Epigenetic erosion of H4K20me1 induced by inflammation drives aged stem cell ferroptosis

Abstract

Aging is characterized by a decline in the functionality and number of stem cells across the organism. In this study, we uncovered a mechanism by which systemic inflammation drives muscle stem cell (MuSC) aging through epigenetic erosion. We demonstrate that age-related inflammation decreases monomethylation of H4K20 in MuSCs, disrupting their quiescence and inducing ferroptosis, a form of iron-dependent cell death. Our findings show that inflammatory signals downregulate Kmt5a, the enzyme responsible for depositing H4K20me1, leading to the epigenetic silencing of anti-ferroptosis genes. This results in aberrant iron metabolism, increased reactive oxygen species levels and lipid peroxidation in aged MuSCs. Notably, long-term inhibition of systemic inflammation that is initiated at 12 months of age effectively prevents ferroptosis, preserves MuSC numbers and enhances muscle regeneration and functional recovery. These findings reveal an epigenetic switch that links chronic inflammation to MuSC aging and ferroptosis, offering potential therapeutic strategies for combating age-related muscle degeneration.

Blanc et al. uncover how chronic inflammation triggers an epigenetic switch in aged muscle stem cells, leading to iron accumulation and cell death by ferroptosis—offering insights into muscle aging and potential paths for regenerative therapies.

Main

Aging is a complex biological process characterized by a gradual decline in cellular and tissue function. In skeletal muscle, aging manifests as decreased muscle mass, strength and regenerative capacity, leading to impaired mobility and quality of life in older adults. Muscle stem cells (MuSCs), also known as satellite cells, play a crucial role in muscle repair and maintenance. However, with advancing age, the functionality of MuSCs declines markedly, contributing to the diminished regenerative potential of aged muscle^1–3. This decline in stem cell function has been attributed to both intrinsic changes within MuSCs and extrinsic factors^1,4–6.

Among the intrinsic factors, epigenetic modifications play a major role in regulating gene expression and maintaining stem cell quiescence and activation states^7–11. One such modification is H4K20 methylation, a key histone mark that regulates chromatin compaction and genome stability¹². Dysregulation of H4K20 methylation has been implicated in age-related decline of adult stem cells, including MuSCs^13,14. Precise epigenetic remodeling enables MuSCs to exit quiescence upon injury and manage inflammatory signals during regeneration¹⁵. Epigenetic erosion, particularly of histone modifications, is central to aging^16–18, but its direct relationship with chronic inflammation remains understudied.

Elevated inflammation in stem cell niches is one of the most important extrinsic contributors to stem cell aging^1–4. In skeletal muscles, numerous studies have shown that age-associated inflammation inhibits regeneration. In previous research, we demonstrated that activation of pro-inflammatory CCR2 signaling in aged skeletal muscle hindered the engraftment and regenerative capacity of young MuSCs when transplanted into an aged host, highlighting the dominant inhibitory effect of the inflammatory environment¹. Additionally, recent findings highlighted senescence as a major contributor to age-associated inflammation in the MuSC niche, where its accumulation impairs MuSC regenerative capacity in response to an acute injury^4,11. On the other hand, the contribution of systemic (rather than local) inflammation to MuSC aging is not fully understood.

Systemic chronic low-grade inflammation (‘inflammaging’) involves elevated circulatory pro-inflammatory cytokines and chemokines and contributes to stem cell aging. Heterochronic parabiosis studies have revealed that systemic factors from aged circulation can induce premature aging, whereas young circulation rejuvenates aged tissues^19,20. However, the specific impact of systemic inflammation on MuSC aging remains unclear. In the present study, we investigated how chronic systemic inflammation influences MuSC fate and aging.

Results

Systemic inflammation rewires aged MuSC epigenetic identity

Muscle regeneration by MuSCs is modulated by both cell-intrinsic and cell-extrinsic mechanisms^21–23. In the present study, we aimed to determine how age-related changes in the circulatory system are integrated by MuSCs and the cellular components of the stem cell niche (Fig. 1a). We first examined how the cellular and molecular composition of the blood changes between young and aged mice (Fig. 1a–c). We used a proteome profiler array to screen over 100 cytokines in young and aged mice, followed by secondary validation of the identified candidates using quantitative multiplexing immunoassays (Fig. 1c) to ensure an unbiased approach. Plasma from aged mice displayed elevated levels of circulatory cytokines (TNF, interleukin (IL)-1α, IL-1β IL-4, IL-6, CCL2, CCL7, CCL11 and CCL12) and myeloid cells and a reduction in lymphoid cells (Fig. 1b,c and Supplementary Table 1). Transcriptomic analysis of aged muscle highlighted a pro-inflammatory environment²⁴ (Supplementary Table 1). Computational predictions integrating blood proteomics and muscle transcriptomics identified CCR2 signaling as prominently activated, linking systemic inflammation to skeletal muscle aging (Fig. 1d,e and Supplementary Table 1).

Fig. 1. Age-associated systemic inflammation correlates with chemokine signalling activity in skeletal muscles and chromatin remodeling in MuSCs.

a–f, Multiomics workflow and computational modeling of systemic inflammatory ligand signaling in skeletal muscle. Blood and plasma (n = 4 adult mice, n = 6 aged mice) were used for cell composition profiling (b) and cytokine analysis via proteomics and validation immunoassays (c; n = 3 mice per condition). Muscle transcriptomics were used to infer active signaling pathways (d), which were integrated with cytokine data through ligand–receptor prediction scoring (e). IL-17 and chemokine signaling were highly activated, with CCR2 as a key upstream node. f, Heatmap of differentially expressed chromatin organization genes (GO:0006325) in aged MuSCs. g,h, Representative pictures (g) and quantification (h) of H4K20me1 status in MuSCs on EDL single fibers isolated from adult and aged mice (n = 6 mice per age, with at least 30 Pax7⁺ cells counted per mouse). i,j, Immunoblot (i) and qPCR (j) assessing Kmt5a mRNA and protein levels from freshly sorted MuSCs (n = 3 mice per condition). Low Kmt5a correlated with decreased H4K20me1. For western blots, two aged mice were pulled together for one lane. Quantification is normalized to total H3 and relative to young values. k–m, Flow cytometry analysis for MACS pre-sorted cells stained for VCAM and Zombie dye prior to fixation and then fixed and permeabilized for H4K20me1 staining. Graph displays percentage of cells positive for VCAM and H4K20me1 for n = 3 mice per age. Two-dimensional plot of H4K20me1 in MuSCs for each replicate (l) and mean fluorescence intensity analysis (m) (n = 3 mice per age). For all violin plots, data represent biological replicates. Statistical significance was determined using two-sided Welch’s t-test, and exact adjusted P values are reported in the figure. FC, fold change; FDR, false discovery rate; WBC, white blood cell; MFI, mean fluorescence intensity; TLR, Toll-like receptor; SK.M., skeletal muscle; AGE-RAGE, advanced glycation end products–receptor for advanced glycation end products.

Source data

CCR2 signaling is a hallmark of muscle aging, with elevated circulating CCR2 ligands correlating with frailty and inflammation^4,25–27. To test how systemic inflammation influences MuSCs, we injected young mice with CCR2 ligands (CCL2, CCL7 and CCL8)¹. Although MuSC numbers remained unchanged, treated mice displayed transient systemic inflammation and increased MuSC activation and cell cycling (Extended Data Fig. 1a–j). CCR2-null mice served as negative controls¹.

Extended Data Fig. 1. Acute systemic inflammation prime muscle stem cells and repress Kmt5a. Related to Fig. 1.

a, b) Muscle stem cell counts from adult tibialis anterior 24 hours post IP injection. CCR2-null mice were used as a negative control. c, d) Relative number of muscle stem cells (Pax7 + ) actively cycling (Edu + ) (C) or out of quiescence (Ki67 + ) (D). e-i) Plasma levels of cytokines assessed in the same cohort of mice before and after treatment. j) Percentage of circulatory monocytes in whole blood prior and after treatment measured by whole-blood count. k-m) Cycling status of MuSCs six weeks after treatment. n-q) Features of accelerated muscle regeneration at 4-day post-injury six weeks after treatment. Treated mice had larger fibers (N and O) and a higher number of MuSC progenitors (MyoD + ; P). Regenerating muscles from treated mice also displayed genetic signature of accelerated regeneration (Q). r, s) Epigenetic genes (R, writers; S, erasers) expression in MuSCs one day after treatment. Kmt5a is the only gene being significantly downregulated. t, u) Heatmap for Myogenic (T) and Cell cycle (U) genes expression Log2 fold change six weeks post-treatment. Data are reported as mean ± s.d‥ v, w) MuSC methylation status for Kmt5a-mediated mark H4K20me1 (V), and MyoD (W), six weeks after treatment. x) Flow cytometry mean fluorescence intensity for MuSCs six- and height-weeks post-treatment. y) Flow cytometry analysis showing detailed frequency of MuSCs with H4K20me1 low or high in relation to Ki67 status six weeks post-treatment. Welch’s t-test (L, M, O-S), one-way ANOVA E-J), or two-way ANOVA (B-D, X, Y). exact p values are reported in the figure.

Source data

Next, we tested whether the transient inflammatory response had long-lasting effects on MuSC activation capacity. We injected mice with CCR2 ligands and waited 6 weeks until the molecular and cellular circulatory signature returned to baseline. Once at baseline, we tested the physiological MuSC activation capacity by injuring the tibialis anterior (TA) of the mice and assessing the contralateral uninjured muscle. MuSCs that are located distal to an acute muscle injury can re-enter the cell cycle faster in the event of a secondary insult^28,29. When compared to mice treated with a vehicle, we found that MuSCs from animals exposed to acute inflammation 6 weeks before injury retained enhanced activation capacity (Extended Data Fig. 1k–m). We found a correlation between enhanced activation and signs of accelerated regeneration, evidenced by larger muscle fibers and increased MyoD⁺ progenitor (Extended Data Fig. 1n–p) numbers at 4 days post-injury (dpi), which was further supported by transcriptional profiling (Extended Data Fig. 1q). MuSCs in mice that have experienced a prior injury have long-term enhanced regenerative capacity³⁰. These long-term functional changes led us to think that MuSCs can adapt to acute inflammatory signals that are indicative of epigenetic reprogramming^31,32. Next, we identified the differentially expressed genes predicted to be regulated by CCR2 signaling in aged MuSCs (Fig. 1e) and performed term enrichment analysis. ‘Chromatin organization’ was the most enriched biological process term in aged MuSCs for differentially expressed genes predicted to be regulated by CCR2 signaling activation (Fig. 1f). Among the most significant epigenetic genes silenced in ‘Chromatin organization’ was the lysine methyltransferase Kmt5a. Kmt5a-mediated catalysis of H4K20 monomethylation is required for subsequent dimethylation and trimethylation required to establish constitutive heterochromatin in virtually every cell. The dynamic regulation of both Kmt5a and H4K20me1 is also critical for proper cell cycle progression, which is intriguing given that MuSCs predominantly exist in a quiescence state and that aged MuSCs often enter irreversible fates upon activation^5,33. In aged MuSCs, Kmt5a gene and protein levels decreased (Fig. 1i,j). Kmt5a silencing was functionally reflected by the decreased methylation of its main substrate, H4K20me1 (Fig. 1g–i). Further analysis using flow cytometry revealed a shift in H4K20me1 intensity in the aged MuSC population, whereby a significant portion of the aged population displayed decreased total levels of H4K20me1 compared to adult MuSCs (Fig. 1k–m).

