Multiomic profiling reveals that Prostaglandin E2 reverses aged muscle stem cell dysfunction leading to increased regeneration and strength

Summary:

Repair of muscle damage declines with age due to the accumulation of dysfunctional muscle stem cells (MuSCs). Here we uncover that aged MuSCs have blunted Prostaglandin E2 (PGE2) - EP4 receptor signaling, which causes precocious commitment and mitotic catastrophe. Treatment with PGE2 alters chromatin accessibility and overcomes the dysfunctional aged MuSC fate trajectory, increasing viability and triggering cell cycle re-entry. We employ neural network models to learn the complex logic of transcription factors driving the change in accessibility. After PGE2 treatment we detect increased transcription factor binding at sites with CRE and E-box motifs and reduced binding at sites with AP1 motifs, overcoming the changes with age. We find that short-term exposure of aged MuSCs to PGE2 augments their long-term regenerative capacity upon transplantation. Strikingly, PGE2 injections following myotoxin- or exercise-induced injury overcome the aged niche leading to enhanced regenerative function of endogenous tissue-resident MuSCs and an increase in strength.

eTOC Blurb

Wang et al. uncovers that aged muscle stem cells have blunted PGE2-EP4 signaling that is critical for their regenerative function. Treatment with exogenous PGE2 reverses age-related defects in muscle stem cells and imparts an epigenetic response that enhances their capacity for repair after injury and transplantation.

Graphical Abstract

Introduction

With aging, there is a progressive loss of muscle mass and strength that is exacerbated by impaired function of the stem cells required for muscle maintenance and muscle repair in response to exercise or injury. Sarcopenia, the drastic loss of skeletal muscle function that afflicts the elderly, has now been recognized as a disease by the World Health Organization¹. People with sarcopenia are at an increased risk for other disabling conditions, such as osteoporosis, heart failure, and cognitive decline. Despite the great need, there are no approved therapies for countering the stem cell dysfunction in regeneration, myofiber atrophy and loss of strength that plague the elderly.

Muscles benefit from tissue-resident, lineage-committed adult stem cells (MuSCs), that are poised to respond to injuries and regenerate skeletal muscle throughout life². During aging, MuSCs progressively acquire intrinsic defects which leads to a decline in the proportion of functional cells that can regenerate and repair the muscle^3–9. This is compounded by changes in the aged microenvironment, which drive altered receptor expression and signaling in MuSCs, causing further dysfunction^2,8,10–13. Consequently, aged MuSCs display reduced self-renewal, mitotic defects, and become senescent. This decreases the production of myogenic progenitors needed to regenerate muscle fibers^7,14,15. Overcoming the dysfunction of MuSCs with aging has the potential to accelerate recovery after injury¹⁶, and thus represents a potential regenerative approach to counter sarcopenia.

We recently identified PGE2, a membrane lipid derived metabolite as a component of the body’s natural acute response to muscle injury¹⁷. A transient surge in PGE2 signaling is required for MuSC function in young mouse muscle regeneration. MuSCs lacking the prostaglandin receptor EP4 (EP4-cKO) proliferate minimally in culture and fail to engraft upon transplantation. Moreover, muscle repair is delayed in EP4-cKO mice or after treatment of wild type mice with NSAIDs that block PGE2 synthesis, highlighting an essential role for PGE2 signaling in MuSC-mediated regeneration.

With aging, PGE2 levels decline in skeletal muscle due to increased expression of the catabolic enzyme 15-hydroxylprostaglandin dehydrogenase, 15-PGDH¹⁸, which we designated a gerozyme¹⁹. Inhibition of gerozyme activity restores PGE2 levels and increases mitochondrial biogenesis, metabolism, and autophagy to restore proteostasis in aged muscle fibers. In addition, PGE2 signals to motor neurons and facilitates the repair of neuromuscular junctions that become dysregulated in aged muscles^19,20. Yet, how deficient PGE2 signaling with aging contributes to MuSC dysfunction in the regeneration of aged muscles remains unresolved.

Here we uncover that PGE2 treatment overcomes cell intrinsic defects of aged MuSCs and augments their regenerative function. Strikingly, a single exposure of aged MuSCs to PGE2 enhances the longterm self-renewal capacity of either transplanted aged MuSCs or tissue resident endogenous MuSCs, overcoming a reduction in EP4 signaling that occurs with aging. Notably, a transient treatment with PGE2 during a downhill running exercise regimen leads to a robust increase in muscle strength. Using single cell transcriptomics, we reveal that aged MuSCs enter an altered trajectory that favors direct commitment over cell cycle. Further interrogation of the epigenetic changes imparted by PGE2 using neural network (BPNet) analysis of ATAC-seq data reveals an enhancement of CREB binding and reversal of age-related changes in chromatin accessibility at AP1 and E-box motifs. The transcriptomic response of aged MuSCs to PGE2 is in agreement with timelapse microscopy tracking of cell fate, which shows a striking increase in cell cycle re-entry and decrease in cell death. Together our data demonstrate that PGE2 has therapeutic potential, as it promotes the multiomic rejuvenation of aged MuSCs by epigenetic and transcriptomic restoration of stem cell viability, proliferative and regenerative capacity.

Results

PGE2 signaling maintains muscle mass and strength throughout life

Previous studies established that muscle injury triggers a surge of PGE2 which signals through the EP4 receptor and is essential for MuSC function during regeneration¹⁷. In aging, both PGE2 levels are decreased¹⁸. Thus, we sought to determine if the loss of PGE2-EP4 signaling in MuSCs impacts muscle during aging.

To assess the role of PGE2-EP4 signaling in MuSCs and muscle during the aging process, we employed a Pax7^creERT2;EP4^f/f mouse model to delete EP4 receptors in MuSCs of young mice (2 mo) and aged these mice for 14 months (Figure 1A). The EP4 cKO mice exhibited a significant ~20% reduction in strength via plantar flexion torque compared to littermate controls (Figure 1B). In accordance, mass of both tibialis anterior (TA) and gastrocnemius (GA) muscles in EP4 cKO were significantly reduced (Figure 1C). These results indicate that deletion of EP4 in MuSCs leads to an accelerated aging phenotype characterized by muscle atrophy and a striking loss of strength. Since EP4^−/− MuSCs contribute to myofibers during this aging experiment, the reduced strength and mass likely results from diminished PGE2 signaling in both MuSCs and myofibers.

Figure 1. PGE2 signaling is necessary and sufficient for the maintenance of muscle mass and strength in aging.

(A-C) Pax7-specific EP4 conditional knockout mice (Pax7^CreERT2;EP4^fl/fl, EP4-cKO) or control littermates (Pax7^CreERT2;EP4^+/+) were treated with tamoxifen (TAM) at 8 weeks of age and assayed for plantar flexion tetanic force at 14 months of age; (n=8 males per condition). (A) Experimental scheme. (B) Plantar flexion tetanic torque. (C) Tibialis anterior (TA) muscle mass and Gastrocnemius (GA) muscle mass. (D-G) Muscle contractile force after downhill treadmill run; (n=10 for vehicle or n=8 for dmPGE2 treated, with 5 technical replicates each). (D) Experimental scheme. (E) Representative twitch force (upper) and tetanic force (lower). (F) Specific muscle twitch force and (G) specific muscle tetanic force. Mann-Whitney test (B, C, F, G). *P<0.05. Means+s.e.m. Abbreviation: mo, months; EP4, Ptger4.

Physical exercise is a major benefit to health^21,22 and triggers muscle remodeling through hypertrophy and MuSC-mediated repair^23–25. To assess whether PGE2 can overcome age-related deficits in exercise-mediated muscle remodeling, we treated aged mice with a non-hydrolyzable PGE2 analog (16,16-dimethyl PGE2; dmPGE2) for 5 days and subjected them to daily bouts of downhill running, a form of eccentric exercise (Figure 1D). DmPGE2 is resistant to degradation by increased 15-PGDH activity in aged muscle¹⁸. Strikingly, two weeks after the exercise training regimen, we found that both specific twitch and tetanic forces were significantly increased in dmPGE2-treated GA muscles compared to vehicle-treated controls (Figure 1E-F). These results suggest that a short-term dmPGE2 treatment, concurrent with exercise, overcomes the aged muscle microenvironment and confers a long-term increase in muscle strength.

An acute exposure to PGE2 augments regenerative function of aged MuSCs

PGE2 signaling has diverse effects on MuSCs¹⁷, myofibers¹⁸, and motor neuron regeneration¹⁹. To identify how PGE2 impacts aged MuSC function, we challenged endogenous aged murine MuSCs with injury and transplantation.

To determine if an acute exposure to PGE2 can lead to improved function of endogenous MuSCs within the aged muscle tissue, we delivered a single bolus of dmPGE2 into the TA muscles of aged mice two days after notexin injury (Figure 2A), timed to emulate the natural surge of PGE2 after muscle injury in young mice^17,26. Immunofluorescence assessment of regenerated muscles two weeks after injury revealed a significant increase in Pax7⁺ aged MuSC numbers (Figure 2B,C) and an increase in myofiber cross sectional area (Figure 2D,E) following dmPGE2 treatment. Remarkably, a single dmPGE2 treatment led to a robust regenerative response that culminated in increased muscle mass and strength in male and female aged mice (Figure 2F-H, S1A-B). Together these data indicate that intramuscular dmPGE2 delivery promotes the expansion of endogenous stem cells culminating in increased myofiber size without depleting the stem cell pool. DmPGE2 has a half-life of 4.5 mins in vivo²⁷, which suggests that the effect of the treatment is propagated to MuSC progeny long after PGE2 has been metabolized.

Figure 2. PGE2 enhances the regenerative and transplantation capacity of aged MuSCs in vivo.

(A-G) Muscle regeneration of hindlimb muscles in aged mice (>25 months) treated with vehicle or dmPGE2 24h post-cardiotoxin (CTX) injury. (A) Experimental scheme. (B) Representative Tibialis anterior (TA) cross-section. DAPI, blue; LAMININ, green; PAX7, red. Arrowheads indicate PAX7⁺ MuSCs. Bar=40 μm. (C) PAX7⁺ muscle stem cells per 100 fibers. (D) Myofiber cross-sectional areas (CSA) in vehicle and dmPGE2 treated TAs. (E) Distribution of small (<1,000 μm²) and large (>1,000 μm²) myofibers. (F) Specific tetanic force. (G) Strength (Torque) and (H) tibialis anterior (TA) muscle mass, 2 weeks post-injury (n=4–7 mice per sex and treatment group). (I, J) Engraftment of GFP/luc-labeled aged (18–20 months) MuSCs (250 cells) coinjected with vehicle or dmPGE2 into TA muscles of young mice; Notexin injury was performed 4 weeks post-transplant; (n=3 mice per condition). (I) Transplantation scheme. (J) BLI signals post-transplant expressed as average radiance (p s⁻¹ cm⁻² sr⁻¹) (left). Representative BLI images for each condition (right). Bar=5 mm. Data are representative of two independent experiments. Mann-Whitney test (C, E). Paired t-test (G, H); ANOVA test for group comparison endpoint difference by Fisher’s test (J). *P<0.05, **P<0.001, ****P<0.0001. Means+s.e.m.

