Chemically induced reprogramming to reverse cellular aging

Nucleocytoplasmic compartmentalization (NCC) is disrupted in fibroblasts from old individuals and senescent cells

To identify small molecules that ostensibly reverse the effects of aging and senescence, we developed an efficient high-throughput system. Rather than relying on a limited set of genes that exhibit age-related changes and to ensure reliability and applicability across various cell types, we sought to develop an age-dependent assay that acted as a surrogate for cellular health and youthful gene expression patterns. To increase scalability and ease of use, we sought a fluorescence-based system that could be quantified in millions of cells per experiment via automated microscopy.

One of the most well-conserved physiological hallmarks of aging is deterioration of nucleocytoplasmic compartmentalization (NCC), which can be visualized as the leakage of nuclear proteins into the cytoplasm and failure of proteins to be imported into the nucleus [34, 35]. In neurons and astrocytes directly converted from fibroblasts of old humans, as well as old nematodes and rat brain tissue, the nuclear pore complex is deteriorated, leading to increased nuclear permeability and cytosolic protein aggregation [34–36].

To monitor age-associated alterations in nuclear permeability, we introduced the NCC reporter system into human fibroblasts from a 22-year-old donor (Figure 1A). mCherry and eGFP were linked to nuclear localization signal (NLS) and nuclear export signal (NES), respectively. In healthy young fibroblasts, the cellular localization of these proteins distinctly separated, whereas in fibroblasts from either a 94-year-old donor or from a 14-year-old Hutchinson-Gilford progeria syndrome (HGPS) patient, the number and intensities of cytoplasmic mCherry puncta were higher than in fibroblasts from the normal 22-year-old donor (Supplementary Figure 1). Despite the difference, Z-factor analysis indicated that the system was not sufficiently robust for large-scale screening purposes, leading us to seek an alternative [37].

Figure 1. The NCC reporter system to monitor cellular senescence. (A) The NCC reporter system integrated in human fibroblasts. (B) NCC signals in quiescent fibroblasts. (C) Fluorescence intensity profiles corresponding to the path of the arrow in (B). (D) The colocalization of mCherry and eGFP signals in quiescent fibroblasts by Pearson correlation. (E) NCC signals in senescent fibroblasts. (F) Fluorescence intensity profiles corresponding to the path of the arrow in (E). (G) The colocalization of mCherry and eGFP signals in senescent fibroblasts by Pearson correlation. (H) Pearson correlation of quiescent and senescent fibroblasts. Data are mean ± SD. ^****p < 0.0001. Two-tailed Student’s t test.

Cellular senescence is accompanied by substantial reorganization of the nuclear envelope and a breakdown in nucleocytoplasmic trafficking, including altered expression and degradation of Lamin B1, and the formation of cytoplasmic chromatin fragments (CCFs) [38–45]. Thus, we reasoned that senescent fibroblasts might produce a strong signal in the NCC reporter system, one that could be used for the screening of molecules to reverse epigenetic aging. Senescence can be induced in a variety of ways, including telomere erosion, oncogene expression, and DNA damage [13, 46]. Because replicative senescence advances the DNA methylation clock but DNA damage-induced senescence does not [46, 47], we reasoned that replicatively senescent cells might be more robust and reliable to find epigenetic age reversal cocktails than other types of senescent cells.

To avoid unintended false-rejuvenation effects caused by the expansion of a small percentage of replication-capable cells in the senescent population, all experiments were performed in low serum conditions that completely suppressed cell division [48]. In non, senescent, quiescent control fibroblasts, the mCherry and eGFP signals were clearly distinguishable (Figure 1B–1D). Senescent fibroblasts were generated by passaging ~40 times, each time with a 1:3–1:5 dilution with fresh media, until there was a complete absence of growth for two weeks, morphology changes characteristic of senescent cells, a dramatic increase in transcripts from the cell-cycle regulator p21 (CDKN1A), and other senescence-associated changes in gene expression (Supplementary Figure 2B, 2C). In the senescent fibroblasts, mCherry was aggregated in the cytoplasm and colocalized with eGFP (Figure 1E–1G), consistent with a previous report [34, 35]. Colocalization of the signals, as measured by Pearson correlation, was significantly higher in replicatively senescent cells compared to quiescent cells (Figure 1H). These experiments indicated that the NCC system could discern non-senescent from replicatively senescent cells, essentially in real time.

