Profiling the epigenomic landscape of late embryonic and adult mouse hind limb muscles

Abstract

Skeletal muscles are essential for movement, supporting a wide range of locomotor behaviors. Muscle tissue is composed of multiple cell types including “fast” and “slow” myofibers, whose contractile properties are largely influenced by selective expression of myosin heavy chain (MyHC) isoforms. While ‘super-enhancers’ regulating MyHC gene clusters have been identified, the cis-regulatory elements (CREs) controlling non-MyHC genes important to myofiber physiology remain less defined. Here, we profile the regulatory landscape of two pairs of mouse hind limb muscles differing in MyHC expression at a late embryonic (E18.5) and adult time point to identify candidate CREs that may regulate genes important to myofiber type. Gene expression and chromatin accessibility analyses revealed that epigenetic differences at E18.5 largely reflect limb patterning, whereas adult differences reflect myofiber differentiation. We identified thousands of differentially accessible regions that may regulate genes important for muscle development, muscle biology, and myofiber identity. Among these, twelve conserved, muscle-specific CREs associated with myofiber type were tested for regulatory activity. Nine enhanced and three reduced gene activity in vitro, although their phenotypic effects remain unknown. By profiling multiple muscles across two time points, our study extends current understanding of conserved, muscle-specific CREs that regulate gene expression during myogenesis.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-32705-4.

Introduction

Skeletal muscle accounts for a substantial portion of mammalian body mass^1,2 and plays a fundamental role in locomotion^3–5, thermoregulation⁶ and nutrient storage^7,8. Each muscle contains a heterogenous population of multinucleated muscle cells known as myofibers, which collectively contribute to muscle function^9–12. Myofibers can be classified into distinct types based on a range of criteria, including physiological properties^13–17, protein content^9,10, developmental timing during myogenesis¹⁸, and the myosin heavy chain (MyHC) isoforms expressed within thick filaments^19,20. Because myofiber physiology and contractile function are heavily influenced by MyHC isoform expression, MyHC isoforms, and the genes that encode them, are commonly used as molecular markers of myofiber type^21,22.

Within a single muscle, the expression of MyHC genes and isoforms varies across developmental stages and between myofiber types. In mammals, developing myofibers first express embryonic MyHC-emb (Myh3), followed by neonatal MyHC-neo (Myh8)¹⁶. Shortly before birth, the expression of these developmental isoforms declines in favor of the MyHC isoforms that distinguish mature myofiber types, largely driven by muscle innervation^18,23–27. After birth, mature myofibers are classified as either slow (i.e., slow-twitch or type-I) or fast (i.e., fast-twitch or type-II), based on the distinct contractile kinetics in the motor domain (S1 region) of MyHC’s globular head²⁸. Slow-contracting myofibers predominantly express slow-oxidative MyHC-I/β (Myh7), whereas fast-contracting myofibers primarily express a range of MyHC-II subtypes, including MyHC-IIa (Myh2), MyHC-IIx/d (Myh1), and MyHC-IIb (Myh4). Slow myofibers are highly fatigue-resistant, partially due to slower motor domain kinetics that reduce the rate of ATP hydrolysis, and a metabolic preference for synthesizing ATP via oxidative phosphorylation^{15–17,29,30}. In contrast, fast myofiber generate velocity and power (force x velocity) several-fold greater than slow myofibers. Their high rate of ATP hydrolysis and tendency to rely on glycolysis to synthesize ATP create a metabolic environment rich in inorganic phosphate and protons, making them more susceptible to fatigue^31–34.

Individual muscles contain a mix of fast and slow myofibers, and variation in this composition can impact the contractile and metabolic properties of both muscle bundles and whole tissue^35,36, ultimately affecting locomotor performance. Across mammals, muscle myofiber composition varies in relation to body size and locomotor strategy^3,37. Humans, sloths and slow lorises are unusual among terrestrial mammals in possessing muscles with a high proportion of fatigue-resistant slow myofibers (‘slow-biased’ muscles), which support long-duration locomotor behaviors such as long-distance walking and running, extended tree hanging, and slow climbing^3,37–39. In contrast, the muscles of mice and other small mammals are dominated by fast myofibers (‘fast-biased’ muscles), which enable higher-powered, shorter-duration locomotor behaviors such as sprinting and frequent bursts of movement. Experimentally altering the physiological properties of muscle tissue in mice has been shown to improve muscle fatigue-resistance (i.e., endurance) and average daily travel distance⁴⁰, although factors like oxygen transport also influence endurance capacity⁴¹.

Underlying shifts in myofiber composition between muscles, and more broadly, between species, are modifications to developmental gene regulation. Mutations in cis-regulatory elements (CREs), such as enhancers and silencers, can alter patterns of gene expression at key developmental stages in a tissue-specific manner, leading to phenotypic variation and evolutionary divergence between species^42–54. Therefore, evolutionary selection acting on developmental CREs that regulate the expression of genes influencing muscle myofiber composition may be critical to understanding the physiological differences between myofiber types in a body, and the same muscle between species³⁷. While MyHC ‘super-enhancers’ that regulate the expression of genes within the fast and slow MyHC gene clusters have recently been identified^55,56, CREs that control the expression of other genes important to the determination of myofiber physiology, such as those involved in mitochondrial biogenesis or MyHC signaling cascades, are less understood. Identifying such elements would provide deeper insight into muscle development and biology and offer valuable targets for investigating species-specific differences in muscle myofiber composition, such as those observed between humans and African apes^3,37.

Here, we used a mouse model to profile the regulatory landscape of two pairs of hind limb muscles (two calf and two thigh) that differ in slow myofiber content at a late embryonic (E18.5) and an adult time point, expanding previous studies focusing on a single time point^55–64, or pooled muscles at an embryonic and postnatal time point⁶⁵. Patterns of gene expression (RNA-seq) and chromatin accessibility (ATAC-seq) largely reflect differences due to time point, with limb patterning driving variation in E18.5 muscles and myofiber physiology driving variation in adult muscles. By targeting non-coding regions with greater specificity to muscle tissue, we identified thousands of candidate CREs that may shape muscle development and identity, including nine associated with myofiber type that enhance gene activity in vitro. Our results reinforce the importance of spatiotemporal gene regulation during myogenesis and suggest that muscle-specific CREs controlling genes involved in myofiber metabolism and contractile function are promising targets for investigating species-specific differences in muscle myofiber composition.

Methods

Mice

All mouse procedures were performed in accordance with relevant guidelines, regulations and protocols approved by the Harvard University and University of Oregon Institutional Animal Care and Use Committees (Capellini protocol: 13-04-161-2; Sterner protocol: AUP-20-04). All mice were maintained in accordance with the Guide for the Care and Use of Laboratory Animals⁶⁶ at the Harvard University Biological Research Infrastructure Mouse Facility (Capellini laboratory mouse room). This study is reported in accordance with ARRIVE guidelines⁶⁷.

Adult FVB/NJ mice were obtained at 3–4 months of age. Mice were free of disease or injury at the time of sample collection and were not previously involved in biological research. Male and female mice were used to establish timed matings, and at embryonic stage (E) 18.5, a stage relevant to our studies^18,68, pregnant females were euthanized to acquire embryos. Adult mice were exposed to CO₂ at a constant rate for 5 min and secondary euthanasia was performed via cervical dislocation by a trained investigator. E18.5 embryos were dissected under a microscope in 1X PBS on ice, decapitated to ensure rapid euthanasia, and anatomically sexed⁶⁹. To minimize variation attributed to X-chromosome silencing factors (e.g., Xist, Tsix), only muscle samples from male mice were used for ATAC-seq and RNA-seq analysis, consistent with prior studies^{12,56,57,62,64,70}.