Next, we assessed whether CCR2 activity was linked to Kmt5a repression. In response to the systemic delivery of CCR2 ligands, Kmt5a was the only epigenetic gene significantly repressed in MuSCs 1 day after treatment (Extended Data Fig. 1r,s). CCR2-mediated Kmt5a silencing was accompanied by long-term transcriptional changes in both myogenic and cell cycle genes at 6 weeks after treatment (Extended Data Fig. 1t,u). Notably, we observed the upregulation of activation genes (Myf5 and Dek) and the downregulation of quiescence genes (Hes1 and Hey1). In addition, MuSCs displayed long-term erosion of H4K20me1 and sustained an increase in MyoD⁺ cells, lasting for at least 6 weeks after injection and returning to baseline 8 weeks after injection (Extended Data Fig. 1v–y). Flow cytometry analysis identified that CCR2 ligand treatment promoted a shift in H4K20me1 intensity, with the population of MuSCs derived from treated animals displaying lower H4K20me1 levels (Extended Data Fig. 1x,y). These results suggest that CCR2-mediated inflammation may trigger long-term epigenetic remodeling in MuSCs. We hypothesized that these epigenetic changes are mediated by Kmt5a and H4K20 methylation. Thus, we investigated the role of Kmt5a in MuSCs to better understand its effect on muscle aging.

Kmt5a maintains quiescence and post-activation survival

To assess the role of Kmt5a in quiescent MuSCs, we generated inducible Kmt5a MuSC-specific knockout mice (Pax7^CreERT2/+;Kmt5a^fl/fl (Kmt5a^KO)) (Extended Data Fig. 2a). Immediately after the deletion of Kmt5a, we did not observe a change in MuSC numbers in vivo, despite the vast majority of Kmt5a^KO MuSCs lacking Kmt5a (>95% efficiency) (Extended Data Fig. 2b–d). Next, we cultured MuSCs from both wild-type (Pax7^+/+;Kmt5a^fl/fl) and Kmt5a^KO mice to assess myogenic potential, cell proliferation and survival. Freshly isolated MuSCs (Pax7⁺) were stained with myogenic markers to assess myogenesis progression toward commitment (MyoD) and terminal differentiation (MyoG) over time¹. We used Ki67 and 5-ethynyl-2′-deoxyuridine (EdU) to assess cell cycle re-entry (that is, activation) and active cell proliferation, respectively. When we challenged the cells outside of their niche, cultured Kmt5a-null cells displayed enhanced activation (Pax7⁺Ki67⁺), directly correlating with the loss of H4K20me1 as early as 24 hours after plating (Extended Data Fig. 2e–i). Examining actively dividing (EdU⁺) MuSCs at 72 hours after isolation, we observed a near-complete loss of H4K20me1, which was in line with the cell cycle regulation of H4K20me1. This decrease was accompanied by impaired myogenic terminal differentiation and cell expansion (Extended Data Fig. 2e–k). Further analyses revealed that Kmt5a^KO MuSCs displayed aberrant morphological phenotypes (for example, blebbing, pycnotic and multiple nuclei) as well as a molecular signature suggestive of genomic instability (Extended Data Fig. 2l,m). Our molecular analysis of 72-hour cultured MuSCs supported this premise, as mutant cells displayed elevated γH2AX and serine 15 phosphorylation of P53 (Extended Data Fig. 2n).

Extended Data Fig. 2. Loss of Kmt5a in muscle stem cells leads to genomic instability features. Related to Fig. 2.

a-d) Strategy for Kmt5a genetic deletion in muscle stem cells to assess immediate contribution to homeostasis and regeneration. Muscle stem cell counts from adult tibialis anterior muscles immediately after final tamoxifen injection (B, C). White arrows indicate Pax7+Kmt5a+ double-positive cells, while the orange arrows point to Pax7+Kmt5a- cells. e-k) Cell culture kinetics of wildtype and knockout muscle stem cells with quantification of H4K20me1 (F), cell survival (G), cell cycle features (H, I) and myogenic status (J, K). Cell density starts to decrease in the knockout as muscle stem cells loose H4K20me1, which correlates with terminal differentiation deficit. l, m) Relative number of muscle stem cells with aberrant morphologies and nuclear phenotypes (red arrows). n, Immunoblotting profiling of DNA damage response signature. Red dots represent females and blue dots represent males; we did not observe obvert sexual dimorphism. Welch’s t-test or one-way ANOVA. p values are reported in the figure.

Source data

When challenged by acute sterile injury, the contribution of Kmt5a^KO MuSCs to regeneration was disrupted, resulting in the apparent absence of muscle regeneration, as we did not observe centrally nucleated fibers (CNFs) among the mutants at 21 dpi (Fig. 2a). We also noted a macroscopic decrease in injured muscle size at this timepoint (Supplementary Fig. 1a,b). To better index regeneration, we used embryonic myosin heavy chain (eMHC), a marker of immature (regenerating) skeletal muscle fibers, throughout the regeneration process (Extended Data Fig. 3a). As expected, wild-type regenerating fibers were nearly all positive for eMHC at early regeneration timepoints (4 dpi and 7 dpi), followed by the generation of large CNFs, a hallmark of regeneration. In contrast, we did not observe any regenerating fibers in Kmt5a^KO mice past 7 dpi (Extended Data Fig. 3b). By 60 dpi, most of the injured area appeared to be reduced to remnants of extracellular matrix and mononucleated cells. Additionally, virtually no surviving MuSCs were observed after regeneration (Fig. 2b and Extended Data Fig. 3c). When we investigated the fate of these MuSCs and their derived progenitors, we found a sharp decline in the number of myoblasts (MyoD⁺) in the mutant at 4 dpi (Extended Data Fig. 3d–g). At this time, we saw a brief increase in the number of terminally committed progenitors (MyoG⁺) before they virtually disappeared at 7 dpi (Extended Data Fig. 3e,f). Because MuSCs and derived progenitors were lost between 4 dpi and 7 dpi, we assessed the proliferative and myogenic capacity at 4 dpi in both injured and uninjured contralateral limbs, where MuSCs are expected to enter a primed state (Extended Data Fig. 3a)²⁹, allowing us to assess the fate of Kmt5a^KO MuSCs in response to direct and indirect environmental pressures. We found that MuSC numbers in Kmt5a^KO mice rapidly declined in both injured and contralateral muscles (Extended Data Fig. 3c,i–n), but MuSCs derived from the contralateral limb displayed primed features, including swifter entry into the cell cycle (Extended Data Fig. 3l–n), accelerated myogenic progression (Extended Data Fig. 3e–g) and activation of mTORC signaling as shown by high pS6 levels (Extended Data Fig. 3o–q)^28,29,34.

Fig. 2. Kmt5a is required for quiescent MuSC homeostasis.

a,b, Schematic representation of the regeneration experiments and MuSC self-renewal. Mice were subjected to injury either immediately after tamoxifen injections or 3 weeks later. Injured muscles were harvested at 21 dpi, and MuSCs were counted by staining for Pax7. The relative number of MuSCs per area of regeneration is shown in b (n = 6 WT and n = 3 KO mice per timepoint; blue, male, and red, female). c,d, Schematic representation of long-term Kmt5a deletion and quantification of MuSC pool maintenance at homeostasis. Muscles were harvested at different times after knockout induction, and the number of Pax7⁺ cells was counted. The stochastic decline of MuSCs in the mutant mice is shown in d (n = 23 WT mice and n = 4–6 KO mice per timepoint; blue, male, and red, female). e, Percentage of activated MuSCs quantified by pS6 immunostaining 3 weeks and 6 weeks after knockout induction (n = 15 WT mice and n = 8–15 KO mice per timepoint). f–h, Relative number of MuSCs actively cycling (g; EdU⁺) or out of quiescence (h; Ki67⁺) in uninjured muscles (n = 6 WT mice and n = 6 KO mice). i, Quantification of precocious terminally committed MuSC progenitors (Pax7⁺MyoG⁺) 6 weeks after deletion of Kmt5a in MuSCs (n = 4 WT mice and n = 4 KO mice). j,k, Molecular profiling of DNA damage and DNA repair signatures in MuSCs 6 weeks after knockout induction. TUNEL was used to detect DNA strand breaks. The results are reported as the mean ± s.d. of Pax7⁺TUNEL⁺/Pax7⁺ (percentage) (j), and immunoblotting was used to measure the p53 signaling response (p-P53) to DNA damage (γH2AX) (n = 5 WT mice and n = 8 KO mice). For all violin plots, data represent individual biological replicates. Statistical significance was determined using two-sided Welch’s t-test (g–j) or one-way ANOVA (b,d,e), and exact P values and adjusted P values (q) are reported in the figure. qMuSC, quiescent MuSC; regen, regeneration; Tmx, tamoxifen; wks, weeks; WT, wild-type.

Source data

Extended Data Fig. 3. Kmt5a-null muscle stem cells prematurely enter an alert-like state. Related to Fig. 2.

a) Strategy to assess muscle stem cell myogenic fate and survival in response to a proximal injury (injured muscles), or to assess Galert and quiescence features in response to a distal injury (contralateral), 4-day post injury. Injuries were performed immediately upon last tamoxifen injection. b) Progression of muscle regeneration was indexed using embryonic myosin heavy chain (eMHC) at 4-, 7-, 30-, and 60dpi. Quantification of eMHC+ fiber per area of regeneration are reported on the representative pictures as mean ± s.e.m. c-e) Myogenic cell fate analysis throughout regeneration. Pax7 (C) was used for MuSCs post-regeneration, MyoD (D) quantified the number of myoblast, and MyoG (E) showed the number of terminally committed progenitors. f-k), Myogenic and cell cycle status of activated muscle stem cells in response to an acute injury at 4dpi. Myogenic progression was assessed by Pax7, MyoD and MyoG, while their cell cycle status was measured by EdU incorporation over 24 h prior to harvesting. l-q, Quiescence and activation status of muscle stem cells in uninjured muscles in response to a distal injury. Number of muscle stem cells was assessed with Pax7, while their activation status was assessed for active cycling (L, N; EdU), Galert (O, P; white arrow shows Pax7 + pS6+ (activated) cell, green arrow shows Pax7 + pS6- (quiescent) cell, orange arrow shows Pax7- pS6+ cell), and quiescence exit (Q; Ki67). Welch’s t-test or one-way ANOVA. p values are reported in the figure.

Source data

H4K20me1 levels in MuSCs decreased sharply after their first cell division, consistent with its cell-cycle-dependent regulation^29,35 (Extended Data Fig. 2e). We hypothesized that loss of Kmt5a impairs H4K20me1 deposition, compromising MuSC quiescence. Indeed, deletion of Kmt5a led to a progressive stochastic decline in MuSC numbers over several weeks (Fig. 2c,d), suggesting that, after Kmt5a deletion, the loss of H4K20me1 does not occur immediately and simultaneously in every cell. Notably, beginning at 3 weeks, Kmt5a^KO MuSCs displayed features of quiescence exit, which was depicted by elevated levels of pS6 (Fig. 2e), an increased number of Ki67⁺ cells (Fig. 2f,h) and a reduced number of actively cycling cells (EdU⁺) (Fig. 2f,g).

Next, we investigated the fate of the Kmt5a^KO MuSCs in vivo. We observed a decline in survival and genomic instability in cultured Kmt5a^KO MuSCs, which correlated with the loss of H4K20me1 (Extended Data Fig. 2e–n). We then tested whether Kmt5a^KO MuSCs exited quiescence and entered cell death. We found that a significant proportion of mutant cells showed signs of cell death (~28% TUNEL⁺) (Fig. 2j), along with an elevated DNA damage response, as indicated by increased γH2AX and p53 signaling (Fig. 2k). Although these cells are relatively rare (<3 cells per 300 fibers), the elevated frequency of MuSCs with DNA damages suggests that their detection and clearance by immune cells could be compromised.

