Progression of irradiated mesenchymal stromal cells from early to late senescence: Changes in SASP composition and anti‐tumour properties

Abstract Genotoxic injuries converge on senescence‐executive program that promotes production of a senescence‐specific secretome (SASP). The study of SASP is particularly intriguing, since through it a senescence process, triggered in a few cells, can spread to many other cells and produce either beneficial or negative consequences for health. We analysed the SASP of quiescent mesenchymal stromal cells (MSCs) following stress induced premature senescence (SIPS) by ionizing radiation exposure. We performed a proteome analysis of SASP content obtained from early and late senescent cells. The bioinformatics studies evidenced that early and late SASPs, besides some common ontologies and signalling pathways, contain specific factors. In spite of these differences, we evidenced that SASPs can block in vitro proliferation of cancer cells and promote senescence/apoptosis. It is possible to imagine that SASP always contains core components that have an anti‐tumour activity, the progression from early to late senescence enriches the SASP of factors that may promote SASP tumorigenic activity only by interacting and instructing cells of the immune system. Our results on Caco‐2 cancer cells incubated with late SASP in presence of peripheral white blood cells strongly support this hypothesis. We evidenced that quiescent MSCs following SIPS produced SASP that, while progressively changed its composition, preserved the capacity to block cancer growth by inducing senescence and/or apoptosis only in an autonomous manner.

spite of these differences, we evidenced that SASPs can block in vitro proliferation of cancer cells and promote senescence/apoptosis. It is possible to imagine that SASP always contains core components that have an anti-tumour activity, the progression from early to late senescence enriches the SASP of factors that may promote SASP tumorigenic activity only by interacting and instructing cells of the immune system.
Our results on Caco-2 cancer cells incubated with late SASP in presence of peripheral white blood cells strongly support this hypothesis. We evidenced that quiescent MSCs following SIPS produced SASP that, while progressively changed its composition, preserved the capacity to block cancer growth by inducing senescence and/or apoptosis only in an autonomous manner.

| BACKGROUND
In multicellular organisms, many cell types, including stem cells, are in quiescent state, which is a reversible exit from the cell cycle with an absence of proliferation. This temporary cell cycle exit is a fundamental event for regulation of tissue development and homeostasis, indeed, an impaired control of cell cycle with aberrant proliferation or permanent cell cycle arrest may promote cancer and aging. 1,2 Quiescence delays cell senescence by reducing damage to macromolecules deriving from DNA replication, metabolic activity, gene transcription and translation.
Quiescent cells are, however, highly susceptible to extrinsic genotoxic damage, since they have reduced expression of repair factors and may live for a long time, thus accumulating multiple injuries. 2 Following an extrinsic genotoxic stimulus, cells, including those in quiescence, may either properly repair the DNA damage, or alternately unrepaired/ misrepaired DNA may trigger apoptosis or senescence. Senescent cells permanently exit the cell cycle, losing their primary functions and acquiring new ones. Senescent cells can contribute to organismal aging but also play a role in tissue development and wound healing. Senescence may counteract cancer but can also promote tumorigenesis. 3,4 The pleiotropic activities of senescent cells are due to an intrinsic dynamic state of senescence, which does not have a static endpoint. Following noxious stimuli, a cell entering senescence will have genetic and epigenetic changes that initially lead to stable cell-cycle arrest, sustained by P53 and retinoblastoma pathways (early senescence). Senescent cells change their metabolism and lysosomal activity and secrete a plethora of factors, collectively indicated as senescence-associated secretory phenotype (SASP). Senescent cells accomplish their divergent tasks by SASP autocrine/paracrine activity.
In our body, senescent cells are cleared by the immune system, but some of them may escape this patrolling phenomenon and may remain for a long time in tissues. As time goes by, senescent cells enter into a late stage (late/deep senescence) and modify the SASP composition, which will be enriched in pro-inflammatory factors. 4,5 Generally speaking, the beneficial effects of senescence (anti-cancer properties, wound healing, contribution to tissue development) are due to SASP produced by senescent cells not into their final stage, while negative senescence outcomes (cancer promotion, aging) are mainly due to pro-inflammatory SASP activity. It is then important to evaluate as SASP changes its composition over time and we need to see how these modifications affect its biological activities.
Mesenchymal stromal cells (MSCs) are a heterogeneous population containing stromal cells, fibroblasts, progenitor cells and stem cells that can differentiate into mesodermal derivatives. MSCs are present in the stroma of many tissues and through secretion of cytokines, growth and survival factors may contribute to tissue healing and homeostasis. 6 The tumour microenvironment mimics an injured tissue and hence tumour growth often appears associated with several types of stromal cells in a manner that overlaps wound-healing and tissuerepair processes. In tumours, MSCs and other stromal cells are recruited to establish a supportive stroma. 7 Any senescence phenomenon affecting MSCs may have a role in tumour growth and survival.
In this context, we wanted to evaluate the effects of SASP released by early and late senescent MSCs on cancer cell biology.
Several findings on senescence have been performed on proliferating cells that exit cell cycle and become senescent cells following a stressful stimulus. Nevertheless, in our body many cells are in a quiescent state and hence it is important to evaluate senescence of cells that were already resting when exposed to genotoxic stimuli.
In this context, in quiescent MSCs we induced senescence by x-ray irradiation and evaluated changes in SASP composition through progression from early to late senescence. We also evaluated the SASP anti-tumour properties on several colon cancer cell lines, chosen for differences in their genetic and epigenetic background, which affect their biological properties. 2.2 | Induction of stress-induced premature senescence (SIPS) with x-ray exposure Quiescent cells were exposed to 2Gy x-ray by using a Mevatron machine (Siemens, Milan, Italy) operating at 6 MeV at room temperature (RT). Following irradiation, cells were further cultivated in alpha-MEM with 0.5% ES-FBS for 10, 30 and 60 days (10D, 30D and 60D, respectively).
The non-irradiated quiescent sample (control) was incubated for 10 days in alpha-MEM with 0.5% ES-FBS.