Four hind limb muscles were chosen for dissection based on previously described differences in slow myofiber content^71–78. The soleus (30.6–60% slow)^{71–74,76,77} and vastus intermedius (2.4–45% slow)^71,74,76 represent slow-biased muscles (i.e., muscles with a greater proportion of slow myofibers compared to surrounding muscles), and the gastrocnemius (0–8% slow)^{71,73–75,78} and vastus lateralis (0% slow)^71,74,75 represent fast-biased muscles. Muscles were identified using a published reference⁷⁹. To dissect these muscles, the leg was skinned and the superficial muscles connecting the pelvis to the tibia or fibula (lateral side: biceps femoris, medial side: gracilis, semimembranosus, and semitendinosus) were removed using sharp dissection scissors and forceps by cutting along the dorsal side of the tibia with dissection scissors and peeling the muscles away. Forceps or a dissection probe were used to remove any fascia and separate the tissues before each was removed by cutting near tendons at the origin(s) and insertion(s). In the calf, this process revealed the gastrocnemius and the underlying soleus, which was clearly distinguished by its smaller size and the darker red color at both stages. For the thigh muscles, the four muscles of the upper thigh were distinguished (superficial: rectus femoris, lateral: vastus lateralis, medial: vastus medialis, deep: vastus intermedius) before the vastus lateralis was carefully collected by cutting away this muscle. Afterwards, the rectus femoris and vastus medialis were removed to reveal the vastus intermedius, which runs along the femur. Similar to the soleus, the vastus intermedius is slightly redder than the surrounding muscles at both stages. E18.5 dissections were conducted under a dissecting microscope at low magnification.

RNA-seq

Soleus, gastrocnemius, vastus intermedius, and vastus lateralis samples were collected from adult male (n = 6) and male E18.5 mice (n = 6) and preserved in TRIzol reagent (Invitrogen)⁷⁹. After collection, muscle samples were pulverized using an electric homogenizer and subjected to phenol-chloroform extraction. Total RNA was isolated and purified using a Zymo Direct-zol RNA Microprep Kit following the manufacturer’s instructions. RNA was eluted in nuclease-free water and evaluated for quantity and quality. RNA concentrations were measured using a Qubit fluorometer, RNA purity was assessed using a NanoDrop spectrophotometer, and RNA integrity was assessed using an Agilent TapeStation. Only samples with RNA Integrity Number (RIN) scores above 7.0 and matched across all four muscles from the same individual were retained for library preparation and sequencing.

RNA-seq libraries were prepared by the Harvard Bauer Core Facility (HBCF) using the Kapa mRNA HyperPrep Kit and automated on a PerkinElmer Sciclone NGS Workstation. RNA was reverse transcribed into cDNA using random primers. Total RNA concentration was measured using the Quant-iT RiboGreen RNA Assay Kit, and total RNA integrity was assessed using an Agilent BioAnalyzer 2100 RNA Nano Kit. Libraries were pooled and assessed for quality using qPCR. Deep sequencing was performed on an Illumina NovaSeq6000 platform using 2 × 100 bp paired-end reads, targeting a minimum depth of 30 million reads per sample⁸⁰. RNA-seq sample details, quality control metrics, and read processing information are provided in Supplementary Table S1.

Demultiplexed sequencing data were processed using the University of Oregon’s high-performance computing cluster, Talapas, and analyzed in RStudio⁸¹ using R (v.4.5.0)⁸² following the Encyclopedia of DNA Elements (ENCODE) RNA-seq guidelines (www.encodeproject.org/about/experiment-guidelines/). Sequence quality, GC content, and adapter contamination was assessed using FastQC (v.0.11.5)⁸³. Adapter sequences and low-quality bases (Phred score < Q20) were trimmed using Trimmomatic (v.0.39.0)⁸⁴. To enrich for mRNA and reduce background noise, residual ribosomal RNA reads were removed using RiboDetector (v.0.3.1)⁸⁵. Processed reads were aligned to the mouse reference genome (mm10) using STAR (v.2.7.11)⁸⁶.

To assess sample relationships, Principal Component Analysis (PCA) was performed using the plotPCA function in R, and Multidimensional Scaling (MDS) analysis was performed using the R package PoiClaClu (v.1.0.2.1)⁸⁷ (Supplementary Fig. S1). One adult vastus intermedius sample (“A Vast Int 7/27/21 RNA”) was excluded from downstream analysis due to clustering anomalies. Therefore, n = 5 adult vastus intermedius samples and n = 47 RNA-seq samples in total.

Analysis of differentially expressed genes (DEGs)

To quantitatively compare gene expression (relative read count) across conditions, a raw read count matrix was generated from processed RNA-seq data using the featureCounts function from Subread (v.2.0.6)⁸⁸. Differential expression analysis was performed using the R package DESeq2 (v.1.44.0)⁸⁹ to identify genes differentially expressed across time points (E18.5 vs. adult), myofiber biases (fast vs. slow), anatomical regions (calf vs. thigh), and muscle identities (soleus vs. gastrocnemius vs. vastus intermedius vs. vastus lateralis), as well as genes differentially expressed within each time point (E18.5 Fast vs. E18.5 Slow; Adult Calf vs. Adult Thigh) using a false discovery rate (FDR)⁶³ threshold of < 0.05. Genes with fewer than five read counts across all samples were excluded, leaving 20,889 genes for analysis. Genes were considered significantly differentially expressed (DEGs) if they met the thresholds of |log₂ fold change| > 0.6 and adjusted p < 0.05. For visualizations, library-normalized counts were transformed based on the mean-variance relationship across genes using the variance-stabilized transformation function from DESeq2 to account for differences in variance. Visualizations were created using the R package ggplot2 (v.3.5.1)⁹⁰.

Cell type-specific marker sets from mouse muscle scRNA-seq studies^59,91 were used to estimate per-sample cell type signature enrichment using the R package GSVA (v.2.4.1)⁹². Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed using the R package clusterProfiler (v.4.12.3)⁹³. The simplify function was applied to reduce redundancy among enriched GO terms. Gene set enrichment analysis (GSEA) was performed using the R package Fast Gene Set Enrichment Analysis (FGSEA, v.1.30.0)⁹⁴. Human orthologs were retrieved using the mouse2human function from the R package HomoloGene (v.1.4.68.19.3.27), and human Disease Ontology (DO) enrichment analysis was performed using the R package Disease Ontology Semantic and Enrichment Analysis (DOSE, v.3.30.5)⁹⁵. To reduce redundancy in DO enrichment results, a semantic similarity matrix was computed using the doSim function followed by hierarchical clustering. The resulting dendrogram was cut at a height of 0.3 to define semantically similar clusters, from which the term with the lowest adjusted p-value (padj) was retained. All statistical analyses applied a q-value threshold of < 0.05.

ATAC-seq

Additional samples of the soleus, gastrocnemius, vastus intermedius, and vastus lateralis were collected from adult male (n = 3) and male E18.5 mice (n = 3 pooled litters). To ensure a minimum of 50,000 cells per embryonic muscle sample for ATAC-seq analysis^96,97, each E18.5 biological replicate consisted of pooled muscles (both left and right sides) from up to five male embryos from the same litter. Muscle nuclei were isolated from undigested tissue using a previously established protocol⁹⁸. Cell viability was assessed using trypan blue exclusion and quantified with a hemocytometer under an inverted microscope. Only samples with < 10% cell death were subjected to an optimized ATAC-seq protocol⁹⁹. Briefly, muscle nuclei were lysed to release DNA and treated with Tn5 transposase to tag transcriptionally accessible fragments of DNA not bound to histones (i.e., potential non-coding regulators of gene expression such as CREs). Transposed DNA was purified using the Zymo DNA Clean and Concentrator Kit following the manufacturer’s instructions. DNA fragments were then barcoded with indexed primers and PCR-amplified to generate ATAC-seq libraries. Libraries were size-selected using the OMEGA Bead Purification Protocol, and concentrations were determined using the KAPA Library Quantification Complete Kit.

