NF‐κB/IKK activation by small extracellular vesicles within the SASP

Abstract Cellular senescence plays an important role in different biological and pathological conditions. Senescent cells communicate with their microenvironment through a plethora of soluble factors, metalloproteases and extracellular vesicles (EV). Although much is known about the role that soluble factors play in senescence, the downstream signalling pathways activated by EV in senescence is unknown. To address this, we performed a small molecule inhibitor screen and have identified the IκB kinases IKKε, IKKα and IKKβ as essential for senescence mediated by EV (evSASP). By using pharmacological inhibitors of IKKε, IKKα and IKKβ, in addition to CRISPR/Cas9 targeting their respective genes, we find these pathways are important in mediating senescence. In addition, we find that senescence activation is dependent on canonical NF‐κB transcription factors where siRNA targeting p65 prevent senescence. Importantly, these IKK pathways are also relevant to ageing as knockout of IKKA, IKKB and IKKE avoid the activation of senescence. Altogether, these findings open a new potential line of investigation in the field of senescence by targeting the negative effects of the evSASP independent of particular EV contents.


| INTRODUC TI ON
The presence of senescent cells has been identified in many physiological and pathological conditions such as during embryonic development, cancer, ageing and several age-related diseases (Di Micco et al., 2020;He & Sharpless, 2017;Munoz-Espin & Serrano, 2014).
Senescence is characterized by the activation of a complex and heterogeneous cellular phenotype presenting a stable cell cycle arrest.
In spite of the lack of proliferative potential, senescent cells are highly metabolic and present elevated levels of transcription and translation. Furthermore, senescent cells have a very active secretome that has been denominated the senescence-associated secretory phenotype (SASP). Thus, intercellular communication between senescent cells and their microenvironment is varied and can be mediated by soluble factors (Coppe et al., 2010). Other less characterized means of communication are extracellular vesicles, metabolites and ions, in addition to cell-to-cell contact, cell-extracellular matrix interaction and cell fusion and cytoplasmic bridges . As these different means by which senescent cells communicate with their surroundings is context dependent, there is need for a better understanding of the molecular mechanisms and downstream pathways implicated.
In the last years, there has been an increasing interest in the role that extracellular vesicles (EV) play in the context of senescence, ageing and cancer . Many groups have identified different individual components enriched in the EV SASP (evSASP) such as microRNA (miR) (Jeon et al., 2019;Terlecki-Zaniewicz et al., 2018), specific proteins (Basisty et al., 2020;Borghesan et al., 2019;Kavanagh et al., 2017;Takasugi et al., 2017) or nucleic acids . As with the soluble SASP (sSASP), the evSASP has detrimental effects on the microenvironment by inducing paracrine senescence in human primary fibroblasts (Borghesan et al., 2019), in proliferative chondrocytes in osteoarthritis (Jeon et al., 2019) or by promoting tumour progression . However, EV can also induce tissue repair by ameliorating senescence and ageing-related tissue dysfunction Yoshida et al., 2019).
In order to identify the molecular mechanisms and downstream pathways implicated in senescence mediated by the evSASP independent of single EV-components, we performed a small molecule inhibitor screen and identified different signalling pathways including the NF-κB pathway. NF-κB proteins are implicated in a wide range of innate and adaptive immune responses that regulate cellular processes including senescence. NF-κB is regulated by phosphorylation of one of its cytoplasmic inhibitors, IκB, mediated by IκB kinases (IKK), allowing NF-κB nuclear translocation and transcriptional activation (Taniguchi & Karin, 2018). IKKα-IKKβ heterodimer complexes regulate the DNA binding proteins implicated in the canonical NF-κB pathway, including p50-p65, while IKKα-homodimers regulate p52-RELB dimers implicated in the alternative NF-κB pathway (Perkins, 2007;Taniguchi & Karin, 2018). On the other hand, IKKε can contribute to the regulation of NF-κB but also type I interferon signalling (Shen & Hahn, 2011). Here, we find that pharmacological inhibition of IKKα, IKKβ and IKKε and knockout of their respective genes using CRISPR/Cas9 technology prevent senescence induced by the evSASP. We observe that the evSASP activates transcription factors mainly implicated in the canonical NF-κB pathway and that this effect is dependent again on IKKα, IKKβ and IKKε. Furthermore, two independent RNAi targeting the NF-κB canonical transcription factor p65 prevent the evSASP functionality. Altogether, we show that the evSASP activation of senescence is dependent on IKKα, IKKβ  (Borghesan et al., 2019;Rapisarda et al., 2017). We isolated sEV from iC and iRAS cells by serial ultracentrifugation as performed previously (Borghesan et al., 2019;Fafian-Labora et al., 2019;Thery et al., 2018) and treated recipient proliferating HFFF2 simultaneously with these sEV and a variety of pharmacological drugs inhibiting different signalling pathways for 6 days ( Figure 1a). As we can see in Figure 1b, sEV isolated from iRAS treated with DMSO induced a cell cycle arrest measured by quantifying the percentage of cells incorporating BrdU. Media supplemented with EV-depleted FBS was used as a control (FBS) and no differences were found between this condition and sEV from iC cells. However, HFFF2 treatment with inhibitors targeting p38MAPK, mTOR (mammalian target of rapamycin) F I G U R E 1 Small molecule inhibitor screen identifies different signalling pathways mediated by the evSASP (a) Details of the experimental settings to unveil novel signalling pathways mediated by the evSASP. Senescence was induced in iRAS donor cells by addition of 200 nM 4-hydroxytamoxifen (+4OHT). HFFF2 recipient cells were treated with a panel of small molecule inhibitors in addition to sEV isolated from iC and iRAS cells. Drugs were maintained for 6 days and washed out. (b) Box and Whiskers plot representing the data obtained from the screen measuring proliferation by quantifying the percentage of HFFF2 that stain positive for BrdU. The graph indicates the targets of the small molecule inhibitors used: 40 µM PD98059 (targeting MEK1/2), 20 µM SB202190 (p38MAPK), 4 µM TGFB-R1 (TGFBR1 kinase), 8 µM VEGFR2 (VEGFR2), 150 nM GSK429286A (ROCK1/2, Rho-associated kinase), 50 nM CPD22 (ILK, integrin-linked kinase), 1 µM CPG (MNK1/2), 100 nM TORIN2 (mTOR, mammalian target of rapamycin), 1µM RUXOLITINIB (JAK1/2 inhibitor), 40µM AG-490 (JAK2/3 kinase), 45 µM JANEX1 (JAK3 kinase), 1 µM AG-879 (protein Tyrosine Kinase), 2 µM IMATINIB (tyrosine kinase; TK1), 20 µM CAY10576 (IKKε, IκB kinase epsilon), 1.5 µM SUNITINIB (multi tyrosine kinase, multi-TK). Data show the min to max Box and Whiskers plot of 4 independent experiments. Dunnett's multiple comparison test analysis was performed to determine statistical significance. p-values are shown. (c) Representative IF pictures for BrdU staining in HFFF2 cells treated with or without 20 µM CAY10576 (IKKε inhibitor) and sEV from iC or iRAS. Scale bar: 50 µm. (d) Representative IF images (left panels) and quantification (right panels) of cells staining positive for senescenceassociated beta galactosidase activity (SAβ-Gal) in HFFF2 fibroblasts treated with 20 µM CAY10576 and sEV derived from iC or iRAS cells. Scale bar of IF pictures: 50 µm. All data represent the mean ± SEM of 4-5 independent experiments. Data were normalized to sEV from iRAS sample. One-way ANOVA analysis was performed with multiple comparisons to the sEV iRAS control sample and IKKε (IκB kinase epsilon) pathways prevented the proliferation arrest induced by sEV as part of the SASP (termed evSASP herein) (Figure 1b,c). IKKε is an essential regulator of innate immunity by regulating both the interferon and NF-κB signalling pathway (Shen & Hahn, 2011). While we have previously identified the importance of the interferon pathway by the evSASP (Borghesan et al., 2019), the role that IKKε plays is unknown. To confirm the implication of the evSASP on the IKKε pathway, we determined additional markers of senescence and confirmed that IKKε inhibition using 20µM CAY10576 treatment for 6 days also reduced the high levels of senescence-associated β-galactosidase activity (SAβ-Gal), a common marker of senescence (Gorgoulis et al., 2019), as shown by IF pictures (left panel) and its corresponding quantification (graph) levels ( Figure 1d). Next, we decided to determine the levels of another important hallmark of senescence, the cell cycle inhibitor p16 INK4A .
When looking at the number of cells staining positive for p16 INK4A by IF, we could also observe a decrease in these cells when treated with CAY10576 shown by IF pictures (above) and its relative quantification (graph below) ( Figure S1a). To confirm that the results observed were not due to cytotoxicity by CAY10576, we determined whether CAY10576 was inducing apoptosis in the donor cells by staining with Annexin V and quantifying the number of cells staining positive by FACS ( Figure S1b). Donor cells were used for this analysis to test pharmacological toxicity in proliferative versus senescent cells.
Treatment of HFFF2 with 3 µM Cisplatin for 72 h was used as a positive control. Cytotoxicity was also determined in HFFF2 recipient cells simultaneously treated with sEV and CAY10576 by determining total cell number after treatment where no changes between DMSO and CAY10576 treatment was observed ( Figure S1c). Altogether, these data suggest a role for IKKε in mediating senescence by the evSASP.