In a second experimental paradigm, we tested whether PGE2 could improve the function of aged MuSCs, independent of the aged microenvironment^8,10–13. We treated freshly-isolated aged MuSCs with dmPGE2 and injected them into injured muscles of young recipient mice. To monitor engraftment non-invasively over the timecourse of an experiment by bioluminescence imaging (BLI), we labeled MuSCs with a lentivirus encoding luciferase (Figure 2I). Recipient muscles were challenged with a subsequent notexin (NTX) injury 4 weeks after transplantation to assess the self-renewal and proliferative capacity of engrafted MuSCs. We observed significantly higher BLI signal from PGE2-treated aged MuSCs relative to controls that persisted for a month and further increased in response to injury (Figure 2J). This suggests that PGE2-treated aged MuSCs display enhanced long-term engraftment (7 weeks) by contributing to both the stem cell pool and to myofibers. PGE2-treated aged MuSCs show an engraftment efficiency comparable to vehicle-treated young MuSCs (Figure S1C), but is surpassed by PGE2-treated young MuSCs. Notably, the effect of PGE2 requires the EP4 receptor, as the increase in engraftment was abrogated when EP4 was ablated from MuSCs (Figure S1D). Together, these data show that acute treatment of aged MuSCs with PGE2 overcomes intrinsic defects that prevent proper MuSC expansion in response to injury. Moreover, these results show that a transient exposure to PGE2 has a long-term effect that is retained by generations of progeny with ameliorated regenerative capacity.

Aged MuSCs exhibit reduced PGE2-EP4 signaling

A remarkable ~70% of aged MuSCs are dysfunctional and fail to contribute to muscle regeneration in transplantation^4,8, we hypothesized that the reduced levels of PGE2 present in aged muscles^18,19 could lead to blunted EP4 signaling in MuSCs and alter their response to injury.

To assess whether PGE2-EP4 signaling in MuSCs is perturbed with aging, we isolated myofibers with their associated MuSCs from young (2–4 mo) and aged mice (>18 mo). While nearly all young Pax7+ MuSCs expressed the PGE2 receptor EP4, only 70% of aged MuSCs expressed EP4 (Figure 3A, S1E). Moreover, the levels of EP4 receptor expression in EP4+ aged MuSCs were reduced by ~50% compared to young (Figure 3A). Alternative PGE2 receptors do not compensate for the reduced EP4 levels on aged MuSCs (Figure S1F). In accordance to reduced EP4 expression and levels of PGE2 in aged muscle^18,19, phosphorylation of the cyclic-AMP response element binding protein CREB downstream of EP4 was nearly absent in aged MuSCs (Figure 3B). Overall, these findings suggest that the compounding effect of reduced PGE2 levels and lower EP4 expression lead to a significant impairment of CREB signaling in aged MuSCs.

Figure 3. Aged MuSCs display reduced EP4 signaling and altered regeneration trajectory.

(A) Representative confocal images of uninjured/resting EDL myofibers of young (2 months) and aged (25 months) mice (n>30 myofibers from n=3 mice) showing EP4 receptor immunostaining in Pax7⁺ MuSCs (left). Scale bars: 20 μm. Percentage of EP4⁺ MuSCs and quantification of EP4 intensity in single MuSCs (right). (B) Representative confocal images of uninjured resting EDL myofibers (n >30 myofibers from n=3 mice) showing phospho-CREB (pCREB) immunostaining in young and aged Pax7⁺ MuSCs (left). Scale bars: 20 μm. Quantification of pCREB intensity in single MuSCs (right). *P<0.05, **P<0.001, ****P<0.0001. Mann-Whitney test (A, B) Means±s.e.m.

PGE2 signaling primes the MuSC response to injury

To interrogate the dysfunctional response of aged MuSCs during regeneration and whether PGE2 signaling plays a role, we performed single cell transcriptome profiling²⁸ of young and aged MuSCs in response to muscle injury. MuSCs and myogenic progenitors were isolated from uninjured young adult (8 weeks) and aged (>18 months) mice at 3, 6, and 10 days after notexin-induced injury (Figure 4A). Myogenic cells were clustered and annotated according to expression of hallmark genes of myogenic states (MuSC, cell cycle, commitment, fusion, and differentiation) and myogenic regulatory factors (MRFs: Myod1, Myog, Myf5, Myf6) (Figure 4B-C, S2A-D). Notably, the larger diameter microfluidic channels in Drop-seq captured myocytes and regenerating myotubes that are rarely captured by methods due to their size.

Figure 4. Single cell transcriptomic analysis of young and aged MuSCs isolated regenerating muscle.

(A) Experimental scheme for MuSC isolation after injury. α7-integrin⁺ myogenic cells from TA and GA muscles of young (8wk) or aged (>18 mo) mice 0, 3, 6, 10 days post-notexin injured and analyzed by droplet-based single cell RNA sequencing. (B) Clustering of myogenic cells into 7 states, annotated according to myogenic gene expression; UMAP coordinates of 16,502 cells. (C) Relative expression of MuSC, cell cycle, commitment, fusion and myofiber genes in each cluster. (D) Schematic and relative expression of PGE2-EP4 signaling components in each cluster. (E) Bubble plot of EP4 receptor (Ptger4) expression in young and aged MuSCs. (F) Trajectory inference and pseudotime estimation of myogenic cells. Color scale represents the relative pseudotime of each cell. Expression of (G) cell cycle genes, and (H) PGE2-response genes in cycling and direct commitment trajectories across pseudotime. (I) Proportion of aged myogenic cells relative to total cells across pseuodotime trajectories.

We analyzed the expression patterns of the EP4 receptor and its effectors to gain insights into the activation of PGE2-EP4 signaling in myogenic cells during regeneration (Figure 4D). MuSCs expressed the highest levels of EP4 (Ptger4), which decreased in cycling MuSCs and committed myoblasts, suggesting that PGE2 signals occur in stem cells and its effects are propagated to their cycling progeny. Moreover, aged MuSCs expressed lower levels of EP4 mRNA than young MuSCs (Figure 4F). These transcriptomic changes were consistent with the reduced EP4 protein expression on committed progenitors relative to MuSCs in culture (Figure S2E) and on aged MuSCs relative to young on isolated myofibers (Figure 3A). An analysis of downstream components of PGE2 signaling revealed that membrane-bound proteins responsible for signal transduction including EP4, its associated G_αs-protein (Gnas) and members of the adenylyl cyclase complex (Adcy2/6/7), cAMP responsive cytosolic mediators such as protein kinase A (Prkacb) and CREB (Creb1) were enriched in cycling MuSCs (Figure 4G). Putative target genes of CREB such as Odc1, Spp1, and Nurr1 (Nr4a2), were enriched in myoblasts. This transcriptional signature suggests that MuSCs are primed for PGE2 signal transduction, and that active PGE2-EP4-cAMP signaling is characteristic of both cycling MuSCs and proliferative myoblasts.

Single cell profiling reveals an altered aged MuSC trajectory in response to injury

We hypothesized that diminished PGE2 signaling in aged MuSCs would alter their fate during regeneration after muscle injury. In support, clustering analysis of single cell RNAseq data detected two distinct MyoG-expressing myocyte populations (Figure 4B, S4A). The Myocyte 1 cluster expressed genes characteristic of cytoskeletal remodeling, characteristic of myotube formation (Figure S3A-C). The Myocyte 2 cluster expressed the cell fusion proteins (Mymk, Mymx) but also AP1 family of transcription factors (Jun and Fos), which are expressed during exit of MuSCs from quiescence²⁹, and genes involved in cell death and apoptosis (Figure S3C). The Myocyte 2 population was enriched in aged samples, suggesting that aged MuSCs produce intrinsically altered commited progeny (Figure S3D).

Trajectory analysis further suggested that MuSCs can either enter the cell cycle, proliferate as myoblasts, and then commit as Myocytes 1 (cycling trajectory), or directly transition into the Myocyte 2 state (direct commitment trajectory) (Figure 4F, S3E). In both trajectories, MuSCs respond to injury by downregulation of quiescence genes (e.g. Pax7 and CD34, etc.) and eventually differentiate by upregulation of myogenic genes (e.g. Myod1, Myog, etc.) (Figure S3F). Cell cycle-related genes including cyclins (cyclin D1, Ccnd1) and proliferation markers (Mki67) were expressed only in the cycling trajectory (Figure 4G). In accordance with our prior findings that PGE2 stimulates proliferation and expansion in young MuSCs¹⁷, we observed that CREB1 and the PGE2 target gene Odc1 are enriched in the cycling trajectory (Figure 3J), suggesting that PGE2 signaling selectively promotes the cycling trajectory.

The relative proportions of myogenic subsets are skewed in aged compared to young MuSCs (Figure 4H). While there were similar proportions of young and aged quiescent MuSCs, progression of aged MuSCs stalled at early pseudotime stages and were biased toward the direct commitment trajectory as they progressed toward later pseudotime points. Fewer cycling cells (myoblasts and their progeny) were found in aged samples, suggesting that the capacity of aged MuSCs to enter the cell cycle are restricted (Figure S3G). These results point to diminished cell cycle capacity as a major culprit in the regenerative dysfunction of aged MuSCs.

These data provide a molecular basis for the previously observed reduced regenerative response of aged MuSCs^7,14,15. They suggest that intrinsic changes in aged MuSCs predispose them to adopt a direct commitment trajectory that bypasses the cell cycle and culminates in a reduction in the production of transiently amplifying myoblasts. Notably, PGE2 signaling genes are enriched in the cycling trajectory characteristic of young MuSCs. In contrast, aged MuSCs sustain expression of AP1 transcription factors, which are normally downregulated during cell cycle entry.

PGE2 treatment restores EP4 signaling in aged MuSCs

We postulated that exposure of aged MuSCs to PGE2 could overcome their intrinsic decifits and drive a molecular program that favors a cycling trajectory. To test this, we exposed aged MuSCs to PGE2 for 24 h on single isolated myofibers and found that EP4 expression was restored to levels found on young MuSCs (Figure 5A). Indeed, this 1 day transient exposure of aged MuSCs to PGE2 led to the persistent expression of its receptor in culture for 7 days (Figure S3H). In accordance, PGE2 elicited an increase in CREB phosphorylation in aged MuSCs which was blocked by the EP4 antagonist ONO-AE3–208 (ONO) (Figure 5B).

Figure 5. PGE2 reverses age-related epigenetic changes in MuSCs.

(A) Representative confocal images of 24 hr PGE2 or vehicle treated EDL myofibers of aged (25 months) mice showing EP4 receptor staining in Pax7⁺ MuSCs (left). Scale bars: 20 μm. Percentage of EP4⁺ MuSCs after 24 hr PGE2 or vehicle on EDL myofibers (n >30 myofibers from n=3 mice; right). (B) Representative microscopy images of 2 hr vehicle, PGE2, ONO, or PGE2+ONO treated MuSCs from aged (25 months) mice showing phosphorylated CREB (p-CREB) staining in Pax7⁺ MuSCs (left). Scale bars: 20 μm. Percentage of p-CREB⁺ MuSCs after 2 hr treatment in culture (N=3 animals; right). (C) Experimental scheme for ATAC-seq analysis and ChromBPNet modeling of cis regulation. (D) Neural network identification of combinatorial transcription factor grammar. (E) Motif analysis from ATAC-seq of MuSC from young mice (2–4 mo) or aged mice (24 mo) treated with veh or PGE2 for 2h (left). Consensus motifs enriched in differentially accessible peaks of young and aged MuSCs (right). (F) Clustering of differentially accessible peak reveals combinatorial regulation associated with transcription factor motifs in aged MuSCs treated with PGE2. Average number of archetypal motifs of each peak cluster (left) and fold change in accessibility for peaks within each cluster (right). (E) Predicted effect of PGE2 treatment on transcription factor-mediated chromatin accessibility on aged MuSCs. *P<0.05, **P<0.001. Mann-Whitney test (A, B); Means±s.e.m.