ATAC-seq library quality was assessed by the HBCF using the Agilent Bioanalyzer 2100 DNA High Sensitivity Kit and the Agilent TapeStation 4200 (D1000 ScreenTape assay). Libraries were pooled and evaluated for quality using qPCR and the Illumina MiSeq Nano system. Library pools were adjusted based on the Illumina MiSeq Nano results to optimize distribution and final loading concentration. Deep sequencing was performed on an Illumina NovaSeq6000 platform using 2 × 100 bp paired-end reads, targeting a minimum depth of 50 million reads per sample⁸⁰. ATAC-seq sample details, quality control metrics, and primer sequences are provided in Supplementary Table S2.

Demultiplexed sequencing data were processed using the University of Oregon’s high-performance computing cluster, Talapas, and analyzed in RStudio following ENCODE ATAC-seq guidelines (www.encodeproject.org/about/experiment-guidelines/). Sequence quality, GC content, and adapter contamination was assessed using FastQC (v.0.11.5)⁸³. Adapter sequences and low-quality bases (Phred score < Q20) were trimmed using Trimmomatic (v.0.39.0)⁸⁴, and reads were aligned to the mm10 genome using Bowtie2 (v.2.4.4)¹⁰⁰. To target more high-value portions of the genome, reads mapping to the mitochondrial genome were removed using SAMtools (v.1.14.0)¹⁰¹, and duplicate reads were removed using Picard (http://broadinstitute.github.io/picard). To evaluate variation in accessibility across samples, an analysis of variance (ANOVA) with Tukey’s post hoc test and a Kruskal-Wallis test were performed. Statistical significance was defined as p < 0.05.

Analysis of differentially accessible peaks of open chromatin

Filtered BAM files were assessed for read alignment enrichment (i.e., peaks) using MACS2 (v.2.1.1)¹⁰² with the callpeak parameters: -B -f BAMPE -g mm -p 0.05. Resulting MACS2 peak output files were merged into a non-redundant consensus peak set yielding 928,789 peaks. A matrix of peak accessibility across samples was generated using the countOverlaps function from the R package GenomicRanges (v.1.60.1)¹⁰³. Differentially accessible regions were identified using DESeq2 to compare peak accessibility between time points (E18.5 vs. adult), myofiber biases (fast vs. slow), and anatomical regions (calf vs. thigh), applying an FDR threshold of < 0.05. Peaks accessible in fewer than three samples were excluded, leaving 252,701 open chromatin regions for analysis. For visualizations, filtered peak counts were then transformed using the vst() function from DESeq2.

Reproducible peaks across biological replicates were identified using an irreproducible discovery rate (IDR)¹⁰⁴ threshold of < 0.05. Differences in peak count and mitochondrial read enrichment were assessed using ANOVA with Tukey’s post hoc test, Kruskal-Wallis test, and paired t-test, with statistical significance defined as p < 0.05. Overlap between peak sets was evaluated using pairwise Jaccard index analysis.

To reduce background noise and exclude peaks likely representing pleiotropic CREs active across multiple tissue types¹⁰⁵, IDR peaks overlapping mm10 blacklist regions¹⁰⁶ and an IDR peak set generated from embryonic mouse brain tissue¹⁰⁷ were removed using BEDTools (v.2.25.0)¹⁰⁸. Although some open chromatin peaks shared between brain and muscle may contribute to muscle biology, this conservative filtering approach produced ‘brain-filtered’ IDR peak sets predicted to contain CREs with increased specificity to muscle tissue. We note that we had previously used additional tissues to filter out shared sequences, but this resulted in only slight reductions in shared peaks⁹⁸.

To examine inter-species conservation, phyloP scores for 60 mammalian species (phyloP60ways)¹⁰⁹ were obtained from the University of California Santa Cruz (UCSC) Genome Browser (http://genome.ucsc.edu)¹¹⁰ and used to extract per-base conservation scores for IDR peak sets before and after brain-filtering. To control for variation in peak size and make peaks comparable for these analyses, all peak coordinates were assigned to a fixed width of 400 bp, centered on the middle of a called peak. Given that base-pair modifications can alter enhancer activity^111,112 and that selection likely acts on individual transcription factor binding sites (TFBSs) rather than across the full length of multi-TFBS enhancers¹¹³, we aggregated per-base conservation scores to compare sequence conservation across IDR peak sets, as described previously¹¹⁴. Within each IDR peak set, per-base conservation scores were averaged across all fixed-width (400 bp) peaks, yielding 400 data points, which were then averaged across each IDR peak set to generate an overall peak set conservation score. Conservation scores before and after brain-filtering were compared using a paired t-test and an ANOVA with Tukey’s post hoc test. See Supplementary Table S3 for MACS2 and IDR peak counts, as well as mean per-base conservation scores. To assess whether observed conservation scores were significantly higher than expected by chance, the average phyloP scores of each peak set was compared to distributions generated from 1,000 randomly sampled genomic regions (permutation backgrounds). For each permutation, the mean phyloP score was calculated, and an empirical p-value for each peak set was computed as the fraction of permutation means greater than or equal to the observed mean (one-sided test). To account for multiple hypothesis testing across the nine peak sets, p-values were adjusted using the Benjamini–Hochberg procedure.

Unfiltered and brain-filtered IDR peak sets were annotated to genomic features using the R package ChIPseeker (v.1.36.0)¹¹⁵, defining promoters as ± 1,000 bp around the transcription start site (tssRegion = c(−1000, 1000)). Significant differences in genomic feature representation were assessed using a G-test and an ANOVA with Tukey’s post hoc test. Brain-filtered IDR peak sets were intersected using the multiinter function from BEDTools to identify regions reliably called as peaks across multiple samples. This generated nine non-overlapping open chromatin peak sets, referred to as ‘condition-specific’ peak sets. The E18.5 condition-specific peak set includes peaks found in all four muscles (soleus, gastrocnemius, vastus intermedius and vastus lateralis) at the E18.5 time point but not in any muscles at the adult time point. Conversely, the Adult condition-specific peak set includes peaks present across all four muscles at the adult time point but not in any muscles at E18.5. Several condition-specific peak sets further distinguish open chromatin accessibility by time point and myofiber bias. The E18.5 Fast and E18.5 Slow condition-specific peak sets contain peaks found exclusively in embryonic muscles, but specific to fast-biased muscles (gastrocnemius and vastus lateralis) or slow-biased muscles (soleus and vastus intermedius), respectively. Similarly, the Adult Fast and Adult Slow condition-specific peak sets contain peaks found exclusively in adult muscles, but specific to fast-biased or slow-biased muscles, respectively. The Fast condition-specific peak set includes peaks found exclusively across fast-biased muscles regardless of time point; the Slow condition-specific peak set includes peaks found exclusively across slow-biased muscles regardless of time point. The Universal condition-specific peak set includes peaks accessible across all muscles at both time points. Significant differences in condition-specific peak set enrichment were compared using a negative binomial regression.

Condition-specific peak sets (i.e., Adult, Adult Fast, Adult Slow, E18.5, E18.5 Fast, E18.5 Slow, Universal, Fast, Slow) were annotated to genomic features using ChIPseeker and significant differences in feature representation were compared as described previously. Each condition-specific peak set was also annotated to nearby genes and analyzed for functional enrichment using the Genomic Regions Enrichment of Annotations Tool (GREAT, v.4.0.4)¹¹⁶. Transcription factor motif enrichment was assessed using Hypergeometric Optimization of Motif EnRichment (HOMER, v. 4.10.4)¹¹⁷ on 400 bp windows centered ± 200 bp from the midpoint of each peak. The following Homer findMotifsGenome.pl parameters were used: -size given -mask. To evaluate inter-species conservation, phyloP scores were obtained from the UCSC Genome Browser and used to extract per-base conservation scores for each fixed-size (400 bp) condition-specific peak set, as described above. Average conservation scores for each condition-specific peak set were then compared to a randomly generated background using a Wilcoxon Rank-Sum test.

