Small extracellular vesicles and their miRNA cargo are anti-apoptotic members of the senescence-associated secretory phenotype

Abstract

Loss of functionality during aging of cells and organisms is caused and accompanied by altered cell-to-cell communication and signalling. One factor thereby is the chronic accumulation of senescent cells and the concomitant senescence-associated secretory phenotype (SASP) that contributes to microenvironment remodelling and a pro-inflammatory status. While protein based SASP factors have been well characterized, little is known about small extracellular vesicles (sEVs) and their miRNA cargo. Therefore, we analysed secretion of sEVs from senescent human dermal fibroblasts and catalogued the therein contained miRNAs. We observed a four-fold increase of sEVs, with a concomitant increase of >80% of all cargo miRNAs. The most abundantly secreted miRNAs were predicted to collectively target mRNAs of pro-apoptotic proteins, and indeed, senescent cell derived sEVs exerted anti-apoptotic activity. In addition, we identified senescence-specific differences in miRNA composition of sEVs, with an increase of miR-23a-5p and miR-137 and a decrease of miR-625-3p, miR-766-3p, miR-199b-5p, miR-381-3p, miR-17-3p. By correlating intracellular and sEV-miRNAs, we identified miRNAs selectively retained in senescent cells (miR-21-3p and miR-17-3p) or packaged specifically into senescent cell derived sEVs (miR-15b-5p and miR-30a-3p). Therefore, we suggest sEVs and their miRNA cargo to be novel, members of the SASP that are selectively secreted or retained in cellular senescence.

Keywords: cellular senescence; exosomes; microRNA (miRNA); senescence-associated secretory phenotype (SASP); small extracellular vesicle (sEV).

Figure 1

Stress-induced premature senescent (SIPS) fibroblasts mirror hallmarks of cellular senescence. (A) Quantification of SA-ß-Gal staining shows a significant increase of ß-Gal in SIPS HDF compared to young proliferating cells at both time points post stress treatment. Representative pictures show SA-ß-Gal staining of donor HDF161 in SIPS on D21 (bottom) compared to young proliferating control (top - HDF161 in population doublings PD15). 15 pictures were taken randomly at a magnification of 100 X and counting was performed in a blinded fashion. Scale bar = 200 µm. Percentages of SA-ß-Gal positive cells from all pictures were calculated. (B) Expression of CDKN1A confirms senescence of SIPS HDF at both time points. mRNA expression levels of CDKN1A (p21) were detected by qPCR. After normalization to GAPDH, fold changes of SIPS HDF relative to quiescent (Q) control cells from D7 were calculated. (C) SIPS treatment induces permanent cell cycle arrest. Incubation with the nucleoside derivate BrdU for 24 hours followed by FITC immunolabelling for flow cytometry shows no significant incorporation of BrdU into the DNA of Q and SIPS samples compared to young dividing HDF at both time points. (D) SIPS cells show flattened and enlarged morphology. Representative pictures from donor HDF161 Q and SIPS on D21 post H₂O₂ treatment. Scale bar = 200 µm. (E) Repeated H₂O₂ treatment does not induce apoptosis. SIPS and Q control cells do not show a substantial increase in percentage (%) of total apoptotic cells at both time points compared to a positive-control (+), treated with 300 nM staurosporin for 24 hours. (A-E) Stress-induced premature senescence (SIPS) of primary human dermal fibroblasts (HDF) derived from three different donors was triggered by chronic H₂O₂ treatment on nine consecutive days. Hallmarks of cellular senescence were confirmed after seven (D7) and 21 days (D21) post last stress treatment. Averages from three biological triplicates are shown +/- SEM from raw values (n = 3). Statistical analysis was performed using 2-way RM ANOVA tested for condition and day following Bonferroni post test. n.s ≥ 0.05; *p < 0.05; **p < 0.01.

Figure 2

sEVs are members of the senescent-associated secretory phenotype (EV-SASP). (A) NTA reveals a vesicle population below 220 nm. Size distribution of vesicles determined by NTA shows percentage (%) of total counted particles against size presented in categories. (B) Media values (X50) from sEVs range from 65 to 80 nm. X50 values from peak analysis of NTA are indicated +/- SEM. circle: Q, squares: SIPS. Statistical analysis using one-way ANOVA was performed: not significant (n.s) p > 0.05. (C) Representative transmission electron microscopy image of sEVs isolated from HDF. Vesicles are around 100 nm in size and are surrounded by a double lipid membrane (arrows). Scale bar = 100 nm. A representative image of sEVs purified from HDF85 at D7 after the stress treatment is shown. (D) Representative Western blot shows expression of TSG101 (top) and GAPDH (below). Representative Western blot of total cell lysates (left) and sEVs (right lanes) from Q and SIPS HDF of donor HDF85 are shown. Total protein content of total cell lysates and purified sEV was analyzed by BCA assay and equal amounts of protein were loaded onto the gel (20 µg). (E) Senescent cells secrete more sEVs per cell than quiescent controls. Total concentration of tracked particles was normalized to the total cell number used for secretion into conditioned media. Fold changes of total particles secreted per cell, relative to Q control cells from D7, +/- relative SEM, are shown. Statistical analysis was performed using 2-way RM ANOVA tested for condition (p < 0.0001) and day (p = 0.28) following Bonferroni post test. **p < 0.01; ***p < 0.01. (A-B and E) Averages from three biological triplicates (n = 3) and two different time points each SIPS and Q, were measured in technical triplicates (n = 18) +/- relative SEM.