ChromBPNet reveals epigenetic mechanism elicited by PGE2 in aged MuSCs

We reasoned that a “molecular memory”, or change in the chromatin landscape that is propagated to MuSC progeny, underlies the change in regenerative function observed for weeks following transient exposure of aged MuSCs to PGE2. To test this hypothesis we employed neural network analysis of ATAC-seq data obtained two hours after PGE2 treatment of freshly isolated aged MuSCs and compared them to young and aged vehicle-treated MuSCs (Figure 5C). ATAC-seq probes the accessibility of chromatin and generates a high-resolution genome wide map of cis regulatory elements, that are bound by transcription factors to determine gene expression^30–32. To gain insights into the combinatorial cis regulation by transcription factors, we trained ChromBPNet models to predict chomatin accessibility as a function of the underlying genomic sequence³².

ChromBPNet is a dual-headed convolutional neural network model with a locked branch trained on GC-matched non-regulatory genomic elements to denoise bias introduced by Tn5 transposase in ATAC-seq (Figure S4A) which allows the main branch to learn the cellular transcription factor logic that determines DNA accessibility. Using a trained ChromBPNet model, we were able to evaluate the contribution of single base pairs to local accessibility and reveal sequence motifs that correspond to transcription factor (TF) binding sites that act in cis to regulate any given gene (Figure 5C, S4B).

This approach overcomes the limitations of antibodies and the imprecise nature of ChIP-seq. We first validated ChromBPNet against known cis-regulatory elements in genes with a role in MuSC specification and commitment. Our model accurately identified a combination paired- and homeo-box motif that drives accessibility of the −57kb regulatory element upstream of Myf5 (Figure S4C), which exactly matched the Pax3/7 binding site revealed by ChIP-seq and mobility shift assay results³³. Moreover, our model confirmed the contribution of two “GC”-centered E-box motifs (“CAGCTG”, Ebox-GC) bound by MRFs, as well as the AP1 motif that are known regulators of the Myod1 core enhancer (Figure S4D)³⁴. Furthermore, ChromBPNet was more precise in mapping TF binding sites than ChIP-seq. Whereas ChIP-seq for CTCF sites in MuSCs yielded large >200bp peaks³⁵, our model pinpointed the exact CTCF site(s) and revealed combinatorial binding with other TFs (Figure S4E). Thus, trained ChromBPNet neural network models can accurately predict binding sites of a broad range of TFs and combinatorial TF interactions.

PGE2 rejuvenates the chromatin landscape in aged MuSCs

We trained ChromBPNet models on ATAC-seq data from young, aged, and PGE2-treated aged MuSCs and interrogated the combined syntax of TF binding sites at genomic loci that change accessibility (Figure 5D, S4B). In addition, we created a motif finding algorithm that matches genomic sequences with high contribution scores from ChromBPNet with a compendium of known archetypal TF motifs³⁶, and generated genome-wide annotations of TF binding sites for each MuSC dataset. Specifically, we focused on regions of the genome that are differentially accessible in aging compared to young or following PGE2 treatment in order to decipher the regulation of TF activities and syntax.

We found that genomic regions that were more accessible (open) in aged MuSCs were enriched in AP1 “TGAGTCA” and C/EBP “TTGCGCAA” consensus motifs (Figure 5E). In contrast, genomic regions that were less accessible (closed) in aged MuSCs were enriched in “GC”-centered E-boxes. This finding suggested that with aging, the activity of MRFs of the MyoD family, which are also necessary for MuSC expansion during injury³⁷, is blunted in early activated MuSCs, whereas AP1 and C/EBP transcription factor activity are increased.

To understand this TF co-regulation, we clustered accessible cis regulatory sites by the combination of TFs that contribute to their accessibility (Figure 5D), which revealed that PGE2-mediated increases and decreases in chromatin accessibility are explained by distinct TF syntaxes. Our analysis uncovered cAMP response element motifs (“TGACG” or “TGACGTCA”), that pCREB binds to, at sites that increase accessibility upon PGE2 treatment in young and aged MuSCs (Figure 5F, S5A-C). In contrast to CREB motifs, we found that peaks with one or more AP1 motif exhibited decreased accessibility upon PGE2 treatment in both young and aged MuSCs. A subset of AP1-containing chromatin sites that close with PGE2 treatment also contained nuclear factor 1 (NFI) motifs, suggesting co-regulation by AP1 and NFI in MuSCs similar to previous reports in astrocytes³⁸. Some PGE2-mediated changes in accessibility were only seen aged MuSCs, for instance closing of sites bound by both AP1 and CTCF (Figure 5F, S5D) and opening at sites associated with a “GC” E-box (to which MRFs typically bind) and “CC” E-box sites (to which the SNAI family of TFs bind and suppress MyoD activation of differentiation gene³³).

PGE2 suppresses AP1 activation in aged MuSCs

Our analyses identified AP1 as a precociously activated TF in aging that is suppressed by PGE2 treatment. Transient activation of AP1 licenses MuSCs for cell cycle reentry²⁹. However, it is also activated in response to cellular stress, tissue damage and inflammation^39,40. Sustained AP1 (Jun, Fos) expression was a hallmark of the direct commitment trajectory (Figure 6A, S6A), suggesting that temporal regulation of the AP1 family of transcription factors could control the trajectory taken by MuSCs. Single cell RNA-sequencing analysis of MuSCs from young and aged human muscle biopsies⁴¹ revealed that the increase in JUN and FOS with aging is conserved across species (Figure 6B). Consistent with our findings in mouse, AP1 family members JUN, JUNB, FOS and FOSB displayed a transient activation from quiescent to a primed state in MuSCs from young and aged human donors. However, the decrease in AP1 family member abundance upon differentiation was observed only in young and did not occur in aged MuSCs. Thus, persistent AP1 activation appears to be a characteristic of the precocious commitment trajectory that is conserved in aged MuSCs across species.

Figure 6. PGE2 suppresses age-related AP1 activation and rejuvenates the transcriptome.

(A) Expression of AP1 genes (Jun and Fos) in cycling and direct commitment trajectories across pseudotime trajectories. (B) Expression of AP1 family genes in young and aged human MuSCs. Data analyzed from Lai et al. 2024. (C) Representative microscopy images of 2 hr vehicle, PGE2, ONO, or PGE2+ONO treated MuSCs from aged (25 months) mice showing phosphorylated JUN (p-JUN) staining in Pax7⁺ MuSCs (left). Scale bars: 20 μm. Percentage of p-JUN⁺ MuSCs after 2 hr treatment in culture (N=3 animals; right). (D) Experimental scheme for RNA-seq analysis. (E) Overlap of differentially expressed genes in aging and PGE2-treatment in aged MuSCs. (F) Expression of rejuvenated genes (overlap from E) in veh. and PGE2-treated young and aged MuSCs. (G) Ingenuity pathway analysis (IPA) of skeletal muscle function enrichment from gene set in E. (H) Fold change in Jun family members (Jun and Jund) in aged MuSCs after 24 hr of PGE2 treatment. (I) Fold change in genes clusters enriched in either cycling (cluster 1) or direct differentiation (clusters 2–6) in aged MuSCs after 24 hr of PGE2 treatment. Gene clusters from Figure S6B. (J) Fold change in key genes of the necroptosis pathway in aged MuSCs after 24 hr of PGE2 treatment. (K) ChromBPNet importance scores identifies an AP1 binding site as a driver of accessibility at the proximal promoter of Mlkl in MuSCs.

We postulated that the rapid suppression of AP1 activity by PGE2 is regulated at the post-translational level. Phosphorylation of JUN (p-JUN) at serine 63 by Jun N-terminal Kinase (JNK) in response to stress signals is known to increase its transcriptional activity⁴². When we quantified the phosphorylation status of JUN in aged MuSCs after PGE2 treatment by immunostaining (Figure 6C), we found that the proportion of aged MuSCs with p-JUN is reduced after 2h of PGE2 treatment. Moreover, the reduction in p-JUN after PGE2 treatment of aged MuSCs was blocked by ONO, an antagonist of the EP4 receptor, suggesting that a key outcome of PGE2-mediated signaling is the suppression of AP1 activity.

PGE2 elicits a rejuvenated transcriptional response

We sought to determine if a short PGE2 treatment could alter the transcriptomic response of aged MuSCsRNA sequencing (RNA-seq) of young and aged MuSCs treated with vehicle or PGE2 for 24 hours on hydrogels that maintain stemness⁴³ (Figure 6D). A comparison of the young and aged vehicle-treated MuSCs revealed 511 differentially expressed genes (Figure 6E). A comparison of PGE2 and vehicle-treated aged MuSCs revealed 556 differentially expressed genes. Of PGE2 responsive genes, 95 genes (~19% of differentially expressed aging genes) exhibited expression levels similar to young (Figure 6E-F, Supplementary Table S1). Ingenuity Pathway Analysis (IPA) analysis of this PGE2-responsive, rejuvenation gene signature revealed strong enrichment of cell survival genes that were upregulated, primarily encoded advantageous properties such as metabolism, migration, motility and negative regulators of cell death (Figure 6G, Supplementary Table S2).

We hypothesized that AP1 family members would be downregulated in response to PGE2 treatment. Jun and Jund were increased by 1.5-fold and 5-fold, respectively, in aged MuSCs relative to young. Notably, this increase was reversed by PGE2 treatment for 24 hr in culture on hydrogels (Figure 6H). To resolve the role of these changes on MuSC fate choice, we analyzed the transcriptomic changes of PGE2-treated aged MuSCs for gene sets enriched in either of the two trajectories, cycling and direct commitment (Figure S6B). We found that in PGE2-treated aged MuSCs, there was a significant increase in the expression of genes characteristic of the cycling trajectory (Figure 6I, Supplementary Table S3).

IPA analysis led us to interrogate the genes downregulated by PGE2 for GO terms associated with programmed cell death and necroptosis (Figures 6G). We found that PGE2 treatment of aged MuSCs suppressed the expression of key necroptosis regulators Mlkl, Rbck1, Casp8, and Cav1 (Figures 6J). ChromBPNet further linked AP1 to the cis-regulation and activation of the necroptotic gene Mlkl (Figure 6K). The AP1 motif located in the proximal promoter (~300 bp upstream) of Mlkl exhibited high contribution scores, suggesting that AP1 drives the accessibility of this site in aged MuSCs, but is reduced by PGE2 treatment.

These results reveal that PGE2 induces a rejuvenated gene expression signature in aged MuSCs. Notably, they show that PGE2 acts as a transcriptional repressor of AP1 and necroptosis genes, and restores of the cycling trajectory.

