Senescence promotes in vivo reprogramming through p16INK 4a and IL‐6

Summary Cellular senescence is a damage response aimed to orchestrate tissue repair. We have recently reported that cellular senescence, through the paracrine release of interleukin‐6 (IL6) and other soluble factors, strongly favors cellular reprogramming by Oct4, Sox2, Klf4, and c‐Myc (OSKM) in nonsenescent cells. Indeed, activation of OSKM in mouse tissues triggers senescence in some cells and reprogramming in other cells, both processes occurring concomitantly and in close proximity. In this system, Ink4a/Arf‐null tissues cannot undergo senescence, fail to produce IL6, and cannot reprogram efficiently; whereas p53‐null tissues undergo extensive damage and senescence, produce high levels of IL6, and reprogram efficiently. Here, we have further explored the genetic determinants of in vivo reprogramming. We report that Ink4a, but not Arf, is necessary for OSKM‐induced senescence and, thereby, for the paracrine stimulation of reprogramming. However, in the absence of p53, IL6 production and reprogramming become independent of Ink4a, as revealed by the analysis of Ink4a/Arf/p53 deficient mice. In the case of the cell cycle inhibitor p21, its protein levels are highly elevated upon OSKM activation in a p53‐independent manner, and we show that p21‐null tissues present increased levels of senescence, IL6, and reprogramming. We also report that Il6‐mutant tissues are impaired in undergoing reprogramming, thus reinforcing the critical role of IL6 in reprogramming. Finally, young female mice present lower efficiency of in vivo reprogramming compared to male mice, and this gender difference disappears with aging, both observations being consistent with the known anti‐inflammatory effect of estrogens. The current findings regarding the interplay between senescence and reprogramming may conceivably apply to other contexts of tissue damage.

The power of cellular senescence in inducing tissue remodelling has been further extended to processes of cellular reprogramming in vivo. The transgenic expression of the four transcription factors abbreviated as OSKM (Oct4, Sox2, Klf4, and c-Myc) (Takahashi & Yamanaka, 2006) in adult mice induces dedifferentiation and cellular reprogramming within multiple tissues (Abad et al., 2013;Ohnishi et al., 2014). However, in addition to reprogramming, the activation of OSKM also results in cellular damage and senescence, both in vitro (Banito et al., 2009) and in vivo (Chiche et al., 2017;Mosteiro et al., 2016). Therefore, OSKM induces two opposite cellular fates, namely senescence and reprogramming, that coexist in vivo in separate, but proximal, subsets of cells (Chiche et al., 2017;Mosteiro et al., 2016). Importantly, it has been demonstrated that senescence plays an active role in facilitating in vivo reprogramming through the paracrine action of the SASP, being interleukin-6 (IL6) a critical mediator (Chiche et al., 2017;Mosteiro et al., 2016). Of note, IL6 plays an important role also during in vitro reprogramming (Brady et al., 2013;Mosteiro et al., 2016). Moreover, the concept that senescence promotes cellular plasticity has been further extended to the activation of somatic stem/progenitor cells. In particular, the SASP can confer somatic stem/progenitor features onto proximal epithelial cells in several tissues (Ritschka et al., 2017).
The tumor suppressor genes p53, p21, Ink4a, and Arf act as cell-autonomous barriers for cellular reprogramming (Banito et al., 2009;Kawamura et al., 2009;Marion et al., 2009;Utikal et al., 2009;Zhao et al., 2008). These barriers are conceivably activated by cellular damages associated to reprogramming, most notably replication stress (Marion et al., 2009), which result in proliferation arrest and, consequently, inhibit reprogramming (Hanna et al., 2009). At the same time, p53 and the genetic locus Ink4a/Arf also affect reprogramming, although in opposite directions, through cell extrinsic mechanisms (Mosteiro et al., 2016). In the absence of p53, the induction of OSKM leads to exacerbated damage and senescence in tissues, which results in high levels of IL6 that further enhance reprogramming (Mosteiro et al., 2016). The Ink4a/Arf locus plays a complex role in reprogramming: it promotes reprogramming through the paracrine influence of senescence (Mosteiro et al., 2016), and, at the same time, it is a cell-autonomous barrier for reprogramming . In vivo, the absence of Ink4a/ Arf severely impairs OSKM-senescence, IL6 levels are modestly increased, and reprogramming is very inefficient (Mosteiro et al., 2016). Therefore, the positive cell-autonomous impact of Ink4a/Arf deficiency is completely obscured in vivo by the absence of senescence and IL6 secretion (Mosteiro et al., 2016). The emerging picture is that tissue damage and senescence provide a tissue microenvironment that is critical for OSKM reprogramming in vivo.
In this report, we dissect genetically the role of Ink4a, Arf, p53, p21, and Il6 on in vivo reprogramming.