Figure 3

sEV-miRNAs as part of the SASP were identified in a preliminary and final qPCR screening. (A) miRNA profiling of the preliminary screening detects in total 542 (72%) secreted miRNAs. Categorization of Ct-values shows 368 miRNAs with an average signal < 38 in one or both conditions (Q, SIPS) tested. (B) The preliminary screening detects in total 386 miRNAs in both conditions tested. (C) The final qPCR screening detects 369 miRNAs with Ct-values below 38. 375 miRNAs were tested in all conditions and time points. % and number of total miRNAs detected in the screening experiment are shown. Categorization according to Ct-values. MiRNAs with an average Ct-value < 31, between 31 and 35, between 35 and 38, > 38 and not detectable are displayed. (D) The final qPCR screening detects 81% of all screened miRNAs in three donors. Averages from D7 and D21 are presented. 81% (285) of miRNAs were detected in all three donors SIPS and Q. 10% (37) of miRNAs were detected in at least two donors and 9% (30) of miRNAs were detected in one donor. (E) Principal Component analysis of sEV-miRNAs from SIPS and Q control cells from day 7 (D7) and day 21 (D21) after the treatment. The expression matrix shows the clustering of 12 samples and 369 miRNAs. Ellipses indicate a confidence level of 95% that a new observation will fall into it. Illustrated 2D-biplot explains a variance of 73.3% in principal component 1 and 7.9% in principal component 2, respectively. Exploratory analysis was done with ClustVis. Green: Q; Purple: SIPS; light colors and rectangular D7; dark colors and circle D21. (F) sEV-miRNAs are higher secreted from SIPS cells compared to Q controls. Heatmap and hierarchical clustering of 369 sEV-miRNAs after D7 and D21 (n = 12). Unit variance scaling was applied and rows are centered. MiRNAs were clustered according to correlation distance and Ward linkage method. Samples in columns are clustered using Euclidean distance and Ward linkage method. Green: Q; Purple: SIPS; light colors and blue D7; dark colors and red D21. Colors in matrix: red = upregulated, blue = downregulated. (A-B) Magnitude of secreted sEV-miRNAs was assessed in a preliminary screening using Q control and SIPS HDF of one cell strain (HDF76) and from one time point (D21). 752 miRNAs were screened using the qPCR ready to use panels supplied by Exiqon. (C-F) Final screening was performed with customized qPCR panels using three different HDF cell strains (n = 3) each Q and SIPS from two different time points (D7 and D21).

Figure 4

Senescent cell derived sEVs confer anti-apoptotic activity. (A) Barchart of the top 20 most highly secreted sEV-miRNAs. To cell count normalized Ct-values from Q and SIPS from two time points were averaged and are plotted +/- SEM derived from all 12 samples. (B) Experimental setup to test the biological effect of the EV-SASP. Recipient fibroblasts were pre-exposed to sEVs for 24 hours followed by an acute stress treatment for 2 hours with 200 µM H₂O₂ and fresh sEVS were added_. On the next day a second stress treatment with 400 µM for 2 hours was performed followed by a recovery time of 3 hours. Annexin-V-PI staining and flow cytometric measurement was used to determine % total number of apoptotic cells. (C) The EV-SASP reduces the amount of apoptotic cells of oxidatively stressed recipient cells. sEVs of SIPS and Q control cells of three different donors between 2 to 4 weeks of recovery post SIPS treatment were freshly harvested and applied before and after acute stress treatments. Human primary dermal (n = 3) and foreskin fibroblasts (n = 3) were used as recipient cells. Averages from 6 independent experiments +/- SEM are shown. Statistical analysis (n = 6) using 2-way RM ANOVA identified the factor 'EV/no EV' as a significant subject (p = 0.014) and the factor 'no stress/stress' as a significant factor (p = 0.00014). Groups were compared by Bonferroni post test, n.s ≥ 0.05; **p < 0.01, ***p < 0.01. (D) Representative pictures of recipient fibroblasts of all conditions tested prior Annexin-V-PI staining. Representative flow cytometric data are shown. Scale bar = 200 µm.

