The unique functions of Runx1 in skeletal muscle maintenance and regeneration are facilitated by an ETS interaction domain

ABSTRACT

The conserved Runt-related (RUNX) transcription factor family are master regulators of developmental and regenerative processes. Runx1 and Runx2 are expressed in satellite cells (SCs) and in skeletal myotubes. Here, we examined the role of Runx1 in mouse satellite cells to determine the role of Runx1 during muscle differentiation. Conditional deletion of Runx1 in adult SCs negatively impacted self-renewal and impaired skeletal muscle maintenance even though Runx2 expression persisted. Runx1 deletion in C2C12 cells (which retain Runx2 expression) identified unique molecular functions of Runx1 that could not be compensated for by Runx2. The reduced myoblast fusion in vitro caused by Runx1 loss was due in part to ectopic expression of Mef2c, a target repressed by Runx1. Structure-function analysis demonstrated that the ETS-interacting MID/EID region of Runx1, absent from Runx2, is essential for Runx1 myoblast function and for Etv4 binding. Analysis of ChIP-seq datasets from Runx1 (T cells, muscle)- versus Runx2 (preosteoblasts)-dependent tissues identified a composite ETS:RUNX motif enriched in Runx1-dependent tissues. The ETS:RUNX composite motif was enriched in peaks open exclusively in ATAC-seq datasets from wild-type cells compared to ATAC peaks unique to Runx1 knockout cells. Thus, engagement of a set of targets by the RUNX1/ETS complex define the non-redundant functions of Runx1 in mouse muscle precursor cells.

Summary: Runx1 non-redundantly regulates a gene network in mouse satellite cells that is required for muscle maintenance and repair, which is mediated through interactions with Etv4 at ETS:RUNX composite motifs.

INTRODUCTION

The Runt-related transcription factors (RUNX1, RUNX2, and RUNX3) play essential roles in diverse tissues regulating homeostasis, cell proliferation, lineage specification, and cell fate determination. RUNX1 and RUNX2 proteins bind to the same consensus DNA sequence (TGTGGT) through a highly conserved (>95%) RUNT DNA-binding domain at the N terminus, and to the CBFβ transcription factor through the proline-rich (PY) motif (Kamikubo, 2018). RUNX family members act redundantly in many developmental and disease processes. For example, loss of the pro-oncogenic activity of RUNX1 in acute myeloid leukemia can be compensated for by elevated expression of RUNX2 and RUNX3. Only simultaneous loss of all RUNX activity by reagents designed to bind consensus RUNX-binding sequences or by shRNA lentiviruses targeting all three RUNX orthologs provided the desired anti-acute myeloid leukemia effect (Morita et al., 2017). By contrast, Runx1 has non-redundant activity in T cells (Taniuchi et al., 2002; Wong et al., 2011) and engages specific targets in mK4 cells (Hass et al., 2021), whereas Runx2 regulates osteogenesis.

Immunostaining of muscle with antibody for Runx1 detected it in adult mouse satellite cells (SCs) (Umansky et al., 2015b), a resident stem cell population in skeletal muscle. Skeletal muscles account for ∼40-50% of adult human body weight and comprise a multinucleated syncytium formed during development by fusion of mononucleated progenitor cells. Muscles have a remarkable capacity for repair and regeneration in response to exercise, injury, aging, or disease, driven by quiescent SCs. In response to injury, SCs are activated, re-enter the cell cycle, and divide to replenish the pool of quiescent stem cells and produce a transiently amplifying population of committed progenitors or myoblasts. The myoblasts expand, then differentiate and fuse with each other or with existing fibers to regenerate a functional muscle (Brack and Rando, 2012; Lepper et al., 2011; Pawlikowski et al., 2015). Quiescent SCs require the paired box 7 (Pax7) protein (Gunther et al., 2013; von Maltzahn et al., 2013) to maintain responsiveness of a gene expression network to muscle injury that enables regeneration. Transition to myoblasts is controlled by the sequential expression of myogenic regulatory factors (MRFs), a family of basic helix-loop-helix transcription factors that includes Myf5, MyoD (Myod1), followed by myogenin and Mrf4 (Myf6) (Penn et al., 2004). The MRFs collaborate with Mef2 family members to drive expression of structural and metabolic genes required in mature muscle fibers (Blais et al., 2005; Braun and Gautel, 2011; Molkentin and Olson, 1996; Potthoff and Olson, 2007).

It was thought that Runx1 is activated in quiescent SCs after injury to skeletal muscle and genetic analyses demonstrated it is essential for regeneration in an MDX mouse model of muscle injury (Umansky et al., 2015b). Genome-wide analysis of Runx1-occupied regions coupled with gene expression analyses revealed enrichment for the RUNX, MyoD and AP-1/c-Jun motifs in primary myoblasts (Umansky et al., 2015a), and identified key targets, including Mef2c and Cdkn1c (p57). Moreover, loss of Runx1 exacerbates the muscle-wasting phenotype of MDX mice leading to early lethality (Umansky et al., 2015b). However, the role of Runx1 in maintenance of muscle SCs was not analyzed, in part because Runx1 was not detected in uninjured wild-type (WT) muscle by the investigators, which led the authors to focus on the effect of Runx1-deficiency during muscle regeneration in MDX mice (Umansky et al., 2015b). Although Myf5 is expressed in SCs, Myf5-Cre has been shown to be inefficient at low expression levels and thereby likely fails to delete floxed Runx1 in SCs to permit muscle maintenance in the absence of injury (Comai et al., 2014). Alternatively, two progenitors populations of SCs have been identified: one that is Myf5/Pax7 double positive and fast cycling, and a second slower cycling population that is Pax7 positive but Myf5 negative (Beauchamp et al., 2000; Picard and Marcelle, 2013). Targeting Runx1 for deletion with Myf5-Cre, even if it was sufficiently expressed to delete Runx1 in fast cycling SCs, would spare the slow cycling SC progenitors and thus confound the analyses.

We investigated the role of Runx1 protein in SC maintenance and regeneration in WT animals using a Pax7-CreERT2 that targets all SCs and provides temporal control over the deletion timing (Lepper and Fan, 2010, 2012). Conditional deletion of Runx1 in SCs with Pax7-CreERT2 impaired regeneration of injured skeletal muscle, confirming the essential role of Runx1 in WT muscle and alleviating the concern that Myf5-negative SC somehow confounded the interpretation of the previous study (Umansky et al., 2015b). Additionally, this model allowed us to show that Runx1 plays a role in maintenance of quiescent SCs: adult uninjured muscle deteriorated over time due to loss of Runx1.

To understand the unique pro-myogenic function of Runx1 and explore the basis for non-redundancy between Runx1 and Runx2 in myoblasts, we used CRISPR to remove Runx1 from C2C12 myoblasts (termed Runx1KO), which retain expression of Runx2. Upon transition to differentiation media, Runx1-deficient C2C12 cells activated myogenic differentiation genes but failed to fuse into myotubes. The crucial SC transcription factor Pax7 was decreased in quiescent Runx1KO cells, suggesting that Runx1 contributes to Pax7 maintenance. Conversely, Mef2c expression increased in differentiating Runx1KO C2C12 cells, and overexpression of Mef2c in otherwise WT C2C12 cells was sufficient to mimic the fusion defect seen in Runx1KO myoblasts. To further investigate the molecular basis for the functional differences between Runx1 and Runx2, we generated Runx1/2 chimeras and tested their ability to rescue differentiation of Runx1KO C2C12 myoblasts. This assay identified the Runx1 MID domain (MID1), which contains an ETS-interaction helix as a critical region required for the unique role of Runx1 in myoblast fusion. Co-immunoprecipitation experiments confirmed that Runx1 (but not Runx2) can form a complex with the ETS factor Etv4 in C2C12 cells. Finally, analysis of C2C12 ATAC-seq data as well as published chromatin immunoprecipitation assay with sequencing (ChIP-seq) datasets from Runx1- or Runx2-dependent tissues revealed significant enrichment for a composite ETS:RUNX site specifically in Runx1-dependent tissues, but not in Runx2-dependent tissue such as bone. Altogether, our findings identify Runx1 as a key molecule reliant on complex formation with Etv4 to regulate Pax7 and Mef2 abundance, SC maintenance, and skeletal muscle regeneration.