PGE2 treatment promotes aged MuSC proliferation and survival

We reasoned that the epigenetic and transcriptomic changes would result in a phenotypic response in vitro. Specifically, we hypothesized that that PGE2 treatment of aged MuSCs would stimulate the cell cycle trajectory and overcome intrinsic deficits. To test this, aged MuSCs were prospectively isolated by FACS and then treated with PGE2 for 24 hr. PGE2-treated aged MuSCs exhibited a robust proliferative response and increased 3-fold in cell number compared to vehicletreated controls (Figure 7A).

Figure 7. PGE2 treatment promotes expansion of cultured aged MuSCs in live cell timelapse imaging.

(A) Quantification of MuSC numbers in aged mice (18–20 months) after 24 hr treatment with vehicle or PGE2 (10 ng/ml), and subsequent culture on hydrogel until day 7 (n=15 mice in 5 independent experiments). (B) Representative trajectories of an aged (18–20 months) MuSC clone treated with vehicle (top) or PGE2 (bottom) in a hydrogel microwell by time-lapse microscopy for 48 hr. (C) Live cell counts in aged MuSC (18–20 months) clones tracked by time-lapse microscopy after vehicle (left, n=32 clones) and PGE2 treatment (right, n=45 clones). (D) Time to first cell division calculated from the tracked aged clones in (C) in comparing to MuSC clones isolated from young mice (2 months). (E) Quantification of young and aged MuSCs treated with vehicle (n=4) or dmPGE2 (n=3) after 7 days of culture. (F) Genealogy tracing of a single representative clone from vehicle or PGE2-treated condition. Red arrowheads indicate mitotic cell death events with both daughter cells dying within 2 hr of division. (G) Dead cell counts in aged MuSC (18–20 months) clones tracked by time-lapse microscopy after vehicle (left, n=32 clones) and PGE2 treatment (right, n=45 clones). (H) Proportions of clones and mitotic events that result in cell death events within 2 hr of division. (I) Flow cytometry analysis of the percent necrotic or dead (Annexin V^- or Annexin V⁺, 7AAD⁺) geriatric MuSCs (>24 months) after 24 hr treatment with vehicle or PGE2 and analyzed 7 days later after culture on hydrogels; n=12 mice performed in 3 independent experiments. (A-G) Data are representative of two independent experiments. *P<0.05, **P<0.001, ****P<0.0001. ANOVA test for group comparison endpoint difference by Fisher’s test (D). Mann-Whitney test (D). Fisher’s exact test (H). ANOVA test for group comparison with Bonferroni correction for multiple comparisons (I); (D, E, I) Means±s.e.m.

To gain insights into the effects of PGE2 on aged MuSCs, we performed clonal cell fate-tracking analysis. Using time-lapse microscopy, we tracked the cell division behavior and fate of individual PGE2 or vehicle-treated aged MuSCs in hydrogel microwells as they exited quiescence and proliferated over a 48-hr period. Cell cycle entry, division, and death events were recorded within cell lineages. Aged MuSCs treated with PGE2 exhibited a notable increase in cumulative cell numbers spanning 6 generations for the most robust clones (Figure 7B, C and S7A,B and Movies S1-2). The increase in proliferation was due, in part, to a markedly accelerated time to first division of PGE2-treated aged MuSCs compared to vehicle-treated aged MuSCs (Figure 7D). PGE2 has similar effects on young and aged MuSCs, increasing cell numbers by ~60%, which overcomes the deficit in proliferative capacity of aged MuSCs (Figure 7E).

Most striking was the marked reduction in overall cell death of aged MuSCs upon PGE2 treatment compared to vehicle-treated MuSCs (Figure 7F-H). Live cell tracking revealed that mitotic events in vehicle-treated aged MuSCs were followed by the synchronized deaths of both daughters (Figure 7F, arrowheads). Aged MuSCs displayed increasing rates of cell death with the number of generations (Figure 7G). The number of cell death events was blunted with PGE2 treatment, suggesting reduced cell cycle defects. We quantified the incidence of death of both daughters within 2 h of a division as a surrogate for mitotic catastrophe¹⁴ (Figure 7H). These events were never observed in young MuSC clones, but occurred in 36% of aged MuSC clones, accounting for 4.24% of mitotic events over the 6 generations. With PGE2 treatment, the rate of death in aged MuSCs decreased by ~3-fold, affecting 13% of clones in 1.25% of mitotic events.

We sought to test if the cAMP-PKA-CREB signaling pathway is required for the pro-survival effect of PGE2 that is specific to aged MuSCs. We assessed the cell death of aged MuSCs to determine if it resulted from apoptosis or necroptosis by flow cytometry quantification of Annexin V and 7AAD immunostaining, respectively. PGE2 treatment increased survival of aged MuSCs by significantly decreasing the proportion of 7AAD+ necrotic/dead cells, an effect that was mitigated in the presence of the Protein Kinase A (PKA) inhibitor, Rp-cAMPs (Figure 6I, S7C). These results confirm that PGE2 promotes MuSC survival via CREB signaling and suppression of cell death via necroptosis.

Discussion

We have discovered that the immunomodulator PGE2 induces epigenetic rejuvenation of aged MuSCs which leads to functional improvements in muscle repair and strength. Our previous data showed that PGE2 signaling through the EP4 receptor is necessary and sufficient for young MuSC function in regeneration¹⁷. In aged muscles, 70% of MuSCs are dysfunctional and fail to efficiently regenerate damaged muscle^4,8. We find that EP4 expression is decreased in aged MuSCs and its downstream effector, phosphorylated CREB, is also diminished. This loss of function is compounded by the loss of neuromuscular connectivity with aging, which leads to increased expression of the gerozyme 15-PGDH and consequent reduction in PGE2 levels^18,19. Here we show by single cell mapping of regenerating muscle that dysfunctional aged MuSCs bypass the cell cycle trajectory essential to stem cell expansion following injury and become prematurely committed. This could be due to division-independent differentiation⁴⁴ or senescence^45,46. We also find that aged MuSCs are diminished in number because they undergo necroptosis, in accordance with the increase in rates of mitotic catastrophe reported by others¹⁴. Strikingly, a short exposure to PGE2 overcomes the aged niche microenvironment and augments aged MuSC function via epigenetic and transcriptomic changes. PGE2 promotes cell cycle entry and suppresses cell death to increase the production of myogenic progenitors and meet the needs for regeneration. We demonstrate that restoration of PGE2 signaling in vivo augments engraftment of transplanted aged MuSCs, and promotes recovery from myotoxin-induced injury and eccentric exercise injury paradigms that challenge aged MuSCs, by bolstering their numbers and regenerative capacity.

Using single cell and multiomic approaches, we reveal the mechanism by which PGE2 ameliorates aged MuSC function. PGE2 restores the level of CREB phosphorylation to that seen in young and reduces the levels of active, phosphorylated JUN. These TF changes lead to decreased chromatin accessibility via AP1 interactions with NFI and CTCF and increased accessibility at motifs containing “GC”-centered E-boxes bound by MRFs. The suppression of AP1 is further reinforced by the transcriptomic downregulation of Jun mRNA, which points to the autoregulation of AP1 as a feed-forward mechanism in aging. PGE2 imparts a potent rejuvenating effect on the transcriptome, restoring 20% of aging-dysregulated genes to levels found in young MuSCs. Notably, these genes are not affected by PGE2 in young MuSCs. Increased MuSC death is another characteristic specific to aged MuSCs, which PGE2 overcomes. PGE2 promotes the survival of aged MuSCs by preventing mitototic catastrophe and necroptosis. Mitotic defects in aged MuSCs have been attributed to reduced Notch signaling¹⁴. Notch represses AP1-mediated transactivation in multiple cell types⁴⁷, suggesting that notch and PGE2-mediated suppression of mitotic catastrophe could converge via reduced AP1 activity. Thus, our findings suggest that in contrast to its effects on young MuSCs¹⁷, in aged MuSCs, PGE2 activates a multiomic molecular response that culminates not only in increased proliferation, but also increased survival.

Here we show that in response to injury, due either to a myotoxin or exercise, a short-term treatment regimen with a PGE2 analog enhances endogenous MuSC regenerative capacity, leading to increased muscle mass and strength in aged mice. These results provide evidence for yet another mechanism by which PGE2 restores muscle function. Our previous reports showed that restoration of neuromuscular innervation and muscle hypertrophy occurred in homeostasis^18,19 following sustained inhibition of the 15-PGDH gerozyme and elevation of PGE2 levels in aged mice. Functionally, we find that a short-term exposure to PGE2 has a profound long-term effect on the regenerative capacity of aged MuSCs that is maintained over a period of weeks. A single PGE2 injection into aged mice at the site of injury, leads to a pronounced increase in myofiber crosssectional area, muscle mass, and strength two weeks later. Moreover, if acute PGE2 treatment is combined with eccentric exercise, regeneration of aged muscles is bolstered, culminating in a dramatic gain of strength. These experiments show that rejuvenating the function of tissue-resident stem cells can have long-term benefits to the health status and function of aged tissues. Moreover, they suggest that treatments to boost PGE2 and muscle regeneration, can be combined with exercise, to promote recovery of the elderly after surgeries, falls and traumatic injuries.

Our regeneration and transplantation experiments suggest that the inflammatory metabolite, PGE2, imparts a “molecular memory” in aged MuSCs that is propagated to their progeny and is evident as enhanced self-renewal and enhanced production of myogenic progenitors. A single exposure to PGE2 facilitates the long-term engraftment of aged MuSCs, independent of their aged microenvironment. Moreover, engrafted PGE2-treated MuSCs retain their stem cell properties and are more responsive to a subsequent injury 1 month later. This molecular memory is reminiscent of the inflammatory memory described by Fuchs and colleagues that sensitizes skin epithelial stem cells and accelerates their response to a second assault, or damage to the skin barrier⁴⁸. In this paradigm, AP1 acts as an epigenetic bookmark that imparts cellular experiences of inflammation in skin stem cells long after the inflammation subsides⁴⁹. Our study suggests that changes in MuSC chromatin architecture underly the rejuvenating memory of PGE2, however, in aging, chromatin accesibility by AP1 is aberrant and persists. It is possible that over a lifetime, this bookmarking process leads to pervasive AP1 chromatin opening that hijacks sites that would otherwise be bound by cell identity transcription factors, as observed across multiple aging cell types⁵⁰. This is exemplified in muscle by the repression of “GC”-centered Ebox accessibility by the MyoD family of MRFs in aged MuSCS that we observe here. Our findings demonstrate that PGE2 reverses AP1-mediated epigenetic aging. We find that PGE2 treatment reduces accessibility at sites containing both AP1 and CTCF motifs, suggesting that PGE2 treatment could alter long range chromatin interactions mediated by CTCF which would explain the extensive changes in the transcriptomic signature. Notably, CREB and AP1 are both basic leucine zipper proteins that bind to “TGACTCA” and “TGACGTCA” motifs⁵¹. CREB family members compete with AP1 for binding⁵² and heterodimerization between CREB and Jun/Fos can alter their binding specificity⁵³. The loss of CREB phosphorylation in aged MuSCs can be attributed to the combination of reduced PGE2 level in aged muscle and downregulation of EP4 in MuSCs. CREB is a potent transcription factor that promotes MuSC expansion and selfrenewal^54,55. cAMP signaling can also lead to MuSCs quiescence via a PKA-YAP1 pathway^56,57, which may mediate the longterm engraftment of aged MuSCs after PGE2 treatment. Thus, the interplay between cAMP. CREB, and AP1 is critical for the activation and self-renewal of MuSCs and becomes dysregulated in aging.

