High levels of oncomiR-21 contribute to the senescence-induced growth arrest in normal human cells and its knock-down increases the replicative lifespan

Cellular senescence of normal human cells has by now far exceeded its initial role as a model system for aging research. Many reports show the accumulation of senescent cells in vivo, their effect on their microenvironment and its double-edged role as tumour suppressor and promoter. Importantly, removal of senescent cells delays the onset of age-associated diseases in mouse model systems. To characterize the role of miRNAs in cellular senescence of endothelial cells, we performed miRNA arrays from HUVECs of five different donors. Twelve miRNAs, comprising hsa-miR-23a, hsa-miR-23b, hsa-miR-24, hsa-miR-27a, hsa-miR-29a, hsa-miR-31, hsa-miR-100, hsa-miR-193a, hsa-miR-221, hsa-miR-222 and hsa-let-7i are consistently up-regulated in replicatively senescent cells. Surprisingly, also miR-21 was found up-regulated by replicative and stress-induced senescence, despite being described as oncogenic. Transfection of early passage endothelial cells with miR-21 resulted in lower angiogenesis, and less cell proliferation mirrored by up-regulation of p21CIP1 and down-regulation of CDK2. These two cell-cycle regulators are indirectly regulated by miR-21 via its validated direct targets NFIB (Nuclear factor 1 B-type), a transcriptional inhibitor of p21CIP1, and CDC25A, which regulates CDK2 activity by dephosphorylation. Knock-down of either NFIB or CDC25A shows a phenocopy of over-expressing miR-21 in regard to cell-cycle arrest. Finally, miR-21 over-epxression reduces the replicative lifespan, while stable knock-down by sponges extends the replicative lifespan of endothelial cells. Therefore, we propose that miR-21 is the first miRNA that upon its knock-down extends the replicative lifespan of normal human cells.


Introduction
Cellular senescence is an irreversible growth arrest of normal somatic cells that is caused by telomere shortening, oxidative stress, DNA damage or by hyperoncogenic signalling (Ben-Porath & Weinberg, 2005). Thus, cellular senescence has been proposed to be a potent tumour suppressive mechanism that prohibits potentially damaged and genetically instable cells from further division via either p16 INK4A or p21 CIP1 as key executers of cell-cycle arrest (Ben-Porath & Weinberg, 2005). Indeed, senescence has been found to act as a tumour suppressor in mouse models in vivo (Cosme-Blanco et al., 2007;Feldser & Greider, 2007). Similarly, hyperoncogenic signalling by H-RAS has been found to induce senescence in normal cells unless a second cooperating oncogene is present (Serrano et al., 1997).
However, antagonistic pleiotropic effects of senescence have been reported as well. As senescent cells seem to persist and accumulate with age in different tissues (Campisi & Sedivy, 2009), their altered functional profile including a pro-inflammatory secretory phenotype changes the tissue microenvironment in ways that can promote both cancer and aging phenotypes (Krtolica & Campisi, 2002). This view is supported by the findings that removal of senescent cells (Baker et al., 2011) or re-elongation of telomeres (Jaskelioff et al., 2011) in transgenic mice can delay the onset of age-related pathologies. These functions of senescence have recently been summarized as the 'four faces' of senescence (Rodier & Campisi, 2011).
To better understand senescence and how the senescent phenotype of cells might contribute to aging or age-associated pathologies, we have been investigating the cellular senescencerelated changes in gene and protein expression profiles of various cell types (Chang et al., 2005;Laschober et al., 2010). We found that similar to fibroblasts also endothelial cells acquire a senescenceassociated secretory phenotype (eSASP), secreting high levels of proinflammatory cytokines like, for example interleukin-8 (Hampel et al., 2006). With regard to how the senescent phenotype is induced and maintained, only a few reports, however, have focused on the role of miRNAs (reviewed in Smith-Vikos & Slack, 2012;Schraml & Grillari, 2012) but still little is known on how miRNAs interact with cell-cycle regulators and functionally contribute to the senescent phenotype.
Here, we report the expression profiling of miRNAs in endothelial cell senescence using human umbilical vein endothelial cells (HUVEC) from 5 individual donors. We identified 12 miRNAs to be up-regulated in senescence, comprising hsa-miR-23a, hsa-miR-23b, hsa-miR-24, hsa-miR-27a, hsa-miR-29a, hsa-miR-31, hsa-miR-100, hsa-miR-193a, hsa-miR-221, hsa-miR-222 and hsa-let-7i. Surprisingly, also miR-21 was consistently up-regulated in replicative senescence as well as in oxidative stress-induced premature senescence (SIPS). Mimics of miR-21 introduced into early passage cells induce a cell-cycle arrest by targeting CDC25A and NFIB that in turn down-regulate CDK2 and induce p21 CIP1 . Since miR-21 is one of the first oncomiRs that have been reported (Si et al., 2007), these data suggest that in normal somatic cells miR-21 alone is sufficient to provide a hyperoncogenic signal limiting aberrant proliferation.

