ROC1 was overexpressed with a high frequency in lung cancer tissues. Furthermore, we found by IB analysis that ROC1 was highly expressed in many human cancer cell lines tested. The list included lung cancer lines (H1299, A549, H460, H1355), breast cancer lines (MDA-MB-461, MDA-MB-231), colon cancer lines (HCT116, DLD1), cervical carcinoma line (HeLa), pancreas cancer line (Panc-1), and glioblastoma lines (U87, U251; see below; data not shown). The ubiquitous expression of ROC1 at the high levels in diverse primary human tumors and multiple cancer cell lines suggests that ROC1 could play an essential role in tumor cell proliferation and survival.

ROC1 silencing inhibited the growth of human cancer cells. To determine the role of ROC1 in regulation of cancer cell proliferation and survival, ROC1 expression was knocked down by lentivirus-based siRNA targeting ROC1 (LT-ROC1), which coexpresses the Green flourescent protein (GFP). At 72 hours postinfection, GFP expression was clearly observed in about 95% cells infected with either LT-ROC1 or scrambled control siRNA, LT-CONT (data not shown). Compared with LT-CONT, LT-ROC1 infection significantly reduced ROC1 expression in U87 and H1299 cells (Fig. 2A  and B, top).

As shown in Fig. 2A and B (middle), ROC1 silencing remarkably reduced the growth of U87 and H1299 cells as measured by ATPlite proliferation assay and by cell counting analysis (data not shown). Clonogenic survival assay also showed a 5- to 10-fold reduction of colony numbers in LT-ROC1–infected cells (Fig. 2A, and B, bottom). Furthermore, the ability of H1299 cells to grow on soft agar was remarkably inhibited up to 4-fold upon ROC1 silencing (Fig. 2C). In addition, ROC1 silencing remarkably inhibited the growth of other human cancer cell lines, including HCT116, PANC-1, and HeLa cells as measured by clonogenic analysis (data not shown). These results showed that ROC1 silencing had a broad inhibitory effect on proliferation and survival of human cancer cells.

ROC1 silencing induced cell senescence in p53/p21- and p16/pRB-independent manner. We next investigated the nature of growth suppression induced by ROC1 silencing. Morphologic observation revealed that LT-ROC–infected U87 cells were larger in size with flattened shape (Fig. 3A,, top left), a feature of senescence (21). To address whether ROC1 silencing indeed induced cell senescence, the expression of SA-β-gal, a classic biochemical marker of senescence (20), was determined by SA-β-gal staining. Indeed, ∼25% of LT-ROC–infected cells, but <2% of the control cells, were positively stained with SA-β-gal (Fig. 3A , bottom left).

p53/p21 axis is a major senescence-triggering pathway (22). We first evaluated the effects of p53/p21 on ROC1 silencing–induced cell senescence in U87 cells harboring a wild-type p53 (23). As shown in Fig. 3A (top right), ROC1 silencing induced neither p53 nor p21. Furthermore, in p53-silenced (via LT-p53) U87 cells (Fig. 3A,, bottom right), ROC1 silencing still induced senescence as determined by both cellular morphologic observation (Fig. 3B,, top) and SA-β-gal staining (Fig. 3B , bottom), suggesting ROC1 silencing induced cell senescence is p53/p21 independent in U87 cells.

To further confirm this, we used H1299 lung cancer cells, a p53-null line (24). Absence of p53 and p21 expression in this cell line was first confirmed by immunobloting (Supplementary Fig. S1A). However, LT-ROC1–infection still caused H1299 cells to display a senescent morphology with enlarged size and flatten shape (Supplementary Fig. S1C). Moreover, ∼17% of LT-ROC1–infected cells, but only 0.3% of control cells, were positively stained with SA-β-gal (Supplementary Fig. S1D). We finally confirmed the p53/p21 independency by using H1299-p53^ts cells, which express a temperature-sensitive mutant p53. The p53 adapts a wild-type conformation when cells were grown at 32°C, but a mutant p53 conformation when cultured at 37°C (24). As shown in Supplementary Fig. S1B, p21, a p53 target protein, could be detected in H1299-p53^ts cells cultured at 32°C but not at 37°C. Consistently, ROC1 silencing was still able to induce senescence in cells cultured at 37°C (Supplementary Fig. S1E). Taken together, these results strongly indicate that ROC1 silencing–induced senescence is independent of p53/p21.