To elucidate the basis for the molecular memory imparted by PGE2 in aged MuSCs, we utilized ChromBPNet, a potent neural network model that learns the cis regulatory logic from ATACseq and predicts chromatin accessibility at single base resolution. ChromBPNet models were able to identify transcription factor binding sites in ATAC-seq data, without prior knowledge of transcription factor motifs and with higher resolution than molecular assays using ChIP-seq. Importantly, interrogation of ChromBPNet models allowed us to identify combinatorial TF grammars in response to aging, and drug treatment. Application of this deep learning approach has broad implications for mechanistic studies and for drug discovery.

PGE2 has been shown to be a pote nt immunomodulator which can act on different stem cell types and promote regeneration^58–60. The aging immune system and the upregulation of the gerozyme 15-PGDH in multiple tissues leads to dysregulated PGE2 levels¹⁸. Here we show that despite the chronic inflammatory milieu typical of aged tissues, an acute treatment with dmPGE2, which is resistant to the gerozyme 15-PGDH in aged muscles and mimics the endogenous PGE2 surge seen in young muscle after injury, elicits a robust regenerative response in aged MuSCs. A short-term exposure to dmPGE2 markedly improves aged MuSC regenerative capacity and boosts recovery of strength, suggesting that transient treatments with small molecules that boost PGE2 levels can produce therapeutic effects and counter muscle aging. Moreover, PGE2 enhances stem function in young hematopoietic, liver and colon regeneration^58–60 and CREB activation has been implicated in the rejuvenation of cognitive functions in aged mice by young blood and in parabiosis experiments of young mice conjoined with aged mice⁶¹. Thus, further analysis of PGE2 signaling could reveal if, as in muscle, PGE2 can bolster the regenerative capacity and rejuvenate the functions of stem cells in other tissues with aging.

Limitations of the Study

Our study provides evidence for dysregulated PGE2-EP4 signaling in MuSC aging. These studies have been performed using aged mice. Further investigation of PGE2’s effect on the epigenome and function of aged MuSCs from human are needed to translate these findings for clinical applications. Moreover, use of advanced epigenomic editing tools that target specific CREB and AP1 binding sites identified here could uncover regulatory mechanisms and genes responsible for the improved aged MuSC function.

Resource availability

Lead contact:

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Helen M. Blau (hblau@stanford.edu).

Materials availability:

This study did not generate new, unique reagents.

Data and code availability:

Single-cell RNA, bulk RNA, ATAC sequencing data have been deposited at GEO as GSE145297, GSE191190 and are publicly available as of the date of publication.

All original code has been deposited on Github and is publicly available at doi.org/10.5281/zenodo.15377238 as of the date of publication.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mouse Strains

We performed all experiments and protocols in compliance with the institutional guidelines of Stanford University and Administrative Panel on Laboratory Animal Care (APLAC). Aged (18–20 mo. or >24 mo.) mice C57BL/6 were obtained the US National Institute on Aging (NIA) for aged muscle studies, and young (2–4 mo.) wild-type C57BL/6 mice from Jackson Laboratory. Mouse transgenic strains were purchased from The Jackson Laboratory (Pax7^CreERT2 No. 017763; EP4^flox/flox No. 028102). Double-transgenic Pax7^CreERT2;EP4^flox/flox (EP4^f/f) were generated as described previously¹⁷. For Pax7^CreERT2; EP4^f/f mice experiments, we treated 8-week-old male mice with five consecutive daily intraperitoneal injections of tamoxifen to activate Cre expression under the control of the Pax7 promoter. We validated these genotypes by appropriate PCR-based strategies. Studies were performed with female and male mice unless specified.

Muscle injury

We used an injury model entailing intramuscular injection of 10 μl of notexin (10 μg ml⁻¹; Latoxan) or cardiotoxin (10 μM; Latoxan) into the Tibialis anterior (TA) muscle. When indicated, 48 hr or 72 hr after injury either 16,16-Dimethyl Prostaglandin E2 (dmPGE2) (13 nmol, Tocris, catalog # 4027) or vehicle control (PBS) was injected into the TA muscle. The contralateral TA was used as an internal control, except for the force measurement experiments where each mouse had both legs injured with the same condition and independent mice were used for each condition.

For Pax7^CreERT2; EP4^flox/flox mice experiments, we treated mice with five consecutive daily intraperitoneal injections of tamoxifen to excise the EP4 allele in Pax7 expressing cells. As control mice, Pax7^+/+; EP4^flox/+ littermates of the same sex were used.

Mouse muscle stem cell isolation

We isolated and enriched muscle stem cells as previously described^4,17,43,62. Briefly, mouse hind-limb muscles were isolated and dissociated using the gentleMACS Octo Dissociator with a modified manufacturer protocol (Miltenyi Biotech). Dissociated muscle was digested with 0.2% collagenase type 2 (Worthington) for 60 min, followed by collagenase/dispase (0.04 U ml⁻¹; Thermo Fisher Scientific) digestion for 30 minutes. Mononucleated cells were liberated by syringe dissociation with an 18G needle. For mouse muscle stem cells, single cell suspension were incubated with biotinylated antibodies against CD11b (1:800), CD45 (1:500), Sca1 (1:200) and CD31 (1:200), followed by incubation with streptavidin magnetic beads (Miltenyi Biotech), streptavidin-APC-Cy7, α₇-integrin–PE (1:500) and CD34-eFluor660 (1:67). The cell mixture was depleted for hematopoietic lineage expressing and nonmuscle cells on a magnetic based selection column (Miltenyi) for biotin-positive cells. The remaining cell mixture was then subjected to FACS analysis to sort for CD45^-CD11b^-CD31^-Sca1^-CD34⁺integrin-α₇⁺ MuSCs with >95% purity (DIVA-Van, Becton-Dickinson). We generated and analyzed flow cytometry scatter plots using FlowJo v10.0. For wild-type MuSC sorts, we pooled together MuSCs (~5,000 each) from at least three independent age- and sex-matched donor mice. For Drop-seq analysis, magnetic activated cell sorting was used to enrich for lin- a7-integrin+ myogenic cells as previously described⁶³. Uninjured and notexin injured muscles were dissected, digested, and depleted of lineage-positive cells as described above for FACS methods. Flow through cells from magnetic selection lineage depletion were incubated with anti-α₇-integrin-magnetic beads (Miltenyi) and positively selected by magnetic selection column. Flow cytometric and Drop-seq analysis results show ~75% of the cells are myogenic. Cells from 2 mice of uninjured and injury time points were collected and pooled. 250,000 cells from each batch were used for five Drop-seq analysis (2 sets of young and 3 sets of aged samples).

Muscle stem cell transplantation

We transplanted 250 MuSCs into the Tibialis anterior (TA) muscles of recipient mice as previously described^4,17,43,62. We transplanted cells from aged mice (18–20 months) that were transduced with a luc-IRES-GFP lentivirus (GFP/luc virus) for a period of 24 hr before transplantation, as previously described⁴ (see below “Muscle stem cell culture, treatment and lentiviral infection” section for details). Prior to transplantation of muscle stem cells, we anesthetized Nod scid gamma (NSG) recipient mice with ketamine (2.4 mg per mouse) by intraperitoneal injection. We then irradiated hindlimbs with a single 12 Gy dose, with the rest of the body shielded in a lead jig. We performed transplantations within 2 d of irradiation. We resuspended either mouse MuSCs in 0.1% gelatin/PBS and then transplanted them (250 mouse MuSCs per TA) by intramuscular injection into the TA muscles in a 15 μl volume. We coinjected cells with 13 nmol of 16,16-Dimethyl Prostaglandin E2 (dmPGE2) (Tocris, catalog # 4027) or vehicle control (PBS). One month after transplant, we injected 10 μl of notexin (10 μg ml−1; Latoxan, France) to injure recipient muscles and to re-activate MuSCs in vivo. We compared cells from different conditions by transplantation into the TA muscles of contralateral legs in the same mice. Eight weeks after transplantation, mice were euthanized and the TAs were collected for analysis.

Bioluminescent Imaging

We performed bioluminescence imaging (BLI) using a Xenogen-100 system, as previously described^4,17,43,62. Briefly, we anesthetized mice using isofluorane inhalation and administered 120 μL D-luciferin (0.1 mmol kg⁻¹, reconstituted in PBS; Caliper LifeSciences) by intraperitoneal injection. We acquired BLI using a 60s exposure at F-stop=1.0 at 5 minutes after luciferin injection. Digital images were recorded and analyzed using Living Image software (Caliper LifeSciences). We analyzed images with a consistent region-of-interest (ROI) placed over each hindlimb to calculate a bioluminescence signal. We calculated a bioluminescence signal in radiance (p s⁻¹ cm⁻² sr⁻¹) value of 10⁴ to define an engraftment threshold. This radiance threshold of 10⁴ is approximately equivalent to the total flux threshold in p/s reported previously. This BLI threshold corresponds to the histological detection of one or more GFP+ myofibers. We performed BLI imaging every week after transplantation.

Hydrogel fabrication

We fabricated polyethylene glycol (PEG) hydrogels from PEG precursors, synthesized as described previously⁴³. Briefly, we produced hydrogels by using the published formulation to achieve 12-kPa (Young’s modulus) stiffness hydrogels in 1 mm thickness, which is the optimal condition for culturing MuSCs and maintaining stem cell fate in culture. We fabricated hydrogel microwell arrays of 12-kPa for clonal proliferation experiments, as described previously⁴³. We cut and adhered all hydrogels to cover the surface area of 12-well or 24-well culture plates.

Muscle stem cell culture, treatment and lentiviral infection

Following isolation, we resuspended MuSCs in myogenic cell culture medium containing DMEM/F10 (50:50), 20% FBS, 2.5 ng ml⁻¹ fibroblast growth factor-2 (FGF-2 also known as bFGF) and 1% penicillin-streptomycin. We seeded MuSC suspensions at a density of 500 cells per cm² surface area. We maintained cell cultures at 37 °C in 5% CO₂ and changed medium every other day. For PGE2, we added 10 ng/ml Prostaglandin E2 (Cayman Chemical) (unless specified in the figure legends, 10 ng/ml was the standard concentration used) to the MuSCs cultured on collagen coated dishes for the first 24h. The cells were then trypsinized and cells reseeded onto hydrogels for an additional 6 days of culture. All treatments were compared to their solvent (DMSO) vehicle control. We performed all MuSC culture assays and transplantations after 1 week of culture unless noted otherwise.

