Depending on the cell type and other factors p53 activation can result in apoptosis, reversible (quiescence) and irreversible (senescence) cell cycle arrest [^1-8]. While the choice between apoptosis and cell cycle arrest has been intensively scrutinized, the choice between quiescence and senescence was not systematically addressed and remains elusive. In order to observe whether p53 activation causes either senescence or quiescence, others and we employed nutlin-3a. Nutlin-3a, a small molecular therapeutic, inhibits Mdm2/p53 interaction and induces p53 at physiological levels without causing DNA damage [^9-11]. It was reported that nutlin-3a caused senescent morphology and permanent loss of proliferative potential [¹²,¹³]. However, in other cell lines nutlin-3a caused quiescence so that cells resumed proliferation, when nutlin-3a was removed [^14-16]. Moreover, we recently reported that in human fibroblasts (WI-38tert) and fibrosarcoma cells (HT-1080-p21-9), in which nutlin-3a caused quiescence [¹⁶], p53 acted as a suppressor of senescence [¹⁷]. Thus, ectopic expression of p21 in these cells caused senescence, while simultaneous induction of p53 converted senescence into quiescence [¹⁷]. In agreement with previous reports [^18-20], we found that p53 inhibited the mTOR pathway [¹⁷]. Importantly, the mTOR pathway is involved in cellular senescence [^21-26]. We suggested that p53-mediated arrest remains reversible as long as p53 inhibits mTOR. If this model is correct, then senescence would occur in those cells, in which p53 is incapable of suppressing mTOR. Here we provide experimental evidence supporting this prediction and demonstrate that irreversibility of p53-mediated arrest may result from its failure to suppress the mTOR pathway.

The role of p53 in organismal aging and longevity is complex [^28-32], indicating that p53 may act as anti-aging factor in some conditions. We have recently demonstrated that p53 can suppress cellular senescence, converting it into quiescence [¹⁷]. In these quiescence-prone cells, p53 inhibited the mTOR pathway, which is involved in senescence program (Figure 6A). Still p53 induces senescence in numerous cell types. Here we showed that in those cell types, in which nutlin-3a caused senescence, it failed to inhibit the mTOR pathway (Figure 6B). The role of active mTOR as a senescence-inducing factor in these cells was demonstrated by using rapamycin, which partially converted nutlin-3a-induced senescence into quiescence (Figure 6B, lower panel). This indicates that rapamycin-sensitive mTOR activity is necessary for senescence during nutlin-3a-induced cell cycle arrest. And vice versa, in quiescence-prone cells, depletion of TSC2 converted quiescence into senescence (Figure 6A, lower panel). Taken together, data suggest that activation of the mTOR pathway favors senescence (Figure 7). In agreement, Ras accelerated senescence in nutlin-arrested cells [¹³]. Similarly, activation of Ras and MEK in murine fibroblasts converted p53-induced quiescence into senescence [³³]. Interestingly, p53 levels did not correlate with the senescence phenotype, suggesting that factors other than p53 may determine senescence [³³]. These important observations are in agreement with our model that senescence requires two factors: cell cycle arrest caused by p53 and simultaneous activation of the growth-promoting mTOR pathway (Note: Ras is an activator of the mTOR pathway). And vice versa it was observed that induction of p53 maintains quiescence upon serum starvation, without causing senescence [³⁴]. In agreement, our model predicts that, by deactivating mTOR, serum starvation prevents senescence.