| Harvest of conditioned media
Conditioned media (CM) from irradiated MSC cultures were harvested 10, 30 and 60 days post-x-ray exposure. To this end, cells were extensively washed with PBS1X and plated into a chemically defined serum-free culture medium for overnight (ON) incubation. Afterwards, CM containing the MSC secretomes were collected and filtered through 0.2 micron Nalgene sterile syringe filters (Thermo Fisher, Waltham, MA, USA). Then samples were centrifuged at 10,000 rpm for 10 min at RT. After this, we discarded the pellets, and collected CM. This procedure was performed to eliminate cellular debris and apoptotic bodies. The collected supernatants were carefully evaluated by Leica DM IL inverted phase contrast microscope (Leica, Wetzlar, Germany) to exclude the presence of residual debris. After this, the collected CM were stored at À80 C until use. As control, we collected CM from unirradiated quiescent MSCs grown for 10 days in alpha-MEM with 0.5% ES-FBS. For evaluation of secretome effects on cancer cell biology, the cancer cells were incubated in the proper culture media supplemented with 50% CM obtained from irradiated MSCs (10D, 30D and 60D) or from control cells (CT) for 10 days at 37 C in a humidified, 5% CO 2 atmosphere. The media were replaced three times during this incubation period.

| Peripheral blood mononuclear cells (PBMCs) isolation
PBMCs were obtained from healthy voluntary male donors (25-40 years of age) after informed consent. The procedure was approved by Campania University Ethical Committee (Prot. 0029471/i 14/10/2021). Cells were isolated according to our previously published protocol. 8 In detail, the cells were separated on a Ficoll density gradient (GE Healthcare, Milan, Italy), and the mononuclear cell fraction was collected and washed in PBS. The resulting cell pellet (PBMCs) was resuspended in a culture medium containing DMEM-HAM F12 (HiMedia, Einhausen, Germany) supplemented with 1% FBS (EuroClone, Pero, Italy) with or without CM (60D) and incubated at 37 C in a humidified, 5% CO 2 atmosphere O.N. before their use in Caco-2 cultures.