For pCREB and pJUN signaling analysis, freshly isolated MuSCs (as described above) were resuspended in MuSC media prepared with charcoal-stripped FBS, then briefly incubated in collagen-coated 96-well plate (Greiner Bio-One, Cat No. 655090) for 30 minutes to permit MuSC adhesion to the bottom of the wells. Once adhered, growth medium was replaced with MuSC growth medium spiked with either PGE2 (10ng/mL, Cayman Chemical), EP4 antagonist ONO-AE3–208 (1uM, Cayman chemical), both PGE2 and ONO (10ng/mL and 1uM, respectively), or vehicle (DMSO), incubated for 2hours (37°C, 5%CO₂), then fixed (0.5%PFA, 10 min). In side-by-side wells, MuSCs for a same age-treatment grouping were stained for either phospho-CREB (1:50, Cell Signal Technology, 9198S, Lot 19) or phospho-cJUN (1:50, Invitrogen, Cat No. MA5–27760, Lot Yl4047345E) and counterstained with DAPI. The abundance of phospho-CREB positive or phospho-cJUN positive MuSCs were normalized to the total amount of DAPI positive MuSCs within the respective well, and represented as a percentage. Student’s t-test for significance was used to evaluate significance amongst different age-treatment cohorts.

Myofiber isolation and culture

EDL myofibers were isolated as previously described⁶⁴. Briefly, the extensor digitum longus (EDL) was dissected and digested in 0.2% Collagenase B (Roche) in DMEM at 37°C for 1 hour. Single myofibers were isolated by triturating the digested EDL muscle with polished Pasteur pipettes. Myofibers were cultured in 12-well plates in DMEM (4.5 g/l glucose) supplemented with 20% FBS, 1% sodium pyruvate, 1% chicken embryo extract and 1% penicillin-streptomycin containing vehicle (DMSO) or 10 ng/ml PGE2 (Cayman Chemicals) for 24 hours.

Clonal muscle stem cell proliferation and fate analyses

We assayed clonal muscle stem cell proliferation by time-lapse microscopy as previously described^{4,17,43,65,66}. Briefly, we sorted MuSCs from C57Bl/6 mice (2–4 months), plated them on collagen-coated plates and treated them PGE2 (Cayman Chemical) or vehicle (DMSO) for 24hr. Cells were then trypsinized and reseeded at a density of 500 cells per cm² surface area in hydrogel microwells with 600 μm diameter. For time-lapse microscopy we monitored cell proliferation for those wells with single cells for 38h after seeding and recorded images every 3 min at 10× magnification using a PALM/AxioObserver Z1 system (Carl Zeiss MicroImaging) with a custom environmental control chamber and motorized stage. We analyzed time-lapse image sequences using the Baxter Algorithms for Cell Tracking and Lineage Reconstruction to identify and track single cells and generate lineage trees.

Viable and dead cells were distinguished in time-lapse sequences based on phase-contrast boundary and motility maintenance or loss, respectively. The proportion of live cells in each generation (G1-G6) at all timepoints is shown as cell number normalized to a starting population of 100 single MuSCs. The data analysis was blinded. The researchers performing the imaging acquisition and scoring were unaware of the treatment condition given to sample groups analyzed.

We found that the rates of proliferation (division) and death in the two conditions varied over time. Therefore, we estimated the rates for the first and the second 24 hour intervals separately. The values were estimated using the equations described in Gilbert et al., and found in Table 1. We denote the proliferation rates in the two intervals p_24 and p_48 and the corresponding death rates d_24 and d_48. As an example, the proliferation rate in the treated condition during the second 24 hour interval is 5.38% per hour. Table 1 (below) shows that the rates of proliferation and death in the two conditions are similar in the first time interval, and that the difference in cell numbers at the end of the experiment is due to differences in both the division rates and the death rates during the second time interval. The modeled cell counts in the two time intervals are given by:

$$c(t)=\left\{\begin{array}{c}{c}_0\exp \left(\left(p_{24}-d_{24}\right)t\right)0\le t\le 24\\ c(24)\exp \left(\left(p_{48}-d_{48}\right)(t-24)\right)24<t\le 48\end{array}\right.$$

where $c_0$ is the number of cells at the onset. The modeled curves are plotted together with the actual cell counts. The proportion of live cells in each generation (G1-G6) at all timepoints is shown as cell number normalized to a starting population of 100 single MuSCs.

Table 1.

Estimated proliferation and death rates per hours.

	𝑝24	𝑝48	𝑑24	𝑑48
DMSO	0.0488	0.0403	0.0045	0.0112
E2	0.0475	0.0538	0.0067	0.0012

The data analysis was blinded. The researchers performing the imaging acquisition and scoring were unaware of the treatment condition given to sample groups analyzed.

Proliferation assays

To assay proliferation, we seeded MuSCs on flat hydrogels at a density of 500 cells per cm² surface area. We counted cell number by using a hemocytometer, we collected cells at indicated timepoints by incubation with 0.5% trypsin in PBS for 5 min at 37 °C and quantified them using a hemocytometer at least 3 times. Additionally, we used the VisionBlue Quick Cell Viability Fluorometric Assay Kit (BioVision, catalog # K303) as a readout for cell growth in culture. Briefly, we incubated MuSCs with 10% VisionBlue in culture medium for 4h, and measured fluorescence intensity on a fluorescence plate reader (Infinite M1000 PRO, Tecan) at Ex= 530–570nm, Em=590–620nm. Data analyses were blinded, where researchers performing cell scoring were unaware of the treatment condition given to sample groups analyzed.

Quantitative RT-PCR

We isolated RNA from MuSCs using the RNeasy Micro Kit (Qiagen). For muscle samples, we snap froze the tissue in liquid nitrogen, homogenized the tissues using a mortar and pestle, followed by syringe and needle trituration, and then isolated RNA using Trizol (Invitrogen). We reverse-transcribed cDNA from total mRNA from each sample using the SensiFAST^™ cDNA Synthesis Kit (Bioline). We subjected cDNA to RT-PCR using SYBR Green PCR Master Mix (Applied Biosystems) or TaqMan Assays (Applied Biosystems) in an ABI 7900HT Real-Time PCR System (Applied Biosystems). We cycled samples at 95 °C for 10 min and then 40 cycles at 95 °C for 15 s and 60 °C for 1 min. To quantify relative transcript levels, we used 2−ΔΔCt to compare treated and untreated samples and expressed the results relative to Gapdh. We analyzed Ptger1, Ptger2 using SYBR-based Green qRT-PCR. For SYBR Green qRT-PCR, we used the following primer sequences: Gapdh, forward 5′-TTCACCACCATGGAGAAGGC-3′, reverse 5′-CCCTTTTGGCTCCACCCT-3′; Ptger1, forward 5’ GTGGTGTCGTGCATCTGCT-3′, reverse 5’-CCGCTGCAGGGAGTTAGAGT-3′, Ptger2, forward 5′-ACCTTCGCCATATGCTCCTT-3′, reverse 5′-GGACCGGTGGCCTAAGTATG-3′. TaqMan Assays (Applied Biosystems) were used to quantify Pax7, Myogenin, Nurr1, Ptger3 and Ptger4 in samples according to the manufacturer instructions with the TaqMan Universal PCR Master Mix reagent kit (Applied Biosystems). Transcript levels were expressed relative to Gapdh levels. For SYBR Green qPCR, Gapdh qPCR was used to normalize input cDNA samples. For Taqman qPCR, multiplex qPCR enabled target signals (FAM) to be normalized individually by their internal Gapdh signals (VIC).

Flow cytometry

We assayed Annexin V as a readout of apoptosis for MuSCs after 7 days in culture on hydrogels, after an initial acute (24 hr) treatment of vehicle (DMSO) or PGE2 (10 ng/ml). We used the FITC Annexin V Apoptosis Detection Kit (Biolegend, cat # 640914) according to the protocol of the manufacturer. We analyzed the cells for Annexin V on a FACS LSR II cytometer using FACSDiva software (BD Biosciences) in the Shared FACS Facility, purchased using an NIH S10 Shared Instrument Grant (S10RR027431–01).

In vivo muscle force measurement

The peak isometric torque (N•mm) of the ankle plantarflexors was assessed as previously described^18,67. Briefly, the foot of anesthetized mice was placed on a footplate attached to a servomotor (model 300C-LR; Aurora Scientific). Two Pt-Ir electrode needles (Aurora Scientific) were inserted percutaneously were inserted subcutaneously over the tibial nerve, just posterior/posterior-medial to the knee. The ankle joint was secured at a 90° angle. The peak isometric torque was achieved by varying the current delivered to the tibial nerve at a frequency of 200 Hz and a 0.1-ms square wave pulse.

Downhill treadmill running and in situ muscle force measurement

For in situ force measurement experiments, aged mice (18 mo.) were subjected to downhill treadmill run for 2 consecutive weeks. During week 1, mice ran daily for 5 days and rested on days 6 and 7. Two hours after each treadmill run during week 1, each (lateral and medial) gastrocnemius (GA) muscle from both legs of each mouse was injected with a dose of either PBS (vehicle control) or 20 μg dmPGE2 (experimental group). During week 2, mice were subjected to 5 days treadmill run only. The treadmill run was performed using the Exer3/6 (Columbus Instruments). Mice ran for 10 minutes on the treadmill at 20 degrees downhill, starting at a speed of 7 meters/min. After 3 min, the speed was increased by 1 meter/min to a final speed of 14 meter/min. 10 minutes run time was chosen, as exhaustion defined as the inability of the animal to remain on the treadmill despite electrical prodding, was observed at a median of 12 minute in an independent control aged mouse group. Force measurements were on the GA muscles at week 5 based on a protocol published previously^4,17. Briefly, for each mouse, an incision was made to expose the GA. We severed the calcaneus bone with intact achilles tendon and attached the tendon-bone complex to a 300C-LR force transducer (Aurora Scientific) with a thin metal hook. The muscles and tendons were kept moist by periodic wetting with saline (0.9% sodium chloride) solution. The lower limb was immobilized below the knee by a metal clamp without compromising the blood supply to the leg. The mouse was under inhaled anesthetic (2% isofluorane) during the entire force measuring procedure and body temperature was maintained by a heat lamp. In all measurements, we used 0.1-ms pulses at a predetermined supramaximal stimulation voltage. The GA muscles were stimulated via the proximal sciatic nerve using a bipolar electrical stimulation cuff delivering a constant current of 2 mA (square pulse width 0.1 ms). GA muscles were stimulated with a single 0.1-ms pulse for twitch force measurements, and a train of 150 Hz for 0.3 s pulses for tetanic force measurements. We performed five twitch and then five tetanic measurements on each muscle, with 2–3 min recovery between each measurement with n=5 mice per group. Data were collected with a PCI-6251 acquisition card (National Instruments) and analyzed in Matlab. We calculated specific force values by normalizing the force measurements by the muscle physiological cross-sectional areas (PCSAs), which were similar between the control and the experimental PGE2 treated group. PCSA (measured in mm²) was calculated according to the following equation (Burkholder et al., 1994):

$$\mathrm{PCSA}\left({\mathrm{mm}}^2\right)=\left[\mathrm{mass}\left(\mathrm{g}\right)\times \mathrm{Cos}\theta \right]÷\left[\rho \left(\mathrm{g}/{\mathrm{mm}}^3\right)\times \mathrm{fiber}\mathrm{length}\left(\mathrm{mm}\right)\right]$$

where $\theta$ is pennation angle of the fiber and $\rho$ is muscle density (0.001056 g/mm³).