miR-21 is up-regulated in senescent HUVECs
To compare miRNA expression of early passage vs. replicatively senescent HUVECs Exiqon LNA microarray technology was used. We found 12 miRNAs (hsa-miR-21, hsa-miR-23a, hsa-miR-23b, hsa-miR-24, hsa-miR-27a, hsa-miR-29a, hsa-miR-31, hsa-miR-100, hsa-miR-193a, hsa-miR-221, hsa-miR-222 and hsa-let-7i) that were consistently up-regulated in the senescent cells of all donors (Fig. 1A), whereas only three miRNAs of the 17-92 cluster were down- Northern blot analysis of HUVEC strains (4 out of 5; for one donor RNA amounts were limiting) confirms miR-21 induction in senescent cells on the mature and pre-miRNA level. (D) Quantification of Northern blot analysis of the different senescent HUVEC strains normalized to 5S rRNA, and relative to early passage cells. (E) miR-21 is up-regulated in senescent human liver sinusoidal endothelial cells (HLSEC; left panel) and in human dermal fibroblasts (HDF; right panel) as analysed by qPCR. (F) miR-21 is up-regulated in stress-induced senescence. Senescence was induced in HUVEC by repeated tert-butyl hydroperoxide (tBHP) treatment (left panel), and in HDF by psoralen and UVA treatment (PUVA; right panel). All experiments were independently repeated at least 3 times. Y, early passage; S, senescent; Q, quiescent (**P < 0.01 t-test). Error bars represent mean AE standard deviation. Statistical analysis was performed using student's t-test. In cases where three groups were compared, statistical analysis was performed using ANOVA with Bonferroni as post hoc test. P-values indicated represent *P < 0.05; **P < 0.01; ***P < 0.001. regulated (Fig. 1A). Data have been submitted and are available at array express (E-MEXP-3596), and the senescent status of all 5 HUVEC strains was extensively characterized (Fig. S1). Microarray expression analysis was confirmed using independent methods ( Fig. 1B-D and S2). These findings are consistent with results of HUVEC strains from four different donors published recently (Hackl et al., 2010). Since more miRNAs were up-regulated than down-regulated, we also tested if this shift towards upregulation of miRNA expression in senescence might be due to increased expression of the miRNA-processing protein Dicer in senescent cells. This does not seem to be likely, since Dicer mRNA was down-regulated in senescent cells by 2-fold (Fig. S3A).
From the 12 up-regulated miRNAs, we decided to follow-up on miR-21 being the most prominent miRNA up-regulated in senescence. First, we confirmed that expression levels were significantly (P < 0.05) higher in senescent HUVECs by qPCR in all 5 donors (Fig. 1B) and by Northern blot analysis (Fig. 1C,D). The latter shows that the pre-miR-21 is also up-regulated, suggesting transcriptional control of miR-21 in senescence. To see whether such an up-regulation of miR-21 in senescence is a general phenomenon, we tested other replicatively senescent cell types like human liver sinusoidal endothelial cells (HLSEC) and human dermal fibroblasts (HDF; Fig. 1E). Since miR-21 levels in quiescent cells resemble the levels of exponentially growing cells ( Fig. 1B-E), the observed up-regulation seems senescence specific in all cell types tested. In addition, stress-induced senescence of either HUVECs after repeated tBHP (tert-Butyl hydroperoxide) treatment or fibroblasts after PUVA (psoralen plus UVA) treatment resulted in up-regulation of miR-21 (Fig. 1F). Four weeks after PUVA treatment, when the senescent growth arrest was permanently established, miR-21 expression was significantly elevated. This trend was already observed 72 h after treatment, although it was not statistically significant at that time point.
These data suggest that miR-21 might be generally up-regulated in cellular senescence and might play a role in either its induction or maintenance. The observed increase in miR-21 expression was in all cases statistically significant.