RESULTS

Deletion of Runx1 in SCs impairs skeletal muscle regeneration in otherwise WT mice

A previous study on the role of Runx1 in muscle did not analyze its function in uninjured adult muscle since it was not detected in SCs with reagents available at the time (Umansky et al., 2015b). However, immunostaining of the tibialis anterior (TA) muscle with the rabbit monoclonal antibody 8529 detected Runx1 in Pax7-positive cells, confirming expression in quiescent adult SCs (Fig. S1A, controls). This left open the possibility that Runx1 helps maintain a quiescent SC pool. In order to delete Runx1 in both SC progenitor populations, to gain temporal control of deletion timing, and to examine its role in maintaining adult muscle homeostasis, we used Pax7^+/CreERT2. Pax7^+/CreERT2; Runx^fl/fl mice (referred to henceforth as Runx1cKO) and control Pax7^+/CreERT2; Runx1^+/fl littermates (referred to henceforth as controls; Fig. 1A). Without tamoxifen, no differences were observed between Runx1cKO and controls over their entire lifespan. To examine the impact of Runx1 deletion in adult SCs on regeneration after injury, we administered tamoxifen daily for 5 days to 6-week-old littermates to induce genetic deletion of Runx1 in adult SCs (Fig. 1B). Immunofluorescence staining after tamoxifen administration (but before application of cardiotoxin, see below) showed that most (∼98%) Pax7-positive cells in control muscles express Runx1 protein, whereas very few Pax7-positive cells expressed Runx1 in Runx1cKO muscles (Fig. S1A,B). Moreover, the number of Pax7-positive SCs has significantly declined in un-injured Runx1cKO muscle, suggesting a role for Runx1 in the maintenance of adult SCs. Four days after the last dose of tamoxifen, we injected cardiotoxin (CTX) into the left TA muscle and harvested the injured and contralateral control muscles at day (D) 5, D7, D12, D30 and D60 post injection (Fig. 1B). We imaged the dissected TA muscles to compare the size (Fig. 1C) and weight (Fig. 1D) of the regenerating muscle and found that Runx1cKO muscles were significantly smaller than controls.

Fig. 1.

Deletion of Runx1 in SCs impairs regeneration of skeletal muscle. (A) Breeding scheme to obtain control and Runx1cKO mice (red text). (B) Schematic outline of the strategy for tamoxifen and cardiotoxin (CTX) administration. See text for detail. (C) Representative image of dissected tibialis anterior (TA) muscle from control and Runx1cKO mice at D30 post-injury. (D) Quantification of TA weight in control and Runx1cKO mice (n=5). (E) Quantification of the percentage of Runx1-positive cells in Pax7-expressing cells per area from injured TA muscle cross-sections of control and Runx1cKO mice at 5 days post-injury. n=4. (F) Cross-sections of TA muscles from control and Runx1cKO mice analyzed by H&E staining on days 5, 7, 12, 30 and 60 after saline or CTX injection. Individual channels are shown for D60 in Fig. S2B. Cells double positive for Pax7 and Ki67 are identified as quiescent SCs. Scale bars: 80 μm. Data are represented as mean±s.d. ***P<0.001 (two-tailed unpaired Student's t-test).

Next, we analyzed the regeneration capacity of the injured muscle by histology and immunofluorescence. Hematoxylin and Eosin (H&E) stained sections of control or Runx1cKO TA muscle were indistinguishable in size and tissue architecture in the absence of injury (Fig. 1F). CTX induced extensive muscle damage and inflammatory infiltrate in both control and Runx1cKO muscle after injury (Fig. 1F). Staining for Pax7 and Runx1 of injured Runx1cKO muscle confirmed Runx1 loss from the Pax7-positive SCs (Fig. 1E). Whereas control mice began to recover by D7 and completed recovery by D30 post-injury (Fig. 1F), Runx1cKO muscles still contained degenerating myofibers, fibrotic tissues, and inflammatory cells on D7 post-injury. Very few small regenerating fibers were seen in Runx1cKO TA muscle, perhaps arising from SCs in which deletion of Runx1 failed to occur or was delayed. Staining for Pax7 and Ki67 (Mki67) at D5 showed that control muscle contained many Pax7/Ki67 double-positive cells, whereas the remaining Pax7-positive cells in Runx1cKO muscles were mostly Ki67 negative (Fig. S1C and quantified in S1D). Control muscle was fully repaired by D60 and displayed colocalized Pax7/Runx1 double-positive SCs that were no longer Ki67 positive (i.e. quiescent; Fig. 1F, Fig. S2). By contrast, Runx1cKO muscle tissue was not recovered by D60 in Runx1cKO mice and Pax7 staining was greatly reduced (Fig. S2A). The Ki67-positive cells in the Runx1cKO muscle at D60 (Fig. 1F, Fig. S2B) were likely immune infiltrates, as shown by the increased CD45 (Ptprc)-positive cells (Fig. S2C). Thus, the effects of Pax7-CreERT2-mediated deletion of Runx1 in a CTX injury model mirror the published report utilizing Myf5-Cre in the MDX mouse model (Umansky et al., 2015b).

Runx1 is essential for both embryonic development and adult maintenance of SCs

To determine whether Runx1 is required in SCs developing in the offspring during gestation, tamoxifen injections were administered to pregnant females on embryonic day (E) 13.5 of pregnancy (Fig. 2A), a time chosen based on studies demonstrating exclusive labeling of the muscle linage (Lepper and Fan, 2010). At postnatal day (P) 2, the average body length of Runx1cKO mice was significantly shorter than that of controls. The average body weight was ∼62% of that of the control group (Fig. 2B,C). By P30, the difference between Runx1cKO mice and control was more pronounced (Fig. 2B,C). With induction at E13.5, Runx1 loss in other Pax7-expressing tissue is unlikely to have contributed to the phenotype (Lepper and Fan, 2010). Additionally, deletion of Pax7 from embryonic myoblasts using Myf5-Cre has been shown to produce a very similar phenotype (Gunther et al., 2013), which suggests that the growth phenotype we observed reflects loss of muscle mass due to defects in SC maintenance or differentiation, involving Runx1 (and Pax7, if it were regulated by Runx1 in SCs).

Fig. 2.

