Introduction Tumor progression is a multistep process wherein several defined events are common to cancer cells, such as uncontrolled proliferation and invasion (Hahn and Weinberg, 2002). Cellular senescence is characterized by an irreversible arrest of cell proliferation, so that it can prevent the aberrant and unlimited proliferation of tumor cells (Campisi, 2005). Senescent cells exhibit enlarged morphological changes and less motility than young cells, which may contribute to the suppression of cell migration, invasion, and metastasis (Chen et al., 2000). Oncogene-induced senescence is a cellular response, which can occur in vivo and provides a bona fide barrier to tumorigenesis (Narita and Lowe, 2005). Oncogene-induced senescence was found in premalignant tumors but not in more advanced malignant tumors (Braig et al., 2005; Collado et al., 2005). Therefore, cellular senescence acts as an important barrier to cancer and plays an important role in tumor suppression. microRNAs (miRNAs) are a class of naturally occurring small noncoding RNAs that negatively regulate the stability and translation of target protein–coding mRNAs at the 3′ untranslated region (UTR). miRNAs typically target a cluster of genes rather than one specific gene (Bartel, 2004), a characteristic which allows them to play critical roles in a variety of biological processes such as cell proliferation, differentiation, apoptosis, and carcinogenesis (He and Hannon, 2004). Recently, a growing number of studies have documented the miRNA expression profiles in human cancers (Calin and Croce, 2006), suggesting that miRNAs emerge as novel biomarkers for various cancers. However, there is currently little information about miRNA profiling studies and biological effects of miRNAs in cellular senescence. The senescence program is established and maintained by p53 and retinoblastoma protein (pRb) tumor suppressor pathways. The requirements of p53 and pRb for the induction of cellular senescence vary in their prominence depending on the genetic context, species, and cell type (Adams, 2007; Schmitt, 2007; Haferkamp et al., 2009). Recently, various studies have indicated that some miRNAs, such as miR-34a and miR-20a, induce senescence-like growth arrest through regulating cell cycle genes and senescence-associated genes involved in the p53 and/or pRb pathway (Tazawa et al., 2007; Poliseno et al., 2008; Sun et al., 2008). Such miRNAs play a direct role in senescence and are called senescence-associated miRNAs (SA-miRNAs; Lafferty-Whyte et al., 2009). In the present study, we attempted to screen SA-miRNAs that control cellular senescence in human fibroblasts, and we report here that miR-22 is a novel SA-miRNA that functions in mediating cellular senescence. We studied the role of miR-22 in cellular senescence using human normal cells and cancer cell lines as an in vitro culture system as well as an in vivo mouse breast tumor model. Upon senescence, cells become flattened and enlarged and exhibit biochemical changes such as the increased perinuclear activity of senescence-associated β-galactosidase (SA-β-gal; Dimri et al., 1995; Narita et al., 2003). Another critical event during the cellular senescence process is a decrease in cell growth and cell motility. We found a widespread decrease of miR-22 expression in various human cancer cell lines. Introduction of miR-22 into cancer cells inhibits cell proliferation, accompanied by senescence-like cell morphology and a decrease in cell motility and invasiveness. We predicted the putative direct targets of miR-22 by the computational prediction of targets based on sequence match to the miRNA. We identified three targets, including CDK6, Sp1, and SIRT1, which are directly regulated by miR-22. Furthermore, silencing of these targets resulted in growth arrest and increased SA-β-gal activity, accompanied by pRB dephosphorylation. We confirmed that miR-22 regulated the pRb pathway of cellular senescence through targeting of CDK6 and SIRT1. Ectopic expression of CDK6, SIRT1, or Sp1 could partially rescue the senescence phenotypes in miR-22–transfected cells. Significantly, miR-22 injection suppresses tumor growth and metastasis in vivo by induction of senescence in breast tumor, suggesting that SA-miRNA miR-22 acts as an important barrier to cancer and plays an important role in tumor suppression. Our findings provide new insight