The senescent bystander effect is caused by ROS-activated NF-κB signalling

1It is becoming increasingly clear that cell senescence is an important cause of the loss of tissue functionality and homeostatic capacity that characterizes ageing. Senescent cells accumulate in many, if not all, tissues with age. In mice, the rate of this accumulation predicts mean and maximum lifespan of animals over an almost threefold difference in life expectancy between slow- and fast-ageing cohorts. Selective ablation of senescent cells postponed the development of age-associated tissue dysfunction/disease and extended mean lifespan in mice. Drugs like rapamycin, which extend life expectancy in mice, have been shown to be senostatics, i.e. suppressors of multiple aspects of the senescent phenotype. First experimental drugs that selectively kill senescent cells have been developed and show encouraging results in a number of mouse models of age-related diseases.

2An important role of cell senescence in ageing should not come as a surprise. After all, cell senescence is a major response to stress and molecular damage. While it was originally regarded only as a consequence of ongoing cell division and associated telomere loss, seminal work from Toussaint and others soon showed that both telomere-dependent and independent senescence can be induced by DNA-damaging stress acting via the by now well-defined DNA damage response (DDR) pathway. As was predicted early, the distinction between telomere-dependent (replicative) and telomere-independent stress-induced senescence became more and more blurred with data showing stress-related telomere loss as predictor of transition to senescence at the single cell level as well as the defining role for telomere damage without telomere loss for senescence in mouse cells in vitro and in vivo. Moreover, cell senescence is also the response to stresses that are not primarily genotoxic but impact preferentially on the proteome or the mitochondrion.

3In addition to being a stress response, cell senescence is itself an inducer of stress in the tissues and organisms harbouring senescent cells. This is because of three important properties of senescent cells: i) Senescent cells cannot proliferate. Accordingly, senescence of stem and progenitor cells limits the regenerative potential of tissues and organisms. ii) Senescent cells hyper-produce and secrete a host of bioactive peptides (the so-called Senescence-Associated Secretory Phenotype SASP), prominently including pro-inflammatory cytokines like IL-1, IL-6 and IL-8 under control of the transcription factor NF-κB. These not only stabilize senescence in an autocrine fashion but may also be a significant cause of age-related chronic sterile inflammation and contribute to bystander effects, i.e. enhanced tumorigenicity of transformed cells and induction of DNA damage and senescence in normal bystander cells. iii) The activated DDR in senescent cells also causes Senescence-Associated Mitochondrial Dysfunction (SAMD), i.e. loss of mitochondrial membrane potential, decreased respiratory coupling and increased ROS production, all associated with failing mitophagy. Importantly, if mitochondria were selectively and specifically obliterated in senescent cells, ROS levels returned to normal, indicating SAMD rather than extra-mitochondrial mechanisms as the major source of elevated cellular ROS in senescence. As for SASP, SAMD initiates an autocrine positive feedback loop that enhances ROS-mediated nuclear DNA damage and thus stabilizes senescence. In its turn, ROS release from senescent cells contributes to senescence induction in bystander cells. Finally, SAMD appears as a cause for metabolic insufficiencies in ageing organisms.

4To understand better how senescent cells contribute to ageing, it will be necessary to dissect how these two major senescent phenotypes, SAMD and SASP, interact. It is clear that the two can influence each other: Short telomeres do not only cause senescence in late generation telomerase knockout mice, they also induce widespread inflammation. Conversely, chronic inflammation in mice null for the repressor p50 subunit of NF-κB aggravates ROS production and senescence.

5To further assess the interaction between SASP and SAMD, we treated human fibroblasts in various modes of senescence with either antioxidants or NF-κB inhibitors and monitored the effects on the corresponding phenotype. We also analyzed the relative efficiencies to induce a DDR in bystander cells under SASP- vs SAMD-inhibition. We found that antioxidants suppress activation of NF-κB, while suppression of NF-κB did not reduce SAMD in multiple senescence modes. Both scavenging of ROS and inhibition of NF-κB blocked the bystander effect. We conclude that SAMD is an important inducer of the pro-inflammatory SASP, which then mediates the bystander effect.