Another factor that favors senescence is the duration of cell cycle arrest [¹³,³⁵]. Importantly, the duration of the arrest may exceed the duration of treatment with nutlin-3a because of persistent induction of p21 even after removal of nutlin-3a in some cancer cell lines [³⁵]. Additional pathways may be involved in the senescence program. For example, nutlin-3a induces cytoskeletal rearrangement [³⁶]. We speculate that p53 affects not only rapamycin-sensitive mTORC1 but also the mTORC2 complex, given that mTORC2 controls the actin cytoskeleton [³⁷]. Also, p53 inhibits downstream branches of the mTOR pathway [³⁸,³⁹]. P53 stimulates autophagy [¹⁸,⁴⁰], which in turn is essential for life-extension by pharmacological manipulations (see [^41-44]). Finally, p53 affects cellular metabolism [^45-48] and this effect may contribute to suppression of cellular senescence and synergistically potentate metabolic changes caused by mTOR inhibition. The relative contribution of all these mutually dependent factors needs further investigations. The key role of mTOR in cellular senescence links cellular and organismal aging and age-related diseases.

Cell lines and reagents. HT-p21-9 cells are derivatives of HT1080 human fibrosarcoma cells, where p21 expression can be turned on or off using a physiologically neutral agent isopropyl--thio-galactosidase (IPTG) [¹⁶,^49-51]. HT-p21-9 cells express GFP. WI-38-Tert, WI-38 fibroblasts immortalized by telomerase were described previously [¹⁶,¹⁷]. Melanoma cell lines, MEL-9 (SK-Mel-103) and MEL-10 (SK-Mel-147), were described previously [⁵²,⁵³]. RPE cells were described previously [²¹,²²]. MEF, mouse fibroblasts isolated from 13-day embryos, were provided by Marina Antoch (RPCI) and maintained in DMEM supplemented with 10% FCS. Rapamycin (LC Laboratories, MA, USA), IPTG (Sigma- Aldrich, St. Louis, MO), nutlin-3a (Sigma-Aldrich) were used as previously described [¹⁷].

Lentiviral shRNA construction . Bacterial glycerol stocks [clone NM_000548.2-1437s1c1 (#10), NM_000548.x-4581s1c1 (#7) and NM_000548.2-4551s1c1 (#9)] containing lentivirus plasmid vector pLKO.1-puro with shRNA specific for TSC2 was purchased from Sigma. The targeting sequences are: CCGGGCTCATCAACAGGCAGTTCTACTCGAGTAGAACTGCCTGTTGATGAGCTTTTTG (#10), CCGG CAATGAGTCACAGTCCTTTGACTCGAGTCAAAGGACTGTGACTCATTGTTTTTG (#7) and CCGGCGACGAGTCAAACAAGCCAATCTCGAGATTGGCTTGTTTGACTCGTCGTTTTTG (#9).

pLKO.1-puro lentiviral vector without shRNA was used as a control. Lentiviruses were produced in HEK293T cells after co-transfection of lentivirus plasmid vector with shRNA or control vector with packaging plasmids using Lipofectamine2000 (Invitrogen). After 48h and 72h medium containing lentivirus was collected, centrifuged at 2000g and filtered through 0.22 uM filter. Filtered virus containing medium was used for cell infection or stored at -80 C. Cells were transduced with lentivirus in the presence of 8 mg/ml polybrene and selected with puromycin (1-2 mg/ml) for 4-6 days. Cells were treated with drugs either 24h after transduction or after puromycin selection for infected cells.

Colony formation assay. Plates were fixed and stained with 1.0 % crystal violet (Sigma-Aldrich).

Immunoblot analysis. The following antibodies were used: anti-p53 and anti-p21 antibodies from Cell signaling and anti-actin antibodies from Santa Cruz Biotechnology, rabbit anti-phospho-S6 (Ser240/244) and (Ser235/236), mouse anti-S6, mouse anti-phospho- p70 S6 kinase (Thr389), mouse anti-p21, rabbit anti-phospho-4E-BP1 (Thr37/46) from Cell Signaling; mouse anti-4E-BP1 from Invitrogen; mouse anti-p53 (Ab-6) from Calbiochem.

Beta-galactosidase staining. beta-Gal staining was performed using Senescence -galactosidase staining kit (Cell Signaling Technology) according to manufacturer's protocol.

The authors of this manuscript have no conflict of interests to declare.