for the role of SA-miRNAs between cellular senescence and tumorigenesis. Results miR-22 overexpression induces cellular senescence in human fibroblasts To identify miRNAs that control cellular senescence, we analyzed miRNA expression profiling by miRCURY locked nucleic acid (LNA) miRNA array in young and senescent TIG-3 fibroblasts (Fig. 1 A). We found that a set of altered expression miRNAs has been reported to be involved in cell growth and tumorigenesis (Fig. 1 A and Table S1). Among them, the majority of antigrowth miRNAs were expressed two– to fourfold more in senescent TIG-3 cells than in young cells. To confirm and validate the results, we performed 3D-Gene miRNA microarray to evaluate miRNA expression levels. Two kinds of microarray analysis showed that five miRNAs are uniformly up-regulated by twofold or greater in senescent compared with young cells. These include miR-22, miR-34a, miR-125a-5p, miR-24-2*, and miR-152, several of which, namely miR-34a and miR-125a-5p, are closely associated with senescence-like growth arrest and metastasis in cancer cells (Tazawa et al., 2007; Li et al., 2009; Wang et al., 2009). Here, we focused on miR-22 and evaluated the expression of miR-22 in young and senescent human diploid fibroblast strains. Quantitative (q) RT-PCR analysis confirmed that miR-22 expression was increased in senescent TIG-3 and other fibroblasts and even up-regulated by more than fivefold in senescent MRC-5 cells (Fig. 1 B). These findings suggest that miR-22 up-regulation is universal in senescent human fibroblasts. Figure 1. miR-22 is up-regulated in senescent human fibroblasts, mediating cellular senescence. (A) miRNA expression profile of TIG-3 fibroblasts was analyzed by miRNA microarray, presented as fold changes in miRNA expression between TIG-3S (senescent) and young (Y) cells. A set of altered expression miRNAs is indicated by red columns (see Table S1). RQ, relative quantitation. (B) Relative quantitation of miR-22 expression in different PDLs of fibroblasts was analyzed by qRT-PCR analysis. miR-22 expression levels in human fibroblasts were indicated, relative to those in TIG-3 (42 PDL) set at 1 in the left histogram and MRC-5 (43 PDL) set at 1 in the right histogram. U6 was used as an internal normalization control. The dashed line represents the threshold of expression level (twofold vs. TIG-3 42 PDL). (C and D) MRC-5 cells were transfected with cont miR or mature miR-22 (miR-22) for 6 d at indicated concentration. (C) qRT-PCR analysis shows the relative quantitation of miR-22 expression (vs. cont miR) in each transfection group. miR-22 expression levels in miR-22–transfected MRC-5 cells were indicated, relative to that in cont miR–transfected cells set at 1. U6 was used as an internal normalization control. (D) SA-β-gal activity was presented by the percentage of SA-β-gal–positive cells, which was indicated in different dose groups. (E) Cell proliferation assay was performed after transfection of 10 nM miR-22 or cont miR, and cells were counted for the indicated days. Each value was determined in triplicate. **, P 10 random fields, and the results were shown in the right histogram in contrast to MRC5S (senescent; 58 PDL). Data in all the panels represent mean ± SEM (n = 3). *, P 2 wk with a decreased percentage of SA-β-gal–positive cells (Fig. 2 E), and there appeared to be a significant decrease in cell size and percentage of cells distributed in a large-sized group (Fig. 2 F), compared with senescent anti-C–treated cells. Although stable knockdown of miR-22 exhibited neither the promotion of cell proliferation nor the extension of the life span (not depicted), this might be a result of irreversible growth arrest in senescent cells. These findings suggest the requirement of miR-22 in mediating senescence and indicate that miR-22 inhibition is indeed an obstacle for the progression of senescence in fibroblasts. miR-22 overexpression induces growth suppression and senescence-like phenotypes in human breast epithelial and cancer cells Senescence has been most widely studied in fibroblasts in vitro but is also well defined in other cell types, such as epithelial cells which are the origin of most carcinoma (Narita et al., 2003). Expression of human telomerase reverse transcriptase (hTERT) in certain cell types has been shown to extend cellular life span without malignant transformation. To