To visually assess the epigenetic landscape surrounding peak sets, MACS2 narrowPeak and condition-specific peak set BED files were imported into the UCSC Genome Browser. The following annotation tracks were loaded: Base Position, GENCODE VM23, NCBI RefSeq, ENCODE cCREs, ORegAnno, ENCODE regulation, CpG Islands, JASPAR Transcription Factors, FANTOM5, RefSeq Func Elems, ReMap ChIP-seq, Placental Mammal Basewise Conservation by PhyloP (PlacentalCons), and RepeatMasker. To target regions that may function as CREs, additional mouse muscle ChIP-seq datasets^55,57,65,118 and MyHC gene super-enhancer annotations^56,59 were included as custom tracks. An independent ATAC-seq dataset (GEO SuperSeries accession GSE123879)⁵⁷ was processed using our peak-calling pipeline and imported into the UCSC Genome Browser to compare open chromatin signals across a wider range of muscles, strains, and library preparation methods.

Luciferase reporter vector construction

To assess whether condition-specific peaks could influence gene expression in vitro, and thus function as CREs, a subset of peaks was selected for functional validation. Candidate CREs were prioritized based on multiple criteria, including chromatin accessibility patterns across ATAC-seq datasets, strength of regulatory signal from UCSC Genome Browser tracks (e.g., ENCODE cCREs) and mouse ChIP-seq datasets^57,65,118, mammalian conservation scores (phyloP60ways), nearby gene function, and nearby gene expression patterns derived from RNA-seq. High priority was given to candidates exhibiting strong regulatory signals and high evolutionary conservation located within ± 100 kb of DEGs known to influence fast and/or slow myofiber phenotypes. Of particular interest were candidates accessible in Slow and Adult Slow peak sets and located near slow myofiber genes up-regulated in slow-biased muscles, as well as candidates accessible in Fast and Adult Fast peak sets and located near fast myofiber genes up-regulated in fast-biased muscles (see example in Supplementary Fig. S2). This approach yielded 260 candidates (Supplementary Table S4). From these, 12 candidate CRE sequences were selected for functional validation: nine associated with fast-biased muscles and three with slow-biased muscles. The expression of genes residing ± 500,000 bp from each candidate sequence are reported in Supplementary Table S5.

Candidate DNA sequences were narrowed to < 500 bp regions based on phyloP conservation scores and screened for restriction enzyme cut sites using NEBcutter (v.3.0.17)¹¹⁹. The twelve candidate CRE sequences were cloned upstream of the firefly luciferase reporter gene (luc2) promoter in the pGL4.23 [luc2/minP] vector (E8411, Promega) by Azenta Life Sciences. All reporter vector constructs were verified by Sanger sequencing¹²⁰. Candidate CRE sequences, nearby gene functions, conservation scores, and restriction enzyme cut sites are provided in Supplementary Table S6.

Cell culture

Mouse C2C12 myoblasts (ATCC, CRL-1772) were cultured in growth medium consisting of Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin, on 0.1% gelatin-coated cultureware at 37 °C in a humidified atmosphere with 5% CO₂. Growth medium was replaced every 2–3 days, and myoblasts were subcultured when reaching ~ 80% confluency. For transfection experiments, undifferentiated C2C12 myoblasts were seeded into 0.1% gelatin-coated 48-well plates and cultured for 24–48 h prior to transfection.

Transfection and luciferase reporter assays

C2C12 myoblasts were transfected in 48-well plates using Lipofectamine 3000 (Invitrogen) following the manufacturer’s protocol. Three independent transfection experiments (n = 3) were conducted, each with at least six technical replicates per reporter vector construct. An empty pGL4.23 vector (E8411, Promega) was used to determine baseline luc2 expression. All reporter vector constructs were co-transfected with a pGL4.74 [hRluc/TK] Renilla vector (E6921, Promega) to control for transfection efficiency. After optimization, cells were harvested 48 h post-transfection and assayed for luciferase activity using the Dual-Luciferase Reporter Assay System (E1910, Promega) on a GloMax Navigator microplate luminometer (GM2010, Promega), following the manufacturer’s instructions. Luciferase activity was calculated as the ratio of firefly (luc2) to Renilla (Rluc) luminescence, averaged across the technical replicates. The mean luc2/Rluc ratio for each reporter vector construct was then normalized to the empty pGL4.23 control vector to confirm regulatory activity in C2C12 myoblasts. Significant differences in normalized luciferase activity were assessed using an unpaired two-tailed Student’s t-test, and Fisher’s combined probability test was used to integrate p-values across the three transfection experiments. Data are presented as mean ± SEM (standard error of the mean) and statistical significance is indicated as follows: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0001 (****).

Results

Time point shapes gene expression and chromatin accessibility patterns in mouse hind limb muscles.

Given the dynamic nature of myogenesis and the changing cellular composition of muscle tissue across development^12,59,121, we hypothesized that differences in gene expression and chromatin accessibility would be largely driven by time point. To test this, we generated gene expression (RNA-seq) and chromatin accessibility (ATAC-seq) profiles for two pairs of slow- versus fast-myofiber biased muscles at a late embryonic (E18.5) and adult time point (see Methods, Fig. 1a) and assessed sample similarity by time point, anatomical region, and myofiber bias using PCA. Both datasets separated by time point along PC1 (Fig. 1b,c), and by adult myofiber bias along PC2 (Fig. 1d,e). Interestingly, RNA-seq data clustered by anatomical region at E18.5 (Fig. 1f) and by muscle identity at the adult time point (Fig. 1g). These patterns were not observed in the ATAC-seq data, likely due to low peak overlap across samples (Jaccard index < 0.3; Supplementary Fig. S3), as expected for this data type¹²².

Fig. 1.

Gene expression and chromatin accessibility in mouse hind limb muscles at a late embryonic (E18.5) and adult time point. (a) Experimental schematic. (b–g) PCA of RNA-seq and ATAC-seq data. PCA of normalized transcriptome data is colored by (b) time point, (d) myofiber bias, (f) anatomical region, and (g) muscle identity (n = 47). (c, e) PCA of MACS2-called ATAC-seq peaks colored by (c) time point and (e) myofiber bias (n = 24). (h) Volcano plot showing DEGs between E18.5 (green) and adult (purple) muscles. Gray dots represent genes that did not meet the significance threshold (|log₂ fold change| ≥ 0.6; adjusted p ≤ 0.05). Gene symbols indicate the top 15 DEGs for each time point. Bold gene symbols highlight significantly differentially expressed MyHC genes.

We next performed differential expression analysis to identify genes significantly up-regulated at each time point. We identified 11,822 DEGs, including 6,589 genes up-regulated in E18.5 muscles and 5,233 up-regulated in adult muscles (Fig. 1h). Notably, embryonic Myh3 and neonatal Myh8 were among the top 15 DEGs in embryonic muscles, while MyHC-IIb Myh4 ranked among the top 15 in adult muscles. As expected, adult muscles showed enrichment for markers associated with differentiated myonuclei, while E18.5 muscles showed enrichment for markers of muscle stem cells and myoblasts, likely reflecting differences in cellular composition (Supplementary Figs. S4, S5). Genes up-regulated at E18.5 were enriched for GO terms and KEGG pathways relating to limb development (Actn1, Hoxd13, Myf5, Myl4, Myl6b, Myo1c, Myh10, Myo15b, Tnnt2, Tpm4), while genes up-regulated in adult muscles were associated with muscle metabolism (Adipoq, Foxo1, mt-Atp8, Ppargc1a, Pparg) and myofiber differentiation (Actn3, Mybpc2, Myh2, Myh7, Myoz1, Tnnc1, Tnni2, Tnnt1, Tnnt3, Tpm3) (Supplementary Table S7). Together, these findings indicate that differences in gene expression and chromatin accessibility are largely driven by time point, and may also reflect changes in gene regulation and cellular composition related to muscle development and identity.