Reversal of characteristics of cellular senescence by epigenetic reprogramming

To assess the applicability of the NCC system for detecting interventions that restore youthful functions and gene expression patterns, we first tested whether it could detect the effects of genetically mediated epigenetic age reversal. Ectopic expression of Yamanaka factors OCT4, SOX2, and KLF4 (OSK) restores youthful gene expression patterns, epigenetic age, and youthful functions to old cells and tissues [7, 8]. Our previously published reverse tetracycline-controlled transactivator (rtTA) module and the polycistronic OSK cistron under the control of a tetracycline-inducible promoter (Tet-on OSK) were transduced using lentivirus to create stable cell lines from the human fibroblasts and passaged until they reached replicative senescence. Treatment with doxycycline was sufficient to activate the OSK cassette in these fibroblasts (Supplementary Figure 2A).

Transcriptomic changes are involved in driving an aging-related decline in function and provide effective biomarkers for predicting biological and chronological age [46, 47]. To verify if these phenotypic changes reflected a more youthful epigenetic signature, we analyzed the transcriptional profile by genome-wide RNAseq. A comparison of quiescent young to quiescent old cells identified 190 genes that were significantly upregulated, and 326 genes that were significantly downregulated. Induction of OSK for four days led to reduced expression in 43.2% (82) of age-upregulated genes and increased expression in 65.3% (213) of age-downregulated genes (Figure 2A–2D and Supplementary Figure 2B). In all, nearly half of the genes changed by senescence were restored by OSK expression (Figure 2B, 2D, Supplementary Figure 2D, 2E). This finding is consistent with our previous findings and those of others, that expressing OSK in a variety of cell types and tissues, including human and mouse fibroblasts, can substantially restore the epigenetic landscape and gene expression patterns of old cells [7, 8, 26]. We call this process the EPOCH method, for epigenetic programming of old cell health.

Figure 2. OSK-mediated partial reprogramming ameliorates features of cellular senescence. (A) Heatmaps for mRNA levels of genes upregulated by senescence (n=3, p-adj < 0.01, FC > 2). (B) Percentage of genes changed by OSK (n=3, p-adj < 0.05) among those upregulated by senescence. (C) Heatmaps for mRNA levels of genes downregulated by senescence (p-adj < 0.01, FC > 2). (D) The percentage of genes changed by OSK (p-adj < 0.05) among those downregulated by senescence. (E) Top 20 gene ontology (GO) processes of genes upregulated by senescence. The red and blue bars indicate upregulation or downregulation by OSK, respectively. (F) Top 20 GO processes of genes downregulated by senescence. Red and blue bars indicate upregulation or downregulation by OSK, respectively. (G) Schematics of the Tet-On OSK system integrated in NCC reporter system fibroblasts. (H) NCC signals and track of the arrows in quiescent, senescent, or senescent + OSK fibroblasts. Scale bar, 50 μm. (I) Fluorescence intensity profiles corresponding to the arrow in (H). (J) EGFP intensities in the cytoplasm. Data are mean ± SD. ^*p < 0.05; ^***p < 0.001. One-way ANOVA-Bonferroni.