| Caco-2 and PBMCs co-cultures
Caco-2 cells were plated in a T25 flask at 4000 cells/cm 2 and O.N. incubated with their growth medium. Then, the culture medium was carefully aspirated, and the Caco-2 cells were cultivated in fresh medium containing PBMCs (40,000 cells/cm 2 ), in this way the PBMCs/Caco-2 cells ratio was 10:1. This ratio was selected according to previous findings. 9 The co-cultures were incubated for 5 days at 37 C with or without 50% D60 secretome supplementation.

| Soft agar assay
We evaluated the morphological transformation of cells with soft agar assay for colony formation, which is an anchorage-independent growth assay in soft agar, according to Wallert and Provost Labs. 10 In detail, the cells were seeded in 0.5 ml DMEM supplemented with 0.35% agarose (Sigma-Aldrich, St. Louis, MO, USA) and 20% FBS (EuroClone, Pero, Italy) in 35 mm Petri dishes pre-treated with 0.5% agar (Sigma-Aldrich) in DMEM with 20% FBS. After 21 days of incubation, the cells were centrifuged at 2000 rpm. Then the pellets were fixed in 100% methanol for 10 min at À20 C. Colonies were then stained with 0.01% (w/v) crystal violet (Sigma-Aldrich) in 25% methanol for 30 min. Subsequently, the cells were washed with PBS three times and resuspended in 100% methanol for 30 min. Colony number was determined by manual inspection.

| Migration/invasion assay
The migration experiment was carried out according to Zhu and co-workers. 11 In detail, 4 Â    2.12 | Immunocytochemistry (ICC) and senescence-associated beta-galactosidase

| Cell cycle analysis
We carried out a beta-galactosidase assay with ICC against Ki67 and pRPS6, as previously described. 12 Briefly, 2 Â 10 4 cells per well were

| Sample preparation for mass spectroscopy
Irradiated and control MSC cultures were incubated in serum-free media for 24 h. Subsequently, culture media (secretomes) were collected from each sample. Culture debris were eliminated by centrifugation at 10,000 g for 10 min, and supernatants were used for the StartaClean beads protein pooling according to a procedure we have already described. 13 2.17 | LC-MS/MS (liquid chromatography-mass spectrometry/mass spectrometry) analysis
PANTHER allowed the GO analysis by classifying protein contents according to "biological process" ontological terms. 14 For PANTHER analysis, we used the statistics overrepresentation, which compares classifications of multiple clusters of lists with a reference list to statistically identify the over-or underrepresentation of PANTHER ontologies. Significance was set to a p value of 0.05.
The secretome proteins were mapped to specific pathways with REACTOME analysis. We carried out an overrepresentation and a pathway-topology analysis. Overrepresentation analysis determines whether certain specific REACTOME pathways are enriched in the submitted protein dataset. This analysis produced a probability score, wherein the false discovery rate (FDR) was corrected for using the Benjamini-Hochberg method. We followed the developers' instructions for running a REACTOME analysis. 15,16 The building of a protein network with NETWORKANALYST allows a clear resolution of the biological context of the analysed proteins. The most relevant biological pathways and molecules that interact with the protein list of interest are assembled into a functional whole. 17 We generated protein networks with a Minimum IMEx Interactome Network, which curtails the networks and holds only seeds and their connecting nodes. Within the generated networks, we selected only nodes having a node degree >20 and betweenness >200.

| Statistical analysis
Statistical significance was evaluated using ANOVA followed by Student's t and Tukey tests. All data were analysed using the GraphPad Prism version 5.01 statistical software package (GraphPad Software, San Diego, CA, USA).

| RESULTS
Several findings on senescence have been performed on proliferating cells that become senescent cells following a stressful stimulus. Nevertheless, in our body many cells are in a quiescent state and hence it is important to evaluate effects of genotoxic clues on resting cells.
In a previous finding, we demonstrated that, following 48 hours serum starvation, the great majority of MSCs in culture left cell cycle and entered quiescence. 12 We induced quiescence of MSCs with serum starvation and then irradiated them with 2 Gy x rays to cause senescence. Cells were further cultivated up to 60 days in conditions that preserved quiescence status but provided minimum nutrients and healthy conditions (0.5% serum supplement).

