Protein and chemotherapy profiling of extracellular vesicles harvested from therapeutic induced senescent triple negative breast cancer cells

Triple negative breast cancer (TNBC) is an aggressive subtype with relatively poor clinical outcomes and limited treatment options. Chemotherapy, while killing cancer cells, can result in the generation of highly chemoresistant therapeutic induced senescent (TIS) cells that potentially form stem cell niches resulting in metastases. Intriguingly, senescent cells release significantly more extracellular vesicles (EVs) than non-senescent cells. Our aim was to profile EVs harvested from TIS TNBC cells compared with control cells to identify a potential mechanism by which TIS TNBC cells maintain survival in the face of chemotherapy. TIS was induced and confirmed in Cal51 TNBC cells using the chemotherapeutic paclitaxel (PTX) (Taxol). Mass spectrometry (MS) analysis of EVs harvested from TIS compared with control Cal51 cells was performed using Ingenuity Pathway Analysis and InnateDB programs. We demonstrate that TIS Cal51 cells treated with 75 nM PTX for 7 days became senescent (senescence-associated β-galactosidase (SA-β-Gal) positive, Ki67-negative, increased p21 and p16, G2/M cell cycle arrest) and released significantly more EVs (P=0.0002) and exosomes (P=0.0007) than non-senescent control cells. Moreover, TIS cells displayed an increased expression of the multidrug resistance protein 1/p-glycoprotein. MS analysis demonstrated that EVs derived from senescent Cal51 cells contained 142 proteins with a significant increased fold change compared with control EVs. Key proteins included ATPases, annexins, tubulins, integrins, Rabs and insoluble senescence-associated secretory phenotype (SASP) factors. A fluorescent analogue of PTX (Flutax-2) allowed appreciation of the removal of chemotherapy in EVs from senescent cells. Treatment of TIS cells with the exosome biogenesis inhibitor GW4869 resulted in reduced SA-β-Gal staining (P=0.04). In summary, this study demonstrates that TIS cells release significantly more EVs compared with control cells, containing chemotherapy and key proteins involved in cell proliferation, ATP depletion, apoptosis and the SASP. These findings may partially explain why cancer senescent cells remain viable despite chemotherapeutic challenge.


INTRODUCTION
In Ireland,~2883 women will be diagnosed annually with breast cancer, representing the highest incidence of all female cancers in Ireland, with 10-15% described as triple negative breast cancer (TNBC). 1 TNBC is characterised by its aggressive behaviour and lack of expression of the oestrogen, progesterone and HER2 receptors. This results in TNBCs not being amenable to endocrine and trastuzumab-based therapies, with conventional cytotoxic therapies such as the chemotherapeutic paclitaxel (PTX) (Taxol) the only option. 2 TNBC is further characterised by early onset and poor prognosis, 3 with TNBC tumours being at a higher risk for subsequent tumour recurrence 4 owing to its chemoresistant biology. Moreover, when TNBC relapses, it usually presents as distant metastases, which tend to have central nervous system involvement, rather than local, more treatable metastases. 5 The chemotherapeutic PTX (Supplementary Figure 1A) preferentially binds to the beta subunit of tubulin with high affinity and enhances the polymerisation of tubulin to stable microtubules. This stabilisation means that depolymerisation cannot take place, the cells are unable to form the correct mitotic spindle and as a result cells are arrested in the G2/M phase of mitosis. 6 Although the goal of chemotherapy is to induce tumour cell death via apoptosis, paradoxically, tumour cells can maintain viability in response to chemotherapy, by undergoing alternative fates such as therapeutic induced senescence (TIS), resulting in the ability of tumour cells to successfully evade apoptosis. 7 Increased expression of p21, p16 and senescence-associated β-galactosidase (SA-β-Gal) activity are well-established markers of senescent cells. 8,9 Senescent cells although being metabolically active are also proliferatively incompetent. 10 Therefore, chemotherapies such as PTX, with preferential efficacy on dividing cells, are less likely to induce apoptotic cell death. 