| Ink4a is required for OSKM-induced senescence and reprogramming
The Ink4a/Arf genetic locus encodes two tumor suppressor genes, namely p16 Ink4a and p19 Arf , acting on separate pathways: p16 Ink4a is an inhibitor of the CDK46/cyclin D kinases and, therefore, functions as a positive regulator of the Retinoblastoma family of proteins; p19 Arf is an inhibitor of the ubiquitin ligase MDM2, thereby, stabilizing p53 (Gil & Peters, 2006). Previously, we have reported that the activation of an inducible OSKM transgene (abbreviated as i4F for inducible four factors) in mice simultaneously lacking Ink4a and Arf fails to trigger senescence and to produce high levels of IL6, and consequently, reprogramming is severely compromised (Mosteiro et al., 2016). To strengthen the role of IL6 during in vivo reprogramming, we tested whether reprogramming can be rescued in i4F; F I G U R E 1 Ink4a is required, and Arf is dispensable for in vivo OSKM-induced senescence and reprogramming. a. Percentage of dysplasia in the pancreas of i4F and i4F;Ink4a-null mice treated with 0.2 mg/ml doxycycline for 10 days and analyzed at the end of the treatment. Values correspond to the percentage (%) of pancreatic area affected (evaluated blindly). b. NANOG immunohistochemistry and SAbG staining in the pancreas of the same mice as in panel a. Images are representative for at least six mice (n ≥ 6) in NANOG and three mice (n ≥ 3) in SAbG. Values correspond to & of NANOG+ cells and % of SAbG area quantified with an automated software. c. mRNA levels of the indicated genes in the pancreas of the same mice as in panel a. The number of mice tested is: n = 5 in WT and Ink4a-null; n = 8 in i4F; n = 10 in i4F;Ink4anull. d. Percentage of dysplasia in the pancreas of i4F, i4F;Arf-null, and i4F;Ink4a/Arf-null mice treated with 0.2 mg/ml doxycycline for 7 days and analyzed at the end of the treatment. Values correspond to the percentage (%) of pancreatic area affected (evaluated blindly). e. NANOG immunohistochemistry and SAbG staining in the pancreas of the same mice as in panel d. Images are representative for at least 13 mice (n ≥ 13) in NANOG and three mice (n ≥ 3) in SAbG. Values correspond to & of NANOG+ cells and % of SAbG area quantified with an automated software. f. mRNA levels of the indicated genes in the pancreas of the same mice as in panel d. The number of mice tested is: n = 3 in WT, Arf-null and Ink4a/Arf-null; n = 10 in i4F; n = 12 in i4F;Arf-null and n = 6 in i4F;Ink4a/Arf-null. All the mice tested were males of 8-10 weeks of age. All values are expressed as average AE SD Statistical significance compared to WT controls was assessed using the unpaired two-tailed Student's t test with Welch's correction: p < .05, *; p < .01, **. Comparisons of each genotype to i4F controls are indicated in the same manner but using the symbol "#." Comparisons between i4F;Arf-null and i4F;Ink4a/Arf-null are indicated in the same manner but using the symbol "&" Ink4a/Arf-null mice by the exogenous administration of recombinant IL6 (rIL6). Of note, we have previously observed that rIL6 administration elevates in vivo reprogramming in wild-type i4F mice (Mosteiro et al., 2016). We focused on the pancreas because it is the tissue that shows the fastest and more widespread senescence, dedifferentiation, and reprogramming (Mosteiro et al., 2016). We observed that i4F;Ink4a/Arf-null mice treated with rIL6 tend to present a higher degree of reprogramming, assessed by the extent of pancreatic dysplasia, upon 7 days of doxycycline treatment (0.2 mg/ ml) compared to i4F; Ink4a/Arf-null controls ( Figure S1).
To genetically dissect the individual contribution of Ink4a and Arf during in vivo reprogramming, we generated reprogrammable mouse lines with null alleles for either Ink4a (Krimpenfort, Quon, Mooi, Loonstra & Berns, 2001) or Arf (Kamijo et al., 1997). After a short treatment with doxycycline, we quantified the following four readouts: the presence of senescent cells (by staining for senescenceassociated b-galactosidase or SAbG), the mRNA levels of several factors characteristically expressed by senescent cells (collectively known as the SASP), the extent of dysplasia (which we use as a surrogate for cell dedifferentiation and partial reprogramming), and the occurrence of advanced or full reprogramming (by staining for the pluripotency marker NANOG).
We compared i4F;Ink4a-null and i4F mice of mixed genetic background, but derived from the same parents and therefore with the same mixed genetic background. After 10 days of doxycycline treatment (0.2 mg/ml), we observed that the degree of pancreatic dysplasia ( Figure 1a) and reprogramming ( Figure 1b) was strongly decreased in i4F;Ink4a-null mice compared to their i4F counterparts.
In parallel to this, i4F;Ink4a-null pancreas presented limited senes- Next, we evaluated the extent of in situ reprogramming in the pancreas of i4F;Arf-null mice. We compared in parallel the behavior of i4F, i4F;Arf-null, and i4F;Ink4a/Arf-null mice, all in a pure C57BL6 genetic background. Interestingly, i4F;Arf-null mice presented similar levels of pancreatic dysplasia compared to i4F mice after 7 days of doxycycline treatment (0.2 mg/ml), while i4F;Ink4a/Arf-null pancreatic tissue was almost unaffected (Figure 1d). We also observed that the amounts of NANOG+ cells as well as senescent cells in the pancreas of i4F;Arf-null and i4F mice were comparable ( Figure 1e).
Finally, we confirmed that the expression of SASP factors in i4F;Arfnull pancreas was increased to similar levels as those observed in i4F pancreas upon OSKM activation ( Figure 1f). All together, these results demonstrate that Ink4a is required for OSKM-induced senescence and secretion of IL6, while Arf plays a minimal role in these processes. This is in agreement with our previous data showing that