| Quiescent MSCs entered senescence following x-ray exposure
We performed biomolecular analyses at 10, 30 and 60 days postirradiation. Hereinafter, these time points are designated as 10D, 30D and 60D, respectively. As a reference, we used unirradiated cells at 10 days of cultivation in "quiescence medium" (CT). We did not use unirradiated cells at 30 days and 60 days post-quiescence onset, since prolonged cultivation of quiescent cells induces senescence and this can introduce a bias in our reference system. Indeed, long term quiescent cells may enter senescence by reduction of lysosomal function and/or other mechanisms, such as oxidative stress. 1,2,18 We identified cycling, quiescent and senescent cells by the algorithm we set up previously. 12 The cycling cells were positive for Ki67 and phosphorylated ribosomal RP6 (pRP6).
At the same time, they were negative for senescence-associated acid beta-galactosidase (β-gal). The quiescent cells were Ki67(À), pRP6(À) and β-gal(À), while the senescent cells were Ki67(À), pRP6 There are findings showing that within a few hours after genotoxic damage cells show a rapid increase in SA-β-gal activity. This event does not pinpoint a senescent phenotype; rather, it evidences stressed cells, which can either successfully cope with the injury and recover a functional status or wilt and enter senescence. In our previous study we showed that stressed cells were Ki67(+), pRPS6(+) and SA-β-gal(+). 12 The stressed cells show increased SA-β-gal activity but are still positive for the Ki67 cycling marker. In irradiated cultures, we observed an increase in stressed cells, besides augmentation of senescent ones ( Figure 1A).
Following genotoxic damage, MSCs are generally more prone to senescence than apoptosis. 19 Nevertheless, the percentage of apoptotic cells at 10D in irradiated cultures changed significantly compared with the control sample ( Figure 1C). A huge increase of apoptosis was observed at 60D. This augmentation of dead cells may be due to prolonged cultivation with minimal serum supplement, or, alternatively, cultures at 60D were approaching a final post-senescence stage with degradation phenomena and cell death. 20 Senescence is associated with the presence of unrepaired/ misrepaired DNA damage and aberrant persistent activation of DNA damage-repair (DDR) machinery. 5,21 At 10D, in irradiated MSCs, we detected an increase in phosphorylated H2A.X (pH2A.X) in nuclei compared with control ( Figure 2A). The pH2A.X, which marks damaged DNA, persisted at 30D and 60D. Damaged DNA was associated with a persistent DDR signalling through activated ATM in cell nuclei (Figure 2A). Of note, an activated ATM, which is localized in the cytoplasm, can contribute also to the acquisition of F I G U R E 1 Biological parameters of MSCs following x-ray exposure. Panel A: representative images of cells stained to identify nuclei (DAPI), pRPS6, Ki67 and to determine β-gal activity. The graphs show the percentage of cycling, quiescent, stressed and senescent cells. In each graph, the * indicates the statistical difference between CT, chosen as the reference, and the other time points. Data are shown with standard deviation (SD), n = 3 (*p < 0.05, **p < 0.01, ***p < 0.001). We used Leica CTR500 microscope equipped with a DCF3000G digital monochrome camera. The β-gal activity was recorded as a grey-stain with this setting. This experimental approach was used to identify in the same cell, a marker emitting a signal in the visible light (β-gal) together with others expressing fluorescent signals. Panel B: Cell-cycle profiles of samples collected at different time points following x-ray treatment. The * indicates the statistical difference between CT, chosen as the reference, and the other time points. Data are shown with standard deviation (SD), n = 3 (*p < 0.05, **p < 0.01). Panel C: Flow cytometry chart of annexin V assay. The percentage of apoptotic cells is indicated in the associated right histogram. Data are shown with standard deviation (SD), n = 3 (*p < 0.05 and **p < 0.01 indicate statistical significance between the control and treated samples). a proinflammatory SASP. 22 We found an increase in the percentage of cells with activated cytoplasmic ATM in irradiated cells at 30D and 60D (Figure 2A).
The executive senescence program relies on the P53 and/or retinoblastoma pathways depending on cell type and animal species. 23 In irradiated cultures, at 30D and 60D, we observed a strong increase of P53 compared to reference culture ( Figure 2B). The initial stages of senescence relied on RB upregulation that progressively declined at 60D and was accompanied by an increase of  The mRNA levels were normalized to GAPDH mRNA expression, which was selected as an internal control. For every mRNA, the change in the expression level is compared with control culture (CTRL), whose expression was fixed as 1. The symbols ***p < 0.001, **p < 0.01 and *p < 0.05 indicate statistical significance between the control and irradiated samples. associated with SASP enriched in pro-inflammatory factors. De Cecco and co-workers identified some factors (IL6, IFNA, IFNB, CCL2, MMP3) whose mRNA levels can be considered as good markers to ascertain passage to the late-senescence stage. 27 Irradiated MSCs showed a continuous increase in IFNA and MMP3 going from 10D to 60D ( Figure 2C). In contrast, IL6 peaked at 30D and then declined, while IFNB and CCL2 did not reach levels above those observed in reference samples ( Figure 2C). Collectively, these results evidenced that SASP could be enriched in some pro-inflammatory factors during in vitro cultivation of senescent cells.