E18.5 transcriptomes suggest earlier differentiation of thigh musculature

To identify genes driving the time point-dependent clustering of RNA-seq data by anatomical region, we performed differential expression analysis between calf and thigh muscles at each time point and assessed the biological functions of DEGs using GO enrichment analysis. As expected, more genes were differentially expressed between calf and thigh muscles at E18.5 (n = 1,136; Fig. 2a) than at the adult time point (n = 111; Fig. 2c). At E18.5, more genes were up-regulated in thigh muscles (n = 734) than in calf muscles (n = 402), possibly reflecting the earlier regionalization of thigh musculature¹²³ or the greater phenotypic similarity between the vastus lateralis and vastus intermedius compared to the soleus and gastrocnemius¹²⁴. Notably, genes involved in distal/posterior limb patterning (Bmp7, Eya2, Hoxa11, Hoxa13) were up-regulated in E18.5 calf muscles, whereas genes associated with adult fast myofibers (Actn3, Myh2, Myh4, Tnnc2, Tnni2, Tnnt3) were up-regulated in E18.5 thigh muscles, suggesting that more anterior and proximal muscle groups (e.g., thigh) may differentiate earlier^125,126. Indeed, genes up-regulated in E18.5 thigh muscles were enriched for terms related to myofiber development, while those up-regulated in E18.5 calf muscles were enriched for pattern specification (Fig. 2b). Similar trends were observed in KEGG pathway and gene set enrichment analyses (Supplementary Table S7). At the adult time point, genes up-regulated in calf muscles were again associated with limb patterning (Hoxa13, Hoxc11, Hoxd10, Hotair), while those up-regulated in thigh muscles were linked to muscle cell proliferation (Tbx1, Tbx5, Ephb1, Mapk11) (Fig. 2d). Only a small subset of DEGs identified at E18.5 remained up-regulated in adult muscles. Genes up-regulated in calf muscles across time points (n = 15; Fig. 2e) were associated with limb patterning^125,127,128, muscle development^129–132, and the central nervous system^133–136. Those up-regulated in thigh muscles across time points (n = 5, Fig. 2f) were largely involved in regional specification^137–139. These findings suggest that limb patterning may be the primary molecular signal observed at E18.5, with thigh musculature undergoing earlier differentiation than calf musculature.

Fig. 2.

Differential gene expression between hind limb muscles from distinct anatomical regions at a late embryonic (E18.5) and adult time point. (a) Volcano plot showing DEGs between calf muscles (gastrocnemius and soleus; cyan) and thigh muscles (vastus intermedius and vastus lateralis; gold) at E18.5 (n = 24). Gray dots represent genes that did not meet the significance threshold (|log₂ fold change| ≥ 0.6; adjusted p ≤ 0.05). Gene symbols indicate the top 10 DEGs per anatomical region. (b) Top five enriched GO Biological Process terms for genes significantly up-regulated in calf (cyan) and thigh (gold) muscles at E18.5. (c) Volcano plot showing DEGs between calf and thigh muscles at the adult time point (n = 23). (d) Top five enriched GO Biological Process terms for genes significantly up-regulated in calf and thigh muscles at the adult time point. (e) Venn diagram of genes up-regulated in calf muscles, showing those uniquely up-regulated at E18.5 (light cyan), those uniquely up-regulated at the adult time point (dark cyan), or those up-regulated at both time points (genes symbols listed below). (f) Venn diagram of genes up-regulated in thigh muscles, showing those uniquely up-regulated at E18.5 (light gold), those uniquely up-regulated at the adult time point (dark gold), or those up-regulated at both time points (gene symbols listed below).

Genes that distinguish adult myofiber types are not expressed at the same level in embryonic and adult muscles

We next compared DEGs between fast- and slow-biased muscles during myogenesis (E18.5) and in fully differentiated muscles (adult). Since tissue-specific patterns of gene expression during development are important for the formation of adult phenotypes^140,141, we hypothesized that DEGs distinguishing fast- and slow-biased muscles at each time point would reflect metabolic pathways and molecular markers of myofiber type. As expected, fewer genes were differentially expressed by myofiber bias at E18.5 (n = 56; Fig. 3a) than in adult muscles (n = 5,633; Fig. 3c), likely reflecting the larger myonuclei population observed in adult tissue^65,142 (Supplementary Figs. S4, S5). DEGs at E18.5 were enriched for GO terms associated with broader developmental processes (Fig. 3b) whereas DEGs in adult muscles reflected physiological differences between differentiated fast and slow myofibers (Fig. 3d). Specifically, genes up-regulated in adult fast-biased muscles (n = 1,894) were enriched for GO terms related to fast myofibers (Actn3, Casq1, Mstn, Myh4, Myod1, Tnnc2, Tnni2, Tnnt3) and glycolysis (Gpi1, Pgk1, Pkm, Foxk2, Tpi1, Gapdh, Ldha). Genes up-regulated in adult slow-biased muscles (n = 3,739) were enriched for GO terms related to slow myofibers (Actn1, Myh7b, Pdlim1, Ppargc1a, Tnnc1, Tnnt2) and KEGG pathways related to lipid metabolism (Acat1, Acox1, Hadh) (Supplementary Table S7). As expected, few DEGs distinguishing fast- and slow-biased muscles were shared across time points. Genes up-regulated in fast-biased muscles across time points (n = 5; Fig. 3e) were associated with lipid metabolism^143–145, cell proliferation¹⁴⁶, and oxidative stress responses¹⁴⁷. Those up-regulated in slow-biased muscles across time points (n = 16; Fig. 3f) were associated with muscle development^148–152, cell proliferation^153–155, stem cell renewal^156,157 and adipose tissue^158–160. These results show that genes associated with myofiber pathways are expressed at lower levels in embryonic mouse muscles than in fully differentiated adult muscles.

Fig. 3.

Analysis of DEGs between fast- and slow-biased muscles at a late embryonic (E18.5) and adult time point. (a) Volcano plot showing DEGs between fast-biased (gastrocnemius and vastus lateralis; orange) and slow-biased (soleus and vastus intermedius; blue) muscles at E18.5 (n = 24). Gray dots represent genes that did not meet the significance threshold (|log₂ fold change| ≥ 0.6; adjusted p ≤ 0.05). Gene symbols represent the top 10 DEGs per myofiber bias. (b) Top five enriched GO Biological Process terms for genes significantly up-regulated in E18.5 fast-biased and slow-biased muscles. (c) Volcano plot showing DEGs between fast- and slow-biased muscles at the adult time point (n = 23). (d) Top five enriched GO Biological Process terms for genes significantly up-regulated in adult fast-biased and slow-biased muscles. (e) Venn diagram of genes up-regulated in fast-biased muscles, showing those uniquely up-regulated at E18.5 (light orange), those uniquely up-regulated at the adult time point (dark orange), or those up-regulated at both time points (gene symbols listed below). (f) Venn diagram of genes up-regulated in slow-biased muscles, showing those uniquely up-regulated at E18.5 (light blue), those uniquely up-regulated at the adult time point (dark blue), or those up-regulated at both time points (gene symbols listed below).

Spatiotemporal regulation of MyHC gene expression

To investigate the molecular composition of myofiber development and differentiation, we quantified the expression of eleven MyHC genes (Myh16, Myh15, Myh7b, Myh6, Myh7, Myh13, Myh3, Myh8, Myh4, Myh1, and Myh2) in E18.5 and adult muscles (Supplementary Fig. S6). Ten MyHC genes were detected, as the ancient masticatory gene Myh16^161,162 was lost in the mouse lineage¹⁶³ (Fig. 4a). As expected, extraocular muscle-associated genes Myh15 and Myh13 were the least abundant MyHC gene transcripts in mouse hind limb muscle. While the extraocular-specific slow gene Myh15¹⁶⁴ was up-regulated in adult slow-biased muscles (adjusted p ≤ 0.0001; Fig. 4b), the expression of extraocular fast Myh13 did not significantly vary by time point or myofiber bias (Fig. 4k). As expected, slow MyHC genes Myh7b, Myh6, and Myh7, which encode the ancient slow tonic, cardiac, and slow β/I MyHC isoforms, respectively, were up-regulated in adult slow-biased muscles (adjusted p ≤ 0.0001; Fig. 4c–e). Notably, Myh7b was down-regulated in adult vastus lateralis and gastrocnemius, while Myh6 and Myh7 were down-regulated in adult vastus lateralis, alone. Developmental MyHC genes Myh3 (embryonic) and Myh8 (neonatal) were up-regulated at E18.5 (adjusted p ≤ 0.0001), but remained expressed in adult muscle (Fig. 4f,g). Although MyHC genes were not significantly differentially expressed between pairs of fast- and slow-biased muscles at E18.5, differences in Myh7 and Myh1 expression were detected between E18.5 calf muscles (Supplementary Fig. S7). Among fast MyHC genes, Myh4 (MyHC-IIb) was up-regulated in adult fast-biased muscles (adjusted p ≤ 0.0001; Fig. 4h), whereas Myh2 (MyHC-IIa) and Myh1 (MyHC IIx/d) were paradoxically up-regulated in adult slow-biased muscles (adjusted p ≤ 0.001 and ≤ 0.0001, respectively; Fig. 4i,j), possibly reflecting differences in promoter accessibility (Supplementary Fig. S8). Collectively, these results highlight the time- and muscle-specific regulation of MyHC gene expression in mice.