p16/pRB axis is another major senescence-inducing pathway (22). Because p16 was expressed in neither U87 (Fig. 3C,, top; ref. 25) nor H1299 cells (data not shown; ref. 26), we focused our study on the effects of pRB on ROC1 silencing–induced senescence. Previous study has shown that pRB is subject to degradation by SCF E3 ligase in the presence of E7 oncoprotein or EB virus latent antigen 3C (27, 28). Upon ROC1 silencing, we did not detect pRB accumulation in both U87 cells (Fig. 3C,, top) and H1299 cells (data not shown), suggesting that ROC1 is not involved in pRB degradation in these lines. Furthermore, in pRB-inactivated HeLa cells (29) and pRB-null MDA-MB-468 cells (Fig. 3C,, bottom; ref. 30), ROC1 silencing was still able to induce senescence effectively (Fig. 3D). Taken together, these findings showed that cell senescence induced by ROC1 silencing is independent of p16 and pRB as well.

ROC1 silencing induced apoptosis. Our morphologic observation also revealed that ∼30% of ROC1-silenced U87 and H1299 cells were shrunk in shape and detached from the culture dishes, a reminiscence of apoptosis (data not shown). We confirmed this by FACS analysis in which the sub-G₁ population is indicative of apoptotic cells. As shown in Fig. 4A, 30% to 40% of LT-ROC1–infected cells underwent apoptosis, compared with 5% to 10% of LT-CONT–infected cells. Apoptosis induced by ROC1 silencing was further confirmed by the activation of caspases, as demonstration by (a) the decrease in procaspase forms (casapses 3, 7, 8, and 9), (b) the appearance of cleaved active form (caspases 3 and 7), and (c) PARP cleavage (Fig. 4B , left). Thus, ROC1 silencing also induces apoptosis.

To understand the mechanism by which ROC1 silencing induced apoptosis, we analyzed the expression in U87 cells of proapoptotic proteins (Bax, Bak, Puma, Bim, Bad, and Bid) and antiapoptotic proteins (Bcl-2, Mcl-1, survivin, XIAP, Bcl-XL, cIAP1, and cIAP2). Among the proapoptotic proteins, Puma was moderately up-regulated, coupled with Bid cleavage, an indicator of apoptosis induction (Fig. 4B,, right). Among antiapoptotic proteins, the levels of Bcl-2, Mcl-1 and survivin were significantly reduced, whereas others were unchanged (Fig. 4B , right). The results suggested that ROC1 silencing could induce apoptosis by up-regulation of some proapoptotic proteins (e.g., Puma) and down-regulation of antiapoptotic proteins (e.g., Bcl-2, Mcl-1, and survivin).

ROC1 silencing induced a G₂-M arrest. Our FACS analysis also revealed that among the remaining cell populations, not undergoing apoptosis, ∼50% to 60% of LT-ROC1–infected U87 cells were arrested in the G₂-M phase of the cell cycle, compared with ∼15% to 20% of control cells at 120 hours postinfection (Fig. 4C). Interestingly, ROC1 silencing–induced G₂-M arrest was not observed in other tested cell lines including H1299, HeLa, and HCT116 (data not shown), suggesting that this effect is rather cell-line dependent.

To pursue potential mechanisms, we analyzed the expression of a panel of cell cycle regulatory proteins, including several known substrates of ROC1-SCF E3 ubiquitin ligases, such as cyclin D1, cyclin E1, p21, p27, c-Jun, Cdt-1, and IκB (6, 7, 31). As shown in Fig. 4D, ROC1 silencing had no effect on expression of Cdc25b, Cdc25c, cyclin D1, cyclin E1, c-Jun, Cdt-1, and IκB, whereas p21 and p27 were undetectable, indicating that the degradation of some of these substrates is likely cell-line dependent. Significantly, ROC1 silencing caused the accumulation of 14-3-3σ, a negative regulator of G₂-M progression (32), and the depletion of Cdc2 and cyclin B1, which form a complex to promote the G₂-M progression (33). Thus, accumulation of 14-3-3σ and elimination of Cdc2/cyclin B1 could be responsible for observed G₂-M growth arrest.