Runx1 is essential for SC proliferation and maintenance. (A) Injection scheme for analyzing the role of Runx1 in SCs during development. (B) Representative image of newborn (P2) and weaned (P30) mice exposed to tamoxifen on day 13.5 of pregnancy (E13.5). (C) Average body length and weight at P2 and P30 for each genotype in control and Runx1cKO mice (n=3). (D) TA muscle cross-sections at 7 months after tamoxifen administration stained with H&E or Pax7 antibodies (green) and Hoechst (blue). Scale bar: 80 μm. (E) Top: Quantification of the Pax7-expressing cell fraction of DAPI-positive cells. Bottom: Average CSA (cross-sectional area) in control and Runx1KO fields. Randomized fields presented as mean±s.d. n=3 for each group. (F,G) Bar graphs showing the Runx1-dependent Notch-induced expression changes of Col6a2, Col6a1, and Cdh15 in control versus Runx1KO mK4 cells exposed to control FC ligand or DLL1 ligand for 4 h. (H) Bar graph showing that Cdh15 expression induction by an activated Notch construct (N1ΔE) in cells in which CRISPR Cas9 was used to delete the open intronic region bound by Notch and Runx1 in mK4 and myoblasts. IEKO, Intronic Enhancer Knock-Out. Data in F-H are mean±s.d. (n=3) for each condition. (I,J) Genomic snapshots of the Notch transcriptional targets Col6a2, Col6a1, and Cdh15 showing the normalized signal from ATAC-seq from WT (blue) or Runx1KO (red) mK4 cells; Notch complex binding detected by SpDamID in WT (blue) or Runx1KO (red) mK4 cells; ATAC-seq from WT (blue) or Runx1KO C2C12 (red), and ATAC-seq and Runx1 ChIP from primary myoblasts. Red boxes highlight Runx1-bound accessible genomic regions that are reduced or absent in Runx1KO cells. *P<0.05, **P<0.01; ***P<0.001 (two-tailed unpaired Student's t-test).

To investigate the effect of Runx1 deletion on adult SC cell maintenance, we removed Runx1 from quiescent SC cells by administration of tamoxifen daily for 5 days to 8-week-old control and Runx1cKO littermates. After 7 months, the mice were euthanized, and the TA muscle was harvested for analysis. H&E staining showed gaps in fiber packing and smaller muscle cross-sections (Fig. 2D, quantified in Fig. 2E). Immunofluorescence staining revealed that an average of 62.25% of nuclei were Pax7-positive per field of view, whereas Runx1cKO mice had only 37.5% (Fig. 2E). The loss of Pax7-positive SCs suggests that in uninjured muscle Runx1 acts to maintain quiescent SCs. Alternatively, loss of Runx1 during postnatal life could cause an early proliferative defect that reduces the available SC pool later in life.

Quiescent SC maintenance depends in part on the Notch signaling pathway (Bi et al., 2016; Conboy et al., 2003; Fujimaki et al., 2018; Gioftsidi et al., 2022). Previous studies have shown that Runx1 can affect Notch signaling in T cells (Choi et al., 2017; Wang et al., 2014). We noticed that the Runx1-deficient mouse kidney metanephric mesenchymal cell line mK4 (Hass et al., 2021) displayed significant enrichment for shared open chromatin regions with proliferating C2C12 cells, suggesting that these mesoderm-derived cells have similar regulatory networks. Interestingly, Notch targets important to SC maintenance, namely the SC marker Cdh15 and the extracellular matrix proteins Col6a1 and Col6a2 (Baghdadi et al., 2018), were robustly expressed upon exposure of mK4 cells to the Notch ligand DLL1 (Hass et al., 2021), and we observed this Notch-dependent induction to be abrogated in Runx1KO mK4 cells (Fig. 2F,G). We performed assay for transposase-accessible chromatin with sequencing (ATAC-seq) in control and Runx1KO mK4 cells and observed a dramatic loss of open chromatin in Runx1KO mK4 cells near these genes (Fig. 2I,J). We performed ATAC-seq on WT C2C12 and Runx1KO C2C12 cells, described in detail later, and observed decreased Notch and Runx1 accessibility near Cdh15, Col6a1 and Col6a2 genes, although not at the same magnitude seen in mK4 cells. Additionally, mapping the C2C12 ATAC-seq reads onto to the Runx1-dependent ATAC-seq sites or Runx1 ChIP-seq peaks from mK4 cells further confirmed the existence of substantial genomic regulatory overlap between C2C12 and mK4 cells (Fig. S3A). Furthermore, SpDamID (Hass et al., 2015) demonstrated Notch/RBPj complex binding to regions in Runx1-positive cells, but not in Runx1KO mK4 cells, that are accessible in myoblasts and recoverable by Runx1 ChIP-seq (Umansky et al., 2015a,b; Zhang et al., 2020). To test the assumption that these Runx1-dependent regions are involved in Notch-dependent regulation, we used CRISPR to delete the Runx1-dependent ATAC peak bound by Notch/RBPj in the Cdh15 intronic region shared in mK4 cells and myoblasts (Fig. S3B). Deleting this region abolished the Notch responsiveness of Cdh15 in mK4 cells (Fig. 2H). Collectively, the data suggest that Runx1 is required to maintain accessibility at critical Notch-responsive enhancers regulating the expression of Cdh15, Col6a1 and Col6a2 genes and support the notion that Runx1 functions in SC maintenance are mediated in part by enabling Notch-dependent expression of SC maintenance genes. Runx2 fails to maintain the accessibility of these regions when Runx1 is absent (Hass et al., 2021).

Loss of Runx1 impairs C2C12 cell differentiation via dysregulation of Mef2c expression

To investigate further the molecular mechanisms underlying the role of Runx1 in myogenesis, we used C2C12 cells, which express both Runx1 and Runx2. The Runx1KO C2C12 cell lines we mentioned above were generated using CRISPR-Cas9 and guide RNAs targeting exon 3, which encodes part of the DNA-binding domain. Western blot (WB) analyses showed that Runx2 expression was retained in cells lacking Runx1 protein (Fig. 3A). The growth of Runx1KO cells in growth medium (GM) was similar to that of control cells (Fig. 3B, Fig. S4). To differentiate the cells, we switched them from GM to differentiation medium (DM) when cell density reached ∼70% confluence. We then compared their ability to form myotubes in vitro. After 10 days in DM, robust differentiation into multinucleated myotubes was observed in controls. By contrast, most of the Runx1KO cells failed to fuse (Fig. 3B), as reflected in the differentiation index (Fig. 3C). Importantly, this muscle fusion defect was also observed in Runx1cKO-derived primary myoblasts but not in WT controls (Fig. S5A-C). These results show that Runx2 cannot perform the myogenic functions of Runx1.

Fig. 3.

Loss of Runx1 impairs C2C12 cell differentiation. (A) Western blot analysis of Runx1 and Runx2 in WT and Runx1KO C2C12 relative to β-actin. Blots are representative of 3 experiments. (B) Representative phase contrast microscopy images of WT and Runx1KO C2C12 cells in growth medium (GM) or after 6 days in differentiation medium (DM). Images are representative of 3 samples. (C) Quantification of the differentiation index of WT and Runx1KO C2C12 cells after 6 days in DM. ***P<0.001 (two-tailed unpaired Student's t-test). Data presented as mean±s.e.m. n=4 for each group. (D) PCA of the samples submitted for RNA-seq. Curved arrows represent the muscle differentiation process. (E) Differential gene expression analysis reveals differences in GM (D0) and DM (D1, D3, D6). The affected genes in each group/day are displayed here and in Table S1. (F) Heat maps of the changes in 368 transcription factors: above the clusters, relative Log2 fold change in each column is shown. Below the clusters, the relative abundance (LogCPM) is shown. Members of the Notch pathway, Ets, MRF gene families are indicated above their relative position. (G) GO terms of the up- and downregulated genes in Runx1KO. Details in Table S1.