Fig. 4.

Expression dynamics of the striated muscle MyHC gene family. (a) Phylogenetic tree illustrating the relationships among striated muscle MyHC genes and the isoforms they encode. Genes are classified into three groups: ‘Ancient’, ‘Cardiac’ and ‘Skeletal Muscle’. Branch lengths are not to scale. † denotes gene loss in the mouse lineage. (b–k) VST-normalized expression of MyHC genes across mouse hind limb muscles at E18.5 and adult time points. Vastus lateralis and gastrocnemius represent fast-biased muscles; vastus intermedius and soleus represent slow-biased muscles. Asterisks denote statistical significance: *Padj ≤ 0.05; **Padj ≤ 0.01; ***Padj ≤ 0.001; ****Padj ≤ 0.0001.

Muscle-specific open chromatin regions are enriched for distal intergenic elements

We next used ATAC-seq to identify regions of open chromatin (potential CREs) that may regulate DEGs in a myofiber bias- and time point-specific manner¹⁶⁵. Despite using a protocol to reduce mitochondrial content⁹⁹, adult muscles, specifically those with a higher proportion of slow myofibers, were enriched for regions mapping to the mitochondrial genome (chrM) (χ² = 6.56, df = 1, p = 0.01041; Fig. 5a, Supplementary Table S2), likely reflecting increased mitochondrial abundance^166,167. Although fast- and slow-biased muscles are expected to differ in metabolic properties, no significant difference in chrM mapping was observed at E18.5.

Fig. 5.

Chromatin accessibility across E18.5 and adult muscles varying in slow myofiber content (n = 24). (a) Average number of ATAC-seq reads aligning to the mitochondrial genome (chrM) by myofiber bias at each time point. Data are presented as mean ± SEM (n = 3). (b, c) Proportion of IDR peaks overlapping annotated mm10 genomic features (b) before and (c) after filtering with a mouse brain ATAC-seq peak set. (d) Changes in genomic feature representation following brain filtering. Data are presented as mean ± SEM (n = 8). (e) Average per-bp phyloP60ways conservation scores for fixed-length (400 bp) IDR peaks. Red lines indicate the mean conservation score for each peak set. Positive values reflect increased evolutionary constraint (slower-than-expected DNA sequence change across mammals), while negative values indicate reduced evolutionary constraint (faster-than-expected DNA sequence change across mammals). Asterisks denote statistical significance: *p ≤ 0.05; **p ≤ 0.01. ***p ≤ 0.001; ****p ≤ 0.0001. Abbreviations: V. lat., vastus lateralis; Gast., gastrocnemius; V. int., vastus intermedius; Sol., soleus.

Open chromatin peaks were called using MACS2 and filtered for reproducibility using IDR (Supplementary Table S3). Most peaks overlapped promoters and distal intergenic regions (i.e., CREs), with no significant differences in genomic feature composition across IDR peak sets (Fig. 5b). To enrich for muscle-specific CREs, we excluded peaks showing accessibility in mouse brain¹⁰⁷, thereby removing pleiotropic CREs shared across tissue types¹⁰⁵. As seen in other studies probing for cell-type specific signals using ATAC-seq^98,114, brain filtering here also significantly reduced average peak set size (before: 29,341 ± 10,056; after: 17,122 ± 8,051; t = 15.907, df = 7, p = 0.0000009413), improved GO term enrichment for muscle-related processes (Supplementary Fig. S9), and altered genomic feature distributions (F_8,126 = 123.8, p < 2e-16; Fig. 5c, Supplementary Table S8). After filtering, the proportion of peaks in promoters significantly decreased (F_1,14 = 117.9, p < 3.34e-08), while peaks in distal intergenic regions (F_1,14 = 100, p < 9.33e-08), first exons (F_1,14 = 28.83, p < 9.89e-05), other exons (F_1,14 = 56.81, p < 2.72e-06), first introns (F_1,14 = 85.69, p < 2.41e-07), other introns (F_1,14 = 78.16, p < 4.2e-07), 3’ UTRs (F_1,14 = 16.79, p < 0.0011), and downstream regions (F_1,14 = 22.85, p < 0.00029) significantly increased (Fig. 5d). Genomic feature distributions were consistent across brain-filtered peak sets. Brain filtering also significantly reduced peak conservation scores (before: 0.3331 ± 0.006; after: 0.2714 ± 0.007; t = 18.895, df = 7, p = 2.892e-07; Fig. 5e), suggesting that pleiotropic CREs are under greater evolutionary constraint (DNA sequence conservation due to negative selection) than tissue-specific CREs. Overall, this filtering approach enhanced the identification of CREs relevant to muscle biology, a key step toward identifying CREs that shape myofiber development and physiology.

Condition-specific peak sets are enriched with TF motifs important for myogenesis and myofiber differentiation

To identify candidate CREs that may regulate genes influencing muscle development and identity, we distilled 136,976 brain-filtered peaks into nine non-overlapping ‘condition-specific’ peak sets (potential muscle-specific CREs). These included peaks unique to muscles of a specific time point and myofiber bias (E18.5 Slow, Adult Slow, E18.5 Fast and Adult Fast), peaks shared across all muscles of a given time point (E18.5 and Adult) or myofiber bias (Slow and Fast), and peaks shared across all muscles (Universal) (Fig. 6a). Only 25,436 peaks (18.6%) fell into these categories (Supplementary Table S9), with the majority being exclusive to E18.5 muscles (n = 15,294). More peaks were shared across Adult Fast-biased muscles (n = 3,641) than Adult Slow-biased muscles (n = 1,470), likely reflecting the greater phenotypic similarity between the gastrocnemius and vastus lateralis compared with the soleus and vastus intermedius^71–78. Many peaks were Universal (n = 3,005), which may represent pleiotropic CREs with roles in regulating muscle identity or shared housekeeping functions. Indeed, genes near Universal peaks were enriched for GO terms related to muscle physiology, metabolism, and development (Supplementary Table S10, Supplementary Fig. S10). Few peaks were shared across time points in Slow-biased (n = 78) or Fast-biased (n = 231) muscles, suggesting limited overlap in gene regulation across time points.

Fig. 6.

Characterization of condition-specific peak sets. (a) Venn diagram showing the number of open chromatin regions in each of the nine condition-specific peak sets. (b) Average per-bp phyloP60ways conservation scores for fixed-length (400 bp) condition-specific peaks. Red lines indicate the mean conservation score for each peak set. Positive values reflect increased evolutionary constraint (slower-than-expected DNA sequence change across mammals), while negative values indicate reduced evolutionary constraint (faster-than-expected DNA sequence change across mammals). (c) Average per-bp phyloP60ways conservation scores for each peak set compared to a random background (grey bars). Asterisks denote significant differences from a random background: p ≤ 0.05; **p ≤ 0.01. (d) Top 15 enriched transcription factor (TF) binding site motifs identified in each peak set, grouped by TF family.