Gene ontology (GO) analysis indicated that the top 20 GO biological processes of upregulated genes encompassed key features of aging, including dysregulation of development, localization, and transport [7], eleven of which were reversed by OSK (Figure 2E). Despite the absence of cell division in all conditions, senescence caused subtle but significant changes in cell cycle gene mRNA levels, including p21 (Supplementary Figure 2C) [49]. Numerous cell cycle- related processes were enriched with downregulated genes by senescence, and 19 of the top 20 were reversed by OSK expression (Figure 2F). The net outcome of this was the demonstration that induction of OSK partially counteracts the aging related changes resulting from senescence.

Using the NCC system, we examined the deterioration of nucleocytoplasmic integrity as cells transitioned from quiescence to senescence and the rejuvenative effects of OSK treatment on those senescent cells (Figure 2G, 2H). Cross-sectional intensity profiles of the cells were used to assess the correlation between distributions of fluorescent molecules (Figure 2I). Compared to quiescent cells, senescent cells had a significant increase in the aggregation of mCherry and eGFP, indicating disruption of nucleocytoplasmic integrity (Figure 2J). After four days of OSK treatment, NCC integrity was significantly restored in senescent cells, comparable to the quiescent, non-senescent cell population (Figure 2J). Taken together, these data show that OSK-mediated epigenetic reprogramming substantially reverses senescence-associated pathology and transcriptomic changes and that the NCC reporter system can detect rejuvenation of senescent cells by OSK.

Reversal of senescence-associated NCC changes by reprogramming small molecules

To identify small molecules that rejuvenate old and senescent cells, we curated a list of molecules that have successfully reprogrammed human and mouse somatic cells into chemically induced pluripotent stem cells (CiPSCs) [30, 31] and tested them using the NCC assay. Again, we used fully senescent cells to avoid detecting changes due to the cell cycle or transition to senescence. Epigenetic age reversal is known to occur within a week of OSK (M)-mediated reprogramming, while the epigenetic age continuously decreases until pluripotency, reaching an approximate age of zero [50–52]. To ensure consistency, we initially tested small molecule combinations on cells within the same four-day period required for OSK to rejuvenate cells safely and consistently.

To achieve age reduction without altering cell identity, we focused on small molecules that were likely to work in the early stages of CiPSC formation, including valproic acid (V), CHIR-99021 (C), E-616452 (6), tranylcypromine (T) and forskolin (F). Previous studies of reprogramming efficiency with small molecules demonstrated that either OCT4 alone or SKM, when combined with VC6T or F, respectively, can generate iPSCs, and VC6TF facilitates a mesenchymal-to-epithelial transition, an early stage of reprogramming in mouse cells [31, 53]. Because of the known differences in differentiation between mice and humans, we also investigated molecules that have been reported for the initiation states of generating human CiPSCs including CHIR-99021 (C), E-616452 (6), TTNPB (N), Y-27632 (Y), Smoothened Agonist (S), and ABT-869 (A) [30]. The molecules VC6TF (Cocktail 1: C1) and C6NYSA (Cocktail 4: C4) were used as basal reprogramming cocktails and supplemented with other boosters known to increase iPSC efficiency, including sodium butyrate, basic fibroblast growth factor (bFGF), and alpha ketoglutarate (α-KG) (Figure 3A, 3B, Supplementary Tables 1 and 2) [54].

Figure 3. Reprogramming small molecule cocktails restore NCC alterations in senescent cells. (A) Chemical structures of small molecules of basal cocktails used to generate induced pluripotent stem cells (iPSCs) from mouse (left) or human (right) somatic cells. (B) Correlation heatmaps showing eGFP and mCherry colocalization in human senescent fibroblasts demonstrate the effects of 80 different combinations of small molecules (n=2). (C, D) Validation of six selected cocktails through independent experiments, showing colocalization (C) and representative images (D) of eGFP and mCherry signals. Scale bar, 50 μm. Data are mean ± SD. ^*p < 0.05; ^**p < 0.01; ^****p < 0.0001. One-way ANOVA-Bonferroni.