| Pharmacological inhibition of IKKα and IKKβ prevents senescence mediated by the evSASP
As IKKε contributes to regulate NF-κB signalling pathway (Shen & Hahn, 2011), we next wanted to determine whether two key players, IKKα and IKKβ, were implicated in senescence mediated by the evSASP. For this, we used two commonly studied small molecule drugs that inhibit IKKα and IKKβ, respectively (10 µM BAY11-7082 and 10 µM MLN120B for 6 days). Experiments were performed as in  Figure S2b, c). An additional p65 antibody was used to corroborate staining specificity ( Figure   S2c). Furthermore, to confirm that additional methods for sEV isolation would produce similar results and that sEV internalization was taking place, we isolated sEV from HFFF2 cells expressing an ectopic mCherry construct for human CD63 by size exclusion chromatography (SEC). We next pooled the sEV-enriched fractions (fractions 3-7) ( Figure S2d), performed and additional ultracentrifugation (UC) step and treated HFFF2 recipient cells as before. Both sEV and protein content were determined in 12 different fractions ( Figure S2d). sEV internalization between the different conditions showed slight differences that were not statistically significant ( Figure S2e, f).
We have previously shown that the evSASP induces DNA damage in recipient cells (Borghesan et al., 2019). Thus, to confirm that SEC-isolated sEV activates senescence, we analysed the number of cells with DNA damage by staining with phospho-γH2AX and observed that pharmacological inhibition of IKKα, IKKβ and IKKε prevented the DNA damage response induced by the evSASP with sEV isolated by SEC ( Figure S2g).

| Knockout of IKKα, IKKβ and IKKε using sgRNA prevents senescence mediated by the evSASP
To further determine the implication of IKKα, IKKβ and IKKε in evSASP-mediated senescence, we proceeded to generate a single guide RNA (sgRNA) targeting IKKA, IKKB and IKKE, respectively, in a lentiviral vector containing a construct encoding for Cas9 and the sgRNA. We next transduced HFFF2 cells with 4 sgRNA per gene and after selection proceeded to treat these cells with sEV isolated from iC and iRAS cells for 6 days and determined different markers of senescence by IF ( Figure 3a). As shown in Figure