| The onset and establishment of senescence were associated with significant changes in SASP composition
We then decided to gain more insights into secretome composition obtained from irradiated MSCs to evaluate changes during progression to deep senescence. We performed a LC-MS/MS analysis of protein secretome content in the different experimental conditions. We identified 148 proteins in control quiescent cultures, while in irradiated samples we found 309, 306 and 249 at 10D, 30D and 60D, F I G U R E 3 Gene Ontology and REACTOME analysis. Panel A: Venn diagram to identify common and specific SASP components among the different experimental conditions. Panel B: main GO and REACTOME outcomes. The SASP contained both factors, which were common among control and irradiated samples (see upper boxes), and proteins that were specifically present in a given sample (see lower boxes). Our analysis was carried out on whole secretomes containing soluble factors and molecules encapsulated within extracellular vesicles. Panel C: Network analysis. The top pictures show Minimum IMEx Interactome Networks with reduced complexity obtained by considering only seeds and their connecting nodes. The bottom images show representative seeds (node degree >20 and betweenness >200). Common: indicates seeds that are present in all irradiated samples (10D, 30D, 60D). respectively (Supplementary file S1). We then performed a Venn analysis to identify which factors were exclusively present in a given condition and which were not. Of note, 147 of 148 proteins identified in control samples were not present in irradiated ones. This result indicates that senescence, following irradiation, induced a drastic change in the secretome composition ( Figure 3A, Supplementary file S1). The data is in line with the onset of senescence-specific biological activities. The irradiated samples contained a core of 143 common factors, while the others were specifically present in each of the three irradiated conditions or were in common between two of them ( Figure 3A, Supplementary file S1).
We We previously found these pathways in the secretome of active proliferating adipose tissue-derived MSCs indicating that these signalling factors are part of the major paracrine activities of MSCs, irrespective of cell physiological status and tissue of residence. 28 Two pathways that were exclusively found in the secretomes of 10D and 30D samples evidenced that senescence phenomena completely modified the paracrine signalling of MSCs ( Figure 3B