Based on the fact that iPSCs can also be generated using SKM or O alone [55, 56], we evaluated the effect of the boosters on VC6T (SKM alternative) and F (O alternative). We also assessed the effect of combinations including C6N, because it has been reported that the removal of Y, S, or A from Cocktail 4 (C6NYSA) did not reduce the CiPSC efficiency [30]. Among 80 cocktails tested in the NCC assay, the VC6TF basal cocktail was the most effective at restoring the integrity of nucleocytoplasmic compartmentalization, a key sign of age-reversal (Figure 3B). A recent, unpublished study reported that 6T pre-treatment prevents senescence in human fibroblasts, and 6, T, or 6T extends the lifespan of Caenorhabditis elegans by up to 42.1% [57]. We, however, saw no benefit of F alone or the VC6T cocktail on reversing senescence phenotypes in our system (Figure 3B). Next, we chose six cocktails of small molecules for further investigation, three of which were based on Cocktail 1 as well as two additives (referred to as Cocktail 2 and 3) and the other three based on Cocktail 4 plus additional additives (referred to as Cocktail 5 and 6) (Supplementary Table 2). Sodium butyrate, a histone deacetylase inhibitor, was one of the most effective additives in both human and mouse cocktails (C2 and C5). Basic fibroblast growth factor (bFGF) was used for Cocktail 3, while α-KG was included in Cocktail 6. To better gauge the effect of these compounds on NCC integrity, we used Pearson’s correlation to assess the distribution of fluorescent proteins (Figure 3C, 3D). The six cocktails statistically improved compartmentalization in senescent cells, both in terms of correlation analysis (Figure 3C) and imaging of NCC signaling (Figure 3D).

For nearly two decades, the writing and maintenance of chromatin marks have been known to be critical for reprogramming [58]. For this reason, we incorporated inhibitors of established chromatin remodeling factors in our screen to investigate whether these factors represented barriers or essential drivers of rejuvenation. The rejuvenation pathway(s) initiated by C1 and C4 were both blocked by the inhibition of H3K9 methyltransferase G9a (BIX01294, 0.5 μM) and TGF-β (SB431542, 10 μM), however they were not disrupted when the H3K27 methyltransferase component of PRC2, EZH2, was inhibited (DZNep, 20 nM) (Figure 3B).

Small molecules can reverse the age of the transcriptome with no loss of cell identity

Based on the improvement in NCC integrity, we performed RNA-seq to test the effect of these six cocktails on transcriptomic age. After treatments with the chemicals, we observed a strong overlap between genes affected by the chemical treatments and the switch from quiescence to senescence (Supplementary Figure 3A). We also observed that the two groups of cocktails generally perturbed the same populations of genes (Supplementary Figure 3A). Treatment with the chemical cocktails did not lead to fibroblasts taking on non-specific cell identity markers (Supplementary Figure 3B). Finally, we did not observe the expression of iPSC specific genes or gene modules in the RNA-seq datasets (Supplementary Figure 3C, 3D). Additionally, we performed immunofluorescence looking for signs of expression of pluripotency-related genes such as NANOG and EPCAM following all cocktail treatments but could not see any expression (Supplementary Figure 4). Collectively, these data indicate that chemical-mediated treatments are only partially reprogrammed and not fully reset to pluripotency.

We then tested the effect of these six cocktails on the transcriptomic age (tAge) of the cells using clocks trained on mouse, human, and a combined training data set 52. Relative transcriptional age was assessed using a rodent transcriptomic clock as well as a combined human and rodent transcriptomic clock (Figure 4A, 4B). The change in years of age was determined using a human-specific chronological clock (Figure 4C). Compared to quiescent cells, senescent cells had a significant increase in transcriptomic age, based on the transcriptomic clocks, consistent with previous findings assessing DNA methylation age [46, 47, 59]. Treatment of NCC cells with each of the six chemical cocktails (C1-6) resulted in statistically significant reduction of the transcriptomic age of senescent cells, with those originating from mouse studies (C1-3) generally producing a greater decrease in transcriptional age relative to the human derived cocktails (Figure 4A, 4B). The reported magnitude of the effect of all six cocktails differed between the hybrid and rodent transcriptional clocks, with the hybrid clock indicating a greater decrease in age by all six cocktails, with the rodent clock showing less variability between treatments.