11 Moreover, senescent cells secrete the senescence-associated secretory phenotype (SASP); a secretome known to be associated with cancer-promoting phenomenon such as chronic inflammation, angiogenesis, cell proliferation and cell invasion. 12 Importantly, Demaria et al. 13 have shown using a mouse model that TIS promotes an inflammatory environment conducive to metastases and that removal of senescent cells after chemotherapy can prevent or delay cancer recurrence and metastases. Intriguingly, senescent cells produce more extracellular vesicles (EVs) than non-senescent cells facilitating their continued engagement with the cellular milieu both locally and distally. 14 EVs are released by multiple cell types and can be found in blood, urine, serum and amniotic fluid. 15 The term EVs encompasses a range of different subsets of lipid bilayer vesicles including vesicles of~50-150 nm diameter termed exosomes. Exosomes are released by most cells upon the fusion of multivesicular bodies with the plasma membrane. 16,17 Exosomes are characterised by a variety of markers including the tetraspanins (CD63, CD9 and CD81), heat shock proteins (HSP70) and multivesicular body formation proteins (for example, TSG101). 18 It is well established that tumour exosomes comprise a large population of total EVs in the blood of cancer patients. 19 Therefore, the profiling of these EVs as circulating biomarkers in a patient's liquid biopsy is feasible. In addition, EVs can transmit proteins, nutrients and RNA from one cell to another thereby, having a functional effect on recipient cells. 20 Moreover, EVs have an integral role in intercellular communication in the TME, 21 can propagate the chemoresistant phenotype and establish metastatic niches. 22,23 EVs also facilitate the removal of misfolded proteins or metabolic waste products that are harmful to the cell.
In relation to drug treatments and chemoresistance, EVs have been shown to neutralise targeted antibody based drugs such as trastuzumab/Herceptin, which targets HER2. Specifically, HER2-overexpressing breast carcinoma cell lines release EVs containing the HER2 protein, which preferentially sequesters trastuzumab, leading to a decreased drug concentration and attenuated interaction of trastuzumab with its intended HER2 + cancer cell target. 24 EVs have also been shown to confer drug resistance in a paracrine manner through the EV-mediated transfer of the multidrug resistance protein 1/p-glycoprotein (MDR1/P-gp) from a docetaxel-resistant breast cancer cell line to its sensitive counterpart. 25 Moreover, cancer cells can export chemotherapeutics in EVs, thereby reducing the intracellular drug concentration. In this regard, it has been shown that cisplatinresistant ovarian carcinoma cell-derived EVs contain more cisplatin Figure 1. Confirmation of therapeutic induced senescence (TIS) in Cal51 triple negative breast cancer (TNBC) cells treated with 75 nM paclitaxel (PTX) for seven days. (a,i) Cal51 treated with 75 nM PTX for 1 week seeded at 100 000 cells/well. Cells were stained using the SA-β-Gal staining kit (Cell Signalling) with 5 mg/ml X-gal. Scale bars represent 20 μm. (a,ii) The percentage of positive SA-β-Gal staining was normalised to the cell count in each condition and shown in log scale (β-Gal % positivity 77% ± 5.204). (b) Cal51 TNBC cells seeded at 100 000 cells/ml, treated with 75 nM PTX for 1 week to induce senescence followed by immunohistochemical (IHC) staining for the proliferation marker Ki67, (i) negative IHC control (ii) Ki67 staining evident in proliferating cells and absent in the larger senescent cell (white arrow). Scale bars represent 50 μm. (c) Cal51 cells were treated with 50 nM and 75 nM PTX for 1 week. SDS-PAGE western blotting was performed using a 12% gel for p16, p21 and GAPDH. (d) Cal51 TNBC cells treated with PTX for 1 week undergo a G2/M cell cycle arrest. (d,i) Control Cal51 TNBC cell cycle profile. (d,ii) TIS Cal51 TNBC cell cycle profile. Cells were stained using RNase A 0.2 mg/ml and PI 50 μg/ml. All values are expressed as the mean of three independent experiments ± s.e.m. The limit of ± 1 skewness was set as normal distribution. Skewed data were transformed to Y = Log(Y) to fit a normal distribution. Results were analysed using the Student's t-test with significant differences having a *P ⩽ 0.05, **P ⩽ 0.01, ***P ⩽ 0.001.