We next assessed sequence conservation across condition-specific peak sets, hypothesizing that peaks shared by all muscles from the same time point (E18.5 and Adult) and peaks shared across time points (Universal, Fast, and Slow) would be more conserved than peaks unique to a specific time point and myofiber bias. Average per-bp conservation scores were generally positive (0.220 ± 0.003), indicating some degree of evolutionary constraint across condition-specific peak sets (Fig. 6b). E18.5 (0.290 ± 0.006, p = 0.004), Adult (0.272 ± 0.01, p = 0.004), Universal (0.268 ± 0.008, p = 0.004), E18.5 Fast (0.193 ± 0.007, p = 0.004), Adult Slow (0.164 ± 0.01, p = 0.004), Adult Fast (0.162 ± 0.007, p = 0.004), and E18.5 Slow (0.134 ± 0.005, p = 0.02) sets were significantly more conserved than a random background, with E18.5, Adult, and Universal sets being the most conserved (Fig. 6c). Fast (0.093 ± 0.02) and Slow (0.08 ± 0.04) sets did not significantly differ from the random background.

Finally, we examined the potential regulatory function of condition-specific peaks. Genomic feature distributions were consistent across peak sets, with peaks predominantly residing in distal intergenic regions (30.6–52.3%), introns (26.4–46.2%) and promoters (11.5–25.6%) (Supplementary Table S11). On average, 66.9% of peaks overlapped active enhancer (H3K27ac) marks^57,65,118, consistent with previous studies^107,168,169 (Supplementary Table S12). We next assessed transcription factor (TF) motif enrichment, hypothesizing that condition-specific peak sets would be enriched for TFs involved in myogenesis and myofiber differentiation. E18.5 peak sets were enriched for bHLH and bZIP motifs (Fig. 6d), which regulate muscle development and differentiation^170,171. Adult peak sets were enriched for MADS and NR motifs, which control muscle growth, regeneration, and metabolism^172–175. The Universal peak set, which likely represents pleiotropic CREs active across all muscles across time points, was enriched for both bZIP and MADS motifs, consistent with a role in maintaining muscle cell identity and function throughout life^176–178. These results suggest that condition-specific peaks likely act as CREs that regulate the expression of genes involved in muscle development, differentiation, and identity.

Nine candidate CREs enhance and three suppress luciferase activity in muscle cells

To test whether condition-specific peaks (potential CREs) influence gene expression in vitro, we selected a subset of peaks unique to fast- and slow-biased muscles for functional validation, as these represent compelling candidates for investigating the genetic basis of muscle myofiber composition. To identify candidates, we first used condition-specific peak sets to annotate open chromatin regions surrounding DEGs between adult fast- and slow-biased muscles (see example in Supplementary Fig. S2). Of particular interest were candidates accessible exclusively in Slow and Adult Slow peak sets (n = 1,548) and located near slow myofiber genes up-regulated in slow-biased muscles, as well as candidates accessible in Fast and Adult Fast peak sets (n = 3,872) and located near fast myofiber genes up-regulated in fast-biased muscles. We prioritized peaks with accessibility patterns mirroring the expression of nearby (< 100 kb) DEGs (e.g., peaks accessible only in adult slow-biased muscles near genes up-regulated only in adult slow-biased muscles) and exhibiting elevated per-bp conservation scores, as conservation of sequence often indicates conservation of function^179,180. This approach yielded 260 candidate regions (Supplementary Table S4). From these, we selected 12 candidate CREs for functional validation: nine associated with fast-biased muscles and three with slow-biased muscles (Supplementary Table S6). To evaluate their capacity to drive gene expression in vitro, each candidate was tested using a dual-luciferase reporter assay in an undifferentiated mouse muscle cell line. Results were consistent across experiments (Fisher’s combined probability test, p ≤ 0.000000004; Supplementary Fig. S11). Nine candidates significantly increased luciferase activity relative to the empty control vector (Fig. 7), with four exceeding a 2-fold increase (p ≤ 0.0001): Slow 2 (log₂FC = 4.28), Fast 1 (3.58), Slow 3 (2.38) and Fast 3 (1.66). In contrast, candidates Fast 8, Fast 9, and Slow 1 significantly suppressed luciferase activity (negative log₂FC, p ≤ 0.0001). The direct effect of each candidate on its putative target gene, however, remains unknown. These results suggest that condition-specific peaks can function as CREs capable of influencing gene activity in vitro, with candidates Fast 1–7, Slow 2, and Slow 3 likely acting as enhancers that increase gene expression. Further experiments in differentiated muscle cells are required to assess their regulatory activity in mature myofibers.

Fig. 7.

Functional validation of candidate CREs using a dual-luciferase reporter assay in C2C12 cells. Relative luciferase activity driven by 12 candidate CREs in mouse muscle cells. CREs associated with fast myofiber gene pathways are shown in orange; those associated with slow myofiber gene pathways are shown in blue. Positive log₂ fold change (log₂FC) values indicate enhanced luciferase activity, while negative values indicate suppressed luciferase activity relative to the empty pGL4.23 control vector. Data are presented as means ± SEM (n = 8). Asterisks denote statistically significant differences from the control vector: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001.

Discussion

Here, we used RNA-seq and ATAC-seq to profile the regulatory landscape of two pairs of mouse hind limb muscles differing in MyHC expression (two calf and two thigh muscles) at a late embryonic (E18.5) and adult time point. By integrating patterns of gene expression and chromatin accessibility, we identified key molecular signals associated with dynamic changes in gene regulation across tissues and time points, thereby refining our understanding of the genetic pathways underlying myogenesis and the molecular establishment of muscle identity. Our results extend previous studies^{12,55–57,62,181} that suggest that CREs orchestrate tissue and time-specific patterns of gene expression important for myogenesis and myofiber differentiation. Furthermore, we demonstrate that nine condition-specific peaks (candidate CREs) enhance gene expression in vitro, warranting further investigation in differentiated muscle cells given their compelling specificity to fast- and slow-biased muscles.

Given the distinct biological processes occurring in developing versus fully developed muscle, most variation in gene expression and chromatin accessibility was attributed to time point. Embryonic muscles exhibited significantly more DEGs and open chromatin regions, consistent with previous studies highlighting the dynamic regulatory landscape of developing tissue^182,183. As expected, few genes up-regulated at E18.5 remained so in adulthood, reflecting the dynamic shifts in gene expression and cellular composition that occur over the course of tissue development and differentiation^184–189. Notable exceptions, such as Hoxc6 and Hoxc8 in thigh musculature, highlight the importance of such genes in maintaining muscle identity throughout life^190,191. One of the defining features of embryonic development is limb patterning¹⁹², which relies on the precise spatiotemporal expression of genes that define limb segments and establish muscle identity¹²³. Accordingly, gene expression patterns in embryonic muscles primarily reflected their anatomical location within the hind limb (e.g., calf vs. thigh). In contrast, chromatin accessibility patterns did not show this anatomical specificity, suggesting that limb muscle development may be regulated by a shared set of pleiotropic CREs or super-enhancers^193–195. Alternatively, these patterns may also reflect the greater phenotypic similarity among thigh muscles¹²⁴.

In adult muscles, patterns of gene expression and chromatin accessibility primarily reflected differences in myofiber metabolism and contractile properties. DEGs and accessible chromatin regions were enriched for pathways characteristic of each myofiber type: glycolytic pathways in fast-biased muscles, and mitochondrial biogenesis and lipid metabolism in slow-biased muscles. Adult slow-biased muscles exhibited greater mitochondrial read abundance, consistent with the high mitochondrial content of slow myofibers, which rely on oxidative phosphorylation for energy^16,17,30,196. In contrast, at E18.5, genes associated with glycolysis and lipid metabolism were not significantly differentially expressed between fast- and slow-biased muscles, and mitochondrial read abundance was low, indicating that developing myofibers lack the mitochondrial density and metabolic activity typical of mature myofibers^16,17,30. It is expected that as differentiation and muscle activation patterns progress, mitochondrial content and the expression of genes involved in lipid metabolism will dramatically increase¹⁹⁷, consistent with the enrichment of these pathways in adult muscles.