miR-21 over-expression induces growth arrest
To test whether miR-21 over-expression induces a senescence-like growth arrest, a synthetic miR-21 mimic was transfected into early passage HUVECs. Elevated levels of miR-21 for at least 72 h after transfection were confirmed by TaqMan-assays ( Fig. 2A). The consequences of miR-21 over-expression in early passage HUVECs were a reduced proliferation and an overt change in morphology as compared to cells transfected with a scrambled control miRNA. The phenotypic changes were most obvious 48 h after transfection (Fig. 2B), and at that time, the effect of miR-21 on cell proliferation was quantified. Using MTT assay as a measure of cell numbers, we observed a highly significant reduction by more than 35% as compared to control (Fig. 2C). To verify that the decrease in cell numbers is due to reduced proliferation and not to increased cell death, we determined apoptosis in miR-21 over-expressing cells. While no differences in the levels of apoptotic cells were observed after miR-21 transfection, the use of an antagomiR to inhibit endogenous miR-21 increased in the percentage of apoptotic cells, from 6% to 12% (Fig. 2D). Furthermore, miR-21 transfected HUVECs displayed a decreased neo-angiogenic potential similar to what we observed in senescent cells expressing high levels of endogenous miR-21 ( Fig 2E,F and S1E). Therefore, miR-21 over-expression mimics various aspects of cellular senescence of endothelial cells.