Figure 4. Transcriptomic rejuvenation by reprogramming small molecule cocktails. (A, B) Delta transcriptomic ages (tAgeΔ), as measured by a biological transcriptomic clock built on rodent and human transcriptomic data (A), or rodent data alone (B). (C) Delta ages, as measured by a chronological transcriptomic clock built using human data. n.s.: p > 0.05; ^*p < 0.05; ^**p < 0.01; ^***p < 0.001. (D) Correlation matrix of transcriptomic differences by cocktail treatment, iPSC reprogramming, or aging. ^*p-adj < 0.05; ^**p-adj < 0.01; ^***p-adj < 0.001. Benjamini-Hochberg approach. (E) Enrichment of pathways by cocktail treatment, iPSC reprogramming, or aging. Normalized Enrichment Score (NES) 0.05 < ^p-adj < 0.1; ^*p-adj < 0.05; ^**p-adj < 0.01; ^***p-adj < 0.001. Benjamini-Hochberg approach. (F) Pictograph of study results showing that both induction of OSK and treatment with C1-6 restore NCC integrity, transcript profiles, and reversing biomarkers of health. Created using https://www.biorender.com.

All six reprogramming cocktails also significantly decreased the estimated chronological age of NCC senescent cells by several years (Figure 4C). As observed with clock-based transcriptional age estimates, C1, C2, and C3 produced the greatest effect, reducing the measured age by more than three years after only four days of treatment. For reference, the effect of this four day treatment is comparable to the total change seen after a year of a regenerative treatment described in a landmark study from 2019, which also focused on restoring epigenetic information [60].

To understand the effect of the chemical cocktails on cell identity and function, we assessed overall gene expression patterns of chemically-treated cells and compared them to old human cells [61] and OSK(M)-induced mouse and human induced pluripotent stem cells (iPSC) [52]. We expressed the correlation in gene expression between groups as a heatmap of the Spearman’s ranked correlation (Figure 4D). Despite having different chemical components, the transcriptomic profile of all six cocktails grouped most closely together, with human C4-6 and mouse C1-3-derived cocktails grouping more closely within their groups (Figure 4D). All six of the chemical treatments were positively correlated with the induced pluripotent stem cell (iPSC) populations and were negatively associated with mammalian age-related changes occurring in specific organs, such as kidney and brain, as well as across multiple tissues of mice, rats, and humans. In agreement with the transcriptomic clock analysis, mouse C1-3-derived cocktails produced a more consistent and stronger anti-aging effect on the cellular transcriptome than the human cocktails (C4-6).

Next, we performed gene set enrichment analysis (GSEA) to identify which pathways might be responsible for the similarities and differences between the chemical treatments, signatures of aging, and OSK(M)-induced iPSCs. The KEGG genes database, HALLMARK gene set collection, and Reactome pathways database were included in this analysis (Figure 4E). The anti-aging effects of chemical cocktails, especially mouse-derived ones, were associated with the upregulation of respiration associated pathways, such as oxidative phosphorylation and mitochondrial translation, as well as downregulation of hypoxia and multiple inflammation-associated pathways, such as interferon and JAK-STAT signaling, which are known to be involved in the SASP. The activation of JAK-STAT signaling by interferons and other SASP factors, for example, contributes to the complex interplay between senescent cells and their microenvironment. Together, these data show that the chemical cocktails identified in this study not only reverse the effects of senescence on NCC and make them transcriptionally younger, but they also reverse key transcriptional signatures of senescence (Figure 4F).