| The SASP coming from cells with different senescent stages showed overlapping paracrine activities on cancer cells
Having evidenced significant changes in secretome composition during the progression of senescence induced by irradiation, we aimed to ascertain how these modifications affected the SASP paracrine functions. We focused our attention on the SASP capacity to arrest the growth of tumour cells by inducing senescence and/or apoptosis.
Indeed, there are several findings showing that SASP from deep senescent cells may lose its anti-tumour properties and, on the contrary, may sustain cancer growth through the presence of proinflammatory factors that modulate immune system activity by creating a favourable tumour microenvironment. 3,5,31,32 We evaluated SASP paracrine function on three colorectal cancer cell lines: SW480, Caco-2 and HCT116. The choice was based on the consideration that the mutation patterns of these cell lines are representative of the genetic and epigenetic landscape of this tumour phenotype. 33 The Caco-2 cells had microsatellite stability, chromosomal instability, a wild-type KRAS, a wild-type PIK3CA and an E204X mutation in TP53. The HCT116 cells had microsatellite instability, chromosomal stability, a G13D mutation in KRAS, an H1047R mutation in PIK3CA and a wild type TP53. The SW480 had microsatellite stability, chromosomal instability, a G12V mutation in KRAS, a wild-type PIK3CA, and a R273H, P309S mutations in TP53. 33 We incubated the cancer cells for 10 days in medium supplemented with secretomes of irradiated MSCs (10D, 30D, 60D) and [Ki67(+), pRP6(+) and β-gal(À)] cycling cells declined when incubated with SASP from 30D and 60D samples compared with untreated samples (UT) and with CT secretome ( Figure 4A). In HCT116, the decline in cycling cells was observed only with respect to UT and not CT ( Figure 4A). Data on cycling cells were in concordance with EdU analysis ( Figure 4B). The pattern of quiescent cells was more complex ( Figure 4A). In Caco-2, the percentage of quiescent cells increased in samples treated with 30D and 60D secretome compared to UT and CT, while in SW480, the increase was observed only with 30D and declined with 60D. In HCT116, we did not observe an increase in quiescent cells but rather a decline in samples treated with 10D and 30D compared to UT and CT. These data must be globally evaluated by also considering the percentage of senescent and stressed cells (see below).
In SW480 and HCT116 cells, the percentage of senescent cells was low in control condition and increased with 30D and 60D treatment. In HCT116, the increase in senescence with 60D was lower than that observed with 30D incubation ( Figure 4A). The Caco-2 samples showed a very low number of senescent cells (around 1% in UT and CT), but this percentage increased significantly (around 9%) when cultures were incubated with 30D and 60D ( Figure 4A). The number of stressed cells increased in all three cancer cell phenotypes following incubation with 30D and 60D ( Figure 4A). Of note, HCT116 showed a significant increase in the percentage of apoptotic cells when treated with senescent secretomes, the SW480 evidenced an increase of apoptosis at 10D and 30D and a very significant augmentation at 60D compared with UT ( Figure 4C). The Caco-2 cells showed changes in apoptosis only in 60D treated samples. In some settings, the CT secretome produced some biological effects on cancer cells that are comparable to those induced by 10D, 30D, 60D secretomes.
It must be remembered that CT samples contain some senescent cells that can secrete SASP factors.
Loss of cell adhesion and invasion has been classically viewed as tumorigenic features of cancer cells. We then evaluated the effect of SASP on colon cancer cell capacity to grow on soft agar and to in vitro migrate. Generally, almost all the analysed SASPs negatively affected the migration and unanchored growth of the cancer cell lines we analysed ( Figure 4D,E). Of note, the 60D secretome induced the most significant impairment of these tumorigenic properties.