investigate the effect of miR-22 on cellular senescence in human epithelial cells, we used hTERT-infected HMEC184 (184hTERT) cells that possess the unlimited proliferation capacity of HMEC184 cells and are regarded as immortalized. We found that miR-22 was expressed higher by >2.5-fold in senescent HMEC184 (22 PDL) than in 184hTERT cells that have a similar miR-22 expression level to that of normal young HMEC184 (Fig. 3 A). Compared with immortalized 184hTERT cells, a widespread decrease in miR-22 level was observed in various human cancer cells (Fig. 3 A), indicating that miR-22 may have an intrinsic function in tumor suppression associated with various human malignancies. Figure 3. Overexpression of miR-22 induces senescence-like phenotypes in human breast epithelial and breast cancer cells. (A) qRT-PCR analysis shows relative quantitation of miR-22 expression level (vs. 184hTERT) in human epithelial and various cancer cells. Expression levels of miR-22 in various cells were relative to that in 184hTERT cells set at 1. U6 was used as an internal normalization control. The dashed line represents the threshold of expression level (0.5-fold vs. 184hTERT). (B) Cell proliferation assay was performed after transfection of miR-22, and cells were counted for the indicated days, compared with control cells. Each value was determined in triplicate. *, P 90% according to our observations made using a fluorescence-labeled double-stranded siRNA. For rescue experiments, the expression plasmids used were pcDNA3.1-SIRT1 (a gift from B. Marshall, Gladstone Institute, San Francisco, CA), pCMVneo-CDK6 (provided by S. Van den Heuvel, Utrecht University, Utrecht, Netherlands), and CMV-Sp1 (Addgene). These constructs contain the encoding region of mRNA but lack the 3′-UTR of these genes. Cells were first transfected with miRNA in a 60-mm dish for 24 h and sequentially transfected with 0.25 µg of plasmid DNA using Lipofectamine LTX Plus reagent (Invitrogen) according to the manufacturer’s protocol. The overexpression of SIRT1, Sp1, and CDK6 was confirmed by Western blotting. After a 24-h incubation, cells were seeded to 35-mm-diameter dishes and incubated for 4 d until cell proliferation assay and SA-β-gal assay. Lentivirus infection Lentiviruses were generated by cotransfecting 0.9 µg of lentiviral vector (premiR-22, miRZip, anti–miR-22, or empty vectors; System Biosciences) and 2.7 µg of packaging plasmid mix (1:1:1 for 0.9 µg pPACK-H1-GAG, pPACK-H1-Rev, and pVSV-G) in 293T cells using Lipofectamine LTX Plus reagent. Supernatants were collected 48 h after transfection, filtered through a 0.45-µm membrane, and directly used to infect cells. Cells were observed and images were acquired using a 10× objective with fluorescent microscopy (Axiovert 200M; Carl Zeiss) in combination with a camera (AxioCam; Carl Zeiss) and AxioVision software (Carl Zeiss) at room temperature. Cell proliferation and SA-β-gal assay For cell proliferation assay, 48 h after transfection or infection, 3–5 × 104 cells were seeded in a series of 35-mm-diameter dishes and counted for the indicated days. For cytochemical and histochemical detection of SA-β-gal activity, SA-β-gal staining was performed as described previously (Debacq-Chainiaux et al., 2009). SA-β-gal–positive cells were quantified by counting positive and negative cells at 100× magnification in at least five random independent fields. Pictures were taken with a 10× phase-contrast objective on a light microscope (IMT-2; Olympus) with a camera using Image saver (AE-6905; ATTO) at room temperature. Automated image acquisition SiHa cells were seeded in a 96-well Viewplate (PerkinElmer) and transfected with cont miR, miR-22, or miR-34a for 72 h. To measure cell size and F-actin, cells were fixed by 4% PFA and stained with the actin marker Rhodamine phalloidin (1:40; Invitrogen; provided by S. Kobayashi and H. Kishi, Yamaguchi University, Ube, Japan). For BrdU quantitative analysis, cells were pulse labeled with 10 µM BrdU (Sigma-Aldrich) for 1 h at 37°C, incubated with 5% CO2, and fixed by 70% ice-cold ethanol for 30 min at room temperature. Cells were treated with 2N HCl for 20 min, neutralized with 0.2 M Tris-HCl, pH 7.5, and permeabilized with 0.1% Triton X-100 for 5 min, followed by the mouse anti-BrdU (1:200; Dako) and incubation for 1 h. Cells were then stained with anti–mouse AF488 (1:500; Invitrogen). DAPI (1 µg/ml; Dojindo) images were used for nuclear recognition and cell counting. Images were acquired in a fully automated and unbiased manner using a 10× objective with a spinning disk confocal microscope (Operetta; PerkinElmer) at room temperature. Eight images per well were collected to obtain a sufficient number of cells for reliable statistical analysis. Image correction and analysis were performed using custom-designed image analysis software (Harmony; PerkinElmer). The histogram in Fig. 7 A shows that F-actin was quantified using the texture analysis by Harmony software (PerkinElmer), and F-actin SER Valley represented the occurrence of stress fiber structures within cells. FACS analysis 48 h after transfection, cells were fixed in 70% ice-cold ethanol and stained with PBS containing 50 µg/ml propidium iodide and 100 µg/ml RNase A for DNA content analysis by flow cytometry analysis on a FACSCalibur system (BD). The percentage of cells in the various cell cycle phases was calculated using ModFitLT v2.0 software (Verity Software House). Apoptosis assays Cells were plated in an 8-well CultureSlide (BD), and apoptotic cells were detected with traditional or modified TUNEL assay using the DeadEnd Fluorometric TUNEL System (Promega) according to the manufacturer’s protocol. In modified TUNEL assay, Cy5-dUTP (GE Healthcare) was substituted for fluorescein-dUTP in a standard TUNEL reaction to detect apoptotic cells of GFP-expressed cells. Images were acquired using a 40× objective with fluorescent microscopy (Axiovert 200M) in combination with a camera (AxioCam) and AxioVision software at room temperature. Hybridization protection assay (HPA) and Southern blot analysis Cells were plated in a 100-mm culture dish and transfected, and DNA was extracted with a traditional phenol-chloroform 72 h after transfection. The lengths of total telomere and G-tail were determined using Southern blotting and HPA methods as described previously (Tahara et al., 2005). Western blotting 72 h after transfection or at day 6 after infection, cells were homogenized in lysis buffer (50 mM Tris-HCl, pH 8.0, 120 mM NaCl, 1% NP-40, 100 mM NaF, 0.2 mM Na3VO4, and Complete mini protein inhibitor cocktail [Roche]). 30 µg of proteins in the total cell lysate was separated by SDS-PAGE and transferred to polyvinylidene fluoride membrane. Antibodies to p53 (clone BP53-12; Millipore), phospho-Rb (ser807/811; Cell Signaling Technology), and β-actin (Sigma-Aldrich) were purchased, and anti-CDK6 (C-21), SIRT1 (H-300), and Sp1 (PEP2) antibodies were purchased from Santa Cruz Biotechnology, Inc. The secondary antibodies were HRP-conjugated anti–rabbit (NA 934V) and –mouse (NA 931V) antibodies (GE Healthcare). Immunoreactive bands were visualized using an ECL Plus kit (GE Healthcare), followed by exposure to x-ray film (RX-U; Fujifilm). The density of bands was densitometrically quantified using ImageJ (National Institutes of Health). Luciferase reporter assay The full-length 3′-UTRs of human SIRT1 and Sp1 were amplified by PCR from genomic DNA and cloned at the SacI and XhoI sites into pmirGLO vector (Promega). The 3′-UTR fragments of human CDK6 containing three putative miR-22 binding sites were also amplified from genomic DNA and cloned at the XhoI and SalI sites into pmirGLO vector. The sense and antisense oligonucleotides for the putative miR-22 binding site at the 3′-UTR of potential targets were annealed and cloned at the SacI and XbaI sites into pmirGLO vector. An internal NotI site was added to the oligonucleotides for clone confirmation. A positive control construct contains complete complementary mature miR-22 sequence. PCR primers and oligonucleotide sequences for constructs are provided in Table S2. All the constructs were further confirmed by sequencing. For luciferase activity analysis, each construct was cotransfected with miRNA duplex in a 96-well plate using DharmFECT Duo transfection reagent (Thermo Fisher Scientific) for 72 h, and luciferase assays were performed with the Dual-Luciferase reporter system (Promega) according to the manufacturer’s instructions. Luminescent signal was quantified by luminometer (Glomax; Promega), and each value from firefly luciferase construct was normalized by Renilla luciferase assay. Cell motility observation by fluorescence microscopy Cells were seeded in a 4-well 35-mm dish (Greiner Bio-One) at a density of 