Abstract

   Splicing factor SRSF2 acts as a critical regulator for cell survival,
   however, it remains unknown whether SRSF2 is involved in myoblast
   proliferation and myogenesis. Here, knockdown of SRSF2 in myoblasts
   causes high rates of apoptosis and defective differentiation. Combined
   conditional knockout and lineage tracing approaches show that Myf5‐cre
   mice lacking SRSF2 die immediately at birth and exhibit a complete
   absence of mature myofibers. Mutant Myf5‐derived cells
   (tdtomato‐positive cells) are randomly scattered in the myogenic and
   non‐myogenic regions, indicating loss of the community effect required
   for skeletal muscle differentiation. Single‐cell RNA‐sequencing reveals
   high heterogeneity of myf5‐derived cells and non‐myogenic cells are
   significantly increased at the expense of skeletal muscle cells in the
   absence of SRSF2, reflecting altered cell fate. SRSF2 is demonstrated
   to regulate the entry of Myf5 cells into the myogenic program and
   ensures their survival by preventing precocious differentiation and
   apoptosis. In summary, SRSF2 functions as an essential regulator for
   Myf5‐derived cells to respond correctly to positional cues and to adopt
   their myogenic fate.

   Keywords: apoptosis, exhaustion of myogenic pool, lineage tracing,
   Myf5‐derived cells, precocious differentiation, single‐cell RNA
   sequencing, SRSF2
     __________________________________________________________________

   Splicing factor SRSF2 plays a key role in myoblast proliferation and
   differentiation. Loss of SRSF2 in Myf5‐derived cells causes severe
   myogenic defects. Lineage tracing and single‐cell RNA sequencing shows
   that Myf5‐derived cells have great heterogeneity. SRSF2 regulates Myf5
   cells to enter the myogenic program and ensures their survival by
   preventing immature differentiation and apoptosis.

   graphic file with name ADVS-9-2105775-g003.jpg

1. Introduction

   In the course of skeletal muscle development, the lips of the
   dermomyotome can mature into the myotome that is a segmented structure
   called somites. They contain muscle progenitor cells (MPCs) highly
   expressing the paired box transcription factors Pax3 and Pax7.^[ [44]^1
   ^] It is generally believed that MPCs establish their myogenic fate by
   forming committed myoblasts with expression of one or a few myogenic
   regulatory factors (MRFs).^[ [45]^2 ^] Following myogenesis involves
   withdrawal of myoblasts from the cell cycle and differentiation into
   mononucleated myocytes, and later fusion to form multinucleated
   myofibers. MRFs compose of Myf5, MyoD, Myog and Myf6, which function
   downstream of Pax3 and Pax7. Gene targeting studies revealed that MyoD
   and Myf5 are both considered to be myogenic determination genes,^[
   [46]^3 ^] whereas Myog and Myf6 are regarded as differentiation genes
   that play a critical role in differentiating myocytes and myofibers.^[
   [47]^4 ^]

   Myf5 transcripts are first detected on the embryonic day E8.0 in the
   dorsal cells of the somites, which occurs about 2 days earlier than
   MyoD expression.^[ [48]^5 ^] During this early stage, MPCs that have
   activated expression of Myf5 still remain multipotent, as loss of Myf5
   induces aberrant migration and mislocalization of these cells,
   resulting in altered cell fate.^[ [49]^5b ^] In the absence of Pax3 and
   Myf5, MyoD is not activated and skeletal muscles are ablated in the
   body.^[ [50]^6 ^] All these findings strongly indicate that Myf5 acts
   upstream of MyoD to direct embryonic multipotent cells into the
   myogenic lineage. On the other hand, Myf5^+ cells have been reported to
   contribute to fat formation^[ [51]^7 ^] and a portion of the ribs and
   cervical vertebrae.^[ [52]^8 ^] Moreover, previous research has
   indicated that these Myf5‐derived cells can generate a functional
   neural crest cell type,^[ [53]^9 ^] despite it is still under
   controversy.^[ [54]^10 ^]

   The activity MRFs is tightly coupled to cell cycle control, and
   abnormal regulation of cell cycle can induce an uncontrolled myogenic
   differentiation. The retinoblastoma tumor suppressor protein (Rb) plays
   a key role in controlling cell cycle progression, and Rb‐deficient mice
   exhibit high rates of apoptosis and an almost complete absence of
   myofibers.^[ [55]^11 ^] Induction of cyclin‐dependent kinase (Cdk)
   inhibitors such as p21 and p57 is required for terminal cell cycle
   withdrawal during skeletal muscle differentiation. Mice lacking both
   p21 and p57 fail to form myotubes, display increased proliferation and
   apoptotic rates of myoblasts.^[ [56]^12 ^]

   Genetic lineage tracing is a useful tool for labeling endogenous cells
   and tracking cell fate during embryonic development.^[ [57]^13 ^] In
   this study, we introduced a Cre‐dependent tdtomato reporter gene
   tracking system, in which Rosa26‐tdTomato (tdT) mice were used to cross
   with Myf5‐Cre mice.^[ [58]^14 ^] Combined with single cell RNA
   sequencing (scRNA‐seq) technology, we identified and characterized the
   transcriptome characteristics of Myf5‐derived cells during embryonic
   myogenesis. Surprisingly, we observed that Myf5‐derived cells showed a
   wide range of heterogeneity. While long‐term expression of Myf5 was
   only detected in the skeletal muscle cells, transient activity of Myf5
   contributed to various non‐myogenic cell lineages including mesenchymal
   cells, chondrocytes, neural cells, endothelial cells, osteoblasts,
   Schwann cells, and immune cells.

