Senescence suppresses the integrated stress response and activates a stress-remodeled secretory phenotype

Abstract

Senescence is a state of indefinite cell-cycle arrest associated with aging, cancer, and age-related diseases. Here, we find that translational deregulation, together with a corresponding maladaptive integrated stress response (ISR), is a hallmark of senescence that desensitizes senescent cells to stress. We present evidence that senescent cells maintain high levels of eIF2α phosphorylation, typical of ISR activation, but translationally repress production of the stress response activating transcription factor 4 (ATF4) by ineffective bypass of the inhibitory upstream open reading frames (uORFs). Surprisingly, ATF4 translation remains inhibited even after acute proteotoxic and amino acid starvation stressors, resulting in a highly diminished stress response. We also find that stress augments the senescence-associated secretory phenotype with sustained remodeling of inflammatory factors expression that is suppressed by non-uORF carrying ATF4 mRNA expression. Our results thus show that senescent cells possess a unique response to stress, which entails an increase in their inflammatory profile.

Keywords: ATF4; ER stress; ISR; SASP; integrated stress response; nanopore direct RNA sequencing; proteomics; ribosome sequencing; senescence; senescence-associated secretory phenotype; translation.

Figure 1.. Cellular senescence is characterized by persistent ER stress signaling but no downstream ATF4 activation

(A) Western blot analysis of senescence markers p21 (CDKN1A), p16 (CDKN2A), and nuclear protein Lamin B1 (LMNB1) for etoposide induced senescent (Sen-Etop), quiescent (Qui), and cycling (Cyc) cells. ACTB is used as a loading control. (B) RT-qPCR analysis of CDKN1A and CXCL8 mRNAs from cells analyzed in (A). Significance determined by t-test with n=3 (**= p-value<0.01; *** = p-value < 0.001). Normalized to ACTB. (C) Quantification of senescence-associated beta-galactosidase staining (SA-βGal) of Sen-Etop and Cyc cells (*** = p-value < 0.001). (D) Principal component analysis of mass spectrometry data for cells harvested in triplicate. (E) Volcano plots of Sen-Etop/Cyc and Sen-Etop/Qui differential protein expression analysis, respectively. Senescence markers LMNB1 and CDKN1A are colored in orange. Statistically significant (p-value ≤ 0.05 by two-tailed student’s t-test) log2FC values are colored as follows: log2FC ≥ 0.2, blue; log2FC ≤ −0.2, red; log2FC between −0.2 and 0.2, grey. (F) Scatterplot of protein fold-change values from the Sen-Etop /Cyc (x-axis) and Sen-Etop /Qui (y-axis) comparisons. Dotted lines mark 0.2 and −0.2 log2FC in both comparisons. Quadrants are annotated with the top 5 most changed proteins, with respect to Sen/Cyc. (G) Scatterplot of protein fold-change values from the Sen-Etop /Cyc (x-axis) and Sen-Etop /Qui (y-axis) comparisons for selected ER stress proteins (H) Western blot analysis of ISR markers eIF2α phosphorylation (eIF2α-P) and ATF4; cells analyzed are cycling (Cyc), etoposide-induced senescent (Etop), ionizing-radiation induced senescent (IR), contact-inhibited quiescent (Qui), and cells treated for 3h with 25nM thapsigargin (Tg). (I) Quantification of eIF2α-P signal normalized to eIF2α apoprotein and relative to the ratio in Cyc cells. Statistical significance assessed using two-tailed student’s t-test; nd = no difference. (J-K) Western blot analysis of senescent cells harvested 3d after transfection with siRNA (45 nM) targeting the eIF2AK3 eIF2α kinase. Quantification performed relative to respective cell type transfected with control siRNA (siCtrl) (n=2). (L) Western blot of lysates harvested at indicated time points after initial etoposide treatment. Etoposide treatments were applied at time points marked with arrows; harvests on day 0, 2, and 4 were performed prior to etoposide treatment. eIF2α-P quantified relative to eIF2α apoprotein; eIF2α quantified relative to ACTB.