Because p53 function can be regulated by Kmt5a, we assessed whether MuSC loss in Kmt5a^KO mice involved p53. Single (single mutant; P53^KO) and double (Kmt5a;P53^dKO) knockout mice were generated. Deletion of p53 alone minimally affected MuSCs, whereas double mutants mirrored Kmt5a^KO phenotypes (Supplementary Fig. 2). Thus, MuSC loss in Kmt5a-null mice is p53 independent.

Kmt5a safeguards MuSCs from ferroptosis

To uncover mechanisms underlying MuSC loss in Kmt5a^KO mice, we performed single-cell RNA sequencing (scRNA-seq) over time, revealing dynamic transcriptional changes in MuSC subsets, particularly in quiescence and differentiation genes^34,36,37 (Supplementary Fig. 3). Using precision run-on sequencing (PRO-seq) to identify genes regulated by H4K20me1-mediated transcriptional pausing^38–42, we found significant enrichment for ferroptosis pathways in Kmt5a^KO MuSCs (Supplementary Fig. 4a–e). Ferroptosis, a regulated iron-dependent cell death process, involves reactive oxygen species, lipid peroxidation and inflammatory signaling^43–46. Transcriptomic analyses of Kmt5a-null MuSCs revealed a pro-ferroptotic signature, marked by repression of anti-ferroptotic genes (for example, Gpx4 and Ftl1) and induction of ferroptosis markers (for example, Ptgs2, Hmox1 and Tfrc) (Supplementary Fig. 4a,f). Notably, high promoter-proximal pausing indexes were observed for Gpx4 and Ftl1, suggesting direct transcriptional regulation by Kmt5a.

Based on these results, we hypothesized that Kmt5a-deficient MuSCs would exhibit impaired iron metabolism, aberrant accumulation of intracellular iron and increased levels of lipid peroxidation, resulting in cell death by ferroptosis (Supplementary Fig. 4g). To test this hypothesis, we characterized MuSC sensitivity to ferroptosis using gold standard compounds to induce ferroptosis (erastin and RSL3) along with ferrostatin-1 (Fer1), which traps lipid radicals to rescue the effects of erastin and RSL3 (Extended Data Fig. 4)⁴³. Both erastin and RSL3 induced a canonical ferroptotic response with decreased cell viability, increased reactive oxygen species production and enhanced lipid peroxidation, which were rescued by Fer1 treatment (Extended Data Fig. 4). We found MuSCs to be more sensitive to RSL3 exposure, perhaps because they directly target GPX4, which plays a critical role in preventing lipid peroxidation and, thus, provides a major safeguard against ferroptosis. Gpx4 was among the most repressed genes identified in our scRNA-seq analysis (Supplementary Fig. 3b), particularly in the Kmt5a-deficient MuSCs that exhibited a cell cycle re-entry molecular signature. Additionally, Kmt5a^KO MuSCs were virtually devoid of GPX4 and exhibited iron-rich pockets in the vicinity of the MuSC niche (Fig. 3a–c). Less than 2% of MuSCs from wild-type mice exhibited detectable levels of intracellular iron (Fe²⁺), whereas over 80% of Kmt5a^KO MuSCs exhibited an accumulation of iron foci (Fig. 3c). Electron microscopy further suggested that Kmt5a^KO MuSCs displayed hallmarks of ferroptosis, including cell swelling, plasma membrane blebbing and rupture, increased mitochondrial content and intracellular iron accumulation (Fig. 3d). We then used inductively coupled plasma mass spectrometry to measure the total iron content of freshly sorted MuSCs. We found that Kmt5a^KO MuSCs had approximately 3.4 × 10⁻² ng of iron per cell, nearly 70-fold higher than the wild-type MuSCs (Fig. 3e), and that Kmt5a^KO MuSCs exhibited higher levels of lipid peroxidation (Fig. 3f) that progressively increased over time (Fig. 3g). Likewise, genetic Kmt5a deletion resulted in the progressive rise of a pro-ferroptotic transcriptomic signature in Kmt5a^KO MuSCs, including the repression of Gpx4 and Rgs4 and higher expression of Ptgs2 (refs. ^47,48) (Fig. 3h).

Extended Data Fig. 4. Characterization of muscle stem cell sensitivity to ferroptosis and Kmt5a catalytic inhibition. Related to Fig. 3.

Muscle stem cells were freshly isolated and plated for 4 h prior being treated with the ferroptosis-inducing compounds along with Ferrostatin (Fer1), a potent and selective inhibitor of ferroptosis. a-d, Erastin (A, B) and RSL3 (C, D) dose and time effects on Muscle stem cells. Cell viability (A, C) was quantified by DAPI integration and increased ROS was quantified by MitoSOX CTCF (B, D). e-g, Optimal treatment for Erastin (10uM) and RSL3 (1uM) for 24 h caused decreased viability (E; data reported as mean of individual experiment), increased ROS (F; data reported as individual cell intensity [CTCF] from 3 different experiments) and lipid peroxidation (G; data reported as mean of individual experiment). Fer1 rescued the effects of Erastin and RSL3, as it traps lipid radicals and lipid ROS that are involved in the induction of ferroptosis in Muscle stem cells. h, i) Immunoblot and quantification for GPX4 relative to Actin following treatments. j) Optimal dose for Kmt5a catalytic inhibitor UNC0379 was determined over 24 hours period, with 4uM effectively reducing H4K20me1 beyond detectable levels. k) UNC0379 treatment (4uM) resulted in cell death initiating at 24 h with no detectable surviving cell past 72 h. Data are presented as mean ± standard deviation, n = 3 independent experiments. l) Transcript levels of ferroptosis-associated genes Gpx4, Rgs4, Ptgs2, Hmox1 in response to RSL3. Gene expression was normalized to the average levels of B2M, TBP, and PPIA, and are reported as normalized fold-change ± s.d.

Source data

Fig. 3. Kmt5a deletion in MuSCs causes ferroptosis.

a,b, GPX4 immunostaining in TA muscles of WT (a) and Kmt5a^KO (b) mice (n = 5 mice per group, one experiment). Arrows, MuSCs; yellow arrow, MuSCs in iron-rich pocket (~50%). c, Quantification of MuSCs with high levels of labile iron (Fe²⁺) (mean ± s.e.m., n = 5 mice per condition). d, Representative electron micrographs of a WT quiescent MuSC and a Kmt5a^KO MuSC showing features of activation and ferroptosis. e, ICP-MS quantification of elemental iron in MuSCs. Total iron was normalized to cell numbers. Data points are reported as average of replicate (n = 5 mice per condition). f,g, Quantification of lipid peroxidation in MuSCs. Flow cytometry plot shows a shift in 510-nm signals in mutant MuSCs. Inverted ratiometric signals of 590 nm/510 nm were calculated to report lipid peroxidation in each cell (g); data points are reported as average of replicate (n = 4 mice per condition). h, qPCR for Gpx4, Rgs4 and Ptgs2. i,j, Immunoblot and quantification for KMT5a, H4K20me1 and GPX4 with DMSO or Kmt5a catalytic inhibitor (Kmt5a_i). Kmt5a was normalized to GAPDH; H4K20me1 was normalized to histone H3; and GPX4 was normalized to actin (n = 3 mice per condition). Data are reported as normalized intensity ± s.d. k, qPCR quantification of ferroptosis markers Gpx4, Rgs4, Ptgs2 and Hmox1 in response to Kmt5a_i. For qPCR data, gene expression was normalized to the average levels of B2M, TBP and PPIA and is reported as normalized fold change ± s.d. (n = 3 mice per condition). l,m, Quantification of lipid peroxidation in response to drug treatments. Histogram shows the intensity of ratiometric signal (590 nm/510 nm) for the lipid peroxidation probe in live cells (l). Violin plots represent the inverted ratiometric signal normalized to vehicle (m). For all violin plots, data represent biological replicates. Statistical analyses were performed using two-sided Welch’s t-test (e,j,k) and one-way ANOVA (g,h,l,m), and exact P values and adjusted P values (q) are reported in the figure. wks, weeks; WT, wild-type; im, immediate.

Source data

Next, we wondered if Kmt5a catalytic inhibition could promote ferroptosis. Kmt5a inhibition is accompanied by signs of ferroptosis at both the cellular and molecular levels. Dosage under the inhibitor IC₅₀ ((C) = 4 µM < IC₅₀ = 7.3 µM), although still significantly decreasing H4K20me1, was seemingly lethal in under 24 hours (Fig. 3i,j and Extended Data Fig. 4j,k). Loss of H4K20me1 after treatment was accompanied by decreased GPX4 protein and mRNA levels as well as repression of Rgs4 and increased Ptgs2 and Hmox1 levels, mirroring the effects observed in RSL3-treated cells (Fig. 3i–k and Extended Data Fig. 4h–l). Kmt5a inhibition also significantly increased lipid peroxidation, which was reduced when co-treated with Fer1 (Fig. 3l,m), further supporting the hypothesis that Kmt5a inhibition promotes cell death by ferroptosis.

Finally, we investigated the mechanism by which Kmt5a regulated ferroptosis. Our transcriptomic data revealed that a limited number of gene expression changes were correlated with promoter-proximal pausing, including Gpx4 and Ftl1 (Supplementary Fig. 4f). Subsequent investigation of key ferroptosis regulators (Fig. 3h–j) suggested that Kmt5a was required for Gpx4 expression in a catalytic-dependent manner (Fig. 3i,j). Therefore, we hypothesized that Kmt5a directly regulated Gpx4 expression through H4K20me1 deposition. To test this hypothesis, we transiently silenced Kmt5a using small interfering RNA (siRNA) and found that targeting Kmt5a resulted in decreased levels of Gpx4 protein and mRNA, which were successfully rescued by re-expressing Kmt5a (Extended Data Fig. 5a,b). Next, we assessed the presence of both Kmt5a and H4K20me1 around Gpx4 gene transcription start site (TSS) under the same conditions. In control myoblasts, Kmt5a was enriched at −250 base pairs (bp) from the Gpx4 TSS and across the gene to a lesser extent (Extended Data Fig. 5c), whereas H4K20me1 was mostly found at the TSS and +250 bp from the TSS (Extended Data Fig. 5d), similar to the promoter-proximal pausing (Supplementary Fig. 4f). Silencing Kmt5a resulted in the loss of both Kmt5a and H4K20me1 across the Gpx4 gene, which was rescued upon Kmt5a re-expression (Extended Data Fig. 5c,d). Coincidentally, increased lipid peroxidation was observed only in cells where Kmt5a was silenced (Extended Data Fig. 5e). We conclude that Kmt5a regulated Gpx4 transcription and that Kmt5a is necessary to prevent ferroptosis. Because targeting GPX4 is sufficient to increase lipid peroxidation and trigger ferroptosis⁴⁹, we investigated whether re-expressing Gpx4 could counteract the Kmt5a-mediated induction of ferroptosis. We overexpressed Gpx4 in myoblasts treated with Kmt5a inhibitor and first confirmed that catalytic inhibition of Kmt5a resulted in decreased levels of Gpx4 protein and transcripts (Extended Data Fig. 5f,g), cell death (Extended Data Fig. 5h) and increased lipid peroxidation (Extended Data Fig. 5i). These results were prevented upon re-expressing Gpx4 (Extended Data Fig. 5f–i).