in comparison with cisplatin sensitive ovarian carcinoma cell-derived EVs. 26 In light of the ability of chemotherapy to induce viable TIS cancer cells, and the documented preponderance of EV release from senescent compared with non-senescent cells, the overall aim of this study was to investigate the chemotherapy and protein content of EVs derived from TIS cancer cells and determine whether the resultant profiles may partially explain why cancer senescent cells remain viable despite chemotherapeutic challenge. TIS cells release significantly more EVs than control cells EVs were isolated from cells by differential centrifugation unless otherwise stated. Isolation of EVs was confirmed using nanoparticle tracking analysis (NTA) measuring the concentration and size of the particles (Figure 2a). Control and TIS EVs displayed average modal sizes of 125 nm and 122 nm, respectively. TIS cells released significantly more EV (P = 0.0002) and exosome (P = 0.0007) fractions compared with control cells (Figure 2b). EVs were confirmed to be calnexin-negative (negative control), CD63-and HSP70-positive (positive control) by sodium dodecyl sulphate polyacrylamide gel electrophoresis western blotting ( Figure 2c).

TIS cells maintain viability via the upregulation of MDR1/P-gp
The TIS model was further investigated for the expression of the multidrug resistance/p-glycoprotein (MDR1/P-gp). TIS Cal51 cells Calnexin was used as a negative control as it represents an ER protein whose absence ensures for EV enrichment of endocytic origin. All values are expressed as the mean of three independent experiments ± s.e.m. The limit of ± 1 skewness was set as normal distribution. Skewed data were transformed to Y = Log(Y) to fit a normal distribution. Results were analysed using the Student's t-test with significant differences having a *P ⩽ 0.05, **P ⩽ 0.01, ***P ⩽ 0.001. treated with 75 nM PTX for 7 days resulted in a significant increased expression of MDR1/P-gp in the whole-cell lysates, compared with control Cal51 cells (P = 0.04). For the EV fraction, those derived from TIS cells demonstrated lower amounts of MDR1/P-gp compared with those derived from control Cal51 cells (P = 0.04) ( Figure 3).

TIS cells maintain viability via the removal of key proteins in their EVs
A proteomic study of the contents of EVs isolated from TIS and control cells, demonstrated that EVs derived from TIS Cal51 cells contained 142 proteins with a significant increased fold change compared with EVs derived from non-senescent cells (Supplementary Table 1). Key proteins of interest shown to be significantly higher in abundance in EVs from TIS Cal51 cells compared with control cells are displayed in Table 1. These were annexin V; which has a role in apoptosis, chemoresistance and metastasis, Na/K ATPase; an ion channel involved in the depletion of cellular ATP levels and α-tubulin; a central part of the cytoskeleton targeted by PTX. Western blot analyses confirmed a significant increased abundance of Annexin V (P = 0.03), Na/K ATPase (P = 0.03) and α-tubulin (P = 0.04) in EVs derived from senescent Cal51 cells compared with EVs derived from control cells, corroborating the mass spectrometry (MS) data ( Figure 4). The EV proteomic analysis also identified higher levels of insoluble SASP proteins in the EVs derived from TIS cells compared with controls (Table 1). Specifically, fibronectin, collagens alpha-1 (V) and alpha-2 (IV) and laminin subunits alpha-5 and beta-1. 27 Pathway analysis of the contents of the TIS EVs using Ingenuity pathway analysis (IPA) revealed the following pathways; remodelling of epithelial adherans junctions, EIF2 signalling, breast cancer regulation by stathmin 1 (Figures 5a and c, Supplementary  Figure 2), phagosome maturation and epithelial adherens junction signalling pathways. Interestingly, InnateDB database analysis demonstrated that the stathmin and breast cancer resistance to anti-microtubule agent's pathway was the second most common pathway in which the proteins from our data set are involved. Pathway analysis of the contents of control EVs using IPA identified the following pathways; clatherin-mediated endocytosis signalling, systemic lupus erythematosus signalling, LXR/RXR (liver X receptor/retinoid X receptor) activation, atherosclerosis signalling and FXR/RXR (farnesoid X receptor/retinoid X receptor) activation ( Figure 5b  Calnexin and CD63 were used as loading controls for the whole-cell lysates (WCLs) and EVs respectively. Densitometry was undertaken using Image J software (Fiji). All values are expressed as the mean of three independent experiments ± s.e.m. The limit of ± 1 skewness was set as normal distribution. Skewed data were transformed to Y = Log(Y) to fit a normal distribution. Results were analysed using the Student's t-test with significant differences having a *P ⩽ 0.05, **P ⩽ 0.01, ***P ⩽ 0.001. GW4869 lead to a significant reduction in SA-β-gal staining (P = 0.04) compared with TIS cells treated with control vehicle alone (Figure 7a). There was a non-significant trend (P = 0.12) towards a reduction in the number of exosomes released from TIS cells treated with GW4869, gating specifically the 45-145 nm size ( Figure 7b). TIS cells maintained p21 upregulation in response to GW4869 and EV isolation was confirmed using western blot analyses (Figure 7c). Figure 7d (i) represents a growth curve of all four conditions (control plus GW4869, control minus GW4869, PTX plus GW4869 and PTX minus GW4869) over a 14-day time course. PTX treatment for 7 days induced TIS in Cal51 cells. These cells were subsequentually treated with GW4869 on day 7 and cells were enumerated on days 9, 11 and 14 following this treatment. GW4869-treated TIS cells displayed a significant reduction in cells per ml on days 9 (P = 0.007), 11 (P = 0.04) and 14 (P = 0.05) compared with TIS cells treated with control vehicle alone (Figure 7d (ii) and E).