6There has been sometimes contradictory discussion about the impact of factors such as mode of senescence induction, cell strain, mitochondrial (dys-)function, etc. on SASP composition and potential downstream bystander effects. We therefore tested the interaction between SASP and SAMD in multiple human fibroblast systems, using two independent cell strains (MRC5 and IMR90) and three different modes of senescence. Cellular senescence was induced via either proliferative exhaustion (replicative senescence in MRC5, ‘RS'), IR stress-induced senescence (20 Gy X-ray irradiation in MRC5 and IMR90, ‘IR') or oncogene induced senescence (Ras induction in a stably transfected IMR90 cell line, ' + RAS'). Successful induction of senescence was confirmed by reduction of growth to less than 10% of controls, senescent morphology and Sen-β-Gal positivity in at least 80% of the cells (data not shown). NF-κB activity was measured as the relative nuclear/cytoplasmic intensity of the immunofluorescence signal from the predominant NF-κB transcription factor subunit RelA (Suppl. Fig. S1A). The assay was standardized by showing an increase in nuclear RelA fluorescence following stimulation of NF-κB with 10 ng/ml TNFα (Suppl. Fig. S1B). Fibroblasts in all senescence modes displayed significantly higher levels of nuclear RelA than the proliferating control cells (Fig. 1A). Cellular ROS levels were measured by DHE fluorescence. While we are aware that DHE fluorescence intensity is a marker but not a quantitative measure of cytoplasmic superoxide, and that cytoplasmic ROS originate from multiple sources including non-mitochondrial ones, we have previously shown that the increase in ROS markers associated with senescence can solely be attributed to mitochondria. Therefore, we regard the increase of DHE fluorescence above basal levels in senescent cells as a simple indicator of SAMD. Accordingly, all senescent fibroblasts displayed increased DHE fluorescence (Fig. 1B) in accordance with earlier data. Treatment of senescent cells for 2 days with the antioxidant catalase (CAT) depleted intracellular ROS levels in senescent cells, but had little effect on the low ROS levels in proliferating cells (Suppl. Fig. S2). Importantly, catalase treatment completely rescued the senescence-associated activation of NF-κB (Fig. 1A), suggesting that SAMD-mediated ROS production is necessary and sufficient for NF-κB activation in senescence. In contrast, inhibition of NF-κB with the specific inhibitor, Bay 11-7082 (1 μM for 2 days) reduced NF-κB activation in senescent cells as expected (Suppl. Fig. S3), but had no effect on ROS levels (Fig. 1B), indicating that the SAMD-mediated ROS hyper-production was independent from NF-κB signaling in all senescence modes. To corroborate this finding, we stably over-expressed a dominant negative NF-κB suppressor, ΔIκBα in MRC5 fibroblasts and grew them to replicative senescence. While this was effective in suppression of TNFα-mediated NF-κB activation (Suppl. Fig. S1B), it again did not diminish ROS production in senescent cells (Fig. 1C).

7IL-6 and IL-8 are two important pro-inflammatory cytokines and major components of the SASP in human fibroblasts. They are under control of NF-κB, mediated partly via induction of IL-1. However, other transcription factors, notably C/EBPβ, also determine expression of these cytokines in senescence. Thus, ROS might stimulate the SASP independently of NF-κB activity. Therefore, we next assessed how NF-κB, IL- 1α and ROS determined the amount of IL-6 and IL-8 secreted from senescent cells. As expected, secretion of both IL-6 and IL-8 was enhanced in all modes of fibroblast senescence (Fig. 1D). Interestingly, inhibition of NF-κB had differential effects depending on senescence mode and interleukin species: It was highly effective for both IL-6 and IL-8 in IR-induced senescence (slightly less so for MRC5 than for IMR90) but resulted only in a minor inhibition of secretion of both peptides in MRC5 replicative senescence. In IMR90 + RAS, it was highly effective for IL-8 but did not change IL-6 secretion at all (Fig. 1D). These data indicate variable roles for signaling from the transcription factor NF-κB in SASP, depending on senescence model and specific interleukin. To test whether this variation might arise at the level of IL-1α expression, we knocked down IL-1α in two senescence models, MRC5-IR and IMR90 + RAS, which differed in their IL-6 and IL-8 response to inhibition of NF-κB (Fig. 1D). Transfection with anti-IL1α siRNA reduced its mRNA (Suppl. Fig. S4A) and protein levels (Suppl. Fig. S4B) similarly in both senescence modes. It also impacted equally on IL-6 and IL-8 release in both forms of senescence (Fig. 1E), suggesting that the different responses of these cytokines to NF-κB inhibition are not mediated at or downstream of cytokine IL-1α expression.

8In co-culture with young, proliferation-competent fibroblasts, senescent cells induce DNA damage that ultimately causes senescence in the bystander cells. To investigate the contributions of SASP and SAMD to the bystander effect, we established 2D co-cultures of young MRC5 human diploid fibroblasts (bystanders) with inducer fibroblasts rendered senescent in different modes (Fig. 2A). Bystander cells were discriminated by expression of an mCherry-tagged 53BP1 which in young cells appear as diffuse nuclear red fluorescence with only occasional bright foci, while it forms intense nuclear foci upon induction of DNA damage and in senescence. Already after 2 days of co-culture, ROS levels in young bystander cells were enhanced and stayed above control levels for the duration of the experiment, with levels comparable to senescent cells (Suppl. Fig. S5). After 6 days of co-culture with inducer cells in any senescence mode, young bystander cells displayed increased DDR (Fig. 2B). Average frequencies of 53BP1 foci were significantly increased after co-culture with senescent, but not non-senescent inducers (Fig. 2C), irrespective of the cell strain or the senescence mode (Fig. 2C). A two way ANOVA with post hoc analysis indicated that there was no significant difference between treatment cell line, with all senescent inducers producing a significant increase in DDR compared to their young inducer controls (either MRC5 or IMR90).

9We next performed the co-culture experiments with MRC5-RS or IMR90 + RAS as inducers in the presence of either the antioxidant catalase or the NF-κB inhibitor Bay11-7082. After 6 days of co-culture, the bystander effect from both types of senescence inducers was reduced by catalase, while the NF-κB inhibitor was able to suppress the bystander effect of IMR90 + RAS, but not that of replicatively senescent MRC5 inducers (Fig. 2D). This may be because Bay11-7082 suppresses IL-6 and IL-8 secretion from replicatively senescent MRC5 fibroblasts only weakly (see Fig. 1C). However, when the incubation of bystander cells with replicatively senescent MRC5 fibroblasts was prolonged to 21 days, NF-κB inhibition became as efficient in suppression of the bystander effect as incubation with catalase (Fig. 2E).

10The long-term bystander effect from MRC5-RS fibroblasts was equally inhibited by expressing the NF-κB super-repressor ΔIκBα in the inducer cells only (Fig. 2F). Conversely, pre-treatment of senescent inducer cells with TNFα before co-culture enhanced the bystander effect from wild-type, but not ΔIκBα − expressing inducers (Fig. 2G). Catalase treatment was sufficient to rescue even this enhanced bystander effect (Fig. 2G).

11Insufficient mitophagy is an important cause of mitochondrial dysfunction and ROS production in senescence, and activation of mitophagy by treatment with rapamycin improves mitochondrial function and suppresses the pro-inflammatory SASP. Accordingly, treatment of co-cultured inducer and bystander cells with the specific mTORC1 inhibitor Torin1 was as efficient to suppress the bystander effect as treatment with the NF-κB inhibitor Bay 11–7082 (Fig. 2H). The stress response kinase p38/MAPK14 is a major intermediate for ROS-mediated upregulation of NF-κB, and its inhibition similarly suppressed the bystander effect (Fig. 2H).

12Together these data suggest that in senescence, mitochondrial dysfunction induces ROS, which activate NF-κB via p38 and possibly other pathways. This is responsible for the induction of a pro-inflammatory SASP, albeit to different degrees in different cell types and/or modes of senescence induction. SASP factors appear to be the main mediators of cell-to-cell communication, transmitting the bystander effect.

13A recent series of intervention experiments has clearly pinpointed senescent cells as an important cause of age-associated tissue dysfunction, (multi-)morbidity and, potentially, mortality. Senescent cells impact on their tissue environment by a host of pathways: they secrete bioactive, especially pro-inflammatory and matrix-degrading, peptides (the SASP,), produce and release ROS as consequence of SAMD, shed microvesicles that contain (relatively) specific miRNAs and proteins and release modified lipids. These may act as signaling molecules that spread senescence to bystander cells, thus aggravating the pro-ageing effects of senescent cells. While previous data are in agreement with the idea that SASP factors might play a major role in induction of the senescent bystander effect, other mediators have not been ruled out. In fact, effects from conditioned medium were sometimes weak in comparison to cell–cell contact or ROS-mediated effects. Moreover, specific roles for individual SASP factors in transmission of the bystander effect are not known.

14To understand the interactions between SAMD and SASP in the induction of bystander senescence, we examined the effects of scavenging SAMD-dependent ROS or inhibiting the major SASP transcription factor NF-κB in multiple modes of fibroblast senescence. It is known that ROS can, in a context-depending manner, either activate or inhibit NF-κB. Our results show that hyper-production of ROS in senescence is driving NF-κB activation in all examined modes of senescence. Conversely, inhibition of NF-κB did not change ROS hyper-production in senescence. This was unexpected as earlier work showed that activation of NF-κB signaling by knockout of the NFKB1 gene resulted in accelerated senescence and aggravated ROS production in senescent cells. The probable explanation for this discrepancy lies in the existence of multiple pathways leading to ROS production in senescent cells: the DNA Damage Response drives SAMD and thus increases ROS to a level that is necessary to establish senescence; activation of NF-κB increases ROS further by pathway(s) that involve cyclooxygenase activation among others, but not enhancement of mitochondrial ROS production. Thus, SAMD activates NF-κB via ROS but is not targeted by NF-κB activity, although NF-κB activation contributes to non-mitochondrial ROS production. This interpretation is supported by data showing that the pro-inflammatory SASP is suppressed in those modes of senescence where mitochondrial dysfunction does not lead to enhanced cytoplasmic (and presumably nuclear) ROS levels. In phagocytes, mitochondrial ROS production and NF-κB activation are interconnected by formation of a complex between mitochondrial complex I–evolutionarily conserved signaling intermediate in Toll pathways (ECSIT) and the ubiquitin ligase tumor necrosis factor receptor-associated factor 6 (TRAF6). Whether this complex is also activated in fibroblasts senescence is not clear yet.

15NF-κB is not the only transcription factor driving the pro-inflammatory arm of the SASP. Our data show that its impact on secretion of the two major pro-inflammatory cytokines, IL-6 and IL-8, significantly depends on cell type, senescence mode and specific cytokine. In MRC5 replicative senescence, for instance, the impact of NF-κB signaling on IL-6 and IL-8 secretion was significant but rather small (Fig. 1D). In contrast, a number of interventions were all similarly effective in blocking the senescence-induced induction of a DDR in bystander MRC5 fibroblasts. Although the antioxidant reduced the bystander response earlier than the NF-κB inhibitor, both were able to rescue the bystander effect completely after prolonged incubation, and this was not different from inhibition of signaling through mTORC1 or p38MAPK. We interpret these results as follows:

16ROS hyper-production in senescence is at least partially due, or aggravated by, accumulation of dysfunctional mitochondria, which in turn is mediated by a failure of mitophagy. SAMD can be rescued by inhibition of mTORC1 signalling using rapamycin or torin, activating autophagy. ROS signal through p38MAPK to activate NF-κB in senescent cells, and NF-κB in turn is responsible for some, but not all of the SASP. The fact that, given sufficiently long incubation, inhibitors for all tested components of this pathway effectively and completely rescued the bystander response suggest that peptides transcribed under control of NF-κB are the intercellular mediators of the senescent bystander response. However, neither IL-6 nor IL-8 seem to play major roles in this respect, as they are only to a small degree controlled by NF-κB in senescent MRC5 fibroblasts. Further work will be necessary to identify more potent intercellular signal transducer candidates for the senescent bystander effect.

17Proliferating cells. Embryonic human lung fibroblasts MRC5 were from ECACC. MRC5-mCherry-53BP1 cells were created as described previously. Cells were cultured in DMEM (Sigma Aldrich) supplemented with 10% foetal calf serum (Sigma Aldrich), 100 U/ml penicillin, 100 mg/ml streptomycin and 2 mM l-Glutamine at 20% oxygen, 5% CO2 at 37 °C. Embryonic lung fibroblast IMR90 were cultured in phenol-free DMEM with supplements as above at 3% oxygen, 5% CO2, 37 °C. Stably transfected IMR90ER:RAS (a gift from Peter Adams) were cultured in the same manner as the parental IMR90 plus 300 μg/ml G418 selection at 3% oxygen, sub-cultured every third day.

18The NF-κB super-repressor vector pLenti-ΔIκBα-IRES-EGFP was created by first cloning IκΒα minus the first 40 amino acids from pIκB-EGFP (Clontech), into the MCS of pIRES2-EGFP (Clontech) and then the entire ΔIκBα-IRES-EGFP cassette into the expression site of pLenti6-UbC-V5DEST. Young (PD< x +25) MRC5 were transduced with viral particles produced in Hek293FT cells using Virapower packaging mix as per the manufacturer's instructions (Life Technologies) and selected for 6 days with Blasticidin before utilisation.

19Senescent cells. To generate replicatively senescent cells (MRC5 RepSen), MRC5 cells were cultured in complete medium at 20% oxygen, passaging before confluency, until PD < 0.1 per month. Stress induced senescent MRC5 (MRC5 IR) and IMR90 (IMR90 IR) were created by X-ray irradiation. 6300 cells/cm2 were seeded 24 h prior to 20 Gy irradiation and were used for experiments 7 days later. Oncogene-induced senescent cells (IMR90 OIS) were used on day 7 after induction of RAS overexpression in IMR-90ER:RAS by incubation with 100 nM 4-hydroxy-tamoxifen (4OHT).

20For direct co-culture experiments, inducer senescent cells were plated at 5500 cells/cm2 density either on glass cover slips (No. 1.5) in 12 well plates or in 35 mm glass bottomed dishes (Willco) for live cell microscopy. Twenty four hours later, young bystander cells were plated in 1:1 ratio and co-cultures were maintained for 6 days. For treatment with chemical inhibitors: 1 μM Bay 11–7082, 10 nM Torin1, 1 μM UR-13756 or 100 U/ml catalase or DMSO as a control were added 24 h post seeding of bystanders and re-freshed every 3 days. TNFα was used at 10 ng/ml. Pre-treatment of inducer cells with TNFα was performed for 1 h before replacing with fresh medium and seeding bystander cells.

21Treatment with siRNA was performed 24 h after seeding of bystander cells. Co-cultures were transfected using 5 nM Silencer select Pre-designed siRNA and Lipofectamine™ RNAiMAX (Thermo Fisher) in antibiotic-free medium. Forty-eight hours later, transfections were repeated and cells were fixed after 6 days in co-culture.

22ROS production was assessed in cultured fibroblasts by staining trypsinised cells with 10 μM Dihydroethidium (DHE) and measuring fluorescence by flow cytometry as described previously. For measurement of co-cultured cells, DHE fluorescence was measured by live cell microscopy to robustly separate young bystanders from senescent inducers (see Microscopy section below).