We next analyzed the transcriptome of C2C12 and Runx1KO C2C12 cells in GM at D0, and in DM at D1, D3 and D6 (Fig. 3D, Table S1; accession number GSE248045). Principal component analysis (PCA) showed that the control C2C12 cells change along PC1 during differentiation but are relatively similar along PC2 (Fig. 3D). By contrast, the Runx1KO C2C12 cells diverge from control in GM, and drift further apart in each time point along both axis. All 5990 differentially expressed genes (DEGs) that passed cutoffs [expression >1.7 CPM in at least one replicate, fold change (FC)>1.5, P<0.01 and FDR<0.05; Table S1] were subjected to supervised clustering whereby each cluster contained all DEGs significantly changed at an individual time point, sorted by the Log2FC (Fig. 3E,F, Table S1, tabs 1 and 2). Cluster 5 was then divided by up- and downregulated DEGs to produce the image in Fig. 3E and tab 2 in Table S1 (column A in tab 1 of Table S1 sorts all 5990 DEGs using this supervised clustering). Of note, 1463 genes were differentially expressed in GM (D0; cluster 1 in tab 2 of Table S1; Fig. 3E), reflecting Runx1-dependent changes in myoblasts. The expression of 4018 genes that were not Runx1 dependent at D0, was increased (1978) or decreased (2040) after 24 h or more in DM (D1, cluster 5; Fig. 3E, Table S1, tabs 1 and 2). Of these, 368 were transcription factors (Fig. 3F, Table S1, tab 1, column AV). To complement the supervised analysis, we performed unsupervised clustering using kMeans clustering (kmC, Morpheus), hierarchical clustering, or both. All largely mirrored the supervised clustering results (Table S1, tab 4). ToppFun analysis of the 1978 upregulated genes (Table S1, tab 1, cluster 5 up, tab 3, kM cluster 1) revealed enrichment for the Gene Ontology (GO) term ‘cell cycle’, indicating continued division in Runx1KO cell on D1-D3, whereas the 2040 downregulated genes (Table S1, tab 1, cluster 5 down, tab 3, kM cluster 3) were enriched for general cellular processes. At D6, 236 Runx1KO C2C12 DEGs were enriched for muscle-specific terms (Fig. 3G, Table S1, tab 3, kM cluster 2). The same was true for 1323 downregulated DEGs regardless of expression status in GM. Contrary to the observation in primary myoblasts, we did not see strong elevation in cyclin-dependent kinase inhibitor genes (Fig. S6) and WT and Runx1KO C2C12 cells proliferated at similar rates (Fig. S4).

Notably, the most upregulated transcription factor on D3 was Mef2c, which has also been shown to be upregulated in Runx1-deficient primary myoblasts (Umansky et al., 2015b). Other upregulated factors in this cluster included Nrarp (a universal negative feedback target of Notch signaling; Jarrett et al., 2019), MyoD, Trp63 (which promotes differentiation; Li et al., 2023) and Myb (which promotes proliferation; Farrell et al., 2011). Expression of Col6a1 and Col6a2 was decreased in Runx1KO C2C12 compared to controls, as was observed in the mK4 cells, consistent with a similar regulatory network in these cell lines. Other downregulated genes included Pax7, the ETS factors Etv1 and Etv4, and the repressor Snai1. A previous study identified numerous peaks within and surrounding the Pax7 gene in Runx1 ChIP from myoblasts (Umansky et al., 2015b). The decrease in Pax7 mRNA after onset of differentiation is consistent with Runx1-dependent maintenance of Pax7 and might explain why Myf5-Cre deletion of Pax7 and Pax7-CreERT2-mediated deletion of Runx1 lead to similar morphological results.

To analyze the expression of the MRFs myogenin, Mef2c, MyoD, and the differentiation marker Myh3 (MHC), mRNA was isolated from cells in GM or after 2, 3 and 6 days in DM. Quantitative real-time (qRT-PCR) analysis revealed that MHC mRNA levels in Runx1KO were decreased to only 1.25% of the level observed in WT controls after 6 days in DM, consistent with the observed defect in differentiation (Fig. 4A,B). Whereas the mRNA levels of MyoD and myogenin were not significantly changed at D2 and D3, Mef2c mRNA levels dramatically increased relative to the WT levels in Runx1KO cells after switching to DM (Fig. 4A). Mef2c mRNA continued to accumulate every day in DM media in Runx1KO C2C12 cells, exceeding 1000-fold the levels seen in differentiating C2C12 cells (Fig. 4A). Thus, C2C12 reproduced some of the main defects observed in vivo, justifying further analysis of the mechanisms involved.

Fig. 4.

Runx1 loss blocks C2C12 differentiation via Mef2c. (A) qRT-PCR analysis of myogenin, MyoD, Mef2c and myosin heavy chain gene Myh3 (MHC) expression in WT (black bars) and Runx1KO (red bars) C2C12 cells on D0, D2, D3, and D6. (B) Immunostaining for Mef2c and Myh3 WT and Runx1KO C2C12 cells cultured in DM for 6 days. (C) Quantification of the number of Mef2c-positive cells in B. (D) Western blot analysis of Mef2c expression in WT and Runx1KO cells after 2 days of differentiation. Quantification was carried out using ImageJ and relative densitometric values are shown in red. (E-G) C2C12 cells were transfected with negative control or Mef2c overexpression plasmid (Mef2C^OE) for 24 h in the growth medium and then cultured in DM for 6 days. (E) mRNA expression of Mef2c in WT and Mef2C^OE C2C12 cells. (F) Representative phase contrast microscopy images of WT and Mef2C^OE C2C12 cells in DM for 6 days. (G) Immunofluorescence staining of MHC (red) and DNA (Hoechst, blue) of the cells shown in F. Data presented as mean±s.e.m. n=4 for each group. ***P<0.001 (two-tailed unpaired Student's t-test). N.S., not significant. Scale bars: 80 μm.

Since mRNA changes are not always concordant with protein levels, we examined Mef2c levels by immunofluorescence staining in MHC-positive cells after 6 days in DM. Multinucleated myotubes were evident in control cells but not in Runx1KO (Fig. 4B). We counted the number of Mef2c-positive cells in both WT and Runx1KO cells after differentiation as detailed in Materials and Methods. The number of strongly positive Mef2c nuclei was significantly higher in Runx1KO cells (Fig. 4C). Western blot analysis confirmed the accumulation of Mef2c protein in Runx1KO C2C12 cells grown in DM (Fig. 4D), but to a lesser degree than the mRNA.

Loss of Mef2c impairs muscle differentiation (Liu et al., 2014), despite the presence of Mef2a/d (Potthoff et al., 2007). By contrast, loss of Mef2a or Mef2d is tolerated (Potthoff et al., 2007). It is unclear, however, whether elevated Mef2c levels could do the same. To mimic the effect of losing Runx1 during differentiation, we overexpressed Mef2c in C2C12 cells (see Materials and Methods for detail). Mef2c mRNA levels were greatly increased in transfected C2C12 (Mef2c^OE) cells after transfection (Fig. 4E) to a degree similar to that seen in Runx1KO cells. Both phase contrast images (Fig. 4F) and immunostaining (Fig. 4G) showed that Mef2c can block fusion when overexpressed in WT C2C12 after growth in DM for 6 days. This suggests that the role of Runx1 in supporting myoblast fusion is mediated in part by maintaining the appropriate expression levels of Mef2c. Clearly, additional activities are likely to be involved in maintenance and proliferation of SCs.

Runx1-specific functions are encoded in the MID domain region

Runx2 is present in C2C12 Runx1KO cells and yet it cannot support myogenic differentiation, when overexpressed, it converts C2C12 to osteoblasts (Bae et al., 2001; Yu et al., 2013). We first attempted to rescue Runx1KO C2C12 cells by overexpressing Runx1 and/or Runx2. Transfecting 200 ng of Runx1-expressing pCDNA3.1 plasmid into Runx1KO cells restored differentiation in DM but transfecting 200 ng of Runx2-expressing pCDNA3.1 plasmid did not (Fig. 5A). Differentiation index calculation (Fig. 5B) and myotube length measurements (Fig. 5C) identified no significant difference between Runx1-rescued Runx1KO C2C12 cells and WT cells.

Fig. 5.