As expected, expression of slow MyHC genes Myh6, Myh7 and Myh7b was highest in adult slow-biased muscles (soleus and vastus intermedius), whereas expression of fast Myh4 peaked in adult fast-biased muscles (gastrocnemius and vastus lateralis). Myh4 was the most highly expressed MyHC gene overall, consistent with the enrichment of MyHC-IIb in the muscles of small-bodied mammals like mice^198–202. Although transcripts of both fast and slow MyHC genes were detected in embryonic muscle, their expression levels were lower than in adults, suggesting that the molecular profile of hind limb muscles at E18.5 largely reflects a population of immature myofibers, consistent with the high expression of developmental MyHC isoforms (Myh3, Myh8)¹⁸. Indeed, none of the fast (Myh4, Myh2, Myh1) or slow (Myh7, Myh7b) MyHC genes were significantly differentially expressed between pairs of fast- and slow-biased muscles at E18.5, although differences in Myh7 and Myh1 expression were observed between E18.5 calf muscles, whose adult slow myofiber content is more divergent than that of the thigh muscles used for this study^71–78. External postnatal stimuli (e.g., neural stimulation) are expected to enhance the expression of fast and slow MyHC genes to adult levels after this stage^{23–26,68,203–205}. Interestingly, expression of Myh2 and Myh4 was significantly elevated in E18.5 thigh muscles compared to E18.5 calf muscles, consistent with previous studies suggesting that myofiber differentiation may occur earlier in the thigh^125,126, although this pattern may also reflect the more similar myofiber composition of thigh musculature^71–78. Unexpectedly, Myh2 and Myh1 showed the highest expression in adult slow-biased muscles. Although these isoforms are known to be highly expressed in the mouse soleus²⁰⁶ and may be associated with slow myofiber composition²⁰⁷, this expression pattern likely reflects the substantial proportion of fast myofibers typically observed in slow-biased mouse muscles^27,71. Of note was the elevated expression of Myh15 in adult slow-biased muscles. Although transcripts have been detected beyond extraocular muscles^{161,208–210} this pattern may reflect coordinated sensory-motor function in intact muscle with varying contractile phenotype.

Of particular interest to understanding the observed variation in fast MyHC gene expression is the open chromatin landscape surrounding the fast MyHC gene cluster. While the fast MyHC super-enhancer^55,56 was accessible in both embryonic and adult muscle, individual fast MyHC gene promoter accessibility varied. The Myh4 promoter was broadly accessible across adult muscles with larger populations of fast myofibers (i.e., gastrocnemius, vastus lateralis and vastus intermedius), with a secondary promoter uniquely accessible in the adult gastrocnemius, correlating with elevated Myh4 expression in this muscle. In contrast, the Myh2 promoter was accessible only in the adult vastus intermedius, which also exhibited the highest Myh2 expression. Myh1 promoter accessibility was more widespread, particularly in the adult soleus and vastus intermedius, potentially explaining the elevated expression of Myh1 in these slow-biased muscles. In embryonic muscles, both promoter accessibility and gene expression of Myh2, Myh1, and Myh4 were minimal, indicating that super-enhancer activity alone is insufficient to drive strong gene expression. In comparison, the promoter of prenatal Myh8 was accessible across embryonic muscles, consistent with its elevated expression at E18.5. These findings suggest that accessibility of the fast MyHC super-enhancer alone does not reliably predict gene expression. Instead, expression depends on both promoter accessibility and the binding of regulatory proteins that mediate enhancer-promoter interactions²¹¹. Consequently, species-specific differences in muscle myofiber composition may arise from coordinated changes not only to CRE sequences that affect regulatory protein binding, but to the expression of genes that regulate the accessibility of CREs throughout development via epigenetic modifications^212,213.

Uncovering the genetic basis of muscle myofiber composition requires identifying CREs active during muscle development that govern myofiber differentiation and identity. Of particular interest are CREs unique to specific cell types or tissues, which are predicted to evolve more rapidly than pleiotropic CREs shared across cell types, highlighting their potential to shape species-specific traits²¹⁴. By targeting regions of open chromatin uniquely accessible in adult fast- and slow-biased muscles (Adult Slow, Adult Fast, Fast, and Slow condition-specific peak sets), we identified thousands of candidate CREs that may drive the expression of fast and slow myofiber genes in a time point- and muscle-specific manner. Intriguingly, these peak sets exhibited the lowest conservation scores, suggesting that these regions are under relaxed evolutionary constraint and more susceptible to DNA sequence change that may contribute to myofiber composition variation within and between species. Such regions are especially relevant to studies of locomotor evolution, as mutations in CREs that alter TF binding, and thus the timing and location of gene expression, are a well-established source of evolutionary novelty and phenotypic diversity²¹⁵. Species-specific variation within these CREs may underlie differences in myofiber composition and contribute to the evolution of diverse locomotor strategies that benefit from muscles with distinct proportions of fast and slow myofibers (e.g., humans with slow-biased lower limb muscles; moles with fast-biased shoulder musculature)^37,124. As an example, human-specific mutations that promote slow myofiber pathways by increasing slow CRE activity or repressing fast CRE activity may have been under positive selection in the human lineage. In support of this, we identified twelve muscle-specific CREs associated with fast and slow myofiber pathways capable of influencing gene activity in vitro. It is tempting to speculate that fast myofiber pathways are driven by enhancer activity of candidate CREs Fast 1–7, whereas slow myofiber pathways are promoted by candidate CREs Slow 2–3. Further functional validation in differentiated myofibers is needed to determine whether species-specific variation within these CREs contributes to differences in muscle myofiber composition across species.

Our results offer a snapshot of the regulatory landscape of four hind limb muscles at a late embryonic and adult time point, highlighting the spatiotemporal regulation of genes involved in limb muscle patterning and the refinement of adult muscle identity. We identified nine muscle-specific CREs that enhance gene expression in vitro and represent compelling targets for further analysis in differentiated muscle cells, given their specificity to fast- and slow-biased muscles and potential roles in shaping myofiber pathways and muscle myofiber composition. Testing the regulatory function of these CREs, along with the hundreds of others we identified, in vivo and evaluating the impact of species-specific sequence variation on target gene expression will help elucidate the genetic basis of species-specific differences in muscle myofiber composition.

In this study, we note the following limitations. First, MyHC isoform levels were not directly measured, so we relied on the consistent and widely reported patterns of MyHC isoform expression in murine muscle for our interpretation^71–78. Rodent muscle, which tends to exhibit relatively homogenous MyHC isoform expression across myofibers compared to humans, provided a useful system for investigating how CREs regulate myogenesis and myofiber differentiation. Second, because RNA-seq and ATAC-seq datasets were not matched, direct comparisons between these two data sets are limited. Third, epigenetic profiles were also not generated from single cells^216,217, so signals specific to fast and slow myofibers may be obscured by the diverse cell types present in muscle tissue^218,219. Future studies using matched scRNA-seq and scATAC-seq samples on single myofibers with known MyHC content from both male and female mice, together with in vivo assays^220–224, will be essential to extending these findings. Fourth, our filtering approach for ATAC-seq peaks, designed to identify CREs with high muscle specificity, excludes pleiotropic CREs active in both brain and muscle that may also contribute to myofiber identity and differentiation. Fifth, because fast and slow myofiber pathways are primarily activated in differentiated myofibers, functional studies in these cells will be crucial to validate whether candidate CREs can drive gene activity in a native context.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

SRQ, TDC, and KNS conceptualized the study; KNS and TDC supervised the project; ASO and SRQ performed the experiments; SRQ wrote the manuscript, analyzed the data, and prepared the figures; MCO and DMC substantially revised the manuscript for intellectual content, and all authors edited the manuscript. All authors approved the final version of the manuscript.

Data availability

All ATAC-seq and RNA-seq datasets generated as part of this study have been deposited to the NCBI GEO repository under accession number GSE292230 and GSE292232. All data included in this study are available upon reasonable request by contact with the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.