Extended Data Fig. 5. Kmt5a regulates GPX4 expression through H4K20me1 deposition and prevents ferroptosis in myoblasts. Related to Fig. 3.

(a, b) Quantitative PCR (A) and immunoblot (B) analysis showing the levels of Gpx4 mRNA and protein in myoblasts treated with Kmt5a siRNA, with or without Kmt5a re-expression. B shows the whole WB membrane Data are presented as mean ± s.d. (c, d) Analysis of Kmt5a (C) and H4K20me1 (D) occupancy around the Gpx4 transcription start site (TSS) in control and Kmt5a-silenced myoblasts, with or without Kmt5a re-expression. Data are presented as % input ± s.d. from three independent experiments. (e) Lipid peroxidation analysis in control and Kmt5a-silenced myoblasts. Data are presented as mean ± s.e.m. from three independent experiments. (f-i) Quantitative PCR (F), immunoblot (G), cell viability (H), and lipid peroxidation (I) analysis in myoblasts treated with a Kmt5a inhibitor, with or without Gpx4 overexpression. Statistical analyses were performed using one-way ANOVA (A, E, F, H and I), and exact p-values are reported in the figure.

Source data

These results suggest that Kmt5a activity prevents premature ferroptotic MuSC death by epigenetically regulating the key genes involved in iron metabolism and antioxidative activity.

H4K20me1 erosion links aging to MuSC ferroptosis

Although not previously molecularly linked, increased pro-inflammatory features, loss of MuSC quiescence and premature death are the hallmarks of skeletal muscle aging. Although ferroptosis is an incompletely understood form of programmed cell death, especially in stem cell homeostasis, recent evidence indicates that ferroptosis is implicated in the pathogenesis of various age-dependent disorders⁴⁵. In the present study, we aimed to confirm whether the key findings observed in Kmt5a^KO MuSCs are physiologically relevant to aging. Specifically, we assessed the contribution of Kmt5a and H4K20me1 to age-dependent decline in MuSC number and function. Because aged MuSCs displayed lower levels of both H4K20me1 and Kmt5a compared to the MuSCs from young mice (Fig. 1g–m), we used flow cytometry to assess the activation status of aged MuSCs upon the loss of H4K20me1 at the single-cell level. Our flow cytometry analyses revealed that, although virtually all quiescent MuSCs from young mice exhibited high levels of H4K20me1 (Ki67⁻H4K20me1^high), an age-dependent reduction in this population of MuSCs suggested H4K20me1 erosion (Fig. 5a), which was consistent with the information theory of aging¹⁶. Furthermore, more than 95% of MuSCs positive for the cell cycle entry marker Ki67 exhibited reduced levels of H4K20me1, suggesting that decreased H4K20 methylation is necessary for quiescence exit and serves as a hallmark of early MuSC activation (Pax7⁺Ki67⁺H4K20me1^low) (Supplementary Fig. 5a).

Fig. 5. Identification and characterization of a ferroptotic subpopulation in aged MuSCs.

a–c, scRNA workflow: adult (blue) and aged (orange) MuSCs were identified and plotted with UMAP (a). Unbiased clustering separated adult and aged cells into four clusters (b) prior to running pseudotime analysis to assess cell trajectory (c). d, Enrichment analysis of 500 most altered genes as a function of velocity and pseudotime trajectory highlighting ‘ferroptosis’ and ‘glutathione metabolism’ (red underlined). Statistical significance: Fisherʼs exact test followed by Benjamini–Hochberg correction. e, Dot plot for Louvain metadata clusters of MuSCs showing average expression and percent expression per cell for key genes. f, Donut plot representing a selection of MuSC fate based on flow cytometry analysis (n = 5 mice per condition). Live cells were negative for all markers; senescent cells were SPiDER⁺; apoptotic cells were Annexin V⁺; ferroptotic cells were Lipid Peroxidation^High; other cells were DAPI⁺ but negative for other markers. g, Lipid Peroxidation^high MuSCs (Aged^Fe) were sorted for mRNA analysis to confirm a ferroptotic signature similar to cluster 3. h, Representative picture of an aged MuSC with intracellular labile iron and niche GPX4 (n = 2 separate experiments). i,j, Abundance of total elemental iron per cell (i) and lipid peroxidation (j) in adult and aged MuSCs. To avoid additional introduction of cell stress bias and to validate lipid peroxidation changes in aged cells, we used a glutathione depletion mouse model (Nrf2^KO;GCLC^KO) to naturally increase lipid peroxidation, therefore bypassing the need for a compound (that is, RSL3) or iron overload. k, Lipid peroxidation in MuSCs. Cells were isolated from adult and aged mice (n = 5 male mice per condition), plated for 4 hours and harvested 20 hours after treatment. l–o, MuSC survival (m), myogenic potential (o) and fusion competence (n) in response to radical trapping drug Fer1. For all violin plots, data represent biological replicates. Statistical analyses were performed using two-sided Welch’s t-test (g,i,j), one-way ANOVA (m–o) and two-way ANOVA (k), and exact P values and adjusted P values are reported in the figure. Avg., average; exp., expression; NS, not significant; Veh, vehicle.

Source data

We mapped H4K20me1 in MuSCs from young and aged mice by using CUT&Tag (Supplementary Fig. 5b) to further understand how Kmt5a epigenetically controls MuSC fate. We found that H4K20me1 is a broad epigenetic marker mostly localized near the TSS of genes and spreads over the body of expressed genes (Fig. 4a,b and Supplementary Fig. 5c). We then integrated H4K20me1 maps with our RNA-seq data to gain further insights into H4K20me1 regulation and found that H4K20me1 was associated with transcriptional status regardless of age (Supplementary Fig. 5d). Gene Ontology (GO) term analysis revealed that genes with a loss of H4K20me1 at their TSS in aged mice were associated with the cell cycle and chromatin organization, whereas genes that gained H4K20me1 signals were notably enriched for the activation of cell death (Supplementary Fig. 5e). Furthermore, gene set enrichment analysis identified key processes that were also identified in Kmt5a^KO MuSCs, including Notch signaling, p53 signaling and myogenesis (false discovery rate < 0.1) (Supplementary Fig. 5f–h). The role of H4K20me1 in the transcriptional regulation of ferroptotic genes was found at a few key genes, with Gpx4 exhibiting the most significant reduction in transcript levels (Fig. 4c). Our data highlight that aged MuSCs display signs of a pro-ferroptotic transcriptional signature with repressed glutathione metabolism genes (for example, Gclm and Gclc), enhanced iron import, handling of elevated iron-responsive element genes (for example, Tfrc, Fth1 and Ftl1) and canonical markers of ferroptosis (for example, high Ncoa4, Hmox1 and Ptgs2).

Fig. 4. Loss of H4K20me1 and Kmt5a in aged MuSCs is associated with pro-ferroptosis transcriptional signature.

a,b, H4K20me1 CUT&Tag in adult and aged MuSCs. a, Heatmap of RPKM-normalized sequencing reads centered on TSSs near H4K20me1 sites unique to young and aged MuSCs or shared between them. The numbers of TSSs in each group are labeled. b, Metaplot comparison of RPKM averaged in 50-bp bins around all TSSs. c, Transcriptional and H4K20me1 changes for ferroptosis genes. Heatmaps are organized by descending gene expression in aged MuSCs and are reported as fold change of z-score for RNA-seq and H4K20me1 CUT&Tag. RNA-seq and CUT&Tag were performed using different sets of mice and different times (n = 3 mice per condition, per experiment). d, H4K20me1 signal at Gpx4 gene. FC, fold change; RPKM, reads per kilobase per million mapped reads.

Source data

We then performed high-depth scRNA-seq on MuSCs from adult (4-month) and aged (24-month) mice (Fig. 5a–c) to characterize the possible presence of a pro-ferroptosis MuSC population during aging. We conducted enrichment analysis for the most altered genes and found ‘ribosome’ as the top pathway, indicative of the high transcriptional and translational turnover associated with MuSC activation. We identified ‘ferroptosis’ and ‘glutathione metabolism’ as among the most enriched pathways in MuSCs from aged mice (Fig. 5d). Pseudotime analysis identified a subpopulation of aged MuSCs (cluster 3) displaying hallmarks of ferroptosis, such as low levels of Gpx4, Slc7a11 and Pcbp2 and high Ftl1 and Fth1 (Fig. 5c–e). Both aging clusters (2 and 3) displayed similar aging signatures, such as decreased quiescence gene expression (Pax7, Spry1 and Rbpj). However, unlike the canonical aging cluster 2 (refs. ^11,50), the ferroptotic cluster (cluster 3) differentiated itself by displaying no signs of premature commitment (MyoD and MyoG) or premature senescence (Cdkn2a and Trp53). Kmt5a was also downregulated along with Gpx4, mostly in cluster 3, suggesting that the loss of Kmt5a could directly contribute to enhanced ferrosensitivity (Fig. 5e and Supplementary Fig. 5i,j).

We performed flow cytometry on adult and aged MuSCs and assessed their viability, cell fate and different forms of programmed cell death^5,11 to gain a better understanding of the extent to which MuSC ferroptosis contributes to skeletal muscle aging. As previously stated, only a small percentage of aged MuSCs entered senescence prematurely (Fig. 5f; SPiDER⁺)⁴. Most aged MuSCs that were dying were skewed toward ferroptosis (~42%; Annexin V⁻;Lipid Peroxidation^High), with fewer cells undergoing apoptosis (~27%; Annexin V⁺), indicating that ferroptosis was the dominant mode of regulated cell death in aged MuSCs (Fig. 5f). Upon sorting, ferroptotic-aged MuSCs expressed very low levels of Gpx4 and Kmt5a but high levels of Hmox1 and Ptgs2, which was consistent with a canonical pro-ferroptotic fate response and similar to cluster 3 (Fig. 5e,g). Upon assessing ferroptosis in vivo, we found that less than 10% of aged MuSCs displayed faint levels of cytoplasmic GPX4 in concert with an abundance of intracellular iron (Fig. 5h and Supplementary Fig. 5k), which was not observed in MuSCs from young mice. This suggests that, although aged MuSCs are defective in processing iron and fail to resist ferroptosis, some MuSCs retain the capacity to express GPX4, suggesting a reversible process. Similar to Kmt5a^KO MuSCs, a significant proportion of aged MuSCs displayed high intracellular iron levels (>31%), with the concentration far exceeding that measured in MuSCs from young mice (Fig. 5h,i). Additionally, freshly isolated aged MuSCs displayed a 1.5-fold increase in lipid peroxidation during homeostasis (Fig. 5j). Iron enrichment in aged skeletal muscles is thought to contribute to both functional and regeneration impairments as well as forms of muscle-wasting conditions such as sarcopenia⁴⁴. Furthermore, increased intramuscular iron is associated with increased lipid peroxidation⁵¹, which was consistent with our results.

To avoid stress response bias, we used MuSCs from Rosa^CreERT2;Nrf2^f/f;Gclc^f/f mice^52,53 treated with tamoxifen that produced Nrf2–Gclc double-knockout cells as a control. This approach was effective because the lack of glutathione in these cells naturally increases reactive oxygen species production and lipid peroxidation⁵⁴. Notably, we found ‘glutathione metabolism’ was enriched in aged MuSCs (Fig. 5d), which is consistent with altered glutathione metabolism and is a hallmark of stem cell aging⁵⁵. Upon plating, the rate of lipid peroxidation dramatically increased (>4.5-fold increase) in aged MuSCs compared to younger cells, highlighting the enhanced sensitivity of aged MuSCs to oxidative and replicative stress outside their niche (Fig. 5k). Although RSL3 treatment increased lipid peroxidation to levels similar to those observed in MuSCs from both young and aged mice, co-treatment with Fer1 rescued these effects. Notably, Fer1 treatment prevented the increase in basal levels of lipid peroxidation observed in aged MuSCs (Fig. 5k) and significantly enhanced MuSC viability and myogenic potential (Fig. 5l–o). These results demonstrate that ferroptosis is a dominant contributor to the age-related decline in MuSC numbers, where activation of the ferroptotic program is triggered by the loss of Kmt5a.