miR-21 over-expression induces cellular senescence
To further investigate the effect of miR-21 on the cellular lifespan, we transfected HUVECs with lentiviral particles encoding hsa-miR-21 or a nontargeting control shRNA and selected stable clones. Successful transduction was easily monitored by flow cytometry, since GFP was expressed under the control of the same promoter as miR-21.
We generated 2 independent stable miR-21 over-expressing and 2 miR-scrambled sequence control HUVEC lines (referred to as miR-21 and miR control HUVEC, respectively). Using the same MoI of virus, miR-21 over-expression significantly reduced colony-forming ability of HUVECs in terms of numbers of colonies ( Fig. 3A) as well as size of colonies (data not shown) as compared to HUVECs stably transduced with miR control. To follow-up on this initial indication that stable miR-21 over-expression reduces the proliferative capacity, we determined cell numbers when selection was completed (referred to as PD1pT) and at two subsequent passages. Stable miR-21 over-expression (Fig. 3B) indeed significantly reduced cell proliferation ( Fig. 3C) consistent with the transient over-expression results above (Fig. 2C). Cell growth was further monitored during the entire replicative lifespan, and indeed, miR-21 over-expressing cells underwent replicative senescence earlier than control cells (Fig. 3D). Furthermore, miR-21 over-expression led to an increase in senescent cells, as the majority of cells stain positive for senescenceassociated ß-galactosidase (SA ß-gal) at PD1pT after selection of stable cells (Fig. 3F,G). However, some cells resume growth, undergo additional 19 population doublings and finally reach replicative senescence at PD20pT, while control cells keep proliferating up to PD26pT. Sustained miR-21 over-expression, albeit lower than after selection of the cells, at later PDLs was confirmed by qPCR ( Fig. 3E). This suggests that the amount of miR-21 impacts on the proliferative capacity of the cells.
Then, we tested if knock-down of miR-21 by sequestration using a completely complementary sponge would lead to extension of the replicative lifespan. This is indeed the case (Fig. 3H). The miR-21 sponge was transcribed as a fusion with the 3′-UTR of a GFP cassette. Accordingly, binding of miR-21 results in cleavage of the GFP mRNA allowing to monitor functionality of the miR-21 sponge by GFP suppression (Fig. 3I).
Taken together, these results indicate that miR-21 levels modulate the replicative lifespan of normal human cells.
Stable miR-21 over-expression induces p21 CIP1 and pCDK2 as effectors of growth arrest When we analysed predicted targets of miR-21, we found CDC25A and NFIB, both of which have recently been confirmed as targets of miR-21 (Fujita et al., 2008;de Oliveira et al., 2009;Wang et al., miR-21 triggers growth arrest of endothelial cells, H. Dellago et al. 2009). CDC25A is a phosphatase removing an inhibitory phosphate residue at Thr14 and Tyr15 from CDK2 (Fernandez-Vidal et al., 2008), whereas NFIB is a transcriptional repressor of p21 CIP1 (Ouellet et al., 2006).
Using a dual reporter assay that allows normalization for intraexperimental variations, we confirmed that also in HUVECs CDC25A (Fig. 4A) and NFIB (Fig. 4B) are direct targets of miR-21.
In consequence, the downstream effectors of NFIB and CDC25A, p21 CIP1 and pCDK2, respectively, are up-regulated upon mir-21 over-expression (Fig. 4C). Thus, miR-21 over-expression in early passage HUVECs showed the expected effect on miR-21 targets NFIB and CDC25A, along with a corresponding regulation of their respective downstream targets, 21 CIP1 and phospho-CDK2, reminiscent of that of senescent cells.
These results suggest that the miR-21 induced growth arrest might be independent of well-acknowledged regulators of senescence-like p53 and p16. Indeed, no differential transcription of these mRNAs was observed (Fig. S4A). To confirm independence of p53, we used the same lentiviral particles for miR-21 and control miR with HeLa cells, in which p53 is highly instable due to rapid E6oncoprotein-dependent degradation (May et al., 1991). Strikingly, miR-21 over-expressing Hela cells formed significantly less colonies during selection of stable clones, and colonies were also significantly smaller (Fig.S4B,C), even though transduction efficiency, measured Stable miR-21 over-expression regulates growth potential and senescence. (A) HUVEC were transduced with lentiviral particles over-expressing miR-21 or a nontargeting control miR. Selection pressure was applied and colonies were counted at day 10 post-transduction. Error bars represent mean AE standard deviation obtained from three biological replicates. Statistical analysis was performed using Student's t-test. P-values indicated represent *P < 0.05; **P < 0.01; ***P < 0.001. (B) qPCR TaqMan analysis of miR-21 fold changes of stable miR-21 over-expressing HUVEC at PD1pT. miR-21 expression was normalized to snRNA-U6, and fold change was calculated relative to cells transfected with a nontargeting control shRNA. (C) HUVEC cell number was determined before seeding for transduction and at two or three passages post-transduction for miR-21 and miR control, respectively. (D) Growth curves of stable miR-21 and miR control over-expressing HUVEC. Population doublings at the first three passages were calculated according to cell numbers and, at later passages, according to split ratios. (E) qPCR taqman analysis of miR-21 fold changes of stable miR-21 over-expressing HUVEC at PD7pT. MiR-21 expression was normalized to snRNA-U6 and fold change was calculated relative to cells transfected with a nontargeting control miRNA. as percentage of GFP-positive cells 48-h post-transduction, was higher for miR-21 than for miR-control (Fig. S4D). However, we cannot exclude that miR-21 acts downstream of p53. While p53 is supposed to induce miR-21 in a p53-inducible cancer cell line (Tarasov et al., 2007), no correlation between miR-21 and p53 expression was observed in cancer tissues (Rask et al., 2011;Math e et al., 2012). However, taken together our data suggest a model for miR-21-induced growth arrest via the two branches of CDK2 and p21 CIP1 (Fig. 4D).

CDC25A and NFIB knock-down phenocopies the effect of miR-21 over-expression in HUVECs
To confirm that the growth arrest of miR-21 over-expressing cells is indeed executed via CDC25A and NF1B, we knocked down CDC25A and NF1B to phenocopy miR-21 over-expression. To this end, we transiently transfected early passage HUVECs with siRNAs against NFIB and CDC25A. Indeed, knockdown of both CDC25A and NFIB inhibited proliferation of HUVECs as compared to cells transfected with a control siRNA. Forty-eight hours after transfection a reduction in cell number by 50% was observed by quantification of randomly chosen microscopic fields (Fig. 5A,B). Furthermore, we tested the effect of individually knocking down either CDC25A or NFIB. Surprisingly, either of these knock-downs was sufficient to induce cell-cycle arrest (Fig. 5C,D). Efficient knockdown of CDC25A and NFIB was evident on protein level by reduced CDC25A and NFIB protein (Fig. 5E).
These data support the hypothesis that miR-21 targets two branches of cell-cycle inhibitory pathways that individually are sufficient to reduce cell proliferation. Interfering with both branches enhances the growth inhibitory effect. We suggest that miR-21 by targeting separate branches converging on the same final outcome The suppression is specific to the miR-21 seed region within the CDC25A 3′-UTR or NF1B 3′-UTR, as mutation of the miR-21 target sites alleviated repression by miR-21. Ratios of CDC25A-or NF1B-dependent renilla luciferase signal (rLuc) to a firefly luciferase signal (fLuc) encoded by the same vector are given. *P < 0.05; **P < 0.01; ***P < 0.001. (C) Stable miR-21 or miR control over-expressing HUVEC at PD1 post-transduction were harvested and lysed in SDS loading dye. Samples were subjected to Western blotting with antibodies against pCDK2 and p21 CIP1 , respectively. p21 CIP1 and pCDK2 were increased in stable miR-21 HUVEC. Quantitation of bands was carried out using the Odyssey software (Licor) Values indicated are mean AE standard deviation obtained from three biological replicates. (D) Model of miR-21 dependent proliferation arrest. miR-21 targets 2 complementary pathways that converge on cell-cycle arrest. MiR-21 (left side) knocks-down its direct target CDC25A. Resulting low levels of the phosphatase CDC25A are insufficient for removing inhibitory phosphate residues from CDK2, thus inhibiting cell-cycle progression. In the second branch (right side), miR-21 directly knocks-down NFIB, which is a transcriptional repressor of p21 CIP1 . Low levels of NFIB in turn allows p21 CIP1 protein levels to increase and inhibit cell proliferation.

miR-21 is up-regulated upon stress-induced senescence in vivo
PUVA is combined treatment of the photosensitizer psoralen and UV-A radiation. This treatment is used against psoriasis and various other dermatoses and results in premature skin aging in vivo and in cellular senescence in vitro (reviewed in Wlaschek et al., 2003). Therefore, we exposed C57BL/6mice to PUVA doses reported to induce cellular senescence in vivo (Singh et al., 2010). Forty-eight hours after treatment, we observed a significant induction of miR-21 using at least 5 animals per group (Fig. 6A). The miR-21 upregulation was concomitant with p21 CIP1 up-regulation (Fig. 6B), and the relative expression of p21 CIP1 to miR-21 correlated quite well (r 2 = 0.52). These data are well in line with our previous observation that miR-21 is significantly up-regulated in foreskin of old vs. young donors (Fig. 6D, data re-plotted from supplements of (Hackl et al., 2010) as well as p21 CIP1 levels as published recently (Hackl et al., 2010). Taken together, our data suggest that miR-21 is induced in normal human cells undergoing replicative or stress-induced senescence in vitro and in vivo and contributes to inducing and/ or maintaining the senescence-associated growth arrest.

