Heat shock protein 27 promotes cell cycle progression by down-regulating E2F transcription factor 4 and retinoblastoma family protein p130

Heat shock protein 27 (HSP27) protects cells under stress. Here, we demonstrate that HSP27 also promotes cell cycle progression of MRC-5 human lung fibroblast cells. Serum starvation for 24 h induced G1 arrest in these cells, and upon serum refeeding, the cells initiated cell cycle progression accompanied by an increase in HSP27 protein levels. HSP27 levels peaked at 12 h, and transcriptional up-regulation of six G2/M-related genes (CCNA2, CCNB1, CCNB2, CDC25C, CDCA3, and CDK1) peaked at 24–48 h. siRNA-mediated HSP27 silencing in proliferating MRC-5 cells induced G2 arrest coinciding with down-regulation of these six genes. Of note, the promoters of all of these genes have the cell cycle–dependent element and/or the cell cycle gene-homology region. These promoter regions are known to be bound by the E2F family proteins (E2F-1 to E2F-8) and retinoblastoma (RB) family proteins (RB1, p107, and p130), among which E2F-4 and p130 were strongly up-regulated in HSP27-knockdown cells. E2F-4 or p130 knockdown concomitant with the HSP27 knockdown rescued MRC-5 cells from G2 arrest and up-regulated the six cell cycle genes. Moreover, we observed cellular senescence in MRC-5 cells on day 3 after the HSP27 knockdown, as evidenced by increased senescence-associated β-gal activity and up-regulated inflammatory cytokines. The cellular senescence was also suppressed by the concomitant knockdown of E2F-4/HSP27 or p130/HSP27. Our findings indicate that HSP27 promotes cell cycle progression of MRC-5 cells by suppressing expression of the transcriptional repressors E2F-4 and p130.

Heat shock protein 27 (HSP27), 2 also known as the heat shock protein family B member 1 (HSPB1), belongs to the small HSP family that commonly has a conserved C-terminal ␣-crystallin domain (1). In response to heat shock, HSP27 functions as an ATP-independent molecular chaperone to facilitate the proper refolding of damaged proteins (2). HSP27 is also involved in a wide range of cellular processes, such as cytoskeletal organization, redox metabolism, suppression of apoptosis, and cell proliferation (2)(3)(4)(5)(6). HSP27 also promotes degradation of ubiquitinated proteins by the 26S proteasome (7,8). Furthermore, HSP27 is frequently overexpressed in cancer cells and closely associated with aggressive tumor behavior, metastasis, poor prognosis, and resistance to chemotherapy (9,10).
Although the involvement of HSP27 in stressed cells and cancer cells is now well documented, there are only a few reports on the role of HSP27 in cell cycle progression and cellular senescence. For example, HSP27 reverses the G 2 arrest caused by HIV-1 viral protein R (11). HSP27 promotes G 1 /S transition of cells arrested with serum depletion or with the nitric oxide donor glyceryl trinitrate by promoting ubiquitination and proteasomal degradation of the cyclin-dependent kinase (CDK) inhibitor p27 Kip1 (8). Embryonic fibroblasts derived from Hspb1-deficient mice display reduced entry into S phase with increased expression of the CDK inhibitors p21 Cip1 and p27 Kip1 (12). Taken together, HSP27 may have a promoting role in cell cycle progression. As for cellular senescence, HSP27 knockdown of MCF-7 mammary carcinoma cells led to reduced proliferation and acquisition of a spontaneous secretory phenotype, a feature of cellular senescence (13). Whereas overexpression of HSP27 protects MCF10A human mammary epithelial cells from doxorubicin-induced cellular senescence by inhibiting p53-mediated induction of p21 Cip1 , the major regulator of the senescence program, depletion of HSP27 in HCT116 human colon carcinoma cells caused cellular senescence through activation of p53 and induction of p21 Cip1 (14). Thus, HSP27 may also have a suppressive role in cellular senescence by interfering with p53 activation and/or function.
Previously, we have demonstrated that HSP27 plays an essential role in transforming growth factor-␤1-induced myofibroblast differentiation of normal lung fibroblasts and accordingly in the development of bleomycin-induced pulmonary fibrosis (15). During the same study, we also observed that siRNA knockdown of HSP27 significantly suppressed cell proliferation of normal lung fibroblasts. Thus, in the present study, we further explored the role of HSP27 in cell cycle progression by using MRC-5, a normal human lung fibroblast cell line (16). MRC-5 was serum-starved and -refed to synchronize the cell cycle. Upon refeeding, we found a rapid increase in HSP27 protein along with transcriptional up-regulation of six G 2 /M-related genes: CCNA2 (cyclin A2), CCNB1 (cyclin B1), CCNB2 (cyclin B2), CDC25C, CDCA3, and CDK1 (17,18). We further found that HSP27 knockdown of proliferating MRC-5 induced G 2 arrest together with down-regulation of the same six genes and up-regulation of E2F-4/p130. We demonstrated that concomitant knockdown of E2F-4 or p130 with HSP27 knockdown rescued MRC-5 from G 2 arrest and also prevented the downregulation of the six genes. MRC-5 also underwent cellular senescence 3 days after HSP27 knockdown as evidenced by increases in senescence-associated ␤-galactose positivity and up-regulation of proinflammatory cytokines. The cellular senescence was also prevented by the concomitant knockdown of E2F-4 or p130 with HSP27 knockdown. Collectively, HSP27 plays a pivotal role in cell cycle progression of MRC-5 by down-regulating the expression of E2F-4/p130, whose up-regulation leads to G 2 arrest through down-regulation of the six G 2 /M-related genes, which eventually results in cellular senescence in MRC-5.

HSP27 increases during cell cycle progression of serum-refed MRC-5
MRC-5 is a human diploid lung fibroblast cell line that is widely used as a model of normal human fibroblasts (15,16). In our preliminary experiments, HSP27 knockdown by siRNA transfection significantly suppressed cell proliferation of MRC-5 (data not shown, but see Fig. 2). To test whether HSP27 was involved in cell cycle progression, we used the technique of serum starvation and refeeding to synchronize the cell cycle of MRC-5. After 24 h of fetal bovine serum (FBS) starvation, we refed MRC-5 with 5% FBS to initiate the cell cycle progression. We confirmed that whereas FBS starvation increased cells at G 0 /G 1 phase (G 0 /G 1 ϭ 73 Ϯ 0.6%, S ϭ 6 Ϯ 0.2%, G 2 /M ϭ 20 Ϯ 0.5%), FBS refeeding increased cells at S and G 2 /M phases (G 0 /G 1 ϭ 52 Ϯ 0.7%, S ϭ 20 Ϯ 0.2%, G 2 /M ϭ 28 Ϯ 0.5%) (Fig.  1A). We found that FBS refeeding also increased the protein level of HSP27 with a peak at 12 h (Fig. 1B). No significant increase was seen in the phosphorylation of HSP27 (Fig. 1B). Because cell cycle progression is regulated by cyclins and CDKs (Fig. 1C), we quantified mRNA levels of these molecules together with that of HSP27 (Fig. 1D). We found that HSP27 mRNA was not up-regulated by FBS refeeding. Thus, the increase of HSP27 protein was mostly translationally and/or post-translationally controlled in FBS-refed MRC-5. On the other hand, among the 12 cell cycle-related genes examined,

HSP27 and cell cycle
the six genes (cyclin A2, cyclin B1, cyclin B2, cdc25c, cdcA3, and CDK1) were strongly up-regulated after 24 -48 h of FBS refeeding (Fig. 1D). No such changes were seen in the mRNA levels of cyclin D1, cyclin D2, cyclin E1, CDK2, and the two CDK inhibitors p21 Cip1 and p27 Kip1 . Of note, all of the six upregulated genes are those involved in the G 2 /M phases and contain the cell cycle-dependent element (CDE) and/or the cell cycle gene-homology region (CHR) in their promoters (19); CCNA2, CDC25C, CDCA3, and CDK1 have both CDE and CHR, whereas CCNB1 and CCNB2 have CHR. These elements are known to be regulated by the binding of the E2F and retinoblastoma (RB) family proteins (19,20).

HSP27 knockdown induces down-regulation of the six cell cycle regulatory genes
HSP27 knockdown efficiently decreased not only HSP27 mRNA but also the mRNAs of the six cell cycle regulatory genes that were up-regulated in FBS-refed MRC-5: cyclin A2, cyclin B1, cyclin B2, cdc25c, cdcA3, and CDK1 (Fig. 2C). On the other hand, although cyclin D1 mRNA was significantly up-regulated by HSP27 knockdown, the mRNA levels of cyclin D2, cyclin E1, and CDK2 as well as those of the two CDK inhibitors p21 Cip1 and p27 Kip1 were not affected by HSP27 knockdown. These results suggested that HSP27 was involved in the up-regulation of the six G 2 /M cell cycle progression molecules in MRC-5.

E2F-4 and p130 increase by HSP27 knockdown
As mentioned earlier, the six genes (CCNA2, CCNB1, CCNB2, CDC25C, CDCA3, and CDK1) that were down-regulated by HSP27 knockdown commonly carry the CDE/CHR or CHR element in their promoters (19). Thus, their expression can be regulated by the E2F family proteins and RB family proteins (RB1, p107, and p130) that bind to these elements (19,20). Among the E2F family proteins, E2F-1 to E2F-3 positively and E2F-4 to E2F-8 negatively regulate the cell cycle (19). Because the HSP27 knockdown strongly decreased the expression of the six genes, we quantified the mRNA levels of the repressor E2F family proteins in MRC-5 upon HSP27 knockdown. By using qPCR, we found that the expression of E2F-4 was significantly increased upon HSP27 knockdown (Fig. 3A), whereas the expression of E2F-5 to E2F-8  To examine whether HSP27 could directly interact with E2F-4 and/or p130, we conducted co-immunoprecipitation experiments and found no evidence for the direct binding of HSP27 to E2F-4 or p130 (data not shown). Because HSP27 was reported to enhance ubiquitination and degradation of intracellular proteins such as p27 Kip1 and IB (7,8), we also conducted the protein chase experiment using cycloheximide to determine the effect of HSP27 knockdown on the half-life of E2F-4 and p130. Although we expected slower degradation of E2F-4 and/or p130 by HSP27 knockdown, we actually found enhanced degradation of E2F-4 and p130 by HSP27 knockdown compared with control knockdown (Fig. 3D).

E2F-4 knockdown rescues G 2 arrest by HSP27 knockdown
To determine the role of E2F-4 in the G 2 arrest of MRC-5 by HSP27 knockdown, we performed single and double knockdown experiments. Immunoblot analysis confirmed the efficient knockdown of HSP27 and E2F-4 by HSP27 siRNA and E2F-4 siRNA, respectively (Fig. 4A). Whereas the single knockdown of HSP27 decreased cell number (Fig. 4B) and induced G 2 arrest (Fig. 4C), the single knockdown of E2F-4 had no such effect on cell number (Fig. 4B) or cell cycle progression (Fig.  4C). However, the double knockdown of HSP27 and E2F-4 reversed the effect of HSP27 knockdown on cell number (Fig.  4B) and G 2 arrest (Fig. 4C). Furthermore, among the six G 2 /M regulatory genes that were down-regulated by HSP27 knockdown, the down-regulations of at least cyclin B2, cdc25c, cdcA3, and CDK1 were significantly prevented by the concomitant knockdown of E2F-4 (Fig. 4D). These results suggested that up-regulation of E2F-4 was mostly responsible for the G 2 arrest induced by HSP27 knockdown.

p130 knockdown also rescues G 2 arrest by HSP27 knockdown
We next performed similar experiments for p130. Immunoblot confirmed efficient knockdown of HSP27 and p130 by HSP27 siRNA and p130 siRNA, respectively (Fig. 4E). We also observed that the concomitant knockdown of p130 rescued the effect of HSP27 knockdown on cell number (Fig. 4F) and cell cycle (Fig. 4G). However, because p130 is a general cell cycle repressor, the single knockdown of p130 significantly increased cell numbers (Fig. 4F). Furthermore, by qPCR, we found that single p130 knockdown significantly up-regulated mRNAs for HSP27, cyclin A2, cyclin B1, cyclin B2, cdc25c, and cdcA3 (Fig.  4H). Since the HSP27/p130 double knockdown led to a similar level of HSP27 as control cells, but a lower level of p130 expression, this might allow us to assess the effect of low p130 without a significant change in the level of HSP27. Thus, the reduction of p130 alone led to the up-regulation of the six cell cycleassociated genes.

HSP27 knockdown leads to cellular senescence
We also examined a long-term effect of HSP27 knockdown on MRC-5. After 3 days of HSP27 knockdown, we observed

HSP27 and cell cycle
significant increases in cells positive for senescence-associated ␤-gal (SA-␤-gal) (Fig. 5A) and significant up-regulation of proinflammatory cytokines IL-1␣ and IL-8 (Fig. 5B). These results demonstrated that HSP27 knockdown of MRC-5 induced not only cell cycle arrest but also cellular senescence after day 3. Again the concomitant knockdown of E2F-4 or p130 with HSP27 knockdown effectively prevented the increase of ␤-gal-positive cells. Of note, although HSP27 knockdown significantly decreased the mRNA level of p16, a senescence inducer molecule (22), it did not affect the protein level of p16 (Fig. 5C).

HSP27 knockdown induces cellular senescence in G 2 -arrested cells
Because HSP27 knockdown induced G 2 arrest, we also examined whether increased cellular senescence by HSP27 knockdown could also be associated with the G 2 phase, whose cells can be identified by intense Ki-67-positive nuclear staining and/or cytoplasmic cyclin B1 staining (23,24). In controlknockdown cells, SPiDER-␤Gal-positive senescent cells were negative for Ki-67 (Fig. 5B) or cytoplasmic cyclin B1 (data not shown). On the other hand, in HSP27-knockdown cells, we found that some SPiDER-␤Gal-positive senescent cells had

HSP27 and cell cycle
Ki-67-positive nucleoli or cyclin B1-positive cytoplasm, suggesting that increased senescent cells in HSP27 knockdown were mostly at the G 2 phase (Fig. 5D).

Cell cycle arrest by HSP27 knockdown is not associated with oxidative cellular damage
Cell cycle arrest occurs under unfavorable conditions, such as nutrient deficiency or DNA damage. Because we used MRC-5 maintained in 5% FBS-Opti-MEM in HSP27-knockdown experiments, the nutrient deficiency might not be the likely cause of the cell cycle arrest. On the other hand, because HSP27 has an antioxidant property (2), oxidative damage might be increased by HSP27 knockdown, thereby leading to DNA damage and cell cycle arrest. To test this possibility, we analyzed protein oxidation, protein ubiquitination, and DNA damage in HSP27-knockdown MRC-5. We found no apparent  (n ϭ 4). *, p Ͻ 0.05. D, cell cycle analysis of senescent cells. Control siRNAor HSP27 siRNA-transfected cells were reseeded in an 8-well chamber slide after a 24-h culture. Three days after transfection, cells were fixed with 4% paraformaldehyde and stained with SPiDER-␤Gal (red), followed by immunofluorescence staining with antibody against Ki-67 (blue) or cyclin B1 (blue). Scale bar, 20 m. Arrows, senescence cells.

HSP27 and cell cycle
increases in protein oxidation or ubiquitination by HSP27 knockdown (Fig. 6A). The phosphohistone 2X, a marker of cellular DNA damage, was also not increased by HSP27 knockdown either (Fig. 6B). Thus, oxidative damage was not the likely cause of the cell cycle arrest induced by HSP27 knockdown.

Cell cycle arrest by HSP27 knockdown is independent of p53
We further examined the possible involvement of p53, a key molecule of cell cycle arrest (25). Although p53 was increased by HSP27 knockdown, p21 Cip1 , the CDK inhibitor and one of the major downstream mediators of p53 function, was not affected (Fig. 6C). In addition, pifithrin-␣, a p53 inhibitor (26), did not prevent the decrease in cell number or cell cycle arrest by HSP27 knockdown (Fig. 6D). Pifithrin-␣ also did not prevent the down-regulation of mRNA levels of the cell cycleassociated genes by HSP27 knockdown either (data not shown). We concluded that the cell cycle arrest induced by HSP27 knockdown was mostly independent of p53.

Discussion
The cell cycle is a process composed of four phases: G 1 (pre-DNA synthesis), S (DNA synthesis), G 2 (pre-division), and M
In the present study, we used MRC-5, a normal human lung fibroblast cell line (16), and demonstrated a rapid increase in HSP27 protein with a peak at 12 h in serum-starved and -refed MRC-5 (Fig. 1). Furthermore, the six genes associated with G 2 /M cell cycle progression (CCNA2, CCNB1, CCNB2, CDC25C, CDCA3, and CDK1) were strongly up-regulated in serum-refed MRC-5 (Fig. 1). Of note, all these six genes carry CDE and/or CHR in their promoters, where the complex of a repressive E2F and RB family proteins is known to bind (19,20). By using the siRNA gene-silencing technique, we showed that HSP27 knockdown of proliferating MRC-5 led to G 2 arrest with concurrent down-regulation of the same six genes (Fig. 2) and up-regulation of the transcriptional repressor complex E2F-4/ p130 (Fig. 3). We further showed that the G 2 arrest as well as the down-regulation of the six cell cycle-related genes induced by HSP27 knockdown was prevented by concomitant knockdown of E2F-4 or p130 (Fig. 4). Because HSP27 is known to enhance the intracellular protein degradation through ubiquitination and proteosomal degradation (7,8), we considered that HSP27 might promote the degradation of E2F-4 and/or p130. However, by a protein chase assay using cycloheximide, we observed no slowing down of E2F-4 or p130 degradation by HSP27 knockdown (Fig. 3). Collectively, HSP27 plays a pivotal role in G 2 progression of MRC-5 by up-regulating the G 2 /M-associated genes via down-regulation of E2F-4/p130.
Although the mechanism(s) of down-regulation of E2F-4/ p130 by HSP27 remains unclear at present, HSP27 may have an effect on the promoter activity of these molecules, because HSP27 has been reported to bind some transcriptions factors and enhance their activities (32). Indeed, we detected a significant increase of the p130 promoter activity in HSP27-knockdown MRC-5 (Fig. S2). We also found the presence of HSP27 in the nucleus of phosphorylated histone H3-positive M-phase cells (Fig. S2). These results are consistent with a notion that HSP27 negatively regulates the transcription of p130.
The cell cycle arrest of proliferating cells occurs under unfavorable conditions such as starvation and DNA damage. Starvation induces G 1 arrest (33), whereas DNA damage induces G 2 arrest (34). Furthermore, there are two types of cell cycle arrest: reversible and irreversible. The former is called quiescence, whereas the latter senescence (35). The cellular senescence is controlled by two signaling pathways; one is the p53 to p21 Cip1 pathway, and the other is the p16 to phosphorylated retinoblastoma protein pathway (36). Yet, another signaling pathway that does not involve p53 or p16 has been recently described in fibroblasts (37). In this case, the accumulation of GATA4, which is digested by autophagy in physiological conditions, activates the transcription factor NF-B to initiate the expression of inflammatory cytokines and cellular senescence (37). Furthermore, whereas 2 days of cell cycle arrest in vitro can be reversible, cell cycle arrest may be irreversible after 3-4 days (38,39). Reversible cell cycle arrest is converted to irreversible senescence through a process called geroconversion, a futile growth activity during the cell cycle arrest, which is mainly regulated by mammalian target of rapamycin (mTOR) signaling (39,40). Senescent cells are also known to exhibit senescence-associated secretory phenotype (SASP) by producing inflammatory cytokines, metalloproteinases, and growth factors (41,42). The SASP phenotype is initiated by NF-B and p38 MAPK signaling (43) and maintained in an autocrine fashion by the SASP factor IL-1␣ (44). The production of various SASP factors is considered to be beneficial for the host by promoting immune clearance of damaged cells and preventing proliferation of potential cancer cells (45,46). However, the SASP factors may also exacerbate age-related diseases, such as atherosclerosis (47) and type 2 diabetes mellitus (42).
In the present study, we also observed cellular senescence in MRC-5 on day 3 after HSP27 knockdown, as evidenced by positivity of senescence-associated ␤-gal and production of proinflammatory cytokines (Fig. 5). Although mTOR is considered to play a key role in the conversion from quiescence to senescence (39,40), we detected no significant increase in the mTOR activity by HSP27 knockdown, as shown by immunoblot of the mTOR substances, phospho-p70 S6 kinase and phospho-4E-BP1 (data not shown). We also observed no significant increase of p16 (Fig. 5) or p21 Cip1 (Fig. 6) in HSP27-knockdown MRC-5. Thus, the cellular senescence induced by HSP27 knockdown in MRC-5 may be independent of mTOR, the p53/p21 pathway, or the p16 pathway. Because the cellular senescence induced by HSP27 knockdown was also suppressed by the concurrent knockdown of E2F-4 or p130, the G 2 arrest induced by up-regulation of E2F-4/p130 may eventually lead to cellular senescence in MRC-5 by still unknown mechanisms.
In conclusion, we have demonstrated that HSP27 promotes cell cycle progression of MRC-5 by negatively regulating E2F-4/p130 expression. Increased expression of E2F-4/p130 in HSP27-knockdown MRC-5 leads to G 2 arrest by down-regulating the expression of the six G 2 /M-related genes (CCNA2, CCNB1, CCNB2, CDC25C, CDCA3, and CDK1). The G 2 arrest induced by up-regulation of E2F-4/p130 eventually results in cellular senescence in MRC-5. Thus, HSP27 may indirectly prevent cellular senescence by preventing up-regulation of E2F-4/ p130. By data mining of NCBI's Gene Expression Omnibus microarray database, we indeed found decrease of HSP27 mRNA in some senescence cells: liver stellate cells (48) and premalignant lung cells (49). At present, we do not know how HSP27 negatively regulates the E2F-4/p130 expression, but the transcriptional regulation may be partly involved (Fig. S2). It also remains to be seen whether the promoting effect of HSP27 on cell cycle progression is a general phenomenon or cell typespecific phenomenon; in HSP27 knockdown in normal human lung fibroblast (NHLF) and lung adenocarcinoma cell line A549, we found that p130 mRNA levels were increased in both cell lines, whereas E2F-4 levels of NHLF or A549 were unchanged or slightly increased, respectively (Fig. S4). This suggests that the promoting effect of HSP27 on cell cycle progression is general.