   SRSF2 belongs to the family of SR proteins and acts as a crucial
   regulator of constitutive and alternative pre‐mRNA splicing. Our
   previous work has demonstrated that absence of SRSF2 resulted in large
   number of hepatocyte apoptosis and acute liver failure.^[ [59]^15 ^]
   Knockdown of SRSF2 caused severe death of live cancer cells.^[ [60]^16
   ^] These findings strongly emphasize the critical role of SRSF2 in cell
   survival. In addition, SRSF2 is also reported to interact with E2F1 and
   stimulates its transcriptional control of cell cycle target genes.^[
   [61]^17 ^]

   In this study, we find that Myf5‐cre mice lacking SRSF2 die at birth,
   display high rates of apoptosis of myoblasts and total loss of mature
   myofibers. Significantly, we observe a scattered and disordered
   arrangement of SRSF2‐depleted tdT cells in the limbs and intercostal
   muscles, indicating loss of the community effect for skeletal muscle
   differentiation.^[ [62]^18 ^] scRNA‐seq further reveals that
   non‐myogenic cells are increased at the expense of skeletal muscle
   cells in the absence of SRSF2. Moreover, differential gene expression
   analysis shows that differentiation markers Myog and p21 were
   upregulated, accompanied by abnormal expression of Pax7 and Myf5 during
   the differentiation pseudotime of mutant versus control. Finally, we
   demonstrate that knockdown of SRSF2 causes precocious differentiation
   of myoblasts and leads to exhaustion of the progenitor pool. Taken
   together, SRSF2 is necessary for Myf5 cells to adopt their myogenic
   fate during embryonic myogenesis.

2. Results

2.1. SRSF2 is Critical for Proliferation and Differentiation of Myoblasts In
Vitro

   To explore the role of SRSF2 in myogenic differentiation, we isolated
   primary myoblasts from wild‐type mice at E16.5 and performed
   immunostaining for Pax7, Myf5, MyoD and SRSF2. SRSF2 was expressed in
   the majority of Pax7‐positive or Myf5‐positive myoblasts, as well as in
   ≈80% MyoD‐positive cells (Figure [63] 1A; Figure [64]S1A,B, Supporting
   information). Then we induced myoblasts into differentiation in vitro,
   and performed RT‐qPCR analysis. mRNA levels of SRSF2 declined sharply
   after differentiation, which was highly similar to the expression
   pattern of Myf5 and MyoD (Figure [65]1B). Protein levels of SRSF2 was
   similarly decreased in immortalized C2C12 cells upon differentiation,
   and the differentiation marker Myog‐expressing cells also showed weak
   signals for SRSF2 (Figure [66]1C; Figure [67]S1C, Supporting
   Information).

Figure 1.

   Figure 1
   [68]Open in a new tab

   SRSF2 knockdown caused severe proliferation and differentiation defects
   of both myoblasts and C2C12 cells. A) Representative confocal images of
   Pax7 (red), SRSF2 (green), and DAPI (blue) immunostaining in primary
   myoblasts purified from WT embryos at E16.5. Scale bars, 50 µm. 2 ×
   10^3 cells were seeded on the coverslips and stained with indicated
   antibodies. Qualification of Pax7 and SRSF2 colocalization in myoblasts
   was shown on the right (n = 3). Data were presented as the mean ± SD.
   B) Relative mRNA levels of indicated genes during primary myoblast
   differentiation. The results were plotted as the mean ± SD (n = 3). C)
   Representative WB analysis of MHC, MyoD, and SRSF2 during
   differentiation of C2C12 cells (n = 3). Cells were induced into
   differentiation and whole cell lysates were isolated at indicated time
   and subjected to WB analysis. D) Representative confocal images of MHC
   (green), Myog (red), and DAPI (blue) immunostaining in C2C12 cells and
   primary myoblasts (n = 3). Scale bars, 100 µm. Cells were transfected
   with indicated siRNAs, then induced into differentiation for 8 days
   (C2C12 cells) or 3 days (myoblasts), respectively. E) Representative
   confocal images of Ki67 (red) and p21 (green) immunostaining in C2C12
   cells and primary myoblasts. Scale bars, 100 µm. After transient
   transfection with indicated siRNAs for 48 h, cells were reseeded on the
   coverslips at density of 1 × 10^4 cells for siNC group and 2 × 10^4
   cells for siSRSF2‐#1 group, respectively, followed by immunostaining.
   Qualification of Ki67^+ and p21^+ cells was shown on the right (n = 3).
   Data were presented as the mean ± SD. 2‐tailed Student's t‐test. ***
   indicates p < 0.001. F) Representative WB analysis of C2C12 cells after
   RNA interference as indicated for 48 h (n = 3).

   Next we transfected C2C12 cells and myoblasts with control siRNA (siNC)
   or SRSF2‐specific siRNAs (siSRSF2‐1#, ‐2#) for 48 h, then switched
   cells from growth medium (GM) to differentiation medium (DM).
   Immunostaining with Myog and/or myosin heavy chain (MHC) revealed that
   knockdown of SRSF2 dramatically impaired myogenic differentiation of
   C2C12 cells and myoblasts (Figure [69]1D). