Feedback between p21 and reactive oxygen production is necessary for cell senescence

Abstract

Cellular senescence--the permanent arrest of cycling in normally proliferating cells such as fibroblasts--contributes both to age-related loss of mammalian tissue homeostasis and acts as a tumour suppressor mechanism. The pathways leading to establishment of senescence are proving to be more complex than was previously envisaged. Combining in-silico interactome analysis and functional target gene inhibition, stochastic modelling and live cell microscopy, we show here that there exists a dynamic feedback loop that is triggered by a DNA damage response (DDR) and, which after a delay of several days, locks the cell into an actively maintained state of 'deep' cellular senescence. The essential feature of the loop is that long-term activation of the checkpoint gene CDKN1A (p21) induces mitochondrial dysfunction and production of reactive oxygen species (ROS) through serial signalling through GADD45-MAPK14(p38MAPK)-GRB2-TGFBR2-TGFbeta. These ROS in turn replenish short-lived DNA damage foci and maintain an ongoing DDR. We show that this loop is both necessary and sufficient for the stability of growth arrest during the establishment of the senescent phenotype.

Figure 1

Mitochondrial dysfunction and ROS production are consequences of senescence. (A) MitoSOX, DHR and NAO fluorescence in irradiated MRC5 human fibroblasts at the indicated times after irradiation as measured by flow cytometry (M±s.e.m., n=3). Asterisks indicate significant differences to non-irradiated controls (ANOVA). (B) Representative JC-1 confocal fluorescence images of MRC5 cells (red fluorescence indicates high MMP, green indicates low MMP, bar: 25 μm) and quantification of JC-1 ratios (M±s.e.m., n=3). Differences are significant with P<0.001 (Mann–Whitney rank sum test). (C) Oligomycin-resistant (mitochondrial proton leak) respiration as proportion of basal (grey bars) and maximum (FCCP-) stimulated (black bars) mitochondrial oxygen uptake in young proliferating (YOUNG), deep senescent (SEN) and irradiated (IR) cells (M±s.e.m., n=9–12). IR and SEN are different from YOUNG with P<0.05, but not from each other (ANOVA). (D, E) Doxocycline removal for 8 days (−DOX) in TRF2^ΔBΔM cells increased MitoSOX fluorescence (D) and decreased JC1 red/green ratio (E). Bar: 20 μm. Micrographs are representative for three experiments.

Figure 2

Feedback signalling through TP53-CDKN1A-GADD45A-MAPK14-GRB2-TGFBRII-TGFβ induces ROS production and maintains DDR. (A–C) MRC5 cells were transfected with wtTP53 (pC-p53), empty vector (pCDNA), siRNA as indicated, inhibitor treated or untreated (N). DHR fluorescence intensity (black) and γH2AX foci frequency (red) were measured in unirradiated cells and at 48 h after 20 Gy IR (M±s.e.m., n=3). ^*P<0.04, ^**P<0.01, ^***P<0.001 (ANOVA/Tukey), NS, not significant. (D) Pathways connecting CDKN1A to TGFβ through MAPK14. Edge thickness: LLS for any individual interaction; edge colour: pathway LLS; node colour: log fold mRNA change in senescent versus young MRC5; arrows: genes blocked in (B) and (C); grey: most probable pathway.

Figure 3

A stochastic feedback loop model predicts the kinetics of DDR and growth arrest at the single cell level. (A) Feedback loop model. Uncapped telomeres (red) or unrepaired double strand breaks (black) trigger a DDR activating TP53 and CDKN1A. High CDKN1A levels initiate signalling through GADD45, MAPK14 and TGFβ leading to mitochondrial dysfunction and increased production of ROS, which damage nuclear DNA, thus inducing more non-telomeric DNA damage foci, stabilizing DDR and growth arrest leading to a stable senescent phenotype. (B) Stochastic simulations. IR at t=0, SB203580 from t=6 days. Results are M±s.d., n=500. The vertical line indicates start of SB203580 treatment (inhibition of MAPK14 activity to 60%, compare Supplementary Figure S11). For CDKN1A, a threshold value at 2 s.d. above basal is indicated (dashed line). Experimental data (DHR flow cytometry for ROS, γH2A.X immunofluorescence for foci frequencies, CDKN1A/TUBULIN and p53 S15/total TP53 western, M±s.e.m., n⩾3) are shown in red for comparison. (C) The same simulation as in Figure 3B but completely without feedback after CDKN1A. (D) Effect of SB203580 treatment at 94 h after IR on DNA damage foci frequencies per cell as predicted by the stochastic model. M (black), s.d. (grey), n=500. (E) Confocal time series of an MRC5 cell expressing AcGFP–53BP1c at the indicated times (in h) after 20 Gy IR. SB203580 was added at t=94 h (+SB). Images are compressed stacks (maximum intensity projections), focal depth=4.5 μm, with grey values converted to a colour scale as indicated in the lookup table. Full resolution time series is given in Supplementary Movie SM1. Green arrows indicate the one focus that is persistent over the whole observation period in this cell. All other foci are transient, and red arrows mark examples of these. (F) Frequencies of AcGFP–53BP1c foci after IR and under SB203580 treatment as measured by live cell microscopy. M (black), s.e.m. (dark), s.d. (light), n=22. The vertical bar indicates the start of treatment with SB203580 at t=94 h. (G) Kaplan–Mayer survival curves for AcGFP–53BP1c foci in young, senescent and irradiated MRC5 cells. Numbers of foci analysed are between 61 and 146 per treatment from two independent experiments. ^***P=1.1 × 10⁻¹² (Cox regression). (H) Frequencies of long- and short-lived AcGFP–53BP1c foci per cell (all irradiated with 20 Gy) before (control) and after start of treatment with SB203580. M±s.e.m., n=275–354. ^*P=0.006; NS, not significant (Students’ t-test).

Figure 4

Feedback loop signalling is necessary and sufficient to maintain proliferation arrest during establishment of irreversible cell senescence. (A) ROS levels (DHR fluorescence intensity) in cells at the indicated times after 20 Gy IR and in young (CONT) and replicatively senescent (SEN) cells (M±s.e.m., n=3). Treatments with either SB203580, PBN or siRNA against CDKN1A were started at the indicated times and ROS levels were measured at 48–72 h later. At all timepoints except CONT, P<0.01 (ANOVA) for comparisons to IR-only treated cells. (B) Same experiment as in (A) with DNA damage foci frequencies measured. At all timepoints except CONT, P<0.01 (ANOVA) for comparisons to IR-only treated cells. (C) Same experiment as in (A) with frequencies of KI67-positive cells measured. At all timepoints except CONT and >20d, P<0.01 (ANOVA) for comparisons to IR-only treated cells. (D) MRC5 cells were irradiated, treated at day 6 with SB203580 or PBN and co-stained for Ki67 (red) and CDKN1A (green). Quantification of frequencies of unstained, single- and double-stained cells is on the right. (E) MRC5 cells in bulk culture were irradiated with 20 Gy, left untreated or treated with SB203580 or PBN at day 6 and cell numbers were counted at day 9 to calculate ΔPD (M±s.e.m., n=3). PD did not change significantly in control, but increased under both treatments over controls with P<0.01 (ANOVA). (F) Irradiated cells were plated at day 1 with 1000 cells/well and were left untreated or treated on day 6 with either SB203580 or PBN. Cells were stained on day 21.

Figure 5

CDKN1A knockout rescues oxidative damage in late generation TERC−/− mice. (A) MitoSOX fluorescence at 48 h after IR in MEFs. M±s.e.m., n=3, P=0.029 (Student's t-test) for IR CDKN1A+/+ against IR CDKN1A−/−. (B) MitoSOX, DHR and NAO fluorescence intensities and frequencies of γH2AX-positive MEFs with the indicated genotypes. G4 indicates late generation TERC−/− P<0.0001 (ANOVA/Tukey) for G4CDKN1A+/+ against G4CDKN1A−/− (all parameters). (C) Representative micrographs of MEF nuclei. Red: telomeres; green: γH2A.X; white: significant co-localization according to a Pearson correlation analysis. Pearson correlation coefficients for telomere-foci colocalization in MEFs from the indicated genotypes on the right (M±s.e.m., n=80–90, ^**P<0.0001, ^*P=0.043). MEFs in (A–C) were grown under 3% ambient oxygen concentration. (D–F) Representative micrographs of γH2A.X (D), broad-band autofluorescence (E) and 8oxodG immunostaining (F) in intestinal crypts from mice (aged 12–15 months) with the indicated genotypes. Quantitative data (right column) are M±s.e.m., n=3–5. ^*P<0.009 against G4TERC−/− for all parameters (ANOVA/Tukey). Arrows in (F) show examples of 8oxodG-positive cells. (G) Frequencies of 8oxodG-positive cells versus autofluorescence in the same individual crypts. Linear regression (straight line) and 95% confidence intervals (dotted lines) are indicated. P<0.0001. (H) γH2A.X foci density versus autofluorescence in the same individual crypts from all three genotypes. Linear regression (straight line) and 95% confidence intervals (dotted lines) are shown.