Experimental procedures
Cell culture and treatment MRC-5 was obtained from Riken Cell Bank (Tsukuba, Japan) and maintained in Dulbecco's modified Eagle's medium (Sigma-Aldrich) supplemented with 5% FBS, streptomycin (100 g/ml) and penicillin (100 units/ml) at 37°C in 5% CO 2 . To synchronize the cell cycle, cells were cultured in a 24-well plate to Ͻ70% confluence, washed with PBS, and placed in serum-free Dulbecco's modified Eagle's medium for 24 h. Then FBS was added to the medium to make the final concentration of 5%. DNA damage was induced in cells by exposing to UV light ( ϭ 253.7 nm, 120 microwatts cm Ϫ2 ) for 15 min and culturing for 4 h.

Immunoblot analysis
Cells were washed with PBS and solubilized with CelLytic M (Sigma-Aldrich) containing Protease Inhibitor Mixture Complete (Roche Diagnostics, Mannheim, Germany) and Phosphatase Inhibitor Mixture (Toyobo, Osaka, Japan). After 15 min of mixing at room temperature, cell debris was removed by centrifugation. In some experiments, nuclear and cytosolic fractions were prepared by using the NucBaster TM protein extraction kit (Merck Millipore, Billerica, MA), following the manufacturer's instructions. Samples were electrophoresed on an SDS-polyacrylamide gel under reducing conditions and electrophoretically blotted onto a polyvinylidene difluoride membrane. Membranes were blocked in 5% skim milk and probed with a primary antibody. Mouse antibodies against ␣-tubulin and ␤-actin were obtained from Sigma-Aldrich; goat anti-HSP27 antibody, rabbit anti-E2F-4 antibody, rabbit antiglyceraldehyde-3-phosphate dehydrogenase (GAPDH), rabbit anti-p53 antibodies, and mouse anti-ubiquitin antibody were from Santa Cruz; mouse mAbs against p130, RB1, and p21 Cip1 were from BD Biosciences; and rabbit polyclonal antibodies against phospho-HSP27 (Ser-82), cyclin B1, p130, p16, and histone H3 were from Cell Signaling (Beverly, MA). After washing, membranes were reacted with a horseradish peroxidaseconjugated secondary antibody and developed using Luminata TM Forte Western HRP Substrate (Merck Millipore). Signal intensities were quantified using ImageJ software (National Institutes of Health, Bethesda, MD). Protein carbonylation, a major form of protein oxidation, was determined using an Oxiselect TM protein carbonyl immunoblot kit (Cell Biolabs Inc., San Diego, CA) following the manufacturer's instructions.

siRNA transfection
The siRNAs listed in Table 1 were obtained from Gene-Design (Osaka, Japan). Transfection of siRNA was performed by Lipofectamine RNAiMax (Invitrogen). Briefly, cells were suspended in 5% FBS Opti-MEM (Thermo Fisher Scientific) and seeded in a 12-well plate at 2 ϫ 10 4 cells/ml. After 24 h of culture, Lipofectamine RNAiMax siRNA complex was added to Table 1 Primers and siRNAs used for this study cells. The final siRNA concentration in each well was 20 nM. The HSP27 siRNA that we used in this study has been extensively used by many previous studies as the accepted standard HSP27 siRNA (50 -52) and was also found to be most effective in this study (data not shown). The results obtained by E2F-4a siRNA and p130a siRNA were shown in this paper, whereas the data obtained by E2F-4b siRNA and p130b siRNA reproduced the results by E2F-4a siRNA and p130a siRNA, respectively (data not shown). In some experiments, cells were treated with 20 M cycloheximide (Sigma-Aldrich) or with 10 M pifithrin-␣ (Cayman Chemical, Ann Arbor, MI). The pifithrin-␣ concentration used was predetermined by its protective effect against doxorubicin or 5-fluorourasil-induced cell growth inhibition (53) (Fig. S3). Cell lysates for immunoblot analysis were prepared as stated above.