Runx1-specific functions are encoded in the MID domain region. (A) Runx1KO C2C12 cells were transfected with 200 ng Runx1 or Runx2 plasmids. After 24 h, cells were induced to differentiate for 6 days followed by immunofluorescence staining for MHC (red) and DNA (Hoechst, blue). (B,C) Quantification of the differentiation index of MHC-positive cells (B) and of MHC-positive cell lengths (C) for the experiments shown in A. (D) Schematics of Runx1 and Runx2 chimera proteins. (E,F) Quantification of the differentiation index of MHC-positive cells (E) and MHC-positive cell lengths (F) for the experiments shown in G. (G) Runx1KO C2C12 cells were transfected with 200 ng of five different chimera plasmids (as numbered in D) then 24 h after transfection, cells were induced to differentiate for 6 days followed by immunofluorescence staining for MHC (red) and DNA (Hoechst, blue). Data presented as mean±s.e.m. n=3 for each group. ***P<0.001 (two-tailed unpaired Student's t-test). n.s., not significant. Scale bars: 150 μm.

The ability of transfected Runx1 to rescue differentiation in Runx1KO C2C12 cells enabled structure-function analyses aimed at identifying the domain(s) supporting myogenesis unique to Runx1. We constructed five different Runx1/2 protein chimeras guided by AlphaFold-predicted structure to preserve domain integrity (Fig. 5D, Fig. S7A,B). We then transfected 200 ng of each of the five plasmids into Runx1KO cells and selected transfected cells with puromycin. When the puromycin-resistant cells reached 80% confluence we switched them to DM for 6 days. Substituting the highly conserved N-terminal DNA-binding Runt domain (#1, Runx1^RUNT2) resulted in a small but insignificant reduction of rescue activity, as did the swap of the C-terminal AD1, ID1 domain and VWRPY terminal sequence [#2, Runx1^AD/ID2, assessed by differentiation index (Fig. 5E), fiber length (Fig. 5F), and MHC and Hoechst staining (Fig. 5G)]. The Runx1 RUNT domain did not rescue muscle differentiation when placed into Runx2 (#3, Runx^1/2; Fig. 5E-G). Inserting the Runx2 MID2 domain into Runx1 (#4, Runx1^MID2) abrogated rescue activity (Fig. 5E-G). Importantly, substituting the MID1 domain from Runx1 into Runx2 (#5, Runx2^MID1) rescued fusion (Fig. 5E-G). Our results indicated that the MID1 domain of Runx1 is required and sufficient to support muscle differentiation.

The MID1 domain of Runx1 contains a region known to mediate an interaction with Ets1 (Shrivastava et al., 2014). It has been previously shown that Etv4 (PEA3), an ETS domain-containing transcription factor, accelerates myogenic differentiation when overexpressed in vitro (Taylor et al., 1997). We hypothesized that the Runx1 MID1 domain impacts myogenesis by partnering with Etv4, thereby enhancing the relatively weak intrinsic transactivation of Runx1. To test this possibility, we co-expressed a Flag-ETV4 and Myc-tagged Runx1, Runx2 and Runx1^MID2 in C2C12 cells and performed co-immunoprecipitation experiments using magnetic beads coated with antibodies against FLAG or MYC tags. Only Runx1 protein containing the MID1 domain immunoprecipitated Flag-ETV4 (Fig. 6A).

Fig. 6.

Runx1 maintains chromatin accessibility by binding Etv4. (A) Western blots from co-immunoprecipitation experiments in C2C12 cells transfected with Myc-tagged Runx proteins and Flag-tagged Etv4. (B) Venn diagram of ATAC-seq peaks from C2C12 or Runx1KO cells. Most peaks were shared; 9318 peaks were only found in control C2C12 cells and 2607 peaks were only found in Runx1KO C2C12 cells. (C) Heatmaps showing ATAC-seq reads from WT or Runx1KO C2C12 cells mapped onto these three classes of ATAC-seq peaks (WT specific, shared, and Runx1KO specific). (D) Graph displaying the −log P-value of transcription factor motif enrichment as determined by HOMER. Runx1 motifs (blue) were more highly enriched in WT C2C12 cells, whereas Mef2 motifs (red MADS-box) were more enriched in Runx1KO C2C12. The inset shows that AP1 motifs (purple) were similarly enriched in both WT and C2C12 cells, as expected for a global enhancer binding factor. (E) Bar graph showing enrichment for the ETS:RUNX composite motif in Runx1 ChIP data from myoblasts or T-ALL cells, less enrichment seen in Runx2 ChIP from preosteoblasts. The composite motif enrichment in each dataset is normalized to enrichment rate of the Runx-only motif in the respective ChIP from each cell type.

Runx1 MID interaction with ETS family member Etv4 is key to the unique functions of Runx1

Our experimental results and the predicted protein structures for Runx1, Runx2, and the Runx1/2 chimeras identified a distinguishing ETS-binding helix (Shrivastava et al., 2014) present in the MID domain of Runx1 and absent from Runx2 proteins, raising the possibility that recruitment of Etv4 is key for Runx1 activity in myogenesis. However, a caveat to this conclusion has been the use of un-tagged chimera, which prevented us from assessing whether differences in expression levels could explain our rescue experiments. The observation that the RUNX1/ETS1 complex binds to a composite motif wherein the ETS motif is followed by the RUNX motif (Shrivastava et al., 2014) offers an orthogonal approach to address this concern; namely, testing the prediction that the composite site will be enriched only in Runx1-dependent tissue. To that end, we performed ATAC-seq on triplicate samples of WT and Runx1KO C2C12 grown in GM and DM. Analysis of the ATAC data (accession number GSE248044) identified 24,800 peaks that were consistent in all replicates and shared between control and Runx1KO C2C12 cells (Fig. 6B). Importantly, 9318 ATAC-seq peaks present in all C2C12 triplicates were not found in Runx1KO cells. Conversely, 2607 ATAC-seq peaks were unique to Runx1KO (Fig. 6B). This is demonstrated visually by mapping all ATAC-seq reads from C2C12 and Runx1KO cells to either shared or unique peaks (Fig. 6C). The higher number of ATAC-seq peaks unique to C2C12 and lost in Runx1KO cells points again to Runx1 functioning primarily as a transcriptional activator. Motif enrichment analysis with the software package HOMER showed that the Runx1 motif is highly enriched in C2C12-specific peaks, consistent with a role for Runx1 in keeping these chromatin regions open (Fig. 6D, (Hass et al., 2021)). Genomic snapshots of the Col6a1, Col6a2, and Cdh15 display reduced chromatin accessibility in the Runx1KO C2C12 cells compared to WT C2C12 cells, in the same regions that had reduced accessibility in Runx1KO mK4 cells (Fig. 2I,J).

To determine whether the interaction between Etv4 and Runx1 contributes to chromatin regulation by in C2C12, we tested for presence of the ETS:RUNX composite motif utilizing the MCOT anchored motif analysis, which searches ATAC-seq peaks for any secondary motifs at variable distance and orientation from a RUNX motif (Levitsky et al., 2022). The MCOT program identified significant enrichment of an ETS motif adjacent to a Runx1 motif in C2C12 unique peaks with orientation and spacing identical to the known ETS:RUNX composite motif (P=1×10⁻¹⁴) (Fig. S8A). The ETS:RUNX motif enrichment in the Runx1KO unique peaks was negligible. The enrichment of the ETS:RUNX composite motif in the C2C12 unique peaks was confirmed using the MEME Suite Simple Enrichment Analysis (SEA) tool (P=4.37×10⁻²²). This supports our assertion that the Runx1 and Etv4 are cooperating to keep chromatin open and drive expression of a subset of targets genes. Notably, one of the most enriched motifs among Runx1KO unique peaks was the MADS-box motif that is bound by Mef2c (Fig. 6D), providing orthogonal confirmation that Mef2c is upregulated in the Runx1KO cells and suggesting that Mef2c expression is sufficient for accessibility to chromatin regions to be altered in the absence of Runx1.