RNA Sequencing

For RNA sequencing, α7-integrin⁺CD34⁺ muscle stem cells were isolated as described above, seeded on collagen-coated plates, treated a day later with PGE2 or vehicle (DMSO) and processed after 24 hours of treatment. RNA was isolated using Qiagen RNAEasy Micro kit from 5,000–10,000 cells, and cDNA generated and amplified using NuGEN Ovation RNA-Seq System v2 kit. Libraries were constructed from cDNA with the TruSEQ RNA Library Preparation Kit v2 (Illumina), and sequenced to 30–40×106 1×75bp reads per sample on a HiSEQ 2500 from the Stanford Functional Genomics Facility, purchased using an NIH S10 Shared Instrument Grant (S10OD018220).

Drop-seq Library Construction and Sequencing

Drop-seq was carried out as in ²⁸ with adjustments. Briefly, a dolomite single cell RNA-seq chip 2 (Dolomite, cat: 3200583) was used with 311 units/ul for the beads and cells. Aqueous channels were set to a flow rate of 40uL/min, the oil channel was set to a flow rate of 200 uL/min. cDNA amplification was carried out using KAPA HiFi HotStart ReadyMix (KAPA Biosytems, cat: KK2602) and tagmentation by the Nextera XT DNA Library Preparation Kit (Illumina, cat: FC-131–1096). DNA library concentration was measured using a qubit. Fragment distribution was measured using a bioanalyzer. Sequence read quality was determined through a MiSeq run with V3 150 cycle kits (cat: MS-102–3001). Libraries were sequenced with Read 1 at 26 cycles and Read 2 at 50 cycles. Depletion of mitochondrial DNA was performed using DASH treatment⁶⁸. In brief, 87 single guide RNAs (sgRNAs) targeting the most abundant RNA transcripts were produced by in vitro transcription (IVT) (Supplemental Table S5). sgRNAs were incubated with purified WT Cas9 at 37C for 10 minutes to create a functional ribonucleic nucleic protein (RNP) complex. The RNP complex was incubated with post-tagmentation Drop-seq libraries for 2 hours and then heat inactivated. Libraries were purified using the Nucleospin gel and PCR clean up kit (cat: 740609.50). Libraries were then amplified using primers targeting the ends of the P5 and P7 sequencing adapters. Validation of mitochondria transcript depletion was ~90% (lower to upper quartile range 73.5%–94.9%). Finally, a full sequence run was performed using an Illumina NextSeq or NovaSeq. Libraries were processed into expression matrices using the Dropseq_tools-1.12 and Hisat 2 for alignment.

ATAC Sequencing

ATAC-seq libraries were prepared as described previously³⁰. For 2-hour time point, around 50,000 freshly sorted MuSC were treated with PGE2 (Cayman Chemical 10 ng/ml), 1 μM EP4 antagonist (ONO-AE3–208, Cayman Chemical) or vehicle. Cells were spun down at 500g at 4 C, washed twice with PBS, lysed followed by transposase reaction. Libraries were constructed using Nextera DNA Library Prep Kit (Illumina) and sequenced to 85–172 million paired end reads per sample on NextSeq through the Stanford Functional Genomics facility.

Immunofluorescence and histology

We collected and prepared recipient TA muscle tissues for histology as previously described¹⁷. We fixed transverse sections from muscles using 4% PFA, blocked and permeabilized using PBS/1% BSA/0.1% Triton X-100 and incubated with anti-LAMININ (Millipore, clone A5, catalog # 05–206, 1:200), and anti-PAX7 (Santa Cruz Biotechnology, catalog # sc-81648, 1:50) primary antibodies and then with AlexaFluor secondary Antibodies (Jackson ImmunoResearch Laboratories, 1:200) or wheat germ agglutinin-Alexa 647 conjugate (WGA, Thermo Fisher Scientific). We counterstained nuclei with DAPI (Invitrogen).

For myofibers, myoblasts and MuSCs staining, we performed fixation using 4% PFA, blocking and permeabilization using PBS/1% BSA/0.1% Triton X-100 and staining with primary antibodies anti-EP4 (Santa Cruz Biotechnology, catalog #sc-55596, 1:100), anti-Myogenin (Santa Cruz Biotechnology, catalog #sc-576), anti-PAX7 (Santa Cruz Biotechnology, catalog # sc-81648, 1:50), anti-pCREB (Cell Signaling Technologies, catalog #9198) and then with AlexaFluor secondary Antibodies (Jackson ImmunoResearch Laboratories, 1:500). We counterstained nuclei with DAPI (Invitrogen).

Confocal images of myofibers were acquired on a Marianas spinning disk confocal (SDC) microscopy (Intelligent Imaging Innovations) with a 40x/0.9 N.A. objective to capture multiple consecutive focal planes (myofiber imaging) or using the KEYENCE BZ-X700 all-in-one fluorescence microscope (Keyence) with 20×/0.75 N.A. objectives. We analyzed PAX7 and EP4 positive cells using the MetaMorph Image Analysis software (Molecular Devices), and the fiber area using the Baxter Algorithms for Myofiber Analysis that identified the fibers and segmented the fibers in the image to analyze the area of each fiber. For PAX7 quantification we examined serial sections spanning a depth of at least 2mm of the TA. For fiber area at least 10 fields of LAMININ-stained myofiber cross-sections encompassing over 400 myofibers were captured for each mouse as above. For EP4 staining, we analyzed young and aged MuSCs or myofibers isolated from 3 independent biological replicates. Data analyses were blinded. The researchers performing the imaging acquisition and scoring were unaware of treatment condition given to sample groups analyzed.

Data Reporting

Sample size was determined based on practical and experimental consideration, with no statistical methods used to predetermine sample size. For florescence data, investigators were blinded to allocation for image acquisition and analyses when possible.

QUANTIFICATION AND STATISTICAL ANALYSIS

Single cell RNA-Seq Analysis

Expression matrix for drop-seq results were analyzed by Scanpy and Seurat with similar results. Single cells were filtered by mitochondrial RNA counts (<5%) and total gene counts (200<n<2000). Genes were filtered by expression in a minimum of 3 cells. Expression data was normalized by total UMI counts per cell and transformed as log(x+1). The percentage of mitochondrial RNA counts was regressed out, and the result scaled to max expression of 10. Harmony ⁶⁹; python version used in this analysis https://github.com/slowkow/harmonypy) was used to overcome batch effects and integrate independent Drop-seq experiments. 40 Principal components (PCs) of the integrated Harmony data were used to construct a 10-nearest neighbor connectivity network and reduced to 2-dimensional embedding using UMAP. Cells were coarsely clustered into major lineages using the leiden clustering algorithm ⁷⁰ at 0.5 resolution (Fig. S2A). Clusters were assigned lineages by differentially expressed genes. Overall, 21,550 cells passed quality control after batch correction (Fig. S2B); of these, 16,502 cells (76.6%) were annotated as myogenic (Figure S2C). Myogenic cells were isolated and finely clustered using leiden algorithm at 1.5 resolution and merged by differentially expressed genes into myogenic states (Fig. S2D).

MuSCs are marked by high expression levels of genes essential to the maintenance of the quiescent stem cell state (Pax7, Myf5, Spry1, Cd34, and Egfr, Id3). Cycling MuSCs share expression of these genes but also express cell cycle regulators (Cdk1, Top2a, Ccnb2, Mki67, Cdk4, Ccnd1, Cdk6). Myoblasts continue to cycle but downregulate Pax7, Myf5, Spry1 and express higher levels of Myod1, Cdk4, Ccnd1, and Cdk6. With further myogenic commitment and initiation of cell cycle withdrawal, myocytes express higher levels of Myod1 and initiate expression of Cdkn1a, Mef2a, and Myog. Two clusters of myocytes were identified: Myocytes 1 share gene signatures with proliferative myoblasts. Myocytes 2 express AP1 family members (Jun and Fos) which are normally downregulated in myoblasts. Myocytes 2 are more frequently found in aged muscles and exhibit higher enrichment for cell death gene ontology terms compared to the Myocytes 1 population. The transition from myocytes to myotubes entails fusion and the expression of fusogenic genes such as Mymk and Mymx as well as expression of embryonic and neonatal myosins (Myh3, Myh8). Immature myofibers these early developmental myosins are replaced by mature isoforms of sarcomeric proteins (Myh1, Myh2, Myh4, Tnnc2) and express the transcription factor Mef2c and muscle creatine kinase Ckm, as part of the transcriptional and metabolic changes that accompany differentiation.

PAGA analysis was performed on myogenic states and trajectories were defined as [“MuSCs”, “Cycling MuSCs”, “Myoblasts”, “Myocyte_1”, “Myotubes”, “Myofibers”] and [“MuSCs”, “Myocyte_2”, “Myotubes”, “Myofibers”]. Wishbone analysis ⁷¹ was initialized with the starting cell bearing the cell barcode of “ATCCCCAGGAGA” (the cell with the highest Pax7 expression), 3 components, and 250 waypoints. Differential pseudotime gene expression was determined by comparing the average expression cells of each trajectory across pseudotime bins with (quantile-normalized) width of 0.05 against the average expression of all cells at each pseudotime bin. Gene sets enriched in each trajectory were clustered according to differential expression profiles across pseudotime bins. Gene ontology was performed on differentially expressed genes using g:Profiler ⁷².

RNA-Seq Analysis

For the RNA-Seq analysis, RNA sequences were aligned against the Mus musculus genome using STAR ⁷³. RSEM ⁷⁴ was used for calling transcripts and calculating transcripts per million (TPM) values, as well a s total counts. A counts matrix containing the number of counts for each gene and each sample was obtained. This matrix was analyzed by DESeq to calculate statistical analysis of significance ⁷⁵ of genes between samples.

ATAC-Seq Analysis

The ATAC-seq reads were mapped to mouse genome (mm10) with Bowtie 2.0 ⁷⁶ and reads originating from the mitochondria and those with low mapping quality scores (below 10) were removed. MACS 2.1 ⁷⁷ was used to call peaks; DESEQ2 used tag counts per peak genome-wide to identify regions with differential accessibility upon PGE2 treatment versus vehicle (DMSO) and EP4 antagonist (ONO-AE3–269) treatment. Motif analysis was performed on differentially accessible peaks using HOMER 4.7 with all peaks as the background ⁷⁸.

ChromBPNet Analysis

Neural network models were trained using a ChromBPNet architecture as described in ⁷⁹. In brief, 1D convolutional neural networks were trained to predict unstranded tn5 transposase insertion sites from ATAC-seq reads observed in vehicle or PGE2-treated MuSCs from local genomic DNA sequences. The BPNet architecture used consists of an initial convolutional layer with 512 filters of width 21 bp and 9 dilated convolutions with ResNet style additions with 512 filters of width 3 bp and dilation rates equal to 2^{dilation depth}, the convolution outputs were used to (1) predict the probability profile of observing reads at a particular base (profile) using a deconvolution layer of width 25 and (2) predict the log of total read counts (transposase insertion sites) for the input sequence using a 1D global average pooling and a fully connected layer of 512 neurons. As such, local genomic sequence of length 3088 bp were used to predict profiles and total counts of 1000 bp regions around the summit of each ATAC-seq peak. Dropout layers were implemented after global average pooling and fully connected layers in counts prediction which improved prediction accuracy. ATAC-seq peaks from chromosomes 1, 8, and 10 were used as test set for model evaluation. The remaining peaks were used for training with batch size of 16 peaks for 25 epochs with early stopping (generally stopping after 10–18 epochs). Fully trained models were interpreted using DeepSHAP ⁸⁰ to compute the contribution for each base pair of an input sequence to the prediction output of the model, this quantitatively identifies subsequences that drive local chromatin accessibility. Contribution scores were split into 10 subsets of ~25000 regions and TF-MoDISco was run on DeepSHAP counts contribution scores for each subset with 400 bp windows, generating ~15000 41 bp seqlets with significant contributions for each clustering/discovery phase of TF-MoDISco, resulting in position frequency matrices of sequence motifs driving accessibility.