In summary, PTX induces senescence in Cal51 TNBC cells with a concomitant upregulation of MDR1/P-gp. TIS cells release significantly more EVs compared with control cells, which not only contain the PTX analogue (Flutax-2) but in addition, key proteins involved in cell proliferation, ATP depletion, apoptosis and the SASP. The exosome biogenesis inhibitor GW4869 results in a reduction in the number of SA-β-gal-positive cells and lowers the proliferation rate of TIS cells (Figure 7f).

DISCUSSION
Chemotherapy such as doxorubicin, has been shown to induce TIS in both p53 wild-type and p53 mutant cell lines. 28 Here we demonstrate induction of TIS in the p53 wild-type TNBC cell line Cal51 using the chemotherapeutic PTX. Clinically, TIS occurs in breast tumour biopsies from patients who received neoadjuvant chemotherapy were 41% positive for SA-β-Gal activity. 29 This suggests that TIS is a novel detrimental mechanism of breast cancer cell survival against chemotherapy. Our data show that the MDR1/P-gp drug efflux pump is upregulated in TIS cells accompanied with a marked increase in p21 expression. Similar to previous studies MDR1/P-gp is detectable in our EV fractions with lower levels in TIS EVs, possibly referring to the retention of this efflux pump by the parent cells in their chemotherapeutic stressed condition. Mechanistically, it is known that the E2F-associated phosphoprotein stimulates the MDR1/P-gp promoter, leading to the upregulation of MDR1/P-gp in a p21-dependent manner. 30 Interestingly, MDR1/P-gp has also been shown to transfer resistance to docetaxel via EVs in breast and prostate cancer cells. 25,31 The higher abundance of EVs in the blood of breast cancer patients with stage IV disease has previously been reported. 32 Our study has found that TIS cells release significantly more EVs than control cells, suggesting that increased EV release is a mechanism conducive to cell survival, following chemotherapeutic cell stress. This is corroborated by Lehmann et al. 14 who showed that irradiation-induced senescent prostate cancer cells released more EVs than controls. 4 We hypothesise that the novel ability of cancer senescent cells to remove ATPases, tubulins, Rabs, insoluble SASP factors, integrins and annexins via EVs, in part conserves the senescent phenotype. Intriguingly, 69 of 142 proteins in higher abundance in TIS EVs are predicted activators of cell growth and proliferation, suggesting that the removal of these proteins by TIS cells represents a further potential mechanism to maintain their senescent phenotype.
Our data show eight ATPases present in higher abundance in TIS EVs compared with controls, including the 26S protease regulatory subunits 6B and 10B ( Table 1). The significance of these subunits' abundant expression in TIS EVs is potentially important, implying a mechanism for retaining maximum ATP levels for the survival and metabolic needs of the senescent cell. Moreover, a sodium/potassium ATPase subunit displayed the highest abundance in the MS profiles for the TIS EVs. This ATPase enables the hydrolysis of ATP, 33 suggesting that its removal via EVs may potentially conserve vital intracellular ATP resources crucial for maintaining the metabolic demands of senescent cells and therefore integral for senescent cell survival. 34 In addition, our results show an increased abundance of eight annexins within TIS EVs compared with controls (Table 1). We hypothesise that annexin removal in EVs increases the survival of TIS cells via their interactions in apoptotic, chemoresistant and metastatic pathways. Specifically, Annexins A1 and A5 are inhibitors of phospholipase A2 signalling, 35,36 Annexin A5 is a marker of apoptosis, 37 whereas Annexin A6, A11 and A2 are involved with cancer aggressiveness promoted by EVs, 38 chemoresistance 39 and EV-mediated angiogenesis and metastasis, 40 respectively. In summary, therefore, removal of annexins via EVs is potentially advantageous to TIS survival.