| Canonical NF-κB pathway is activated by the evSASP
To further confirm the implication of NF-κB pathway in mediating evSASP senescence and to discern whether the canonical or noncanonical pathways are implicated, we next performed a transcription factor high throughput assay to quantify binding of NF-κB transcription factors to different oligonucleotides containing NF-κB consensus binding sites. For this, we isolated nuclear extracts from

| siRNA targeting p65 prevents senescence mediated by the evSASP
To further confirm an implication of NF-κB mediated by the evS-ASP, we took advantage of two independent siRNA targeting the canonical NF-κB transcription factor, p65. A non-targeting siRNA was used as a control (Scr). Furthermore, we used previously validated siRNA targeting two well characterized effectors of senescence: and sip53 and sip16. Furthermore, we could also observe that downregulation of p65 siRNA ( §7 and §10) also prevented the stabilization of p53 mediated by the evSASP (Figure 5d). The reduction in the protein levels of p53 and p16 INK4A using their corresponding siRNA, sip53 and sip16, was confirmed by IF (Figure 5c, 5d). IL-8 has been shown to induce senescence and its expression is regulated by p65 (Acosta et al., 2008). We therefore determined the levels of  Figure S4e). Altogether, these data suggest that the evSASP activates senescence through p65.

| IKKs are important for evSASP senescence in human primary fibroblasts from young and old donors
In order to determine the importance of NF-κB and relevant IKKs in ageing, we took advantage of four human primary fibroblasts cell cultures isolated from young (~2 years) and four cultures from old (~70 years) donors. We have previously shown that fibroblasts from old donor express several markers of senescence Rapisarda et al., 2017). All cells were used between passages 2-3 after arrival to prevent the onset of replicative se-

F I G U R E 4
The canonical NF-κB pathway is activated by the evSASP. Nuclear extracts were isolated from HFFF2 fibroblasts treated with (a) different IKKs inhibitors or (b) sgRNA targeting IKK and sEV from iC and iRAS simultaneously for 3 days. 5 ng/ ml lymphotoxinβ (Lymβ) was used as a positive control for 8 h for the activation of the transcription factors p52 and RelB. Binding of canonical NF-κB transcription factors, p65, p50 and c-Rel, to an immobilized DNA fragment is shown upon treatment with sEV from iRAS and lost when treating with CAY10576, MLN120B and BAY11-7082 IKKs inhibitors or in cells expressing sgRNA against IKKs constructs. No binding or changes upon the inhibitors treatments or sgRNA expression could be observed with the non-canonical transcription factors: p52 and RelB. All data show the mean ± SEM of 3-6 independent experiments. Oneway ANOVA analysis was performed with multiple comparisons to the iRAS sEV control sample  Figure 6c). We could also observe that this induction of senescence was prevented by sgIKKA, sgIKKB and sgIKKE. The knockdown efficiency was confirmed by qPCR using two housekeeping genes ( Figure S5). Thus, these data demonstrate that IKKα, IKKβ and IKKε play an important role in senescence mediated by the evSASP in human fibroblasts cultures from old and young individuals highlighting its relevance in ageing.