1,000 cells/well and grown for 48 h in culture medium. Before recordings were initiated, the multidishes were left for at least 30 min on the microscope stage for temperature equilibration. Time-laspe video recordings of live cells were performed for determination of cell motility. In brief, live cells from several nonoverlapping areas were recorded in 30-min or 1-h intervals over a period of 8–24 h using a CFl Plan Apo 10× objective with a fluorescence microscope (BIOREVO BZ-9000; KEYENCE) equipped with a motorized movable microscope stage. Recordings were stored as 8-bit 680 × 512 pixel images. The microscope stage contained a thermostatically controlled heating element and was surrounded by a Plexiglas incubator, thereby ensuring that live specimens could be maintained at 37°C during recordings. In vitro invasion assay 48 h after transfection, cells were resuspended in culture medium without serum and seeded at densities of 1.5 × 105 cells/well in 24-well Transwell inserts (8-µm pores; BD) coated with 50 µg Matrigel (BD). The lower chamber was supplemented with a medium with 10% FBS. After 48 h of incubation, the cells on the upper surface were scraped off, whereas the invasive cells attached to the lower surface of the membrane inserts were fixed and stained with hematoxylin. The invading cells were observed and counted from nine images (including at least 2,000 cells) in three fields of three membranes using a 10× phase-contrast objective under light microscopy (U-PMTVC; Olympus) at room temperature. Tumor imaging in vivo 5-wk-old female C.B17/Icr-scid (Scid/scid) mice (CLEA Japan, Inc.) were inoculated with MDA-MB-231-luc-D3H2LN cells into the fat pad on day 0 as a 1:1 mixture of ECM gel complex and cells at 2 × 106 cells/50 µl/site. The hsa-miR-22 duplex and cont miR with RNA-jetPEI (Polyplus Transfection) complex at the ratio of 1:1 in a volume of 100 µl (20 µg/site) were injected intratumorally every other day from day 13 to 31 after inoculation. The development of subsequent tumor growth and metastasis was monitored once a week by in vivo imaging. In brief, mice were injected with 150 mg/kg D-luciferin (Promega) intraperitoneally and imaged immediately to count the photons from the whole bodies using the IVIS imaging system (Xenogen) according to the manufacturer’s instructions. 10 min later, photons from firefly luciferase were counted. Data were analyzed using LivingImage software (version 2.5; Xenogen). Immunohistochemistry Immunohistochemical staining was performed with anti–Ki-67 monoclonal antibody (1:50; Dako) after antigen retrieval by microwave treatment in citrate buffer, pH 6.0, and detection by a streptavidin–biotin peroxidase system using the LSAB kit (Dako). The sections were incubated with primary antibody at 4°C overnight. A labeling index percentage of Ki-67 was determined by examining at least 500 tumor cells at 200× magnification in three representative and intensely stained areas using a 20× objective under light microscopy (U-PMTVC) at room temperature. The expression of Ki-67 was graded as high (>50% of positive cells) and low (<50% of positive cells). Statistical analysis Significance of differences between the treated samples and controls was determined by two-tailed t tests using Excel (Microsoft). P < 0.05 was considered statistically significant. Online supplemental material Fig. S1 shows the opposite effect of miR-22 overexpression and knockdown on apoptosis in human cancer cells. Fig. S2 shows that miR-22 has no effect on the length of total telomere or G-tail in human cancer cells. Fig. S3 shows an examination of BrdU quantity analysis, cell morphology area, and F-actin formation by automated image analysis. Fig. S4 shows the images of cell motility observation in senescent fibroblasts and Lenti-Pre22–infected MDA-D3 cells. Fig. S5 shows a scheme for the role of miR-22–induced senescence in cancer cells. Table S1 shows altered expression miRNAs identified by miRNA microarray function in cell growth and tumorigenesis. Table S2 shows PCR primers and oligonucleotide sequences for each construct of miR-22 putative targets in luciferase reporter assay. Videos 1–4 show cell motility observation of young and senescent MRC-5 cells, Lenti-C, and Lenti-Pre22–infected MDA-D3 cells, respectively. Online supplemental material is available at http://www.jcb.org/cgi/content/full/jcb.201010100/DC1.