Cell proliferation assay
Cells were seeded in a 96-well plate and transfected with control siRNA or HSP27 siRNA. Cell numbers were quantified by Cell Counting Kit-8 (CCK8; Dojindo Laboratory, Kumamoto, Japan). In brief, we added 10 l of CCK8 reagent to each well and incubated for 1 h. Then the medium from each well was transferred to a new 96-well plate. Optical density at 450 nm was measured on a microplate reader (ARVO, PerkinElmer Life Sciences).

Cell counting and cell cycle analysis by MUSE
Cells in a 12-well plate were transfected with siRNA and cultured for 2 days. Then cells were harvested by trypsin-EDTA and centrifuged at 100 ϫ g for 5 min. Cell pellets were resuspended in 330 l of PBS. A portion of the cell suspension (30 l) was mixed with the Muse Count & Viability Kit (Millipore) to determine cell numbers and viability using the MUSE cell analyzer (Millipore). The remaining cell suspension (300 l) was mixed with 700 l of ethanol and fixed at Ϫ20°C for 4 h. After washing twice with PBS, cells were suspended in the Muse cell cycle kit (Millipore), and the cell cycles were determined using the MUSE cell analyzer.

Quantitative PCR
Total RNAs were extracted from cells using the RNeasy Mini Kit (Qiagen). RNA samples (500 ng each) were reverse-transcribed using the High Capacity RNA-to-cDNA Kit (Applied Biosystems, Foster City, CA). Quantitative PCR was performed using THUNDERBIRD SYBR qPCR mix (Toyobo, Osaka, Japan) and the ABI Step One real-time PCR system with ␤-actin or GAPDH as a reference control. The sequences of the sense and antisense primers used for amplification are listed in Table 1.

Immunocytochemistry
Cells were cultured in a multiwell glass chamber plate (BD Falcon, Franklin Lakes, NJ), washed with PBS, and fixed with ice-cold ethanol at Ϫ20°C for 4 min. After washing, cells were treated with Dakocytomation Protein Block (Dako, Glostrup, Denmark) for 1 h. Then cells were incubated with anti-E2F-4 or anti-p130 antibody. After washing, cells were incubated with Alexa 488 or Alexa 555 fluorescein-labeled secondary antibodies (Invitrogen) and DNA staining reagent To-PRO3 (Invitro-gen). After washing and mounting, images were taken by confocal laser microscope (Carl Zeiss GmBH, Jena, Germany).

Cytochemical staining for ␤-gal
Cells in a 24-well plate were fixed with 4% paraformaldehyde at room temperature for 15 min. After washing with PBS, cells were incubated in a ␤-gal staining solution (100 mM sodium phosphate buffer, pH 6.0, 2 mM MgCl 2 , 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, and 1 mg/ml X-gal). After incubation at 37°C for 18 h, cells were washed with PBS and counterstained with 1 g/ml 4Ј,6-diamidino-2-phenylindole. Four images were taken randomly, and total and ␤-galpositive blue cells were counted by ImageJ software.
To detect whether senescence cells were at the G 2 phase, fluorometric senescence cell detection reagent SPiDER-␤Gal (Dojindo Laboratory) was used. One day after control and HSP27 siRNA transfection, cells were detached, seeded in an 8-well chamber slide (BD Falcon), and cultured an additional 2 days. After washing, cells were fixed with 4% paraformaldehyde and treated with SPiDER-␤Gal, following the manufacturer's instructions. After incubation with this reagent, cells were fixed again and then stained with Ki-67 (Abcam, Cambridge, UK) or cyclin B1 (Cell Signaling) antibodies, which were visualized with Alexa 633 fluorescein-labeled secondary antibody (Invitrogen) and under confocal microscopy.

Statistical analysis
Quantitative data are presented as the mean Ϯ S.E. Statistical analysis was performed using Student's t test. p Ͻ 0.05 was considered statistically significant.