23Cells were incubated in fresh half normal volume of media for 40 h prior to collection. Samples were cleared by centrifugation for 10 min at 4 °C at 8000g and then concentrated 2 times by Cryospin. Protein concentration was measured using DC Protein assay (BioRad) according to the manufacturer's protocol. Concentrated conditioned media were assessed using the respective Sandwich ELISA DuoSet Development System kits (R&D Systems, Abingdon, UK). The kits were tested and suitably modified to be transferred onto a 384-well plate platform with the use of an EpMotion 5075 Liquid Handling robot. All colorimetric measurements were performed at 450 nm with wavelength correction at 550 nm on an Omega microplate reader (BMG Labtech, Aylesbury, UK) applying a four-parameter logistic curve fit to the respective standard curves. Data were normalized to total protein concentrations.

24For end point analysis of mCherry fluorescence, cells were fixed for 10 min with 4% PFA and mounted using Prolong Diamond Antifade with DAPI (Thermo Fisher, UK), or were imaged by live cell microscopy.

25For immunofluorescence, cells were fixed as above and incubated with rabbit anti-p65 (1:400, Cell Signalling) overnight at 4oC. For visualization of the staining, AlexaFluor 488 conjugated secondary antibody was used (Thermo Fisher).

26Imaging was performed using a Zeiss Spinning Disk Confocal system, coupled with Quant-EM 512 SC camera (Photometrics) and Yokogawa CSU-X1 spinning disk head, with a 63 x (NA 1.4) objective, and a temperature and gas regulated incubator (Pecon). Image acquisition was via AxioVision software (Zeiss, UK) and image analysis (counts of mCherry DNA damage foci per nucleus, and RelA nuclear and cytosolic fluorescence intensities) was performed in ImageJ. To ensure accurate comparisons, all imaging within an experiment was performed using the same excitation intensities, exposure times and objectives. Also, to improve quantification, mean nuclear RelA fluorescence was normalised to the cells respective cytosolic RelA fluorescence staining. This removes any differences from staining variability from coverslip to coverslip, or any slight differences in focal plane or laser intensity from field to field. Population-level activation rates were determined by measuring the percentage of cells per field that had predominantly nuclear (i.e. nuclear: cytosolic fluorescence ratio of RelA staining was >1 per cell) RelA fluorescence.

27For measurement of DHE fluorescence by live cell microscopy, MRC5 cells stably transduced with an EGFP-luc expression cassette (pSLIEW) were used as bystanders to identify them from the senescent parental line MRC5 inducers based upon nuclear EGFP fluorescence. Cells were plated as described above and stained with DHE as for flow cytometry measurements, then washed and replaced with Fluorobrite medium (Thermo Fisher) plus 10% FCS. Staining was performed in a serial manner to maintain the same incubation times per dish. Images were captured using a DMi8 widefield fluorescence microscope equipped with a 20 × 0.8NA objective, a Sola-SE LED light source (Lumencor) and a Flash4 sCMOS camera (Hamamatsu) using LASX software (Leica). EGFP, DHE and brightfield images were captured for each field using GFP and TX2 filter cubes (Leica) for the fluorescent channels representing EGFP and DHE respectively. Mean nuclear DHE fluorescence (from TX2 filterset) was determined using ImageJ for at least 100 cells from 3 different dishes per group, using the brightfield image to identify the nucleus. Intensities were background subtracted based upon mean+ 2xSD nuclear fluorescence intensity from TX2 images of young EGFP expressing cells without any DHE staining. In co-cultured dishes, bystander cells were identified by EGFP expression being present throughout the cell (using the GFP filterset image), and were identified by their nuclear EGFP fluorescence to allow discrimination from lipofuscin-based autofluorescence, visible in the cytosol of the senescent inducer cells.

28All analysis was performed using SigmaPlot 12.5 software, using one-way ANOVA test (parametric data) or Kruskal-Wallis ANOVA (non-parametric). Post-hoc tests were performed pairwise within datasets using Dunn’s or Holm-Sidak method. Cut-off of significance was set at p < 0.05 (denoted as * in Figures). Two-way ANOVA test was used for assessment of the relationship between cell line background and mode of senescence or treatments.