Lastly, we assessed whether age-associated death by ferroptosis is a shared fate across other adult stem cells, given that recent studies have demonstrated the selective vulnerability of human hematopoietic stem cells (HSCs) to ferroptosis⁵⁶ and observed the accumulation of labile iron in aged HSCs⁵⁷. As described previously, we found that both short-term and long-term HSCs were enriched in aged bone marrow (Supplementary Fig. 6a,b), with long-term HSCs displaying myeloid-biased CD41, a hallmark of HSC aging (Supplementary Fig. 6c). Upon sorting long-term HSCs from young and aged mice, we found that only aged long-term HSCs displayed enhanced lipid peroxidation (Supplementary Fig. 6d). These data indicate that lipid peroxidation and cell death by ferroptosis are important mediators of stem cell aging and warrant further exploration across tissues⁴⁵.

Chronic inflammation in young mice mimics Kmt5a^KO in MuSCs

We observed that acute inflammation epigenetically reprograms MuSCs through Kmt5a/H4K20me1 for a temporary period (Extended Data Fig. 1). This seems to facilitate physiological readiness in response to secondary injuries, as it correlates with activation features (Extended Data Figs. 1 and 3). We hypothesized that chronic low-grade inflammation drives aberrant epigenetic erosion of H4K20me1, leading to the demise of MuSCs by ferroptosis. This is based on our previous findings that aged MuSCs, which live in a niche enriched with pro-inflammatory factors¹, exhibit low H4K20me1 levels (Fig. 1g–m), which is associated with ferroptotic cell death (Fig. 5).

First, we validated that this effect is indeed mediated by CCR2 ligands, as observed in our acute inflammation model (Extended Data Fig. 1), by administering a chemokine cocktail intraperitoneally to young mice over 3 weeks of age to induce systemic chronic inflammation (Extended Data Fig. 6a)¹. Mice were treated with saline (vehicle), CCR2 ligands (chemokines) or a combination of CCR2 ligands and neutralizing antibodies against Ccl2, Ccl7 and Ccl8 (chemokines+nAbs). We found that chronic chemokine administration significantly reduced the MuSC pool in adult mice compared to controls, whereas the combination treatment with neutralizing antibodies partially rescued MuSC numbers (Extended Data Fig. 6b). Additionally, we found that sorted surviving MuSCs displayed enhanced activation and accelerated cell cycle re-entry (EdU incorporation), with an average first division occurring at 40 hours compared to 64 hours in the controls (Extended Data Fig. 6c). Further analysis revealed that MuSCs derived from mice chronically treated with chemokines presented similar features to aged MuSCs, including low levels of H4K20me1, repression of Kmt5a (Extended Data Fig. 6d,e) and a ferroptotic signature with repression of Gpx4, high levels of Hmox1 and high levels of lipid peroxidation (Extended Data Fig. 6e,f). Moreover, young MuSCs chronically exposed to CCR2 ligands in culture died rapidly, with nearly all cells lost after 72 hours, mirroring the phenotype of the Kmt5a^KO cultures (Extended Data Fig. 6g).

Extended Data Fig. 6. Chronic administration of Ccr2-ligands induces systemic inflammation and accelerates MuSC depletion, activation, and ferroptosis, resembling the aged MuSC phenotype. Related to Fig. 5.

a) Schematic representation of the experimental design. Young mice were administered a chemokine cocktail intraperitoneally over three weeks to induce systemic chronic inflammation. Mice were treated with saline (Veh.), Ccr2-ligands (chemokines), or a combination of Ccr2-ligands and neutralizing antibodies against Ccl2, Ccl7, and Ccl8 (chemokines + nAbs). (b) Quantification of the MuSC pool size in adult mice following chronic chemokine administration. Data are presented as mean ± s.e.m. with n = 5 mice per conditions. (c) EdU incorporation assay showing the accelerated cell cycle re-entry of MuSCs sorted from chemokine-treated mice. Data are presented as mean ± s.e.m. from three independent experiments. (d, e) Analysis of H4K20me1 (D) and qPCR analysis of Kmt5a and ferroptosis markers (E) in MuSCs derived from chronically treated mice. Data are presented as mean ± s.e.m. with n = 5 mice per conditions. (f) Lipid peroxidation analysis showing elevated levels in MuSCs from chemokine-treated mice, mirroring the aged MuSC phenotype. Data are presented as mean ± s.e.m. with n = 5 mice per conditions. (g) Survival analysis of MuSCs in culture exposed chronically –every 12hrs- to Ccr2-ligands. Data are presented as mean ± s.e.m. from three independent experiments. Statistical analyses were performed using one-way ANOVA, and p-values are indicated in the figure.

Source data

To test whether systemic inflammatory factors in aged plasma could mimic MuSC aging, we chronically transfused aged plasma into young mice^19,58,59. This caused MuSC loss and features resembling Kmt5a^KO MuSCs—including H4K20me1 loss, lipid peroxidation and Kmt5a/Gpx4 repression—all of which were rescued by CCR2 ligand neutralization (Extended Data Fig. 7a–e). These data establish a molecular link between CCR2-mediated inflammation and erosion of H4K20me1 in MuSCs, showing that the chronic systemic presence of CCR2 ligands is sufficient to induce ferroptosis in these cells.

Extended Data Fig. 7. Chronic exposure to aged plasma induces premature aging of MuSCs via Ccr2-ligands, recapitulating the Kmt5aKO phenotype. Related to Fig. 5.

a) Schematic representation of the experimental design for aged plasma transfusion. Young mice were intravenously injected with aged plasma over a specified period to mimic chronic exposure. b) Quantification of MuSC pool size following chronic aged plasma exposure. c-e) Analysis of MuSCs sorted from aged plasma-treated mice. Flow cytometry analysis showing the loss of H4K20me1 (C) and repression of Kmt5a (E). (D) Lipid peroxidation analysis indicating increased lipid peroxidation in aged plasma-treated MuSCs. (E) qPCR analysis showing the repression of Gpx4 in MuSCs exposed to aged plasma. All effects were rescued by neutralizing antibodies against Ccr2-ligands. Data are presented as mean ± s.e.m. from three independent experiments with n = 10 male mice per conditions. Statistical analyses were performed using one-way ANOVA and p-values are reported in the figure.

Source data

Anti-inflammatory intervention halts MuSC aging

We found that H4K20me1 levels transiently decreased in MuSCs in response to distal injury and the ectopic induction of acute systemic inflammation. In contrast, aged MuSCs that are naturally exposed to a pro-inflammatory environment or adult MuSCs chronically exposed to systemic inflammatory factors display H4K20me1 erosion that ultimately leads to an aberrant ferroptotic death program. Additionally, we found that neutralizing CCR2 ligands with antibodies prevented this fate.

Chronic low-grade inflammation is a hallmark of aging. However, the exact timing of this process remains elusive, which makes it difficult to employ restorative strategies. Therefore, we aimed to develop a preventive approach. We assessed whether long-term inhibition of CCR2 pro-inflammatory ligands could restore Kmt5a levels in MuSCs, prevent Kmt5a-mediated loss of quiescence and ferroptotic cell death and improve skeletal muscle homeostasis and regeneration during aging (Fig. 6). We confirmed that aged mice treated with an anti-inflammatory drug specifically targeting our pathway of interest (bindarit once per week, intraperitoneal, 30 mg kg⁻¹ (12 to 24–30 months)) exhibited significantly diminished cellular and molecular inflammatory profiles similar to those observed in young mice (Extended Data Fig. 8a–k). As expected, bindarit treatment restored the Kmt5a levels in MuSCs from aged mice to levels similar to adult MuSCs (Extended Data Fig. 8l). The number of MuSCs in vivo was significantly improved at both homeostasis and after regeneration (Fig. 6a–d), and both intracellular iron and lipid peroxidation were significantly decreased (Fig. 6h,i) compared to vehicle-treated aged mice, complemented by enhanced muscle repair and restoration of muscle strength after injury (Fig. 6d–g). Bindarit had no effect in adult mice, consistent with its selective benefit under chronic inflammation^60–62 (Extended Data Fig. 9a–f).

Fig. 6. Long-term inhibition of inflammation rejuvenates skeletal muscle and improves functional recovery.

a–d, Quantification of MuSC numbers at homeostasis (a,b) and at 21 days after regeneration (c,d) in aged mice. Vehicle (Veh) and bindarit (Bin; 30 mg kg⁻¹ per week, intraperitoneal) treated mice were injected weekly between 12 months and 24–30 months of age. e, Grip strength measurement after regeneration normalized to adult mice and histological evaluation of muscle regeneration. f,g, Quantification of the regenerated myofiber size (f) and size distribution (g) in the TA muscle. h,i, Quantification of lipid peroxidation (h) and single-cell average iron content (i). Two-sided Welch’s t-test. Each dot represents a mouse replicate. P values are reported in the figure.

Source data

Extended Data Fig. 8. Inflammatory profiling of aged mice in response to long-term Bindarit treatment. Related to Fig. 6.

a-f) HESKA to quantify whole blood cellular content with monocytes (A), lymphocytes (B), neutrophils (C), eosinophils (D), basophils (E), and total red blood cells (F). g-k, Multiplex quantification of circulatory pro-inflammatory molecules targeted by Bindarit: Ccl2 (G), Ccl7 (H), Ccl8 (I), TNF-α (J), and IL-6 (K). l, Kmt5a levels in muscle stem cells relative to adult cells.

Source data

Extended Data Fig. 9. Short-term Bindarit treatment does not improve adult mouse skeletal muscle regeneration. Related to Fig. 6.

a) Strategy to assess Bindarit effect over adult skeletal muscle regeneration. Mice blood were collected pre-treatment, then post-treatment every week for whole blood count (WBC). Once levels of circulatory monocyte were significantly decreased, we stopped the treatment and proceeded with barium chloride injury. Mice rested for 21days before tissue harvest and analysis. b) WBC before and after 3-weeks Bindarit treatment versus vehicle treated mice. Twenty-four hours after the third round of treatment, Bindarit-treated mice showed significantly lower circulatory monocyte than before treatment. c) TA CSA from contralateral uninjured limb from mice treated for 3-weeks with vehicle or Bindarit. d) Regeneration assay for adult mice treated with vehicle or Bindarit for 3-weeks. e, f) Quantification of MuSC number in uninjured and injured TA (E) and regenerated fiber size average and frequency distribution (F) at 21dpi. For each violon plot, data points represent biological replicates (n = 5 mice per condition). Statistical significance was performed with ANOVA (B) or two-sided Welch’s t-test (E, F) and exact adjusted p values are reported in the figure.