| Absence of p21 increases OSKM-induced senescence and reprogramming
The cell cycle inhibitor p21 CIP1 participates in senescence (Muñoz-Esp ın & Serrano, 2014;Sharpless & Sherr, 2015), and it is a cell-autonomous barrier for in vitro reprogramming (Banito et al., 2009;Kawamura et al., 2009). The protein levels of p21 are elevated in tissues expressing OSKM and this occurs independently of p53 (Mosteiro et al., 2016) and, possibly, through post-transcriptional mechanisms (see above Figure 1c). To investigate the role of p21 during in vivo reprogramming, we generated reprogrammable mice deficient for p21 (Brugarolas et al., 1995). We first evaluated the reprogramming efficiency in the pancreas of i4F;p21-null mice compared to i4F controls, all of pure C57BL6 genetic background, after 7 days of doxycycline treatment (0.2 mg/ml).
Interestingly, in the case of i4F;p21-null pancreas, we observed that the degree of pancreatic dysplasia was significantly higher than in i4F pancreas ( Figure 2a). We also observed an increased accumulation of reprogrammed (NANOG+) and senescent (SAbG+) cells in the pancreas of i4F;p21-null mice compared to i4F controls ( Figure 2b). In agreement, the levels of SASP factors, such as Il6, Tnf, and Pai1, presented the same tendency, that is, i4F<i4F;p21-null ( Figure 2c). Therefore, p21 is a negative regulator of OSKM-induced senescence and reprogramming in vivo.