To determine whether any C2C12-unique ATAC-seq peaks are directly regulated by Runx1 binding, we intersected our C2C12- and the Runx1KO-unique peaks with published Runx1 ChIP-seq data from myoblasts (Umansky et al., 2015a). We found that 3938 of the 9318 (42.3%) of the C2C12 unique peaks were bound by Runx1 in myoblasts, while only 21.2% (552 out of 2607) of Runx1KO unique ATAC-seq peaks were bound by Runx1 (Fig. S9A,B). Consistent with our previous report (Hass et al., 2021), direct binding by Runx1 helps maintain chromatin accessibility at many Runx1 and ETS:RUNX-specific ATAC-seq peaks.

To explore whether combinatorial binding of ETS and Runx1 to the composite ETS:RUNX site could underlie the unique functions of Runx1 in myoblasts, we tested the prediction that the composite ETS:RUNX site will represent a larger fraction of all RUNX sites in cell types that depend on Runx1 relative to cell types that are dependent on Runx2. In a Runx1 ChIP-seq dataset generated in myoblasts (Umansky et al., 2015a), composite motif enrichment near genes related to muscle development is highly significant (P=2.11×10⁻⁹; Fig. S9C). Likewise, in a Runx1 ChIP-seq datasets generated from the T-ALL cell line CUTLL1 (Wang et al., 2014), very significant enrichment for the composite motif near genes regulated during T-cell development was observed (P=2.42×10⁻²¹). By contrast, composite motif enrichment near genes regulating bone development in Runx2 ChIP-seq datasets generated from preosteoblasts (Wu et al., 2014) was modest (P=6.77×10⁻²). Since all three ChIP data sets showed highly significant enrichment of Runx motifs (myoblast Runx1 ChIP, P=4.74×10⁻⁵⁷; T-ALL Runx1 ChIP P=1.95×10⁻²⁸; preosteoblast Runx2 ChIP P=2.61×10⁻³²), we were able to plot enrichment of the composite motif in Runx1 ChIP as a ratio between Runx and the composite motifs (Fig. 6E, Fig. S9D). Taken together, these analyses support the idea that Runx1-dependent tissues rely on the combinatorial function of ETS and Runx1 proteins.

DISCUSSION

The role of Runx1 during muscle regeneration from injury was established previously but the question of what role Runx1 might have in during SC development or in quiescent SCs was left unanswered. The quality of a commercial anti-Runx1 antibody enabled detection of Runx1 in quiescent SCs, motivating us to explore a role outside of muscle regeneration. To gain temporal control and target all SCs, we deleted Runx1 in SCs using an inducible Pax7-CreERT2. Removal of Runx1 from adult SCs (Pax7 positive, Myf5 negative, Ki67 negative) demonstrated that Runx1 is required to maintain SCs throughout life, in part by maintaining accessibility to Notch-dependent enhancers, and in part by maintaining Pax7 expression. Runx1 is thus required for all phases of skeletal muscle development, maintenance, and regeneration. In its absence, Pax7 staining in uninjured Runx1KO muscle (and in C2C12 myoblasts) is reduced. Since Runx1 binds near the Pax7 gene in myoblasts (Umansky et al., 2015b), and because Runx1 sites and the ETS-Runx1 composite motif were highly enriched in Pax7 ChIP from embryonic stem cells differentiating into muscle (Fig. S8B) (Zhang et al., 2020), we surmise that Runx1 is directly regulating Pax7 expression and that Pax7 and Runx1 function together to maintain SCs in a feedback loop.

Deleting Runx1 at E13.5 using Pax7-CreERT2 reduced pup size, as was observed when Pax7 itself is deleted either constitutively or specifically in SCs (Gunther et al., 2013; Seale et al., 2000) and is also observed with muscle SC-specific deletion of other signaling pathways, such as Sox7 (Rajgara et al., 2017), Notch (Lin et al., 2013), or p38α (Brien et al., 2013), which suggests that loss of SC function may be related to overall reduced growth. By contrast, the Myf5-Cre; Runx1^f/f mice were of normal size and lifespan (Umansky et al., 2015b), perhaps because Myf5-Cre does not efficiently delete all floxed alleles in SCs (Comai et al., 2014), which leaves a subpopulation that has not undergone deletion and is sufficient to carryout SC function. Alternatively, the Pax7-positive Myf5-negative SCs (Picard and Marcelle, 2013) may ‘rescue’ the Myf5-Cre; Runx1^f/f mice. Additionally, deletion of Runx1 during early development resulted in reduced number of SCs later in life either through loss of SC maintenance or by reducing proliferation and thus the size of the stem cell pool.

Analysis of Runx1KO C2C12 revealed that the muscle transcription factor Mef2c was upregulated in growth media with further dramatic upregulation during differentiation (noted also during differentiation of Runx1KO myoblasts; Umansky et al., 2015b). Accordingly, we saw an increase in Mef2c-binding sites (MADS box motifs) in unique Runx1KO ATAC peaks. Increasing Mef2c levels artificially in WT C2C12 generated a similar fusion blockade, which suggests that loss of fusion in Runx1KO C2C12 cells may be caused by the inappropriate upregulation of Mef2c. It is possible that Runx1 directly represses Mef2c expression. Alternatively, the repressive role of Runx1 in C2C12 may be indirect, as it is in mK4 cells where Runx1 regulates the expression of the repressors Zeb1 and Zeb2 (Hass et al., 2021). In C2C12, this role may be fulfilled by the Runx1 target Snai1, which encodes a transcriptional repressor of Mef2c in C2C12 (Batlle et al., 2013) that is downregulated in Runx1KO C2C12 and to whose enhancer Runx1 binds in myoblasts (Umansky et al., 2015b).

Despite high degree of conservation and affinity to highly related binding motifs, co-expressed Runx proteins have demonstrable non-redundant functions. Runx1 has been extensively investigated for its role in T-cell development, where it is co-expressed with Runx3, and in muscle development, where it is co-expressed with Runx2. Similar to its multiple roles supporting the common lymphocyte progenitors through their subsequent differentiation to T- and B-cell linages in adult hematopoiesis (Growney et al., 2005; Rothenberg, 2021; Shin et al., 2021), we show here that Runx1 is also involved in SC maintenance in the absence of injury. Neither of these functions is compensated for by Runx2 protein. Deletion of Runx1 in adult SCs reduced their numbers in uninjured mice (indicating a defect in self-renewal). Since differentiating Runx1-deficient SCs fuse poorly, muscle maintenance (and regeneration) fail. The role of Runx1 in myoblasts was previously shown to involve a cooperative function with MyoD (Umansky et al., 2015a,b). However, MyoD is not expressed in quiescent SCs, suggesting that Runx1 functions in SCs depend on other factors with whom Runx2 cannot interact. In T cells, Runx1 interacts with ETS family transcription factors, and, accordingly, a composite ETS:RUNX motif was highly enriched (P=10⁻²¹) in Runx1 ChIP-seq datasets generated from T-ALL cells. We observed that Runx1-dependent ATAC-seq peaks present in C2C12 myoblasts but absent from Runx1KO were also strongly enriched (P=10⁻²²) for the composite motif, with similar enrichment detected in Pax7 ChIP-seq datasets from embryonic stem cells undergoing muscle differentiation, and in Runx1 ChIP-seq datasets generated from myoblasts (P=10⁻⁹). The enrichment of the composite ETS:RUNX motif in Runx2 ChIP-seq datasets generated from preosteoblasts was orders of magnitude weaker (P=10⁻²).