Non-overlapping archetypal motif matching and cis regulation analysis

Genome wide coverage of archetypal motifs from ³⁶ were used to match motifs with high contribution scores from BPNet. Archetype motifs overlapping each ATAC-seq peak were extracted from motif scans of the mm10 reference genome (https://resources.altius.org/~jvierstra/projects/motif-clustering/releases/v1.0/mm10.archetype_motifs.v1.0.bb). A confidence score for each motif within ATAC-seq peaks was calculated as the percent of bases in the motif with contribution scores that are greater than 99^th percentile of contribution scores across the genome multiplied by the match score of the sequence to an archetypal motif. Since consensus motifs can be discontinuous sequences, motifs were kept if >25% of base pairs had 99^th percentile contribution scores or higher. In case of overlapping archetypal motifs with high contribution base pairs, motifs were ranked by their confidence scores and non-overlapping motifs were kept. The number of motifs of each archetypal class was quantified for each ATAC-seq peak. The archetypal motif compositions for peaks with significant change in accessibility upon PGE2 treatment were embedded in a network using UMAP ⁸¹ and common cis regulatory patterns were found by clustering using the leiden algorithm. Resolution of clustering for the leiden algorithm was determined by assessing cluster homogeneity. Accessibility changes were quantified as average log2 fold change in accessibility of peaks in each cluster.

Statistical Analysis

We performed cell culture experiments in at least three independent experiments where three biological replicates were pooled in each. In general, we performed MuSC transplant experiments in at least two independent experiments, with at least 3–5 total transplants per condition. We used a paired t-test for experiments where control samples were from the same experiment in vitro or from contralateral limb muscles in vivo. A non-parametric Mann-Whitney test was used to determine the significance difference between vehicle (control) vs PGE2 treated groups using α=0.05. ANOVA or multiple t-test was performed for multiple comparisons with significance level determined using Bonferroni correction or with Fisher’s test as indicated in the figure legends. Unless otherwise described, data are shown as the mean ± s.e.m. Differences with p value <0.05 were considered significant (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001).

Supplementary Material

2

Table S1. Differential genes between young and aged MuSCs restored by PGE2-treatment related to Figure 5F.

3

Table S2. Ingenuity pathway analysis (IPA) summary and terms related to Figure 5G.

4

Table S3. Genes enriched in cycling (cluster 1) and direct commitment (cluster 2) trajectories related to Figure 5I.

5

Table S4. Oligonucleotide guide-RNA sequences used for mitochondrial DNA depletion using DASH related to STAR Methods

6

Supplementary Movies S1. Time lapse tracking of clonal cultures of vehicle-treated aged MuSCs on engineered hydrogel microwells, related to Figure 6B-G.

7

Supplementary Movies S2. Time lapse tracking of clonal cultures of dmPGE2-treated aged MuSCs on engineered hydrogel microwells, related to Figure 6B-G.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti Pax7	Santa Cruz Biotechnology	Catalog# sc-81648
Rat polyclonal anti-Laminin (Clone A5)	EMD Millipore	Catalog# 05–206
Rabbit monoclonal anti phospho-CREB (Ser133)	Cell Signaling Technologies	Catalog #9198
Rabbit polyclonal anti-Myogenin	Santa Cruz Biotechnology	Catalog #sc-576
Annexin V FITC Apoptosis Detection Kit	BioLegend	Catalog# 640922
Rat monoclonal anti-mouse CD11b (Clone M1/70)	BD Biosciences	Catalog# 553309
Rat monoclonal anti-mouse CD45 (Clone 30F11)	BD Biosciences	Catalog# 553078
Rat monoclonal anti-mouse Sca1 (Clone E13–161.7)	BD Biosciences	Catalog# 553334
Rat monoclonal anti-mouse CD31 (Clone 390)	eBioscience	Catalog# 13–0311-82
Rat monoclonal anti integrin-α7 (clone R2F2), PE conjugated	AbLab	Catalog# 10ST215
Rat monoclonal CD34-eFluor660 (clone RAM34)	eBioscience	Catalog # 50–0341-82
Streptavidin APC-Cy7	BD Biosciences	Catalog# 554063
Streptavidin MicroBeads	Miltenyi Biotec	Catalog# 130–048-101
LS Columns	Miltenyi Biotec	Catalog# 130–042-401
Chemicals, Peptides, and Recombinant Proteins
4’,6 diamidino-2-phenylindole (DAPI)	Life Technologies	Catalog# D1306
7-aminoactinomycin D (7-AAD)	Thermo Fisher Scientific	Catalog# A1310
D-Luciferin	Caliper Life Sciences	Catalog# 122796
Collagenase type 2	Worthington	Catalog# LS004176
Dispase II	Thermo Fisher Scientific	Catalog# 17105041
Collagenase B	Roche	Catalog# 11088815001
Streptavidin Microbeads	Miltenyi Biotech	Catalog# 130–048-101
Prostaglandin E2	Cayman Chemicals	Catalog# 14010
16,16-dimethyl Prostaglandin E2	Tocris	Catalog# 4027
ONO-AE3–208	Cayman Chemicals	Catalog# 14522
Tamoxifen	VWR	Catalog# AAJ63509–03
4-hydroxy Tamoxifen (4OHT)	Cayman Chemicals	Catalog# 17308
Dimethyl Sulfoxide (DMSO)	Sigma	Catalog# D8418
Notexin	Latoxan	Catalog# L8104
Cardiotoxin	Latoxan	Catalog# L8102
Critical Commercial Assays
PGE2 ELISA Kit	R&D Systems	Catalog# KGE004B
Ovation RNA-Seq System V2 kit	NuGen	Catalog# 7102–08
TruSEQ RNA Library Preparation Kit v2	Illumina	Catalog# RS-122–2101
SensiFAST cDNA Synthesis Kit	Bioline	Catalog# BIO-65053
RNeasy Micro Kit	Qiagen	Catalog# 74004
VisionBlue Quick Cell Viability Fluorometric Assay Kit	BioVision	Catalog# K303
Click-iT EdU Alexa Fluor 555 Imaging kit	Thermo Fisher Scientific	Catalog# C10338
cAMP-Glo Assay protocol	Promega	Catalog# V1501
Nextera DNA Library Prep Kit	Illumina	Catalog# FC-121–1030
Deposited Data
Raw and analyzed RNASeq data	This paper	GEO: GSE145297, GSE191190
Experimental Models: Cell Lines
Mouse: Primary myoblasts	This paper	N/A
Experimental Models: Organisms/Strains
Mouse: C57Bl/6J	Jackson Laboratories	000664
Mouse: TgCAG-luc,-GFP L2G85 (GFP/luc)	Jackson Laboratories	008450
Mouse: Pax7tm1(cre/ERT2)Gaka (Pax7CreERT2)	Jackson Laboratories	017763
Mouse: Gt(ROSA)26Sortm1(Luc)Kael (Rosa26-LSL-Luc)	Jackson Laboratories	005125
Mouse: Ptger4tm1.1Matb/BreyJ (Ptger4flox/flox)	Jackson Laboratories	028102
Mouse: NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG)	Jackson Laboratories	005557
Recombinant DNA
Lentivirus Ef1α-Luc-IRES-GFP	Cosgrove, 2014
Lentivirus pLM-CMV-R-Cre	Addgene	27546
Lentivirus pLKO.1-scramble shRNA (shSCR)	Sigma	SHC002
Lentivirus pLKO.1-Nurr1 shRNA (shNurr1)	Sigma	TRCN0000026029
Sequence-Based Reagents
qRT-PCR Primer: Gapdh F: TTCACCACCATGGAGAAGGC R: CCCTTTTGGCTCCACCCT	This paper	N/A
qRT-PCR Primer: Ptges F: GCTGTCATCACAGGCCAGA R: CTCCACATCTGGGTCACTCC	This paper	N/A
qRT-PCR Primer: Ptges2 F: CTCCTACAGGAAAGTGCCCA R: ACCAGGTAGGTCTTGAGGGC	This paper	N/A
qRT-PCR Primer: Ptger1 F: GTGGTGTCGTGCATCTGCT R: CCGCTGCAGGGAGTTAGAGT	This paper	N/A
qRT-PCR Primer: Ptger2 F: ACCTTCGCCATATGCTCCTT R: GGACCGGTGGCCTAAGTATG	This paper	N/A
DASH mitochrondrial DNA depletion	Gu et al., 2016	See Supplemental Table S4
Software and Algorithms
Flowjo	FlowJo, LLC 2013–2016	www.flowjo.com
Baxter Algorithms for Cell Tracking and Lineage Reconstruction and Myofiber Analysis	Magnusson et al., 2015	http://www.codesolorzano.com/Challenges/CTC/KTH-SE_2013.html
Keyence Analysis Software (BZ-X700 Microscope)	Keyence	www.keyence.com
Graphpad Prism	GraphPad Software, Inc, 2016	www.graphpad.com
MetaMorph Image Analysis software	Molecular Devices	www.moleculardevices.com
R	R development core team, 2014	www.r-project.org
STAR	Dobin et al., 2013	https://github.com/alexdobin/STAR/releases
RSEM	Li et al., 2011	https://github.com/bli25ucb/RSEM_tutorial
DESeq2	Anders et al., 2010	http://www.bioconductor.org/packages/DESeq2/
Living Image v.4.5	Perkin Elmer, 2015	Perkinelmer.com
Free-D (3D) v.1.14	Institut Jean-Pierre Bourgin, INRA Versailles, France	http://free-d.versailles.inra.fr/html/freed.html
Ingenuity Pathway Analysis	Qiagen, 2015	https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis/
Metacore Pathway Analysis	Thomson Reuters System Biology	https://portal.genego.com/
Bowtie 2.0	Langmead et al., 20096	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
HOMER	Heinz et al., 2010	http://homer.salk.edu/homer/
GREAT	McLean et al., 2010	http://great.stanford.edu/public/html/
ChromBPNet	Avsec et al., 2021 Nair et al., 2023 Pampari et al., 2025	https://github.com/kundajelab/chrombpnet
Scanpy	Wolf et al., 2018	https://github.com/theislab/scanpy/
g:Profiler	Raudvere et al., 2019	https://biit.cs.ut.ee/gprofiler/gost
Archetypal motif finder	This paper	https://github.com/will_yx/BasePair_motif_finder
Analysis pipelines	This paper	Available upon request

Highlights.

• EP4 expression and phosphorylated CREB are decreased in aged MuSCs

• PGE2 reverses epigenetic and transcriptional changes in aged MuSCs

• AI identification of AP1 and grammar responsible for epigenetic rejuvenation

• PGE2 treatment increases aged MuSC function and strength after injury and with exercise