Figure 2.. ATF4 translation is inhibited at uORF2 in senescent cells.

(A) RT-qPCR analysis of ATF4 mRNA levels; significance assessed with two-tailed student’s t-test. Normalized to ACTB. (B) Western blot of indicated cells 3 days after transfection with siRNA targeting CEBPB. At right, quantification of mRNA levels obtained from RT-qPCR; normalized to ACTB. (C) Western blot of cells transfected with indicated ATF4 mRNA for 6.5h relative to un-transfected Cyc cells. The image is a composite of three gels. (D) Polysome profiling traces on a 10–45% sucrose gradient for lysates from indicated cell types after treatment with 0.1 mg/mL cycloheximide. The ratio of polysome/monosome fractions (P/M) are quantified and shown in inset. (E) RT-qPCR of ATF4 levels for RNA extracted from sucrose fractions of polysome profiling experiment for indicated samples. ATF4 RNA abundance calculated as a percentage of total ATF4 mRNA quantified among fractions (F) Boxplot showing polysome/monosome density (left) and monosome density (right) of indicated cells co-treated with 25 nM Tg/mock and 200nM ISRIB/mock for 2h and subjected to polysome profiling as above. (G) Differential RNA-seq expression analysis for the Sen-Etop/Cyc comparison. Statistically significant (adjusted p-value ≤ 0.05) log2FC values are colored as follows: log2FC ≤-1, red; log2FC ≥ 1, blue; log2FC between −1 and 1, black. Inset indicates log2FC values for senescence markers in the Sen-Etop/Cyc and Sen-IR/Cyc comparisons. (H) Metaplot distribution of RPFs on the ATF4 mRNA (ENST00000674920). Normalized density is calculated as the sum of RPFs at each position normalized to the total RPFs on the transcript. (I) Quantification of relative uORF (20:268) and ORF (366:1338) densities on ATF4 mRNA in (I), normalized to respective window sizes; RPFs in overlap between uORF2 and the ORF (269:365) are excluded.

Figure 3.. Senescent cells have a global decrease in translating ribosomes

(A) Volcano plot of Sen-Etop/Cyc differential TE (see methods) from ribosome sequencing. Statistically significant (adjusted p-value ≤ 0.05) log2FC values are colored as follows: log2FC ≤ −1, red; log2FC ≥ 1, blue; log2FC between −1 and 1, black. (B) Metaplot distribution of ribosome footprints near the CDS start site (-100 to +300). Reads from Sen-Etop (red) and Sen-IR (cyan) are compared to Cyc (grey). (C) Boxplot of log₁₀(Ribo-seq/RNA-seq) reads after RPKM normalization and filtering for protein-coding genes. (D) Linear regression analysis of Ribo-seq~RNA-seq RPKM reads. (E) Comparison of slopes from linear regression analysis described in (D) between Cyc and Sen cells from Ribo-seq data obtained in this work and other analyses of senescent cells^,. (F) Linear regression analysis of Ribo-seq data as above for data from a previous study on aged mice. Analysis of murine samples harvested from liver (left) and kidney (right) at indicated ages in months. (G) Diagram of ribosome content assay using T1 RNAse and cell count normalized lysates. (H) Polysome profiling using lysate derived from 10 million cells pre-T1 treatment (left) and post-T1 treatment (right); the grey boxes in both spectra are expected to represent the same total ribosomes. Post-T1 signal is quantified as the total absorbance within the grey box and summarized as a bar graph inset. (I) Representative images from protein synthesis pulse-label assay (see methods) where time indicates length of pulse. Green, nascent protein; blue, DAPI stain. At right, quantification of translation as brightness detected within nuclear region divided by total DAPI count detected (Brightness/Cell). (J) Boxplot of ribosomal protein abundances from mass spectrometry data. Significance of ribosomal protein reduction is assessed by two-tailed student’s t-test (*= p-value<0.05; *** = p-value < 0.001). (K) Western blot validation of expression for select ribosomal proteins.