| DISCUSS ION
The SASP is one of the main means by which senescent cells communicate with their microenvironment in physiological and pathological conditions (He & Sharpless, 2017;Munoz-Espin & Serrano, 2014). In fact, it is known that chronic inflammation or inflammaging causes tissue damage in ageing. One hypothesis for this is the accumulation of senescent cells and the effect of their associated SASP released to the microenvironment (Faget et al., 2019;Lee & Schmitt, 2019).
While much research focuses on the discovery of novel drugs blocking the SASP (termed senomorphic drugs) (Di Micco et al., 2020), the off-target effects of these drugs on organelle inter-trafficking is understudied. During the maturation of the secretome in senescence, multiple organelles are likely to interact such as lysosomes, phagosomes and ER-Golgi trafficking (Vizioli et al., 2020). In fact, dissecting the release of the soluble SASP (sSASP) from the evSASP proves difficult due to common pathways regulating inter-organelle communication in senescence (Cavinato et al., 2020;Dou et al., 2017;Vizioli et al., 2020). Thus, an alternative approach to prevent the detrimental effects of the evSASP could be to prevent its effect on the microenvironment and neighbouring cells as described in our study, approach which has been successfully used in a model of liver regeneration (Bird et al., 2018). It is interesting that our data show dependence of different pathways to activate senescence by the evSASP. This could be considered as a limitation of our study, as theoretically several pathways would need to be inhibited to prevent the detrimental effect of the evSASP. However, the fact that most pathways identified in this study lead to a particular common pathway, NF-κB, is an advantage (Mowla et al., 2013). Inhibition of NF-κB would not only act as a senomorphic drug but would also prevent evSASP senescence. NF-κB is constitutively activated during senescence and is considered a key regulator of the sSASP, where inhibition or knockout of several components that modulate NF-κB signalling, for example p65, have shown that during the initial activation of senescence, IKKβ is activated by ATM while in the later stages the sSASP is expressed independently of IKKα/β and proteasome degradation of IκB in spite of depending on the NF-κB protein p65 (Kolesnichenko et al., 2021).
Altogether, this suggests that the activation of NF-κB is far more complicated than initially thought in the context of senescence and both the sSASP and evSASP.
The fact that the evSASP activates the IKKα, IKKβ and IKKε pathways could be a reflection of different things. The evSASP could be activating a common upstream regulator of the NF-κB pathway such as IKKβ, resulting in the nuclear translocation of p65, with a direct or indirect contribution of IKKα and IKKε. In fact, when manipulating a single sEV protein component such as the interferon protein IFITM3, theoretically targeting a particular evSASP subpopulation, we only see partial amelioration of the senescence phenotype mediated by the evSASP (Borghesan et al., 2019) indicating that other factors are essential or that other sEV populations can mediate evSASP senescence. The other alternative is that we are isolating a heterogeneous sEV population (Tkach & Thery, 2016) where one sEV subpopulation could be targeting IKKβ while another sEV subpopulation within the same sample could be targeting IKKα/IKKε. Interestingly, in all cases, we see induction of senescence and activation of p65. This is not uncommon as EV has been shown to activate NF-κB via TLR signalling in other contexts resulting in the induction of pro-inflammatory cytokines by human monocytic cells (Bretz et al., 2013) or during parasite infection (Toda et al., 2020).
Importantly, we see that the evSASP employed using the iRAS model of senescence represents what happens using an in vitro model of ageing. Thus, the downstream signalling pathways involving

| siRNA reverse transfection
HFFF2 cells were reverse transfected with 50 nM siRNAs on a well of a 96-well plate. After 24 h, the medium was changed. Two days later, the cells were washed and incubated with the sEV in medium supplemented with 10% (v/v) EV-depleted FBS and 1% (v/v) A/A for 72 h. siRNAs used in this study are below: F I G U R E 5 siRNA targeting p65 prevents senescence mediated by the evSASP. (a) Schematic representation for the experimental settings (left panel) and timings (right panel) of the experiments performed in this figure. Reverse transfection was performed with 50 nM of the indicated siRNA in the recipient cells. After washing the siRNA out, HFFF2 recipient cells were incubated with sEV isolated from either iC or iRAS for 72 h and 3 days later several markers of senescence were determined. (b) The use of two independent siRNA targeting p65 (sip65) - §7 and §10-prevents the cell cycle arrest characteristic of evSASP senescence. A scramble siRNA control (Scr) was used as a negative control while siRNAs targeting CDKN2A (encoding for p16 INK4A ) (sip16) or TP53 (sip53) were used as positive controls. (c) The percentage of cells expressing p16 INK4A protein levels and (d) stabilizing p53 protein was quantified. (e) Representative IF images for IL-8 expression levels in HFFF2 transfected with the indicated siRNAs and treated with sEV isolated from either iC or iRAS HFFF2 cells. Scale bar: 50 µm. Data show the mean ± SEM of 6 independent experiments. Statistical analysis was performed using One-way ANOVA. All conditions were compared to sEV from iRAS control