| The presence of immune cells impaired the anti-tumour properties of SASP
The observation that SASP coming from deep senescent cells preserved its anti-tumour properties is at odds with several findings showing that during progression from early to late senescence the SASP will be enriched in pro-inflammatory cytokines that can foster cancer growth. 4,34 In this context, it is possible to imagine that SASP always contains some core components that have anti-tumour activity,   35 These caveats may render uninformative a quantitative MS study on SASP. We used an experimental approach to minimize such hindrance. Our study aimed at identifying SASP features that were exclusive for every experimental condition. For this reason, the identification of ontologies, pathways and networks was parsed by Venn analysis to identify specific SASP characteristics for every experimental condition (10D, 30D, 60D), rather than quantitative evaluation of changes occurring in the expression of a given factor.
The bioinformatics evaluation of SASPs (GO, REACTOME and Network analysis) showed some common and specific features among 10D, 30D and late 60D SASP. The common ontologies present in the three SASPs refer to biological processes associated with protein depolymerization and organelle organization. This is in line with profound morphological and functional changes occurring in senescent cells compared with healthy ones, as we already evidenced in a previous finding. 36 In 10D SASP, there are ontologies specifically related to cytoskeleton remodelling, stress and chemical response. This data may suggest that in the early stages of senescence cells have to set up a proper and effective reaction to cope with an injury event. In 30D SASP, the specific ontologies belong to chromatin remodelling phenomena as expected for permanent activation of a senescencespecific gene expression profile. 37 In the 30D and 60D SASPs, we found ontologies encompassing metabolic processes involving organic acids, carbohydrates and small molecules. These events are in line with changes in metabolism, which are essential for both inflammatory and anti-inflammatory responses. 38 Merging the GO and REACTOME analyses evidenced that in all the senescent secretomes the most significant pathways belong to four main cellular activities: modulation of gene expression, remodelling of cytoskeleton and ECM, stress response and related ER-phagosome activities and inflammation. It is noteworthy that LC MS/MS may fall detection of specific inflammation-related cytokines (such as IL-1, IL-6, TNF-α). These factors are present in extracellular fluids at very low concentrations, which are below the MS limit of detection. Nevertheless, the bioinformatics tools helped us to overcome this limit. Indeed, the bioinformatics analyses clearly evidenced a key role for inflammationrelated ontologies and networks in SASPs of our experimental model.
In detail, the 30D and 60D SASPs, which are secretomes of late senescent cells, are enriched in interferon pathways, which play a main role in inflammation phenomena. The senescence process relies on an executive program for its implementation. Of note, several factors belonging to pathways that are associated with this program, such as P53 signalling, stress response and inflammation, are also present in the SASP, as evidenced by Network analysis. These results further strengthen the hypothesis that senescence heavily depends on autocrine/paracrine signalling. showing that the inactivation of PTBP1 can impairs the protumorigenic effects of SASP by modulating immune surveillance. 45 The presence of global regulators of inflammation-related gene expression indicates that this phenomenon has a major part in SASP duties of late senescent cells. Globally, these results, and other findings showing that late senescent SASP is enriched in cytokines that could foster cancer growth, are at odds with our data showing 30D and 60D SASPs preserved their anti-tumour properties in our experimental conditions. This contradictory result may be explained by hypothesizing that in our conditions we did not reach a deep senescence state. Indeed, we demonstrated that, 30 days after SIPS, the secretome of MSCs contained pro-inflammatory cytokines and factors associated with deep senescence status. Moreover, 60 days after SIPS, we detected pathways related with IL1-α and IL-β that De Cecco and others showed were associated with the final stage of senescence. 27 In the SASP at 60 days after SIPS, we found the PTBP1 protein, which is involved in the inflammatory and pro-tumorigenic activity of SASP. 43 On the other hand, the cancer cell lines we chose may be unresponsive to SASP's pro-tumorigenic effect. This may occur for one cell line, but it is difficult to hypothesize that three different cell lines were SASP-insensitive. It must be remembered that the colon cancer cell lines we selected had different epigenetic and genetic profiles and recapitulate the most common colon cancer molecular signature. An alternative hypothesis could reconcile the contradiction between our data and other published findings. The pro-tumorigenic activity of SASP has been demonstrated in in vivo experiments. 31,32 The most convincing study on the pro-tumorigenic activity of senescent cells comes from Eggert and colleagues, which found that the SASP of senescent hepatocytes, coexisting with liver cancer cells, may recruit immature myeloid cells and may promote cancer progression by impairing the function of immune cells, such as NK cells. 31 In this context, it is possible to imagine that SASP always contains some core components that have an anti-tumour activity, the progression from early to late senescence enriches the SASP of pro- Generally speaking, we can affirm that the SASP pro-tumorigenic activity is not autonomous. It should also be remembered that, besides interaction with immune cells, signals coming from cancer cells may modify SASP composition and activity. In a previous paper, we found that the SASP of senescent MSCs loses its anti-tumour properties if senescent cells are "primed" by allowing them to interact with cancer cells. 49

| CONCLUSION
We have shown that quiescent stromal cells following SIPS-produced SASP that, while progressively changing its composition, preserved the capacity to block cancer cell growth by inducing senescence and/or apoptosis only in an autonomous manner. This finding adds further complexity to the way senescent cells act through SASP and demonstrates that SASP activities are cell-context dependent. This result must be taken into consideration for effective anti-cancer or anti-aging therapeutic strategies.