Our data (Table 1) demonstrated an increase in nine different subunits of tubulin within TIS EVs compared with controls. Pathway analyses showed that the 'stathmin and breast cancer resistance to anti-microtubule agents' pathway' is upregulated in TIS EVs. Normally, during mitosis, microtubule polymerisation occurs, which is crucial for cell cycle progression. However, stathmin promotes depolymerisation of microtubules. Differential expression levels of stathmin leads to G2/M arrest. 41 Interestingly, protein phosphatase 2A and PP1 (protein phosphatase 1) regulate the phosphorylation of stathmin, 42 with PP1 present in higher abundance in TIS EVs compared with controls (Supplementary Table 1). Of note, the regulatory subunit 15B of protein phosphatase 1 has been identified as a survival factor in breast cancer. 43 A recent study demonstrated that non-small cell lung cancer cells, resistant to PTX, have high expression of stathmin and inhibition of stathmin expression increased sensitivity to this chemotherapeutic. 44 Potentially therefore, the removal of tubulin and PP1 (stathmin regulator) from TIS cells, could suggest that microtubule dynamics have been switched from the stabilisation effect of PTX to the depolymerisation effect of stathmin, crucial in maintaining G2/M arrest and the chemoresistant and senescent phenotypes.
In addition, there was an increase in three integrin subunits; integrin beta-3, integrin beta-1 and integrin alpha-V, within TIS EVs compared with controls. Interestingly, integrin beta-3 has been associated with the progression of glioblastoma 45 with attenuated integrin beta-3 expression resulting in glioblastoma cells becoming senescent. 46 The integrin alpha-V-beta-3 complex regulates cancer cell survival and proliferation by signalling through the ERK1/ERK2 MAPK pathway in MDA-MB-231 breast cancer cells. 47 Moreover, a recent study has shown that integrin beta-3 is a marker and regulator of senescence, activating the TGF-β pathway leading to senescence in human primary fibroblasts. Downregulation of integrin beta-3 has also been shown to result in the reduction in oncogene-induced senescence and TIS. 48 Overall therefore, TIS cells may potentially remove the integrin alpha-V-beta-3 complex in EVs to maintain the senescent phenotype. . SDS-PAGE western blot validation of the increased abundance of Annexin V, Na/K ATPase and α-tubulin in the EVs derived from TIS Cal51 cells compared with EVs derived from non-senescent non-treated controls. EVs isolated from TIS and control cells were immunoblotted to validate the protein expression of annexin V, Na/K ATPase and α-tubulin from the MS data. CD63 and HSP70 were used as loading controls. Densitometry was completed using Image J software (Fiji). All values are expressed as the mean of three independent experiments ± s.e.m. The limit of ± 1 skewness was set as normal distribution. Skewed data were transformed to Y = Log(Y) to fit a normal distribution. Results were analysed using the Student's t-test with significant differences having a *P ⩽ 0.05, **P ⩽ 0.01, ***P ⩽ 0.001.
There was also an increase in 6 Rab proteins within TIS EVs compared with controls (Table 1). Rabs modulate the release of EVs from the multivesicular bodies. Rab-mediated EV trafficking is important for migration, invasion and proliferation. 49 K562 leukaemic cells express higher levels of Rab11 compared with other Rabs, previously shown to modulate EV release. 50 The increased release of EVs containing these Rabs suggests a possible link to the increased release of EVs compared with control cells.