| Retroviral and lentiviral infections
The generation of stable retroviral and lentiviral expression was carried out following previous studies (Acosta et al., 2008;Borghesan et al., 2019;Rapisarda et al., 2017). Briefly, retroviral particles were generated by transfecting ER-HRAS G12V plasmid and retroviral helper plasmids (vsvg and gag-pol) with Polyethylenimine (PEI) in HEK293T packaging cells for 48h. Recombinant lentiviral particles were generated using the second-generation packaging vectors psPAX2 and pMD2.G using PEI in HEK293T. The supernatant was then filtered with 0.45µm filters (Starlab) and applied to HFFF2 cells in the presence of 4 µg/ml polybrene (hexadimethrine bromide; Sigma-Aldrich) following threee rounds of infection. Cells were subsequently selected with the appropriate antibiotic resistance either 0.5 μg/ml puromycin or 300 μg/ml neomycin (Invitrogen).
For infections with the sgRNA, a pool for the 4 sgRNA (5 μg per single sgRNA) targeting a single IKK was generated by transfecting equal amounts of DNA and the packaging vectors psPAX2 and pMD2.G. Infection was performed as described earlier for lentivirus.

| sEV isolation and treatment
To isolate the sEV fraction, the protocol of differential ultracentrifugation (Thery et al., 2018) was modified and adapted (Borghesan et al., 2019;. Briefly, 1 × 10 6 early passage donor cells were plated in a 10 cm dish ( previously validated that the induction of paracrine senescence was due to sEV content rather than sEV number (Borghesan et al., 2019).
A Sorvall 100SE Ultra Centrifuge, with a Beckmann Fixed Angle T865 rotor was used for sEV isolations. The k-factor of the rotor is 2,08.

| Transcription factor binding assay NF-κB
The protocol was performed following the manufacturers instruc-

| Nanoparticle tracking analysis
The NanoSight LM10 (Malvern Instruments) was calibrated using Silica Microspheres beads (Polyscience). sEV were diluted in PBS (Sigma-Aldrich) in order to obtain a particle number between 10 8 -10 9 particles. Three measurements of 60 s were taken per each sample and the mean value was used to determine particle number.
The movement of each particle in the field of view was measured to generate the average displacement of each particle per unit time, which was calculated using the Nanoparticle Tracking Analysis (NTA) 3.0 software (Malvern Instruments).

| Immunofluorescence staining
To determine BrdU incorporation, 50 µM of the thymidine analogue

| High content analysis immunofluorescence
The images of immunofluorescence were acquired using the automated high throughput fluorescent microscope IN Cell Analyzer 2000 (GE Healthcare) with a Å~20 objective. The fluorophores in wavelength settings were used to generate the IF images ('DAPI' for DAPI, 'FITC' for AlexaFluor 488 FITC, and 'Cy5' for AlexaFluor647).
Ten fields per well were acquired to include a minimum of 100 cells per sample well.
High content analysis (HCA) of the images were processed using the IN Cell Investigator v.2.7.3 software as described previously (Borghesan et al., 2019). DAPI was used as a nuclear mask and a top-hat method allowed the segmentation of cells. To detect cytoplasmic staining in cultured cells, a collar of 7-9 nm around DAPI was applied. Nuclear staining in the reference wavelength, that is, all the other wavelengths apart from DAPI, was quantified as an average of pixel intensity (grey scale) within the specified nuclear area.
The Cytoplasmic IF was quantified as a coefficient of variance of the pixel intensities within the collar area in the reference wavelength.
In samples of cultured cells, a threshold for positive cells was assigned above the average intensity of unstained or negative control samples.

| Statistics
Results are represented as the mean ± S.E.M except where specified. Statistical analysis are specified in the figure legend. Samples were compared to the senescent sample unless specified otherwise.

CO N FLI C T O F I NTE R E S T
AO's lab is funded by Starklabs in a project unrelated to these data.
AO is part of Starklabs Scientific Advisory Board.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in the supplementary material of this article.