Discussion
Here, we report that miR-21, one of the most prominent onco-miRs, is up-regulated during replicative and stress-induced senescence of human umbilical cord vein endothelial cells, sinusoidal liver endothelial cells as well as in human dermal fibroblasts. Since senescent endothelial cells are present in vivo at sites of atherosclerosis (reviewed by (Erusalimsky, 2009), this seems well in agreement with high levels of miR-21 in human atherosclerotic plaques (Raitoharju et al., 2011). However, one recent report found down-regulation of miR-21 in senescent human aortic endothelial cells (Weber et al., 2010), and it remains to be elucidated if such differences are due to differences of the endothelial cell type or cultivation conditions.
In any case, several findings connect miR-21 to the cardiovascular system. Mir-21 is up-regulated upon various types of stress in rat vascular smooth muscle cells (VSMCs) (Ji et al., 2007;Lin et al., 2009), in cardiomyocytes after H 2 O 2 exposure , as well as in vivo, in rat carotid arteries after balloon injury (Ji et al., 2007), but also in human VSMCs exposed to hypoxia (Sarkar et al., 2010) or to ischaemic preconditioning (Dong et al., 2009) and in endothelial cells after shear stress (Weber et al., 2010). In all these settings, miR-21 up-regulation inhibits apoptosis and induces cell proliferation of VSMCs, contributing to the formation of neointima thickening in vivo.
Thus, in VSMCs and cardiomyocytes induction of miR-21 expression upon stress leads to cellular phenotypes resembling an anti-apoptotic and pro-proliferative effect that fits miR-21's role as an oncomiR. Indeed, neointimal lesion formation in proliferative vascular disease shares many similarities with cancers (Ross et al., 2001).
In contrast, when testing the effect of miR-21 over-expression on endothelial cell proliferation, we found an antiproliferative effect of miR-21, which is supported by similar findings by Fleissner (Fleissner et al., 2010). Although this observation may at first seem counterintuitive, one has to be aware of the well-known fact that miRNAs can have different effects in different cell types [reviewed by (Calin & Croce, 2006;Esquela-Kerscher & Slack, 2006)].
Since miR-21 is high in senescence, our results suggest that it might contribute to inducing or maintaining the senescent cell-cycle arrest (Fig. 5F). Strikingly, our findings that two separate pathways converging on endothelial cell-cycle arrest by targeting p21 CIP1 and CDK2 as key effectors support such an idea. Indeed, p21 CIP1 is known to be transcriptionally repressed by NFIB (Ouellet et al., 2006), which is an experimentally validated target of miR-21 (Fujita et al., 2008), so that high miR-21 decreases NFIB and p21 CIP1 is de-repressed. The second regulatory branch is by miR-21 directly targeting the phosphatase CDC25A (de Oliveira et al., 2009;Wang et al., 2009). Thereby, the cell-cycle inhibitory phosphorylation on CDK2 cannot be removed and reduces the cell-cycle promoting activity of CDK2 (Shen & Huang, 2012). This scheme is well in line with the recently proposed model of endothelial cell senescence, where a prominent role of reduction of CDK2 protein levels has been postulated (Freedman & Folkman, 2005).
These findings still seem to be very counterintuitive to the a plethora of reports showing that miR-21 over-expression is involved in carcinogenesis and tumour progression . In our opinion, these paradoxical data might be reconciled by the hypothesis that miR-21 acts and/or contributes to induce senescence triggered by hyperoncogenic signalling in normal cells similar to oncogenic Ras (Serrano et al., 1997). This idea is supported by the fact that H-RAS induces miR-21 (Loayza-Puch et al., 2010). It is intriguing to think that in this type of tumour, onco-miR-21-induced cellular senescence might slow down the disease progression, as by now hyperoncogenic-signallinginduced senescence has been shown to be a tumour suppressor mechanism in vivo in mouse models. Another similarity is shared by the type of tumour that is formed in H-RAS/MYC transgenic mice or in miR-21 transgenic mice. Both specifically develop B-cell lymphoma (Langdon et al., 1989;Medina et al., 2010).
Taken together, we suggest that miR-21 is a cellular senescenceinduced miRNA that contributes to induction and/or maintenance of a senescence-like cell-cycle arrest in normal human cells. Since overwhelming evidence identifies miR-21 as a bona fide onco-miR, this cell-cycle arrest might be due to hyperoncogenic signalling similar to the effect of over-expressing oncogenic H-RAS in normal human cells.
Human Liver Sinusoidal Microvascular Endothelial Cells are from Cell Systems (ACBRI 566).

PUVA treatment and stress-induced premature senescence
Subconfluent HDF were pretreated for 2 h with 40 ng mL À1 8methoxy-psoralen (8-MOP) in growth medium. For UVA, irradiation cells were incubated in PBS containing 40 ng mL À1 8-MOP and irradiated with 9 J cm À2 (H€ onle UV-technologie UVA-hand 25065 with a blacklight filter) controlled by a UV-meter (UV-Meter lC Nr.: 607439 FS UVA D1 E110 equipped with a sensor 16401). Nontreated and cells only irradiated with UVA served as a control.
The backs of C57BL/6 mice were shaved 1 day before PUVA treatment with an electric clipper. The mice were painted on their back with either 200 lL of 8-methoxypsoralen (8-MOP) in ethanol (1 mg mL À1 ) or 200 lL of vehicle (95% ethanol). UVA irradiation was performed after 30 min using a Sellamed 3000 UVA-1 therapy lamp (Sellas, Ennepetal, Germany) filtered for the emission at 340-400 nm at a distance of 20 cm from the dorsal skin of the mouse. The UVA dose used was 10 J cm À2 . Mice were sacrificed 24 h and 48 h after a single PUVA exposure. Dorsal skin was excised and immediately frozen in liquid nitrogen for RNA analysis. All mice experiments were performed according to the rules of the Austrian law.
HUVECs (PD 20) were treated with 75 lM tert-butyl hydroperoxide (tBHP) in growth medium for one hour per day on 5 consecutive days as described by Dumont et al. (2000). After each treatment, the cells were washed twice with PBS, and normal growth media were added to the cells. RNA was isolated on day 10, where no proliferation was detected anymore.
Senescence-associated beta-galactosidase SA ß-gal staining was performed according to standard protocols described by Dimri et al. (1995).