Figure 5

Changes in miRNA composition of senescent cell derived sEVs. (A) Volcano plot shows 31 significantly differently present senescence-associated (SA) sEV-miRNAs after normalization to the global means at D7 and (B) 32 SA sEV-miRNAs at D21 after the last H₂O₂ treatment. (C) Venn diagram shows miRNAs more abundantly present in sEVs of SIPS cells. (D) Venn diagram shows miRNAs less abundant in sEVs of SIPS cells. (A-B) Raw Ct-values from each sample were normalized to the respective global mean. Log2FC of SIPS relative to Q control cells were calculated. Values from D7 (panel A) and D21 (panel B) recovery are plotted on x-axis against their individual -log10(p-value) on y-axis. Horizontal dotted lines indicate a separation between miRNAs passing a p-value higher or lower than 0.05. Vertical dotted lines separate secreted miRNAs with log2FC > 1 or log2FC < 1. MiRNAs reaching a p-value < 0.05 are illustrated with green and blue dots and miRNAs with a p-value > 0.05 are shown in black. None reached the 0.05 cut-off value for the FDR of an adjusted p-values. Analysis was performed using three different HDF cell strains (n = 3) each Q and SIPS from two different time points (D7 and D21). (C-D) Log2FC was calculated and significantly regulated (p-value < 0.05) miRNAs from D7 and D21 were compared in a Venn diagram.

Figure 6

Intracellular miRNA analysis by NGS identifies early and deep senescence specific miRNAs. (A) Principal component analysis of SIPS versus Q HDF. Principal components were calculated using singular value decomposition (SVD) for imputation. Rows were scaled by applying unit variance scaling. Confidence level of 95% is indicated by ellipses assuming that a new observation from the same group will fall into it. Expression matrix of principal component 1 shows a variance of 34.8% and 24.6% in principal component 2. (B) Heatmap and hierarchical clustering of samples and miRNAs of SIPS versus Q human dermal fibroblasts. Clustering was done according to Euclidian distance and Ward linkage method. Samples in columns were clustered using correlation distance and Ward linkage method. (colors in matrix: red = highly transcribed = upregulated, blue = low transcribed = downregulated). (C) Volcano plot of differentially transcribed miRNAs in SIPS cells after seven (left D7) and (D) 21 days (right D21) post stress treatment. Log2FC are plotted on x-axis against their individual -log10 (p-value) on y-axis. Horizontal dotted lines indicate a separation between miRNA differences of a p-value higher or lower than 0.05. Vertical dotted lines separate transcribed miRNAs with log2FC > 1 or log2FC < 1. MiRNAs reaching a p-value < 0.05 are illustrated with white dots and miRNAs with a p-value > 0.05 are shown in black. (E) Venn diagram shows upregulated miRNAs of senescent cells on D7 and on D21. 46 miRNAs are commonly upregulated at both time points of senescence. (F) Venn diagram shows downregulated miRNAs of senescent cells on D7 and on D21. 36 miRNAs are commonly downregulated at both time points of senescence. (A-D) Analysis was performed using three different HDF cell strains (n = 3) each Q and SIPS from two different time points (D7 and D21). Differential expression analysis and statistics, calculated with Edge, was done with 432 miRNAs with normalized TPM signals > 5 in all conditions in at least 1 donor. (A-B) Each color and symbol represents another annotation defined by data input file. Green: Q; Purple: SIPS; light colors and rectangular D7; dark colors and circle D21.

Figure 7

Correlation of intracellular and sEV-miRNAs identifies specifically secreted versus retained miRNAs in cellular senescence. (A) Venn diagram of top 20 secreted miRNAs (positive values) from HDF, calculated by Δrank = rank_intra – rank_extra from Q and SIPS separately. (B) Venn diagram of top 20 retained (negative values) miRNAs in HDF, calculated by Δrank = rank_intra – rank_extra from Q and SIPS separately. (C) Selectively senescence-associated secreted (high values) or retained (low values) miRNAs are identified. ΔΔrank and ΔΔratio were correlated and specifically secreted (high values of ΔΔrank and ΔΔratio) or retained (low values of ΔΔrank and ΔΔratio) senescence-associated miRNAs were identified. Spearman correlation R = 0.81 with a 95% confidence interval 0.76 to 0.85 P value (two-tailed) < 0.0001. Bubble size corresponds to quartiles calculated from transformed average Ct-values, whereby the larger the bubble size, the higher the expression value. Dotted lines represent the 25% and 75% percentiles, which define the specifically secreted and retained miRNAs in senescence. ΔΔrank: 25%: 8.0; Median: -0.5; 75%: 9.0; ΔΔratio: 25%: 0.7099; Median: 0.927; 75%: 1.186. (D) Venn diagram of the top 20 specifically secreted senescence-associated sEV-miRNAs. MiRNAs are identified by comparing the top 20 of ΔΔrank and ΔΔratio method. (E) Venn diagram of top 20 specifically retained senescence-associated miRNAs. MiRNAs are identified by comparing the top 20 of ΔΔrank and ΔΔratio method. (F) (A-E) Correlation was performed with 228 miRNAs identified intracellularly (small RNA-NGS) as well as in sEVs (qPCR panels) in samples derived from three different HDF cell strains (n = 3) each Q and SIPS from two different time points (D7 and D21).