Source data

Finally, we evaluated whether the plasma from bindarit-treated mice had true rejuvenating properties that would transfer to young animals (Extended Data Fig. 10a) because we found that chronic exposure to plasma derived from aged mice mimicked aging phenotypes. Unlike the plasma derived from the Aged^Veh group, young mice that received chronic injections of plasma derived from aged mice treated with bindarit for 12 months were virtually indistinguishable from the control group. Bindarit-treated mice had a normal pool of MuSCs (Extended Data Fig. 10b) and proper skeletal muscle regeneration (Extended Data Fig. 10c). Furthermore, MuSCs from the Aged^Bin plasma-treated group showed basal levels of H4K20me1 (Extended Data Fig. 10d), lipid peroxidation (Extended Data Fig. 10e) and Kmt5a and Gpx4 expression (Extended Data Fig. 10f). We describe our proposed model in a schematic (Supplementary Fig. 7).

Extended Data Fig. 10. Plasma from Bindarit-treated aged mice reverses aging phenotypes in young recipient mice. Related to Fig. 6.

a) Schematic representation of the experimental design. Young mice were chronically injected with plasma derived from aged mice treated with Saline or Bindarit for 12 months to assess potential rejuvenating effects. b) Quantification of MuSC pool size in young mice following chronic plasma exposure. Plasma from Bindarit-treated aged mice maintained a normal MuSC pool comparable to control mice, unlike plasma from untreated aged mice. c) Assessment of skeletal muscle regeneration in young mice treated with aged Bindarit plasma, showing normal regeneration similar to controls. d-g) Analysis of MuSCs sorted from young mice exposed to aged Bindarit plasma. (D) Flow cytometry analysis showing levels of H4K20me1. (E) Lipid peroxidation analysis indicating normal lipid peroxidation levels. qPCR analysis showing normalized expression levels of Kmt5a (F) and Gpx4 (G). Data are presented as mean ± s.e.m. Statistical analyses were performed using one-way ANOVA (B-F), and p-values are indicated in the figure.

Source data

Discussion

Our study provides insights into the mechanisms underlying MuSC aging, particularly highlighting the role of inflammation in depleting Kmt5a and its associated histone mark H4K20me1 in aged MuSCs. We demonstrated that Kmt5a-mediated H4K20me1 deposition was critical for maintaining MuSC quiescence in young mice and protecting against ferroptosis, an iron-dependent form of cell death. Our findings, supported by genetic models and multi-omics analyses, reveal how systemic inflammation contributes to MuSC aging through the epigenetic erosion of H4K20me1, leading to the downregulation of genes that protect against MuSC oxidation and eventual ferroptotic cell death.

We found that age-associated systemic inflammation led to a decline in Kmt5a expression in MuSCs, resulting in the loss of H4K20me1. This epigenetic change disrupts quiescence and predisposes MuSCs to ferroptosis. Our data indicate that aged MuSCs exhibit increased markers of ferroptosis, including lipid peroxidation and iron accumulation, along with decreased expression of key anti-ferroptotic genes, such as Gpx4. Notably, our results suggest that ferroptosis, previously uncharacterized in MuSCs, may play an important role in the decline in MuSC numbers and regenerative capacity during aging.

Our findings align with previous studies, showing that inflammation accelerates the aging processes in various tissues⁶³. However, our study extends this knowledge by identifying a specific epigenetic mechanism, H4K20me1 erosion, through which inflammation drives MuSC aging. Iron metabolism is crucial for myogenesis but poses a risk of ferroptosis if misregulated. Tfrc deletion in MuSCs disrupts iron handling, promoting lipid peroxidation and impairing regeneration^51,64. Thus, MuSCs must balance iron uptake and scavenging to support regeneration while avoiding oxidative damage⁶⁵.

The mechanistic axis that we propose, linking inflammation to the Kmt5a-dependent epigenetic regulation of MuSCs, extends recent findings on the role of inflammation in aging. For example, Benjamin et al.^54,55 recently showed that a subset of aged MuSCs is dysfunctional due to high NF-κB-mediated inhibition of NRF2, leading to unbalanced glutathione metabolism, which is critical for limiting lipid peroxidation and ferroptosis. Because bindarit targets NF-κB (ref. ⁶⁶), its benefits may stem from restoring redox balance. Notably, both inflammation and iron accumulation are linked to human aging. Our data demonstrate that inflammation induces epigenetic changes that lead to iron accumulation and enhanced sensitivity to ferroptotic death in stem cells, which could have important implications for human aging. As chronic inflammation is a hallmark of aging, it is plausible that the accumulation of iron in stem cells due to inflammation-induced epigenetic reprogramming contributes to age-associated low-grade inflammation. The release of pro-inflammatory molecules associated with ferroptotic death can perpetuate the inflammatory environment, creating a vicious cycle of inflammation, iron accumulation and ferroptotic death across several aged tissues. Age-related chronic inflammation leads to increased iron accumulation in various tissues, whereas excess iron promotes inflammation through the generation of reactive oxygen species and activation of pro-inflammatory signaling pathways⁴⁵. This reciprocal relationship between inflammation and iron level is thought to contribute to the development and progression of age-related diseases.

Here we present data indicating that age-associated systemic inflammation reprograms MuSCs through epigenetic erosion, leading to premature loss of quiescence and subsequent cell death by ferroptosis. Our findings directly link aging and inflammation to the loss of Kmt5a function, which drives an epigenetic switch that disrupts MuSC quiescence and promotes ferroptosis. Our findings provide evidence to suggest that the inflammation–epigenetic–ferroptosis axis has major therapeutic implications for stem cell rejuvenation, as iron metabolism is already known to contribute to stem cell aging⁵⁷. We present a compelling case for further exploration of the complex interplay among epigenetic erosion-mediated diversity, iron metabolism and ferroptosis in diverse adult stem cell compartments during aging. This research could expand the growing body of evidence linking the involvement of ferroptosis to aging-related disorders, which has so far focused primarily on neurodegenerative diseases^67,68. Our findings underscore the urgent need for regenerative medicine to further explore the fundamental mechanisms driving these processes, with the aim of developing therapeutic interventions to combat age-related diseases beyond neurodegenerative disorders. Unraveling these mysteries may pave the way for a new era of personalized regenerative medicine that substantially improves the quality of life of aging individuals.

Study limitations

Our study used male mice only. Although we focused on H4K20me1, other marks may contribute to ferroptosis. Further research is also needed to define how bindarit exerts its benefits.

Methods

All animal procedures and experimental protocols were conducted in compliance with all relevant ethical regulations and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Rochester Medical Center (URMC) and the University of Wisconsin-Madison (UW-Madison). Protocol oversight and animal care compliance were provided by the Division of Comparative Medicine (URMC) and the Research Animal Resources and Compliance Unit (UW-Madison). Biosafety protocols were reviewed and approved by the Institutional Biosafety Committee at the URMC and the Office of Biological Safety at UW-Madison. This research was conducted in accordance with guidelines set forth by the National Institutes of Health and the National Institute on Aging, where applicable.

Animal studies

Animal experiments in this study were carried out in accordance with guidelines set by the Animal Care and Use Committee at the University of Rochester and UW-Madison. C57BL6 adult (4–6 months; Jackson Laboratory) and aged (24–25 months; National Institute on Aging) mice were used for indicated experiments. Kmt5a^FL/FL;Pax7^CreERT2 were generated by crossing Kmt5a^FL/FL generously gifted by Laurie Steiner (URMC) and Pax7^CreERT2 (no. 017763; Jackson Laboratory). CCR2-KO mice (no. 004999; Jackson Laboratory) were used as a control for chemokine treatments. P53^FL/FL were generously gifted by Stephano Spano Mello (URMC). For regeneration assays, mice were anesthetized with isoflurane. TA muscles were injected with 50–80 µl of a 1.2% solution of BaCl₂ in normal saline. Mice were housed in a specific pathogen-free facility under standard conditions with a 12-hour light/12-hour dark cycle. Ambient temperature was maintained at 20–24 °C with relative humidity of 40–60%. Animals had ad libitum access to standard chow and water. All procedures were conducted in accordance with institutional and national guidelines for animal care and were approved by the IACUC.

Cell isolation, culture and immunostaining

For magnetic activated cell sorting (MACS), skeletal muscles were dissociated using gentleMACS, and satellite cells were freshly isolated using a Satellite Cell Isolation Kit (Miltenyi Biotec). For fluorescence-activated cell sorting (FACS), skeletal muscles were dissociated using gentleMACS, and satellite cells or SC-derived myogenic progenitors were isolated as previously described⁶⁹.

Satellite cell cultures and related experiments, including EdU pulse and fate assay, were performed as previously described⁶⁹. All cells were incubated at 37 °C in 5% CO₂ and humid air conditions.

For ferroptosis induction assay, freshly sorted MuSCs were plated for 4 hours to adhere prior to changing to growth media supplemented with DMSO, RSL3, erastin, Fer1 or UNC0379 at various concentrations and incubation times (Supplementary Data). After these results, we then used 10 µM erastin, 1 µM RSL3, 10 µM Fer1 and 1–5 µM UNC0379 for 18–24 hours.

For expression assays, MuSC-derived myoblasts or C2C12 were transfected with RNAiMAX for siRNA and with Lipofectamine 2000 for DNA plasmids. For Kmt5a knockdown, we used siGenome pools of four siRNAs (no. 1 ACCCGUGGCUGAAGCAUUA, no. 2 GCAACUAGAGAGACAAAUC, no. 3 GAUUGAAAGUGGGAAGGAA, no. 4 GCACGACAUCGACGGCGUA; Dharmacon). For Kmt5a overexpression, we used the Kmt5a (NM_030241) Mouse Tagged ORF Clone (cat. no. MR205297; OriGene) along with pCMV6 as empty vector control. For Gpx4 overexpression, we used Gpx4 (BC106147) Mouse Tagged ORF Clone (cat. no. MG201418; OriGene).

Single myofibers were isolated from extensor digitorum longus (EDL) muscles by enzymatic digestion in 0.2% Collagenase Type I (in DMEM) for 60 minutes at 37 °C. After digestion, muscles were gently dissociated in pre-warmed DMEM containing 10% FBS to release individual fibers, which were then washed multiple times to remove debris. Myofibers were fixed either immediately or after specified culture timepoints using 10% neutral buffered formalin for 12 minutes at room temperature and then transferred to PBS and stored at 4 °C until further use.

Tissue section and immunostaining

For injury studies, TA/EDL muscles were harvested and frozen in isopentane cooled in liquid nitrogen and stored at −80 °C prior to sectioning or prepared for MACS. Frozen muscles were sectioned at 10 μm. Muscle sections were fixed for 10 minutes in 10% normal buffered formalin and permeabilized with 2% NGS/PBS-T (0.2% Triton X-100) for 30 minutes at room temperature and blocked in 10% NGS (Jackson ImmunoResearch) in PBS-T for 30 minutes at room temperature. When mouse primary antibodies were used, sections were additionally blocked in 3% AffiniPure Fab fragment goat anti-mouse IgG(H + L) (Jackson ImmunoResearch) with 2% NGS in PBS at room temperature for 1 hour. Primary antibody incubation in 5% NGS/PBS-T was carried out at 4 °C overnight or 2 hours at room temperature, and sections were incubated with secondary antibodies in 5% NGS/PBS-T for 1 hour at room temperature. DAPI staining was used to label nuclei. All slides were mounted with Fluoromount-G (SouthernBiotech). At least three sections from three slides were analyzed per sample. Sections and cells were imaged on an Echo Revolve Microscope through Echo PRO software (version 4.2) and processed and analyzed in ImageJ2 (versions 1.52, 1.53 and 1.54p). Sample analysis was conducted in a double-blind manner.