In relation to the SASP, it is known that the insoluble components can affect the physical properties of tissue structure relaxing the extracellular matrix support. Moreover, the soluble factors of the SASP in vivo, potentially allow tumour cells to migrate through the extracellular matrix and metastasise. 51 Of note, the insoluble SASP factors (fibronectin, collagen alpha-1 (IV) chain, collagen alpha-2(IV) chain, laminin subunit alpha-5 and laminin subunit beta-1) ( Table 1) 53 were identified in EVs from TIS Cal51 cells. Therefore, our findings support the proposal that EVs derived from TIS cells not only represent a mechanism of trafficking out unwanted cellular cargo but could equally be utilising EVs to transport senescence/tumour promoting proteins to neighbouring/distant sites to further promote the senescent/tumorigenic phenotype. Urbanelli et al. 52 similarly suggest a role for EVs in the maintenance and promotion of the senescent phenotype via the SASP.
MS analysis of EVs from control cells identified 32 proteins with significantly increased fold change when compared with TIS EVs (Supplementary Table 2). We hypothesise that these proteins are crucial for TIS cell survival and therefore not released in EVs. Two key pathways in higher abundance in control EVs were LXR/RXR and FXR/RXR activation. Potentially, activation of these pathways is crucial for the survival of TIS cells. LXR activation is known to inhibit proliferation of breast cancer cells 53 and FXR activation in breast cancer cells results in cytotoxicity. 54 Activated FXR can also counteract the tumour promoting ability of cancer-associated fibroblasts within the breast cancer microenvironment. 55 Therefore, one might hypothesise that the removal of LXR and FXR pathway components in control cell EVs suggests that normal proliferation occurs. In contrast, TIS cells retaining these components may allow them to maintain replicative Images were taken with a × 40 oil objective on a Leica DMI6000B epifluorescent microscope. Individual colour channels were merged and pseudocolours were applied with Image J (Fiji) software. Any contrast adjustments were applied strictly across all channels in both images. Scale bars represent 50 μm. (g) Graphical representation of the differing fluorescence levels detected in EVs derived from control, PTX treated and Flutax-2 (FLU) treated Cal51 TNBC cells. In relation to the concentration of Flutax-2, the manufacturers recommend using × 10 the concentration of paclitaxel (75 nM), therefore, 750 nM was used. https://www.thermofisher.com/order/catalog/product/P22310. All values are expressed as the mean of three independent experiments ± s.e.m. The limit of ± 1 skewness was set as normal distribution. Skewed data were transformed to Y = Log(Y) to fit a normal distribution. Results were analysed using the Student's t-test, with significant differences having a *P ⩽ 0.05, **P ⩽ 0.01, ***P ⩽ 0.001.  (Flutax-2). All values are expressed as the mean of three independent experiments ± s.e.m. The limit of ± 1 skewness was set as normal distribution. Skewed data were transformed to Y = Log(Y) to fit a normal distribution. Results were analysed using the Student's t-test, with significant differences having a *P ⩽ 0.05, **P ⩽ 0.01, ***P ⩽ 0.001. stasis, maintain senescence and evade apoptosis by chemotherapy.
Cal51 cells treated with the fluorescent PTX analogue Flutax-2 demonstrated removal of chemotherapy in EVs. The significantly higher detected fluorescence in EVs derived from Flutax-2-treated cells indicates that the chemotherapy is present within the EVs, demonstrating a potential novel method by which senescent cancer cells can remove chemotherapy. Whether this packaging and removal of chemotherapy by the cells is a passive or active process remains to be elucidated.
Previously, it has been shown that treatment with GW4869 impairs EV formation and subsequent release from macrophages. 56 In our study, TIS cells treated with GW4869 showed significant reduction in SA-β-gal activity a key marker of senescence. We also speculate that TIS cells utilise exosome biogenesis to maintain their proliferative potential, as blocking this with GW4869 lead to a reduction in proliferative recovery. The lack of change in p21 following GW4869 treatment for 48 h may reflect a time-dependent observation.
In conclusion, our study demonstrates that TIS cells release significantly more EVs than control cells. Importantly, these EVs not only contain Flutax-2 (PTX analogue) but also suggests a novel mechanism by which TIS cells remove key proteins involved in cell proliferation, ATP depletion, apoptosis and the SASP. These findings may partially explain why cancer senescent cells remain viable despite chemotherapeutic challenge and indeed can continue to interact with the tumour microenvironment both locally and distally. SA-β-Gal staining assay TIS cells were seeded at 1 × 10 5 cells/ml, cultured for 24 h, fixed and stained using the SA-β-Gal staining kit, 5 mg/ml (Cell Signalling Technology, Danvers, MA, USA, #9860), incubated for 16 h, imaged using a light microscope (Olympus model CKX41, Tokyo, Japan) and enumerated for SA-β-Gal positivity using Image J software (FIJI).