RUNX/ETS interactions observed in the crystal structure of the Runx1 and Ets1 on DNA containing the composite motif are mediated by a helix within a conserved region termed MID. AlphaFold predicted an alpha helix in the Runx1 MID region (MID1), but not in Runx2 MID (MID2). The binding of Runx1 to Ets1 prevents Ets1 autoinhibition leading to enhanced target transcription (Shrivastava et al., 2014). We found that the MID1 domain was both necessary (within Runx1) and sufficient (when inserted into Runx2) to rescue myoblast differentiation/fusion defect(s) in Runx1KO C2C12 cells. Furthermore, we found that Runx1, but not Runx2, is able to co-immunoprecipitate Etv4, an ETS family member that has previously been implicated in myoblast differentiation. Similarly, in mK4 cells, both Runx1 and Runx2 are expressed; yet Runx1 deficiency caused dramatic changes in chromatin accessibility despite Runx2 binding to many of the regions that are closed in the absence of Runx1 (Hass et al., 2021).

RUNX proteins collaborate with many transcription factors at different stages of development, most notably the binding of the transcription factor CBFβ to RUNX proteins through the PY repeat. We propose that non-redundant activity of Runx1 in SCs, T cells, and perhaps additional tissues or developmental stages reflects its ability to form a stable complex with ETS transcription factors via the MID domain. This complex can maintain chromatin accessibility and facilitate transcriptional regulation of essential targets genes via the ETS:RUNX composite motif recruiting additional co-activators (Hollenhorst et al., 2009). However, the role of the ETS:RUNX composite sites is rather complex: the ETS protein Fli1 blocks effector T-cell differentiation (T_EFF) by occluding ETS:RUNX composite sites. loss of Fli1 enables T_EFF differentiation in a Runx3-dependent fashion (Chen et al., 2021). Runx3 is not predicted to form a complex with ETS by AlphaFold, perhaps binding to this site without an ETS partner. By contrast, Runx1 overexpression in Fli1-deficient cells reverses the gains in T_EFF, promoting an alternative T-cell type (Chen et al., 2021), perhaps by recruiting another ETS protein to the composite site. It will be interesting to see whether Runx2 can rescue a Fli1/Runx3-deficient T cell in the experimental paradigm explored by Chen et al., and whether the Runx1MID2 chimera will promote T_EFF, like Runx3, instead of suppressing it.

MATERIALS AND METHODS

Mice

All mice used in this study were on the C57BL/6J/CDI mixed genetic background, between 8 and 12 weeks of age, and had age-matched littermate controls. Both sexes were used. Mice maintained at Cincinnati Children's Hospital Medical Center animal facility were handled following animal care guidelines approved by the Animal Studies Committee of CCHMC (IACUC 2018-0108/0107). After transfer to Fudan University, mice were handled following animal care guidelines approved by the Fudan Animal Studies Committee (IACUC IDM20211021). Runx1^flox/flox mice were crossed with Pax7^CreERT2/+; Runx1^+/flox mice (Lepper et al., 2009; Lepper and Fan, 2010, 2012) to generate Pax7^CreERT2/+; Runx^1fl/fl mice. Tamoxifen (Aladdin) was dissolved in 90% (vol/vol) sesame oil and 10% (vol/vol) ethanol at 20 mg/ml and delivered to both control and Runx1cKO mice at 8 weeks of age by intraperitoneal injection at 0.1 mg/g every day for 5 days. for the generation of control and Runx1cKO pups, 50 µg/g tamoxifen were administrated by intraperitoneal injection on three consecutive days from E13.5 to E15.5. Since tamoxifen was injected in late pregnancy, progesterone was not used and no significant reduction in pups was observed. Genotyping primers are in listed in Table S2.

TA muscle CTX injury

CTX (Latoxan) was dissolved in sterile saline solution to a final concentration of 10 μM, aliquoted, and stored at −80°C. For muscle injury, mice were anesthetized by intraperitoneal injection of 2.5% Avertin (tribromoethanol; Aladdin) at (15 μl/g). The fur on the hind legs was shaved and the legs were swabbed with 75% ethanol. Intramuscular injections into the TA muscle of 50 μl of CTX were performed with a 26-gauge needle. As a control, contralateral TA muscles were injected with the same volume of normal saline. After injection, animals were kept under a warming lamp until the anesthetic wore off. Post-injection mice were anesthetized and euthanized by CO₂ asphyxiation to harvest the TA muscles at 5, 7, 12, 30 and 60 days post-injury. Tissue was fixed in 4% paraformaldehyde (PFA) for 12-18 h, dehydrated in an ethanol/xylene series and embedded in paraffin. Paraffin sections were rehydrated for histological analyses as detailed below.

H&E staining

For the assessment of muscle morphology, 5-μm-thick cross-sections of TA muscles were subjected to H&E staining, which was performed according to standard procedures provided by the H&E staining kit (Servicebio).

C2C12 cell culture

C2C12 myoblasts were purchased from ATCC and proliferated in Dulbecco's Modification of Eagle's Medium (Gibco, DMEM) with 1% Penicillin/Streptomycin (Pen/Strep; Gibco) and 20% (v/v) fetal bovine serum (Sigma-Aldrich) (termed GM) under moist air with 5% CO₂ at 37°C and were regularly tested for Mycoplasma contamination. Post-confluent C2C12 cells were washed with DPBS (Gibco) and incubated with DMEM supplemented with 1% Pen/Strep and 2% (v/v) horse serum (Gibco) (termed DM) under the same conditions for differentiation. Induction of cell differentiation was performed when cells were confluent to ensure the same cell density across samples.

Generation of Runx1KO C2C12 cells

We cloned guide RNAs (sgRNAs; Table S2) targeting exon 3 of Runx1 containing the start codon in the px459 CRISPR/Cas9 vectors. The sgRNA plasmids were constructed using methods and tools described previously (Haeussler et al., 2016; Ran et al., 2013). We split C2C12 cells into control and experimental sibs and transiently transfected pX459-sgRNA plasmids into the experimental sib with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. At 12 h post-transfection, the cells underwent selection with 5 μg/ml puromycin for 2 days. When all cells in the control dish died, clones growing on the transfected were picked ∼1 week later using cloning disks. The clones were tested for exon 3 deletion by the T7E1 assay and for Runx1 protein expression by western blot. These cells were characterized at the Cincinnati Children's Hospital Medical Center for the data shown in Fig. 3; however, Runx1KO-C2C12 were independently recreated by the same investigator (M.Y.) at Fudan University using the same protocol with cells from a different stock of C2C12. The results presented in Figs 4-6 were reproduced at both institutions, and the molecular analysis was performed on the Fudan Runx1KO-C2C12 cells.

Western blot

Confluent C2C12 control or Runx1KO cells were collected in 100 μl of RIPA lysis buffer (Beyotime) with protease inhibitors (Bimake) and PMSF (LIFE-iLAB) plus 20 μl of 6× sample buffer. Protein samples were run on EZ Protein any KD PAGE (LIFE-iLAB) and then transferred to 0.22 μm PVDF membranes (Millipore). Membranes were blocked using 5% non-fat milk in PBS with 0.1% Tween 20 for 1 h at room temperature. Primary antibodies (Table S3) were applied at 1:1000 overnight at 4°C, and then horseradish peroxidase-conjugated secondary antibodies were used at 1:2000 at room temperature for 1 h. The western blot signal was detected using Super ECL Detection Reagent (Yeasen) using a Bio-Rad Chemidoc MP Imaging System.