RNA isolation, quantification and quality control
Total RNA was isolated using Trizol reagent (Invitrogen, Life Technologies Ltd, Paisley, UK) according to the manufacturer's instructions. RNA concentration was determined by absorbance measurement on a Nanodrop ND-1000 spectrophotometer.

RNA labelling, microarray hybridization and scanning
About 2 lg of total RNA was labelled according to the manufacturer's instructions with Hy3 or Hy5 fluorescent dyes (Exiqon miRCURY Hy3/Hy5 labelling kit). Corresponding Hy3 and Hy5 reactions were pooled and reduced to a total volume of 25 lL by vaccum centrifugation. Thereafter, reactions were mixed with 25 lL 2 9 hybridization buffer and hybridization to Exiqon miRCURY LNAarrays (v8.0) was performed for 16 h at 60°C using a Tecan HS-400 station. Subsequently, microarrays were scanned with an Agilent Type B Scanner at 5 lm resolution, 16 bits per pixel and 100% PMT for both channels (532 and 635 nm). Feature intensities were extracted using GenePix 4.1 software, and the resulting GenePix result files (GPR-files) were processed under R 2.9.1 using the linear models for microarray data analysis package [LIMMA, (Smyth, 2004)]. Spot intensities were background corrected using the normexp function and normalized using local weighted linear regression (LOESS). For differential expression analysis, replicate spots of each miRNA probe were correlated, and linear models were fitted to the data (Smyth et al., 2005). Using a contrast matrix, paired comparisons of senescent vs. young HUVECs were calculated for each donor and tested for significance using moderated hypothesis tests (moderated t-statistic) as available in LIMMA. The resulting data were adjusted for multiple testing (p-adj.), by controlling the false discovery rate according to Benjamini and Hochberg (Benjamini et al., 2001). All miRNAs were ranked according to the adjusted P-values and a cut-off of p-adj. 0.05 was imposed.

Northern blot analysis
About 10 lg of total RNA was used as input for Northern blot analysis adapted from (Lau, 2008). In brief, samples were run on a denaturing 15% polyacrylamide gel at 30 W for 1-2 h. Prior to transfer, equal loading was verified by staining with SYBR-Gold (Invitrogen, 1:10 000 in 0.5 9 TBE) for 30 min and visualization on miR-21 triggers growth arrest of endothelial cells, H. Dellago et al. a Typhoon 9400 (Amersham, excitation 488 nm, emission 555 nm). Subsequently, RNA was transferred to positively charged nylon membranes (Pall Biodyne B 0.45 lm) on a semi-dry blotter and immobilized by UV-crosslinking with 100 mJ.
For hybridization, 20 pmol of anti-miR DNA-oligos (Invitrogen) was labelled for 1 h at 37°C with 75-150 lCi c-32P-ATP (6000 Ci mmol À1 ; GE-Healthcare, Uppsala, Sweden) using 1 lL T4 polynucleotide kinase (NEB, Ipswich, MA, USA). The probe was purified using G-25 columns (GE-Healthcare), mixed with hybridization buffer and hybridized to the membrane over night at 50°C. For detection, membranes were exposed to PhosphoScreen (Kodak, Rochester, NY, USA) and scanned with a PhosphoImager (Molecular Dynamics Storm 860). Quantification was performed using Image-Quant 5.2 software after local median background subtraction.  (Livak & Schmittgen, 2001). For quantification of mRNA expression, 100 ng of total RNA was reverse transcribed using Dynamo cDNA synthesis kit (Finnzymes, Thermo Scientific, Waltham, MA, USA). QRT-PCR was performed using SensiMix Plus SYBR-Green Mix (No UNG). Expression values were normalized to GAPDH and indicated as fold changes relative to control.