EV isolation
Cells grown for EV harvesting purposes were cultured prior to isolation in media supplemented with bovine EV-depleted FBS for 48 h. FBS was depleted of EVs by ultracentrifugation at 120 000 × g for 16 h at 4°C. The RCF average was calculated using the Beckmann Coulter rotor calculator for the Beckman Coulter Optima L-100 XP Ultracentrifuge and Beckman Coulter Type 70Ti rotor. EV-depleted FBS media was conditioned for 48 h then collected and centrifuged at 300 × g at 4°C for 10 min followed by 2000 × g at 4°C for 20 min to ensure cellular debris removal. The supernatant was filtered using a 0.22 μm syringe filter (Millipore, Billerica, MA, USA, #SLGP033RS). For isolation using the ultracentrifugation protocol, the supernatant was sealed in an ultracentrifugation tube (Beckman Coulter, Brea, CA, USA, #342414), and subsequently ultracentrifuged twice at 120 000 × g for 1 h 15 min at 4°C. For isolation using Exospin the manufacturer's guidelines were followed (Cell Guidance Systems, #EX-01). EV counts represent values normalised to cell count. NanoSight NS300 EV profiling EV samples were run using the standard measurement procedure at 25°C with a constant syringe infusion rate of 50 as per the NanoSight NTA 3.1 Software (Malvern). The data for each sample were obtained from 10 independent 60 s video captures on the NS300, analysed and normalised to final cell counts for each condition.
Quantitative proteomic profiling by label-free liquid chromatography-MS/MS analysis EV pellets were solubilised in 8 M urea/50 mM NH4HCO3/0.1% ProteaseMax using sonication, with resultant protein levels quantified using a Bradford assay. Following dithiothreitol reduction and iodoacetic acid-mediated alkylation, a double digestion was performed using Lys-C (for 4 h at 37°C) and Trypsin (overnight at 37°C) on 5 μg of protein. The samples were desalted prior to analysis using C18 spin columns (Thermo Scientific) and 500ng was loaded onto an Ultimate 3000 NanoLC system (Dionex Corporation, Sunnyvale, CA, USA) coupled to a Q-Exactive mass spectrometer (Thermo Fisher Scientific). The raw data were analysed using Progenesis QI for Proteomics software (version 3.1; Non-Linear Dynamics, a Waters company, Newcastle upon Tyne, UK) as previously described, 61 with following modifications: the filter mass peaks with charge states from +1 to+6 was applied to the MS/MS data files, peptides were identified using taxonomy: Homo sapiens in the SwissProt database and peptides with XCorr scores41.9 for +1 ions,42.2 for +2 ions and43.75 for +3 ions or more (from Sequest HT) were selected.

Treatment of Cal51 TNBC cells with the exosome biogenesis inhibitor GW4869
GW4869 is a cell permeable, potent inhibitor of the biogenesis of exosomes via the inhibition of N-SMase. 56 Cal51 cells grown in both control and senescence inducing conditions were subsequently treated with control vehicle or GW4869, 5 μM (Sigma-Aldrich, #D1692) for 48 h. Cells were trypsinised and seeded for the SA-β-gal enumeration. Media was collected and EVs isolated with ExoSpin.

Pathway analysis
Pathway, interaction network and functional enrichment analyses were performed with the proteins identified in EVs using IPA (www.ingenuity. com) software (Ingenuity Systems, Qiagen) and InnateDB (http://www. innatedb.com). P-values reported for IPA results are calculated by IPA using a right-sided Fisher exact test for over-representation analysis and Benjamini-Hochberg correction for multiple hypothesis testing correction.

Statistical analyses
All values are expressed as the mean of three independent experiments ± s.e.m. The limit of ± 1 skewness was set as normal distribution. Skewed data were transformed to Y = Log(Y) to fit a normal distribution. Results were analysed using the Student's t-test, unless as otherwise stated in the quantitative proteomic profiling by label-free liquid chromatography-MS/MS analysis and pathway analysis sections. Significant differences are defined as *P ⩽ 0.05, **P ⩽ 0.01, ***P ⩽ 0.001.