RNA extraction and qRT-PCR

RNA from biological triplicate samples were extracted from C2C12 cells with PuroLink kit (Invitrogen) and the Reverse Transcription Kit (Promega) was used to make cDNA following manufacturer's indications. The cDNA was diluted to 20 ng/μl and 2 μl of each sample was added to each qRT-PCR reaction and was amplified using 2× SYBR Green qPCR Master Mix (Bimake) performed on a CFX96^TM Connect Real-Time PCR System from Bio-Rad. Gene expression levels were normalized to mouse Gapdh expression and changes were determined relative to control cells with significance calculated using the two-tailed Student's t-test.

Primary myoblast purification and culture

Isolation of primary myoblasts was performed as described previously (Hindi et al., 2017). Briefly, hind limb muscles were pooled, minced, and digested with 400 U/ml Collagenase II (Yeasen), followed by trituration. Cell suspension was filtered through 70- and 40-μm cell strainers (Biosharp), and primary myoblasts were pelleted after centrifugation at 350 g for 5 min at room temperature. Isolated primary myoblasts were cultured on 10% Matrigel (Corning)-coated 10 cm dishes. Proliferation medium contained Ham's-F10 (Gibco) with 1% Pen/Strep, 20% fetal bovine serum (myoblast growth medium) and 2 ng/ml basic human fibroblast growth factor (FGF2). When cell confluency had reached ∼70%, the media was replaced with differentiation medium (DMEM with 2% horse serum and 1% Pen/Strep).

Immunofluorescence staining

Cells were cultured in Chamber Slides (Thermo Fisher Scientific) or 12-well plates and were fixed in 4% PFA for 15 min and permeabilized in 0.3% Triton X-100 in PBS for three 10 min incubations at room temperature. Isolated TA muscles were fixed in 4% PFA at 4°C overnight, dehydrated by graded ethanol and embedded in paraffin. Paraffin-embedded samples were cut into 5-μm-thick sections using a Leica RM2125RTS rotary microtome. Samples were then blocked in 10% normal donkey serum (Jackson ImmunoResearch) in PBS for 1 h at room temperature followed by incubation with primary antibodies at 4°C overnight. We followed a published Pax7 antigen retrieval protocol (Feng et al., 2018) with the modification of using a high-pressure cooker at 70 kPa and 112°C for 20 min. Subsequently, appropriate fluorescently labeled secondary antibodies (Alexa Fluor 488 or 549) were incubated for 1 h at room temperature. Hoechst 33342 was used to stain the cell nuclei for 15 min. Images were acquired with an Olympus Fluoview FV300 confocal microscope (Olympus). Antibodies are listed in Table S3.

Transfection of plasmids

The mouse Runx1 overexpression plasmid pCDNA3.1-Flag-Runx1 was purchased from Addgene (#64894). The mouse Runx2 plasmid pCMV-Flag-mRunx2 was purchased from Origene. The fragments shown in Fig. 5D were cloned from pCDNA3.1-Flag-Runx1 or pCMV-Flag-mRunx2 using Phanta Super-Fidelity DNA Polymerase (Vazyme) with the primers listed in Table S2. ClonExpress II One Step Cloning Kit (Vazyme) was used to ligate and construct RUNX1/2 chimeras. Sequences were confirmed by Sanger sequencing performed by Beijing Tsingke Biotech Co., Ltd. For the rescue experiments and chimera transfection, 200 ng plasmids (a mix containing the Runx plasmid and pLVX-GFP) were transiently transfected into 12-well plates according to the protocols recommended by the Lipofectamine 2000 manufacturer (Invitrogen). For the Mef2c overexpression experiment, confluent cells were transfected twice with a 24 h interval using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. pCDNA3.1 was used as an empty vector transfection control. Transfection efficiency was monitored using GFP fluorescence.

Differentiation index, quantification and myotube dimensions analyses

C2C12 cells and isolated primary myoblasts were cultured in DM to induce terminal differentiation. Histological parameters were derived from at least four different animals per genotype. Phase-contrast and immunofluorescence images were obtained in four randomly selected fields in each of three biological replicates using a 20× objective and then analyzed using ImageJ software.

MTS proliferation assay

For the proliferation assay, 20,000 WT or Runx1KO C2C12 cells were seeded in each well of a 96-well plate and 4 h after seeding a baseline cell count was determined using Promega's CellTiter 96 Aqueous One Solution Cell Proliferation Assay (MTS) and read on a GloMax at 490 nm. Additional counts were performed after 24 or 48 h and the readings were normalized to the baseline to determine the relative proliferation at each timepoint.

RNA-seq generation and analysis

C2C12 cells were grown and differentiated as described above. RNA was isolated using Invitrogen's Purelink RNA Mini Kit according to the manufacturer's directions. The DNBSEQ Transcriptome libraries were used by BGI Hongkong Tech Solution NGS Lab to produce over 20 million reads per sample. Raw RNA-seq data were processed using a pipeline developed in-house called CSBB (https://github.com/csbbcompbio/CSBB-v3.0). It employs fastqc to check read quality followed by Bowtie2+RSEM for alignment and quantification against the mm10 genome. The final outputs were transcripts per million and count matrices at the gene and isoform level. DEGs were then identified using DESeq2 (Love et al., 2014).

ATAC-seq data generation and analysis

C2C12 cells were grown and differentiated as described above until nearly confluent. Cells in GM were washed and media replaced with lysis buffer [10 mM Tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl₂, 0.5% NP-40] for 10 min on ice, then centrifuged at 500 g for 5 min at 4°C to obtain nuclei. Nuclei were then segmented with Tn5 transposase (Vazyme) at 37°C for 30 min. The ATAC libraries were prepared using TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme, TD501) according to the manufacturer's directions. VAHTS DNA Clean Beads (Vayzme, N411) was used to do the libraries selection. An aliquot was tested by Agilent 2100 Bioanalyzer for average fragment size. Libraries were sequenced on Illumina Novaseq 6000 with the NJNA Biopharmaceutical Public Service Platform.

Paired reads were processed with the nf-core ATAC-seq pipeline v.2.0 (10.5281/zenodo.2634132), using the mm10 genome/blacklist and specifying parameters ‘–read-length=100’ and ‘–narrow-peak’. This encompasses all steps from FASTQ trimming and quality control to final bigwig and peak files. The main tools used were BWA for alignment, MACS2 for peak calling, and HOMER for annotation. To identify differentially accessible peaks, the individual ATAC-seq peak files were intersected using the Bedtools intersect intervals tool on Galaxy (https://usegalaxy.org) to identify peaks that were present in all triplicates. The same tool was then utilized to identify the shared peaks between WT and Runx1KO cells and the peaks unique to each. The peaks from the different classes were analyzed using HOMER to determine enrichment of transcription factor motifs. The heatmaps of the ATAC-seq reads were generated using the Deeptools computeMatrix and plotHeatmaps tools on Galaxy. In Gene Expression Omnibus, the peak bed files for myoblast Runx1 ChIP (GSE56077), preosteoblast Runx2 ChIP (GSE54013), and CUTLL1 Runx1 ChIP (GSE51800) were annotated to nearest genes using the Great Annotation tool and intersected with the gene lists from GO terms ‘muscle differentiation’, ‘bone development’, or ‘T cell differentiation’, respectively. The Runx1 or Runx2 ChIP peaks were subjected to motif enrichment using HOMER and MEME-Suite SEA enrichment tools to determine enrichment of the ETS:RUNX composite site, or submitted to the MCOT program and searched for motifs near the Runx1 anchor.

Statistics

Data are presented as mean±s.e.m. Differences between groups were tested for statistical significance by unpaired, two-tailed Student's t-test (Excel and GraphPad Prism software). P<0.05 was considered significant. For multiple comparison analysis, FDR<0.05 was selected. Analysis-specific cutoffs are described in Table S1.

Peer review history

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202556.reviewer-comments.pdf