Immunoblotting and Luminex

Cell lysates (50 mM HEPES (pH 7.4), 150 mM NaCl and 1% Triton X-100) and freshly added EDTA-free protease and phosphatase inhibitor cocktails (Roche) were immunoblotted with primary antibodies overnight at 4 °C. HRP-conjugated secondaries and ECL from Bio-Rad were used for chemiluminescence detection, and DyLight secondary antibodies were used for infrared detection on a LI-COR Odyssey using LI-COR Image Studio software (version 3.1) and further analyzed on Image Lab (version 6.2), ImageJ (version 1.52) or Image Studio (version 5.5). For Luminex assays, samples were prepared using a customized Mouse Magnetic Luminex Assay (R&D Systems) following the manufacturer’s protocol and using a Bio-Rad Bio-Plex 200 apparatus and Bio-Rad Bio-Plex Manager software (version 6.2).

RNA extraction and qPCR

RNA was extracted using TRIzol reagent (Life Technologies) following the manufacturer’s protocol. RNA was normalized and used for RT–PCR followed by qPCR using EvaGreen (Bio-Rad) and a 7500 Fast Real-Time PCR system (Applied Biosystems) using StepOnePlus software (version 2.3). Transcript levels were normalized to an average of at least two housekeeping genes among TBP, GAPDH and B2M and then to the control condition (ddCT) and reported as fold change of the ddCT.

Blood analysis and transfusion

Blood was collected by submandibular bleed or retro-orbital bleed using heparinized capillary tubes. Approximately five drops of blood per mouse were collected (to limit stress and excessive blood loss) into K2EDTA microtainer tubes (no. 365974; BD Biosciences). Shortly thereafter, blood samples were processed on a HESKA HemaTrue analyzer (HESKA Corporation) to obtain CBC parameters. Remaining blood was spun down for plasma collection and stored at −80 °C or used immediately for molecular analysis. For plasma transfusion, recipient mice were anesthetized, and 200 µl of plasma was administered via tail vein injection.

Recombinant chemokines and bindarit treatment

A cocktail of recombinant Ccl2, Ccl7 and Ccl8 was administered by intraperitoneal injection as previously described¹. Bindarit was freshly prepared prior to injection with a mix of solvents (saline, PEG300, DMSO and Tween 80). A single dose of bindarit at 30 mg kg⁻¹ per week was administered via intraperitoneal injection. As a control, a vehicle containing only the solvents used in bindarit without the drug was injected in a volume matching the maximum amount of the drug administered. The dosage for the experiment was established through a pilot study with the assistance of veterinarians. The lowest effective dose was determined based on its ability to reduce circulatory monocyte and Ccl2 levels for at least 1 week in aged mice. The pilot study was conducted to optimize animal well-being in accordance with University Committee on Animal Resources and IACUC guidelines to minimize stress and invasive procedures for the model animals. Bindarit was selected for its strong anti-inflammatory properties demonstrated in clinical trials. It was chosen because it does not cause complete suppression of the immune system and has not shown any toxicity to either humans or rodents⁶⁶. We initiated the study with 30 mice and used 20–25 mice per condition for the results reported. We used the ANCOVA method to perform a power analysis with age as a covariate and found that power was sufficient for such a repeated measure based on our current and past data.

Antibodies, probes and dyes

The following antibodies were used: mouse anti-Pax7 (Developmental Studies Hybridoma Bank (DSHB) or Santa Cruz Biotechnology, sc-81648); mouse anti-MyoD (BD Biosciences, 554130); mouse anti-myogenin (Abcam, 1835); rabbit anti-myogenin (Abcam, 124800); mouse anti-embryonic myosin heavy chain BF-45/F1.652 (DSHB); rabbit anti-H4K20me1 (Abcam, 9051); rabbit anti-H3, rabbit anti-H3K4me3 (Abcam, 8580); rabbit anti-H4 (Abcam, 10158); rabbit anti-pS6(Ser235/236) (Cell Signaling Technology, 2211); rabbit anti-γH2AX(Ser139) (Cell Signaling Technology, 9718); mouse anti-γH2AX(Ser139) (Millipore, 05-636); rabbit anti-γH2AX(S139)[-AF750] (BioTechne, IC2288); rabbit anti-pP53(Ser15) (Cell Signaling Technology, 9284); rabbit anti-pP53(Ser46) (Cell Signaling Technology, 2521); mouse anti-P53 (Santa Cruz Biotechnology, sc-126); rat or rabbit anti-laminin (Sigma-Aldrich, L0663 or L9393); rabbit anti-SETD8 (Millipore, 06-1304, or Cell Signaling Technology, 2996); rabbit anti-skeletal muscle myosin (Sigma-Aldrich, HPA1239); rabbit anti-GPX4 (Abcam, 125066); mouse anti-GAPDH(HRP) (Abcam, 9482); and rabbit anti-Ki67(488) (Cell Signaling Technology, 11882). The following antibodies were used as previously validated neutralizing antibodies: anti-Ccl2 (R&D Systems, AF479NA); anti-Ccl7 (R&D Systems, AF456NA); and anti-Ccl8 (R&D Systems, MAB281). Neutralizing antibodies were freshly added to the specific treatment prior to injection (intraperitoneal, 2 mg kg⁻¹; intravenous, 800 µg kg⁻¹).

All probes and kits were used according to the manufacturer’s protocol: FerroFarRed (Millipore, SCT037); MitoSOX (Invitrogen M36008); ApopTag (Millipore, S7100); ImageIT Lipid Perox (Invitrogen, C10445); Annexin V (Invitrogen, A23204, A35109 or A35122); FastCellular Senescence Detection Kit SPiDER-βGal (MP Biomedicals, 2690301); EdU-ClickIT (Invitrogen, C10337, C10425 or C10340); and DAPI.

Elemental mass spectrometry (iron quantification)

Freshly sorted MuSC pellets were processed for elemental iron detection using a PerkinElmer NexION 2000 inductively coupled plasma mass spectrometry (ICP-MS) system, equipped with a multi-element detection capability and ppb/ppt sensitivity. In brief, MuSC pellets were digested in 70% HNO₃ at 60 °C for 2 hours, followed by dilution with ultrapure water to achieve a final acid concentration of 2%. Calibration standards for iron were prepared in the same acid matrix to ensure consistency. The samples were then introduced into the ICP-MS system using an autosampler, with detection settings optimized for iron analysis at the ppt level. Data acquisition was performed using Syngistix software (version 4.0), and iron concentrations were quantified against a standard curve generated from serial dilutions of certified iron standards. All measurements were conducted in triplicate to ensure accuracy and reproducibility.

Flow cytometry and FACS

MuSCs were freshly isolated using MACS. Cells were then incubated with VCAM and the probes of interest. For MitoSOX, lipid peroxidation, Annexin V and SPiDER detection, we followed the manufacturers’ instructions, except for lipid peroxidation where incubation was reduced to 15 minutes. Cells were then processed on a BD LSRFortessa using BD Diva (version 8.0) and analyzed in FCS Express (version 7.0). The gating strategy is provided in the Supplementary Data. For lipid peroxidation FACS, we used a ratiometric measurement of 590 nm/510 nm and sorted the highest positive population as seen in Fig. 3f. For intracellular staining, MuSCs were permeabilized with BD Pharminogen Transcription Factor Buffer Set (BD Biosciences, 554714) following the manufacturer’s instructions.

Bulk RNA-seq

Total RNA was isolated using an RNeasy Plus Micro Kit (Qiagen). RNA concentration was determined with a NanoDrop 1000 spectrophotometer, and RNA quality was assessed with an Agilent Bioanalyzer 2100. One nanogram of total RNA was pre-amplified with a SMARTer Ultra Low Input Kit version 4 (Clontech) per the manufacturer’s recommendations. The quantity and quality of the subsequent cDNA were determined using a Qubit Fluorometer (Life Technnologies) and the Agilent Bioanalyzer 2100. Then, 150 pg of cDNA was used to generate Illumina-compatible sequencing libraries with a NexteraXT Library Preparation Kit (Illumina) per the manufacturer’s protocols. Amplified libraries were sequenced using a NextSeq 550 sequencer (Illumina) on the high flow cell to obtain 25–30 million single-end 75-nucleotide reads per sample.

Bulk RNA-seq analysis

Raw reads generated from the Illumina basecalls were demultiplexed using bcl2fastq (version 2.19.1). Quality filtering and adapter removal were performed using FastP version 0.23 (ref. ⁷⁰) with the following parameters: ‘–length_required 35–cut_front_window_size 1–cut_front_mean_quality 13–cut_front–cut_tail_window_size 1–cut_tail_mean_quality 13–cut_tail -y –r’. Processed/cleaned reads were then mapped to the GRCm39/gencode M31 (mouse) reference using STAR_2.7.9a with the following parameters: ‘—twopass Mode Basic–runMode alignReads–outSAMtype BAM Unsorted – outSAMstrandField intronMotif–outFilterIntronMotifs RemoveNoncanonical –outReadsUnmapped Fastx’^71,72. Gene-level read quantification was derived using the subread-2.0.1 package (featureCounts) with a GTF annotation file (GRCm39/gencode M31) and the following parameters for unstranded RNA libraries: ‘-s 0 -t exon -g gene_name’⁷³. Differential expression analysis was performed using DESeq2-1.34.0 with a P value threshold of 0.05 within R version 3.5.1 (https://www.R-project.org/)⁷⁴. A principal component analysis plot was created within R using pcaExplorer (version 2.22.0) to measure sample expression variance⁷⁵. Heatmaps were generated using the pheatmap package (version 1.0.12) and were given rLog transformed expression values. GO analyses were performed using the Enrichr package^76–78 (version 3.1). Volcano plots and dot plots were created using ggplot2 (ref. ⁷⁹) (version 3.3.5).

scRNA-seq methods

Skeletal muscles were dissociated and filtered using MACS protocol to obtain a single-cell suspension prior to having the Genomics Research Center (GRC) core process the samples for scRNA-seq. For aged wild-type and mutant conditions, we pulled three and five mice, respectively, to obtain sufficient cell numbers for the selected depth/capture size. Cell suspensions were processed to generate scRNA-seq libraries using Chromium Next GEM Single Cell 3′ GEM, Library and Gel Bead Kit, version 3.1 (10x Genomics), per the manufacturer’s recommendations, as summarized below. Samples were loaded on a Chromium Single Cell Instrument (10x Genomics) to generate single-cell gel bead-in-emulsions (GEMs). GEM reverse transcription (GEM-RT) was performed to produce a barcoded, full-length cDNA from poly-adenylated mRNA. After incubation, GEMs were broken; the pooled GEM-RT reaction mixtures were recovered; and cDNA was purified with silane magnetic beads (DynaBeads MyOne Silane Beads; Thermo Fisher Scientific). The purified cDNA was further amplified by PCR to generate sufficient material for library construction. Enzymatic fragmentation and size selection were used to optimize the cDNA amplicon size, and indexed sequencing libraries were constructed by end repair, A-tailing, adaptor ligation and PCR. Amplified libraries were sequenced using a NovaSeq 6000 sequencer (Illumina) on the S1 flow cell to obtain 100,000 reads per cell.

scRNA analysis

Samples were demultiplexed using ‘cellranger mkfastq’ within 10x-provided cellranger-7.0.1 using default parameters. Samples were aligned to 10x-provided reference refdata-gex-mm10-2020-A (mouse) using ‘cellranger count’ within cellranger-7.0.1 using default parameters⁸⁰. Clustering was performed using Seurat’s (4.1.0) standard workflow in R (4.1.1)⁸¹. Cell counts were normalized using Seurat’s SCTransform function, regressing out mitochondrial content. Principal component analysis was performed to determine the first 40 principal components within the RunPCA method. Neighbors were called using the first 20 determined principal components within the FindNeighbors method. Uniform manifold approximation and projection (UMAP) was used to visualize dimensional reduction of the first 20 principal components as well using RunUMAP. The dataset was clustered at a resolution of 0.2. Cell markers were then determined for each cluster using Seurat’s FindAllMarkers(). Clusters were subsequently typed using Single Cell Mouse Cell Atlas (version 0.2.0)⁸². Significantly differentially expressed genes (adjusted P < 0.05) were determined between knockout and wild-type in cluster 1 using Seurat’s FindAllMarkers(). Gene set enrichment analysis was performed with the following databases: KEGG_2021_Mouse, GO_Biological_Process_2021, WikiPathways_2019_Mouse and ChEA_2016 using the Enrichr package^76–78 (version 3.1). After population clustering and pre-analysis, a sub-dataset was generated to analyze MuSCs. Data were again filtered out to fit quality control criteria previously established. scRNA-seq data were analyzed using the following packages. Clustering: cells were clustered using the Seurat package (version 4.3) based on the top variable genes. The number of principal components used for clustering was determined using the ElbowPlot function. Integrated data were then clustered with Louvain (0.8) and embedded with UMAP (cosine, 0.3-minute distance). Embedding: robust principal component analysis (RPCA) plot was generated using the Seurat package based on the first 26 principal components. Velocity: RNA velocity was calculated using the velocyto.R package (version 0.6) to determine the directionality of gene expression changes. RNA velocity was cross-verified with PRO-seq datasets when possible. Pseudotime: Monocle 3 (version 1.2.7) was used to perform pseudotime analysis and construct a trajectory plot. The top principal components and RNA velocity were used as input.