Figure 4.. Senescent cells fail to activate the ISR in response to acute stress.

(A) Western blot of Sen-Etop, Sen-IR, Qui, and Cyc cells treated for 3 h with DMSO (none) or 25nM Tg (Tg). The ATF4 Efficiency ratio is measured using intensity-based quantification of gel images after normalizing ATF4 to ACTB and eIF2α-P to eIF2α, and then calculating the resulting ATF4/eIF2α-P ratio. (B) ATF4 Efficiency, calculated as above, related as a percentage of the ATF4 Efficiency calculated for CTg cells (n=3). (C) Western blot of Sen-Etop and Cyc cells treated with 25 nM Tg and harvested after indicated times; eIF2AK3 inhibitor (GSK, 550 nM) added concurrently with Tg when indicated. ATF4 efficiency as described for (A). (D) Differential gene expression for nanopore-based direct RNA sequencing (dRNA-seq) of Sen-Etop and Cyc cells (n=2); green, log2FC ≥ 1, red, log2FC ≤ −1. (E) Scatter plot of differential gene expression for previously described ER-stress activated genes ^–. Dots are colored based on whether they are at least 2-fold increased (above dashed lines) in the indicated comparisons (exclusive or both). (F-G) Western blot of cells starved for indicated times in culture media lacking methionine and cysteine (-Met -Cys). (H) Scatter plot as in (E) for starvation-activated genes ^,.

Figure 5.. Acute stress remodels the senescence associated secretory phenotype.

(A) Comparison of log2FC values for the ETg/Etop and CTg/Cyc comparisons for all detected RNAs; insets show enriched non-disease KEGG pathways for genes in respective quadrant. (B) KEGG pathway enrichment analysis of genes observed to be uniquely enhanced by stress. (C) Heatmap of genes with annotated secretory function; color indicates log2FC for each library relative to cycling cells. (D) Conditioned media collected in 24h intervals for cells treated with DMSO (Mock) or Tg for 3h at 25nM. Cells were then allowed to recover in fresh media for indicated time. Secreted proteins determined by a multiplex ELISA (see methods). (E) RT-qPCR of SASP factors CCL2 and CXCL8 from RNA extracted from cells used in (D). Normalized to ACTB. (F) Volcano plots identifying Stress-induced Secretory Remodeling (SSR) factors (p-adj ≤ 0.05, n=3) from secretory proteins with statistically significant change in Stressed/Non-Stressed Sen-Etop cells. Text labels correspond to genes with a |log2FC| ≥ 1. (G) Dumbbell plot displaying SSR factors from (F) reanalyzed using DEseq2 to Cyc cells. Log2FC of Sen-Etop Non-Stressed/Cyc and Sen-Etop Stressed/Cyc in red and yellow, respectively. (H) Bar plot of Sen-Etop Stressed cells harvested at indicated time points post-Tg treatment (3h, 25 nM) and compared to (3h, DMSO) using DESeq2 (n = 3 for each time point). (I) Western blot of cells with or without Tg treatment (3h, 25 nM). (J) Western blot of cells harvested 20 h after recovery from Tg treatment (3h, 25 nM). (K) SSR factors analyzed as in (G) compared to Sen-Etop Stressed + ATF4 (relative to Cyc + ATF4) in blue.

Figure 6.. Model for translation repression and ISR inhibition in senescence.

(A) Scatter plot of PC1 and PC2 from PCA of dRNA-Seq libraries. (B) Bar plot of percent change in isoform usage and schematic of isoform structure for RPS6. (C) Model of the proposed continuum of ribosome and ternary complex availability that determines ISR activation in cycling, quiescence, and senescence. The progressive imbalance of eIF2α/eIF2α-P ratio and ribosome abundance results in a persistent state of stress desensitization in senescence.