Cell culture

Human fibroblasts derived from a healthy 22-year-old, 94-year-old, and a 14-year-old with HGPS were obtained from the Coriell Institute (GM23976, AG08433, and AG11498). These cells were cultured in DMEM supplemented with 20% fetal bovine serum (FBS), 1% penicillin-streptomycin (P/S), and 0.1 mM β-mercaptoethanol (β-ME). For Tet-On cells, the medium composition was adjusted to include 15% tetracycline (Tet)-free FBS instead of the usual 20% FBS. To induce replicative senescence, fibroblasts were passaged until their growth ceased completely for at least two weeks. Senescence was confirmed through various assessments, including analysis of cell morphology, cell size, and gene expression of gold-standard senescence markers.

The human iPSC (hiPSC) line AG27602 (Coriell Institute) was used as a positive control for staining of iPSC cell markers and was cultured in mTeSR™ Plus (Stem Cell Technologies, #100-1130) according to manufacturer guidelines.

Generation of stable cells

To generate NCC-stable cells, we used FugeneHD (Promega, E2311) for transfection of pLVX-EF1alpha-2xGFP:NES-IRES-2xRFP:NLS (addgene, #71396), psPAX2 (addgene, #12260), and pMD2.G (addgene, #12259) into 293FT cells following the provided instructions. The 293FT cell medium was collected two and four days post-transfection and filtered through a 0.45-micron filter. To facilitate transduction, the collected medium was combined in a tube, concentrated using the Lenti-X™ Concentrator (Takara Bio, #631231), and added to human fibroblast medium with polybrene at a concentration of 5 μg/ml. After 24 hours, the medium was replaced with fresh medium. Following approximately one week, a subset of fibroblasts showing stable expression of mCherry and GFP were observed, and NCC-positive cells were sorted using the BD FACSAria system.

For cloning of the pLVX-UbC-rtTA-hOSK-Neo vector, the Ngn2:2A:EGFP and PuroR cassettes on pLVX-UbC-rtTA-Ngn2:2A:EGFP (addgene, #127288) were swapped with the hOCT4:2A:hSOX2:2A:hKLF4 and NeoR cassettes, respectively. NCC fibroblasts were transduced with hOSK lentivirus using the same procedure as mentioned earlier to achieve stable hOSK expression. Two days post lentiviral transduction, the cells were cultured in DMEM supplemented with 15% Tet-free FBS, 1% P/S, 0.1 mM β-ME, and 200 μg/ml G418. To induce hOSK expression, senescent fibroblasts were treated with doxycycline (2 μg/ml) for stated periods.

Small molecule treatment

The small molecules were dissolved in suitable solvents and carefully stored according to the recommended conditions (Supplementary Table 1). To prepare for the small molecule treatment, the growth medium was changed to a low serum medium with 1% FBS, a day prior to the treatment. Fresh low serum medium was used to prepare the small molecule solution, which was thoroughly mixed before replacing the old medium in the dish. The medium containing the small molecules was completely refreshed every other day until the cells were harvested. To evaluate the alterations in NCC signals resulting from the small molecule treatment, NCC images were captured using the Cytation C10 (Agilent) imaging system. NCC correlation was calculated using Cellprofiler^® colocalization analysis.

Immunofluorescence

Cells were fixed in 3.7% paraformaldehyde (PFA) for 15 minutes and washed three times with 1X PBS. Then cells were permeabilized in 0.1% Triton X-100 in PBS followed by 30 minutes of blocking with 1% bovine serum albumen (BSA) in PBS+ 0.1% Tween-20 (PBST) + 22.52 mg/mL glycine. Primary antibodies were used at the following concentrations in 1% BSA in PBST: NANOG (Invitrogen, PA5-85110) 1:200, and EPCAM (Abcam, ab71916) 1:100. Primary antibodies were incubated for 1 hour at room temperature followed by three washes of PBS. Then secondary antibodies were used at 1:1000 in 1% BSA in PBST (Goat anti-rabbit Alexa Fluor™ 647, Invitrogen A-21244 or Goat anti-rabbit Alexa Fluor™ 488, Invitrogen A-11008), incubated for one hour, and followed by three washes of PBS. Nuclear counterstaining was performed for 15 minutes using Hoechst 33342 (1:2000 in PBS) followed by a final three washes with 1X PBS. Staining was assessed by 10X wide field fluorescence imaging using the IXM-LZR and processed using Metaxpress and ImageJ.