| Absence of p53 renders IL6 production and reprogramming independent of Ink4a
We have previously reported that p53-null mice present massive levels of senescence and reprogramming upon OSKM activation, and both processes are partially inhibited by in vivo treatment with anti-IL6 antibodies (Mosteiro et al., 2016). Given the essential role of Ink4a in senescence and IL6 secretion, we investigated its importance in a p53deficient context and, for this, we generated p53/Ink4a/Arf mice. Interestingly, we observed that, after 7 days of doxycycline treatment (0.2 mg/ml), i4F;p53-null;Ink4a/Arf-null mice had a similar degree of dysplasia as i4F;p53-null mice, all being in a pure C57BL6 genetic background ( Figure 3a). Similarly, the amount of reprogramming assessed by the presence of NANOG+ cells in the pancreas was comparable between i4F;p53-null and i4F;p53-null;Ink4a/Arf-null mice (Figure 3b).
We next evaluated the accumulation of senescent cells and observed that it was decreased i4F;p53-null;Ink4a/Arf-null pancreas compared to i4F;p53-null pancreas (Figure 3b), reinforcing the importance of Ink4a in the induction of OSKM-driven senescence (Mosteiro et al., 2016).
However, no significant differences were detected regarding the upregulation of the SASP between i4F;p53-null and i4F;p53-null;Ink4a/Arfnull mice ( Figure S2). This indicates that in a p53-null context, the SASP becomes independent of Ink4a/Arf. In this regard, it is known that p53 is a negative regulator of the SASP and its deficiency results in a massive upregulation of the SASP (Copp e et al., 2008). Our current observations with i4F;p53-null;Ink4a/Arf-null mice also suggest that the SASP, rather than the extent of senescence, is the key factor responsible for a high reprogramming efficiency. All together, we conclude that the absence of p53 has a dominant effect over the absence of Ink4a Arf p53 Fold induction relative to WT 20 30 40 # F I G U R E 2 p21 limits in vivo OSKM-induced senescence and reprogramming. a. Percentage of dysplasia in the pancreas of i4F and i4F; p21-null mice treated with 0.2 mg/ml doxycycline for 7 days and analyzed at the end of the treatment. Values correspond to the percentage (%) of pancreatic area affected (evaluated blindly). b. NANOG immunohistochemistry and SAbG staining in the pancreas of the same mice as in panel a. Images are representative for at least 9 mice (n ≥ 9) in NANOG and 3 mice (n ≥ 3) in SAbG. Values correspond to & of NANOG+ cells and % of SAbG area quantified with an automated software. c. mRNA levels of the indicated genes in the pancreas of the same mice as in panel a. The number of mice tested is: n = 6 in WT; n = 8 in p21-null; n = 9 in i4F; and n = 10 in i4F;p21-null. All the mice tested were females of 12-16 weeks of age. All values are expressed as average AE SD Statistical significance compared to WT controls was assessed using the unpaired two-tailed Student's t test with Welch's correction: p < .05, *; p < .01, **. Comparisons between i4F and i4F;p21-null are indicated in the same manner but using the symbol "#" Percentage of dysplasia in the pancreas of i4F;p53-null and i4F;p53-null;Ink4a/Arf-null mice treated with 0.2 mg/ml doxycycline for 7 days and analyzed at the end of the treatment. Values correspond to the percentage (%) of pancreatic area affected (evaluated blindly). b. NANOG immunohistochemistry and SAbG staining in the pancreas of the same mice as in panel a. Images are representative for at least four mice (n ≥ 4) except in i4F; p53-null; Ink4a/Arf-null (n = 2). Values correspond to & of NANOG+ cells and % of SAbG area quantified with an automated software. All the mice tested are males of 10-14 weeks of age with regard to cytokine secretion and this is key in determining the efficiency of reprogramming.

| Il6 mutation impairs in vivo reprogramming
Damaged and senescent cells promote the reprogramming of their neighboring cells through secreted factors, being IL6 one of the key mediators of this cross talk (Mosteiro et al., 2016). Moreover, IL6 plays an important role during in vitro reprogramming (Brady et al., 2013) and treatment with anti-IL6 antibodies decreases reprogramming both in vitro (Mosteiro et al., 2016) and in vivo (Chiche et al., 2017;Mosteiro et al., 2016), while exogenous administration of rIL6 enhances in vivo reprogramming (Mosteiro et al., 2016) (see also Figure S1). To further strengthen the role of IL6 during in vivo reprogramming, we generated reprogrammable mouse lines (i4F) combined with mutant alleles of Il6 (Kopf et al., 1994). These groups of mice were of mixed genetic background but were derived from the same parents, thereby sharing a comparable genetic background. We observed that after 7 days of doxycycline treatment (0.2 mg/ml), the pancreatic dysplasia in i4F;Il6mutant mice was much reduced compared to their i4F counterparts ( Figure 4a). In this genetic background and at the time point analyzed, the number of NANOG+ cells was too low to assess differences between genotypes (Figure 4b), but we observed a suggestive reduction in the levels of senescent cells in the pancreas (Figure 4b) together with a residual induction of Ink4a, Arf, and several SASP factors, such as Tnf, Il1a, Il1b, and Pai1 (Figure 4c). This is in agreement with the previously reported role of IL6 in the induction of paracrine senescence (Kuilman et al., 2008). All together, these results reinforce the critical role of IL6 for in vivo reprogramming. The number of mice tested is n = 5 in WT and Il6-mutant and n = 7 in i4F and i4F;Il6-null. All the mice tested were males of 9-18 weeks of age. All values are expressed as average AE SD Statistical significance compared to WT controls was assessed using the unpaired two-tailed Student's t test with Welch's correction: p < .05, *. Comparisons of each genotype to i4F controls are indicated in the same manner but using the symbol "#" noted that, upon OSKM induction, control female i4F mice presented reduced levels of senescence, dysplasia and reprogramming compared to control i4F males. To substantiate this quantitatively, we compiled our current data in young (8- ingly, we observed that aged mice undergo in vivo reprogramming more efficiently than young mice (Mosteiro et al., 2016). Aging also results in lower levels of estrogens in female mice (Nelson et al., 1981). Based on this, we hypothesized that aged female mice should have an in vivo reprogramming efficiency similar to the one in aged male mice. To address this, we compared the degree of pancreatic dysplasia in a group of old females and males (≥54 weeks of age).