PRO-seq

Freshly isolated MuSCs were permeabilized for PRO-seq as previously described⁸³ and shipped on dry ice to the Nascent Transcription Core (Harvard University) for processing. For refinement of gene annotation (proTSScall) using PRO-seq alone, PRO-seq reads were used to identify active TSSs using a custom script: proTSScall is available on the Nascent Transcription Core GitHub (https://github.com/NascentTranscriptionCore/proTSScall). In brief, PRO-seq 3′ read bedGraphs for ‘+’ and ‘−’ strands were separately combined across samples, and the composite read counts were assigned to TSS-proximal windows (TSS to +150 nucleotides) using the same filtered TSS annotation described above. TSSs with ≤9 counts in this window are deemed ‘inactive’, and the remaining TSSs, deemed ‘active’, are collapsed to yield one dominant TSS per gene, defined as the one with the highest TSS-proximal read count—if the highest read count is shared among multiple transcripts, the TSS farthest upstream, in a strand-aware fashion, is called dominant. Dominant TSSs sharing the same start position are deduplicated as follows: (1) if start positions are equal, the TSS with the longest associated annotated transcript is called dominant; (2) if start positions and transcript lengths are both equal, the TSS associated with the lowest Ensembl gene ID (numerical portion) is dominant.

CUT&Tag sequencing

MuSCs were freshly isolated from young (4-month) and aged (24-month) mouse skeletal muscles. For H4K20me1 CUT&Tag, we used three young male mice and three aged male mice as individual replicates; in addition, one young mouse and one aged mouse were used as ‘No Antibody’ control. CUT&Tag was performed following the Henikoff group’s original protocol and using the original enyme⁸⁴. Precipitated DNA was used for a short PCR pulse (five cycles) using one barcode pair for all samples. For quality assessment prior to library construction, PCR product was used for qPCR; libraries were balanced based on qPCR amplification. To generate library, we performed a 13-cycle PCR with individual barcode pairs for each sample, followed by AMPure XP beads (Beckman Coulter) cleanup. Libraries were sent to GENEWIZ for 150-bp paired-end sequencing.

CUT&Tag data analysis

CUT&Tag raw sequencing reads were adapter trimmed using Cutadapt and then aligned to mouse genome (mm10) using Bowtie 2 (version 2.5.1). PCR duplicates were marked and removed using Picard (version 3.0). Peak calling was performed using SEACR (version 1.2) as top 5% of regions by area under the curve with the stringent setting. Individual peaks from each replicate of either young or old mice were merged using BEDTools as union peak for downstream analysis. Read count normalized genome browser tracks were generated using bamCoverage from merged SAM files of triplicates with the following parameters: –normalizeUsing RPKM.

CUT&RUN qPCR

Cells were harvested and subjected to CUT&RUN using the CUTANA kit (EpiCyher). In brief, cells were incubated with protein A-MNase and primary antibodies overnight at 4 °C. Targeted chromatin fragments were released by activating MNase with calcium, followed by DNA extraction. DNA concentration was quantified using PicoGreen. qPCR was performed with primers targeting specific genomic regions of interest to assess enrichment of chromatin marks. Fold enrichment was calculated relative to the IgG control.

Ligand–receptor inflammatory signaling modeling

We first proceeded with identifying receptors with a predicted activation (z-score > 2) using young versus aged skeletal muscle RNA-seq that we published previously using Ingenuity Pathway Analysis (IPA) Upstream Regulator Analysis⁸⁵ (version 20.0). IPA was employed to prognosticate the activation status of receptors based on downstream gene profiles comparing young versus aged mouse skeletal muscles. We then filtered the candidates to only retain receptors associated with inflammation as indicated by IPA. We then transformed these data through a regression model to predict ligand–receptor activity over those activated signaling pathways. We scored the ‘Signaling Activity’ by integrating our ‘Ligands to Receptor Score’ and our ‘Upstream Regulator Receptor Score’ through an adapted version of the Crosstalkr package⁸⁶ (version 1.4, in R version 3.5.1). We validated the predictive score by associating the levels of cytokines with the enrichment of receptor-to-signaling activity using the updated SPIA package⁸⁷ (version 2.52). The SPIA package facilitated the evaluation of enriched inflammatory signaling pathways in tandem with the increase of cytokines during aging. Proteome fold change data were converted into normalized fold change values to better analyze signaling activity. We also confirmed that the ligand–receptor activity prediction score correlated with the appropriate directionality of transcriptomics changes observed in those signaling pathways using a Markov chain reduction model to define the ligand–receptor activity as a high-probability intermediate signaling molecule between circulatory cytokine and downstream gene expression in skeletal muscles (Python version 3.6.4). After validation, we filtered data to only retain receptors that are expressed in skeletal muscles; that have one or more ligands significantly enriched in aged plasma, where the ligand mRNA was not detected in skeletal muscles; and whereby changes of the receptor mRNA levels in skeletal muscles did not affect its activation z-score. Significance was assessed by permuting the populations 100 times and recomputing the interaction scores in permuted datasets. Only activity scores greater than 2 s.d. from the mean of the permuted scores were considered significant and reported⁴.

Transmission electron microscopy

Skeletal muscles were immersion fixed in a combination fixative containing 4.0% paraformaldehyde/2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). After 24 hours of primary fixation, the muscle specimens were rinsed in the same buffer and post-fixed for 2 hours in buffered 1.0% osmium tetroxide/1.5% potassium ferrocyanide, washed in distilled water, dehydrated in a graded series of ethanol to 100% (3×), transitioned into propylene oxide, infiltrated with EPON/Araldite resin overnight, embedded into molds and polymerized for 48 hours at 60 °C. The blocks were sectioned at 1 µm and stained with toluidine blue prior to thin sectioning at 70 nm onto slot formvar/carbon-coated nickel grids. The grids were stained with aqueous uranyl acetate and lead citrate and examined using a Hitachi 7650 transmission electron microscope with an attached Gatan 11-megapixel Erlangshen digital camera and DigitalMicrograph software.

Statistics and reproducibility

Statistical significance was assessed using GraphPad Prism 10 software via two-sided Welch’s t-test (unpaired, 95% confidence interval) or one-way or two-way ANOVA (followed by Tukey post hoc test; 95% confidence interval). Error bars are reported as s.e.m. or s.d. and are displayed in appropriate graphs. P < 0.05 was considered statistically significant. Data distribution was assumed to be normal, but this was not always formally tested. Whenever possible, we used violin plots to show the number of biological replicates with statistical tests and exact P values or adjusted P values (q values) directly reported on the graphs. Mice and samples were assigned to experimental groups without predetermined criteria whenever possible. Where power analyses were performed to determine sample sizes, details are provided in the relevant sections of the Methods. In all other cases, sample sizes were based on prior published studies in the field and are considered sufficient to detect biologically meaningful effects^{1,2,4,6,34,37,50}. No biological replicates were excluded from analysis. For high-throughput datasets such as muscle fiber size measurements, outlier filtering was applied uniformly across all experimental groups using the ROUT method (Q = 1%), a robust statistical approach for outlier detection. This ensured consistency while minimizing bias in the calculation of group averages. Whenever feasible, experiments and subsequent analyses were performed in a double-blind manner and repeated on different occurrences; in this instance, to ensure unbiased data acquisition and analysis, a third person blinded the samples in an arbitrary order, and the individual performing the analysis was blinded to sample identities. The data were then analyzed by a second individual who was also blinded to the sample identities. The final results were unblinded only after the completion of the analyses.

Inclusion and ethics statement

This study was conducted in accordance with the principles outlined by both University of Rochester (prior institution) and University of Wisconsin-Madison updated guidelines regarding inclusion and ethics. We also confirm that all the authors of this manuscript followed these strict principles and were treated equally during all collaborations. Measures were taken to ensure that the study was inclusive and accessible to all, regardless of gender, race, ethnicity, sexual orientation or any other characteristics protected under applicable laws. We also ensured that the study did not involve any unethical practice or procedure. Any concerns regarding the ethical conduct or content of this study should be directed to the corresponding author for review.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

R.S.B. initiated, designed and obtained funding for the project; performed the experiments; analyzed and interpreted the data; wrote the manuscript; and made the figures. N.S. performed all molecular experiments related to the RSL3 and Kmt5a inhibitor treatments. N.A.S.S. analyzed and produced figures for scRNA-seq and performed HSC-related experiments. S.H. performed all the experiments related to GPX4 overexpression. F.W.M. performed, analyzed and produced figures for the CUT&Tag experiments. B.A.Y. and C.A.A. provided guidance and analysis strategies for the CUT&Tag dataset analysis and integration into RNA-seq. A.M. performed experiments related to G_Alert characterization of KMT5a^KO MuSCs. G.P. performed some flow cytometry analysis. J.H.H. and E.E.W. performed, quantified and analyzed pS6 experiments related to the manuscript revisions. J.O.O. shared valuable resources and helped conceptualize and interpret the ferroptosis results. J.V.C. conceptually supported project initiation, notably regarding CUT&Tag data generation, as well as provided financial support for R.S.B. salary and experiments at the inception of this work. P.J.M., L.C., J.F.D. and R.T.D. supported the project, helped interpret data and provided critical feedback on the project and manuscript.

Peer review

Peer review information

Nature Aging thanks So-ichiro Fukada and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

The omics datasets generated during the present study are deposited in the Gene Expression Omnibus repository in accordance with National Institutes of Health guidelines and are publicly available: PRO-seq: GSE227366; RNA-seq bulk and single cell: GSE228013 and GSE228055; and CUT&Tag: GSE227367. All other datasets generated or analyzed during this study are available upon reasonable request. No original codes are reported, but data reanalysis can be accessed upon reasonable request to the corresponding author and various core directors involved in their development (‘Acknowledgements’). Uncropped western blots, unprocessed data underlying the display items in the manuscript in every figure and the flow cytometry gating strategy are available in the Source Data file.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s43587-025-00902-5.

Supplementary information

The online version contains supplementary material available at 10.1038/s43587-025-00902-5.