RNA sequencing and analysis

RNA was harvested from cells using Omega ENZA Total RNA kit and assessed for quality and integrity using an Agilent Tapestation. Library preparation and 150 bp paired-end sequencing was performed on an Illumina Novaseq by Novagene. Fastq read files were processed using FastQC. Illumina adapters were removed using TrimGalore! (Version 0.4.0, Babraham Bioinformatics), and aligned to the mm10 genome using Hisat2 (Version 2.2.1) [76]. Aligned reads were assembled using StringTie (Version 1.3.3b) [77], and expression level and transcripts were estimated. Differential expression was determined using DEseq2 [78], with FDR < 0.05.

Signature association analysis

Association of gene expression log-fold changes induced by chemical C1-6 cocktails in human fibroblasts with established transcriptomic signatures of mammalian aging and OSK (M)-induced iPSCs was examined with Spearman correlation method as described previously [61]. The utilized signatures of aging included tissue-specific liver, kidney, and brain signatures as well as multi-tissue signatures of the mouse, rat, and human [61]. Signatures of OSKM reprogramming included genes differentially expressed during cellular reprogramming of mouse fibroblasts (mouse), and shared transcriptomic changes during OSK(M)-induced reprogramming of mouse and human fibroblasts (mouse and human) [47]. Pairwise Spearman correlations for gene expression changes induced by chemical cocktails and transcriptomic signatures of aging and OSK (M) reprogramming were calculated based on the union of top 300 genes with the lowest p-value for each pair of signatures.

For the identification of enriched functions affected by chemical cocktails, we performed functional GSEA [79] on a pre-ranked list of genes or proteins based on log10 (p-value) corrected by the sign of regulation, calculated as:

$$\log \text{(}pv\text{)}\times \mathrm{sgn}\text{(}lfc\text{)}$$

where pv and lfc are p-value and logFC of a certain gene, respectively, obtained from edgeR output, and sgn is the signum function (equal to 1, −1 and 0 if value is positive, negative, or equal to 0, respectively). HALLMARK, KEGG, and REACTOME ontologies from the Molecular Signature Database (MSigDB) were used as gene sets for GSEA. The GSEA algorithm was performed separately for each cocktail via the fgsea package in R with 5000 permutations. P-values were adjusted with Benjamini-Hochberg method. An adjusted p-value cutoff of 0.1 was used to select statistically significant functions. A similar analysis was performed for gene expression signatures of aging and OSK (M) reprogramming.

Transcriptomic clock analysis

To assess the transcriptomic age (tAge) of fibroblasts treated with chemical cocktails, we applied multi-tissue chronological human, lifespan-adjusted biological rodent (mouse + rat) and hybrid (mouse + rat + human) transcriptomic clocks based on the identified gene expression signatures of aging [52]. For data preprocessing, filtered RNAseq count data was passed to log transformation and scaling. The missing values corresponding to clock genes not detected in the data were imputed with the precalculated average values. Estimated sample tAges were centered around median tAge of control quiescent cells. Pairwise differences between average tAges of senescent untreated cells and either quiescent cells or senescent cells treated with C1-6 cocktails were assessed using independent t-tests. Resulting p-values were adjusted with the Bemjamini-Hochberg method.

iPSC profiler

In order to validate that the cells treated with chemical reprogramming cocktails did not lose their fibroblast cell type identity, transcriptome of all samples were compared against that of human stem cells using SEQUIN iPSC Profiler, as previously described [80].