|
Interestingly, the gender differences observed in young mice disappeared in old mice (Figure 5a).
In summary, the gender-specific difference observed in reprogramming and its attenuation with aging is consistent with the concept that the levels of IL6 strongly influence in vivo reprogramming and suggest that estrogens negatively regulate in vivo reprogramming.

| DISCUSSION
The activation of OSKM in vivo triggers not only the reprogramming of a small population of cells, but also induces damage in many other cells (Chiche et al., 2017;Mosteiro et al., 2016). We have previously identified a positive causal connection between damaged cells and reprogramming in tissues upon in vivo OSKM activation (Mosteiro et al., 2016). In this report, NFjB activity and IL6 production by estrogens (Kalaitzidis & Gilmore, 2017;Naugler et al., 2007). We have previously reported that aged mice are more prone to undergo reprogramming than young ones (Mosteiro et al., 2016) and we proposed that this may reflect the higher levels of systemic inflammation and IL6 characteristic of aged mice (Cevenini et al., 2013;Maggio et al., 2006). Estrogen levels are reduced during mouse female aging (Nelson et al., 1981), and indeed, we have also observed that reprogramming was similarly high in aged female and male mice.
We have also demonstrated that OSKM-induced senescence does not require p53 (Mosteiro et al., 2016). Indeed, tissues lacking p53 present a stronger damage response upon OSKM activation, leading to increased accumulation of senescent cells and increased reprogramming efficiency (Figure 6a-c). We wondered if Ink4a was also critical for senescence and reprogramming in the absence of p53. Triply deficient p53/Ink4a/Arf mice presented levels of senescence clearly lower than p53-null mice, but higher than in Ink4a-deficient mice (Figure 6a). Interestingly, however, p53/Ink4a/Arf triply deficient tissues behaved equal to p53-null tissues regarding their high levels of IL6 and reprogramming (Figure 6b,c). Therefore, the absence of p53 has a dominant effect that renders IL6 production independent of Ink4a. In fact, it is known that p53 is a negative regulator of the SASP and its deficiency results in a massive upregulation of the SASP (Copp e et al., 2008).
To shed some light on other known regulators of senescence and reprogramming, we investigated the role of p21. Mice deficient for p21 present higher levels of OSKM-induced senescence, SASP production, dysplasia, and reprogramming compared to control i4F mice (Figure 6a-c), indicating that it is a barrier for both OSKMinduced senescence and reprogramming. Of interest, p21 protein levels are notably increased upon OSKM activation, both in wildtype as well as in p53-null tissues (Mosteiro et al., 2016), and our current data indicate that this is not due to an upregulation of p21 mRNA levels (Figures 1c,f, 4c and S1). These observations suggest that, upon in vivo OSKM activation, the levels of p21 protein are elevated through post-transcriptional mechanisms that are known to play a role in p21 regulation (Scoumanne et al., 2011;Warfel & El-Deiry, 2013;Zhang & Chen, 2008). In this context, it is worth mentioning that the upregulation of p21 during developmentally programmed senescence is also independent of p53 (Muñoz-Esp ın et al., 2013;Storer et al., 2013).
All together, we have genetically dissected some of the key players in the cross talk between damage and reprogramming in vivo.

| SAbG staining of histological sections
SAbG staining was performed in tissue cryosections preserved in OCT freezing medium using the Senescence b-Galactosidase Staining Kit (Cell Signaling, #9860). Briefly, tissue cryosections of 12 µm were fixed at room temperature 5 min with a solution containing 2% formaldehyde and 0.2% glutaraldehyde in PBS, washed three times with PBS, and incubated 48 hrs at 37°C with the staining solution containing X-gal in N-N-dimethylformamide (pH 6.0). Sections were counterstained with nuclear fast red. Image analysis and quantification were performed in a completely automated manner as indicated above for the immunohistochemistry.

| Statistical methods
Mice were allocated to their experimental groups according to their predetermined genotype, and therefore, there was no randomization.
Investigators were not blinded to the experimental groups. Quantitative PCR data were obtained from independent biological replicates (n values correspond to the biological replicates, that is number of mice; technical replicates of the PCR were not considered in the n value). Data in Figures 1-5 are represented as average AE SD; data in Figure 6 are represented as average AE SEM. Statistical significance was assessed using Student´s t test (two-tailed, unpaired) with Welch´s correction (Pearson correction in the case of Figure 6d).

CONFLI CT OF INTEREST
Authors declare no conflict of interest.