In vitro angiogenesis
The in vitro angiogenesis assay on matrigel (BD, Franklin Lakes, NJ, USA) was performed in 24-well cell culture plates according to the manufacturer's instruction using 5 9 10 4 cells cm À2 . Cells were incubated at 37°C in a humidified atmosphere containing 5% CO 2 . Microscopy was performed 8-h postinoculation onto matrigel. Experiments were conducted in triplicates using random pipetting of the cells onto matrigel to avoid technical bias. Quantitative analysis of tubule complexes was performed by Angioquant software (Niemisto et al., 2005). Angioquant files were extracted with Excel to additionally perform paired Student's t-test to obtain P-values.

Cell proliferation
MTT test was used to monitor metabolic activity as an indirect measure for proliferation. It was performed in 96-well culture plates using 8 9 10 3 cells/0.3 cm 2 . 2 lg lL À1 MTT was added, and cells were incubated 4 h at 37°C. Following, cells were lysed using 100 lL per well 10% SDS in 0.01 M HCL and incubated for 12 h at 37°C. Measurement was taken by 570 nm/690 nm measurement on Tecan Infinite M200 reader (Tecan, Mä nnedorf, Switzerland).
To monitor growth, pictures from randomly chosen microscopic fields were taken, and the cell number was determined. A minimum of 6 microscopic fields and at least 200 cells were counted.

Apoptosis
For staining, the Annexin-V-fluos staining kit (Roche, Basel, Switzerland) was used. Cells were harvested and centrifuged at 170 g for 10 min and the pellet washed with Annexin-V binding buffer (10 mM Hepes/NaOH pH 7,4, 140 mM NaCl, 5 mM CaCl 2 ). After centrifugation, the pellet was resuspended in Annexin-V/PI staining buffer (250 ng mL À1 propidium iodide, 200 ng mL À1 Annexin-V-Fluos, diluted in Annexin-V binding buffer). After 10 min incubation, samples were subjected to flow cytometric analysis. For compensation, cells were stained with either PI or Annexin-V alone. The analysis was performed using an excitation wavelength of 488nm and a 600-nm emission filter for detection of PI (FL-3) and a 535nm filter for Fluos (FL-1).

Dual-luciferase reporter gene assays
For Luciferase assays, HUVEC were used at no later PD than 10. For CDC25A, a 100 bp portion of the 3′UTR comprising the miR-21 target site (either wild-type or mutated) was cut out of a pLuc construct generously provided by Jian Yu (Ph.D., University of Pittsburg) and religated into psiCHECK2 (Promega, Madison, WI, USA). For NF1B, synthetic oligos comprising the miR-21-target site in the 3′UTR (wild-type and mutated) were cloned into psiCHECK2. HUVEC were cotransfected with 1 lg reporter plasmid and 15pmol miR precursor (control or miR-21) using the Nucleofector device (Amaxa) according to the manufacturer's instructions. Forty-eight hours post-transfection, cells were harvested and lysed by resuspending in Dual-Glo lysis buffer (Promega) and incubated for 5 min. Luciferase activity was measured on the Tecan Reader using the Dual-Glo Luciferase Assay System (Promega). Renilla luciferase signal was normalized against Firefly luciferase.

Statistics
If not indicated otherwise error bars represent mean AE standard deviation. When only two groups were compared, statistical analysis was performed using Student's t-test. ANOVA and Bonferroni post hoc tests were performed where indicated. Statistical significance is indicated by asterisks representing P-values (*P < 0.05; **P < 0.01; ***P < 0.001). Data obtained from PUVA-treated mice were analysed using ANOVA and Mann-Whitney U-test.

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site.   S2. Confirmation of miRNA micro-array profiling by QPCR and northern blot for a panel of additional, differentially expressed miRNAs in senescent HUVECs. Fig. S3. Northern blot analysis confirms miR-21 up-regulation, which is not due to dicer up-regulation in senescent cells. Fig. S4. Inhibitory effect of miR-21 on the proliferative potential of cultured cells does not involve p53 or p16 and is observed also in Hela cells. miR-21 triggers growth arrest of endothelial cells, H. Dellago et al.