Sequential induction of G₂-M arrest, apoptosis, and senescence in U87 cells. We next addressed an obvious question as to how ROC1 silencing induces senescence, apoptosis, and G₂-M arrest among U87 cells in the same culture dish. Since above results were obtained at one point, 120 hours postinfection with LT-ROC1, we wonder whether the changes actually occur in a sequential order. To test this, we collected cell samples at 72, 96, or 120 hours post–LT-ROC1 infection, along with LT-CONT control cells, and performed FACS analysis for apoptosis (sub-G₁ population) and G₂-M arrest, SA-β-gal staining for senescence, and IB to correlate the phenotypic changes with the degree of ROC1 silencing. As shown in Fig. 5A and B, the G₂-M arrest occurred early with a 3.5-fold increase over the control cells at 72 hours postinfection, where 20% of ROC1 was silenced (Fig. 5C). Induction of apoptosis started to occur with a 3-fold increase over the control cells at 96 hours postinfection (Fig. 5A and B), where 50% of ROC1 was silenced (Fig. 5C). The senescence also occurred at 96 hours with a 7-fold increase over the control cells (Fig. 5B) but within a minimal cell population (5.8% versus 0.8%; Fig. 5A). The level of senescence was remarkably increased, reaching a 20-fold over the control cells among 26% of cell population at 120 hours postinfection (Fig. 5A and B), where 90% of ROC1 was silenced (Fig. 5C). The percentage of cells arrested at the G₂-M (60%) or undergoing apoptosis (30%) was also reaching the peak at 120 hours, but the fold increase over the control cells remained the same at 2.5- to 3.5-fold higher level because cell populations at the G₂-M or sub-G₁ phase in control cells were also increased after prolonged virus infection (Fig. 5A and B). The findings suggested that ROC1 silencing–induced phenotypic changes occur sequentially with initial induction of G₂-M arrest, followed by apoptosis and senescence in U87 cells.

siROC1 inhibited cancer cell growth by inducing apoptosis, G₂-M arrest, and senescence. To exclude the possibility of off-target effects, we repeated all the experiments using a ROC1-specific siRNA oligonucleotide reported by others (17). As shown in Supplementary Fig. S2A, siROC1 oligonucleotide significantly down-regulated ROC1 expression in both U87 and H1299 cells. Like LT-ROC1, siROC1-mediated ROC1 silencing significantly inhibited cell proliferation in both cancer lines (Supplementary Fig. S2B), induced apoptosis in 40% of ROC1-silenced U87 cells compared with 10% of control cells (Supplementary Fig. S2C), caused G₂-M arrest in 45% of ROC1-silenced U87 cells compared with 9% of control cells (Supplementary Fig. S2D), and induced the senescence in ∼30% of ROC1-silenced U87 cells (Supplementary Fig. S2E). Taken together, these results clearly showed that observed biological consequences are specific for ROC1 silencing, excluding the possibility of off-target effects.

ROC1 silencing induced DNA damage. Stress-induced DNA damage plays an essential role in induction of G₂-M arrest, apoptosis (34–36), as well as senescence (37–40). We, therefore, determined if ROC1 silencing could induce DNA damage in U87 cells by immunofluorescent staining of phosphor-γH2AX (Ser 139) as an index of DNA damage. Etoposide was included as the positive control. As shown in Fig. 6A, positive phosphor-γH2AX staining was apparently observed in ROC1-silenced cells with a similar intensity to that from etoposide-treated cells (top). In images at higher magnification, numerous phosphor-γH2AX–positive foci could be clearly observed in both ROC1-silenced and etoposide-treated cells (second panel). Strikingly, the phosphor-γH2AX in ROC1-silenced cells at 120 hours postinfection, when the senescence population reached the peak (Fig. 5), was remarkably induced with the level similar to that of etoposide treatment (Fig. 6B). To further define how early the ROC1 silencing could activate DNA damage response, we performed a time course study. As shown in Fig. 6C, no difference in the level of phosphor-γH2AX was observed between LT-CONT and LT-ROC1–infected cells at 48 hours postinfection, when no ROC1 silencing was observed. The levels of phosphor-γH2AX started to increase in ROC1-silenced cells at 72 hours and reach the peak of 11-fold induction at 120 hours postinfection correlated well with ROC1 silencing. These results clearly show that ROC1 silencing could induce DNA damage in U87 cells that contributes to the induction of G₂-M arrest, apoptosis, and senescence in p53/Rb-independent manner.

Previous studies have revealed that ROC1 is a growth essential gene in yeast (5, 12), Caenorhabditis elegans (13), Drosophila (14), and mouse (15) as well as for the growth of several human cancer cell lines (16). Here, we showed mechanistically that siRNA silencing of ROC1 dramatically suppressed cell proliferation and survival by induction of senescence and apoptosis in multiple cancer cell lines and of G₂-M arrest in a particular line.

Cellular senescence is a powerful mechanism to restrain proliferation of potentially tumorigenic cells and kill cancer cells (41, 42). Senescence is mainly regulated by two tumor suppressor pathways: p53/p21 and p16/pRB (22, 42, 43). Here, using multiple approaches in multiple cell lines, we clearly showed that ROC1 silencing induces the senescence independent of p53/p21 and p16/pRB. Significantly, we found that ROC1 silencing could induce DNA damage to a level comparable with that induced by a well-known DNA damaging agent, etoposide (Fig. 6). Because senescence, initiated by oncogene activation, telomere dysfunction, or other stimulus, is frequently associated with DNA damage (37–40), ROC1 silencing–induced senescence could result from cellular response to DNA damage. It is very likely that some DNA damage responsive molecule(s) are the substrates of ROC1-SCF E3 ubiquitin ligases, whose accumulation upon ROC1 silencing triggers sustained DNA damage response. Our study, therefore, opens up a new avenue for future study toward the identification of these substrates and elucidation of the novel senescence-inducing pathways that are independent of p53/Rb but regulated by ROC1-SCF E3 ubiquitin ligases.

Induction of apoptosis is one of the most important strategies for anticancer therapy (34). In this study, we found that ROC1 siRNA silencing induced apoptosis in multiple cancer cells. Mechanistic study revealed that ROC1 silencing not only led to moderate accumulation of proapoptotic Puma, but also caused remarkable reduction of antiapoptotic protein Bcl-2, Mcl-1, and survivin, although we cannot exclude the possibility that these later changes were the consequence of apoptosis. Nevertheless, the changes in more than one protein may explain why our effort to rescue apoptosis-inducing phenotype in ROC1-silenced cells by simultaneous silencing of Puma was not successful (data not shown). Given the fact that ROC1 is the RING component of SCF E3 ligases, required for ubiquitination and subsequent degradation of a variety of protein substrates, one could anticipate that alterations of multiple protein substrates would contribute to the induction of senescence and apoptosis upon ROC1 silencing and single gene–based rescue would not be sufficient.

A cell line–dependent feature observed upon ROC1 silencing is induction of the G₂-M arrest in U87 cells, which is associated with significant accumulation of 14-3-3σ and reduction of Cdc2/cyclin B1. It is known that 14-3-3σ sequesters Cdc2/cyclin B1 in the cytoplasm away from its nuclear targets, leading to a subsequent G₂-M arrest (32), whereas the down-regulation of Cdc2 and cyclin B1 could directly impair the formation of Cdc2/cyclin B1 complexes required for G₂-M transition (33). Thus, accumulation of 14-3-3σ and reduction of Cdc2 and cyclin B1 could act in a synergistic manner to trigger the G₂-M arrest upon ROC1 silencing. Given the fact that G₂-M arrest is a common cell cycle checkpoint mechanism in response to DNA damage (35, 36), it is very likely that G₂-M arrest is also initiated by DNA damage upon ROC1 silencing. This notion is supported by our observation that both G₂-M arrest (Fig. 5) and DNA damage (Fig. 6) started to occur at early time point (72 hours post cell infection), when the ROC-1 silencing effect starts to appear.

The major finding of this study is that ROC1 silencing triggers senescence and apoptosis in multiple lines as well as G₂-M arrest in U87 cells. How can three distinct phenotypic changes occur, upon ROC1 silencing, among U87 cells while grown under the same culture dish? One explanation is that these pathways were activated consecutively, with induction of G₂-M arrest occurring first, followed by apoptotic cell death and senescence, as shown and supported by our time course study (Fig. 5). In fact, the process of G₂-M arrest followed by induction of apoptosis and/or senescence has been described in few other systems (44–47). Another more sophisticated answer would be that LT-ROC1 infected individual cancer cells rather randomly with an uneven silencing of ROC levels among cells. Individual cells with different degree of ROC1 reduction would accumulate different subsets of ROC1-SCF substrates that are responsible for induction of different phenotypes, namely G₂-M arrest, apoptosis, or senescence. The future challenge will be to identify these different sets of ROC1-SCF substrates that separately regulate these growth arrest and death pathways. Toward that goal, a recent study using global protein stability profiling has identified over 350 potential SCF substrates that are closely involved in the regulation of cell cycle, apoptosis, and cell signaling (48, 49). Mechanistic characterization of these SCF substrates would broaden our understanding how ROC1-SCF E3 ligases regulate cell proliferation and survival. Finally, the high sensitivity of ROC1-overexpressed human cancer cells to ROC1 silencing suggest that growth of cancer cells is heavily reliant on a high level of ROC1, rendering ROC1 as a promising anticancer target for selective cancer cell killing.

No potential conflicts of interest were disclosed.

Grant support: National Cancer Institute grants CA111554 and CA118762 (Y. Sun), and CA107237 (M.S. Soengas).

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 Danfeng Cai, Monique Verhaegen, and Mary Beth Riblett for their technical support, and Drs. Dafydd Thomas and Thomas Giordano for providing us the primary human cancer tissue microarrays.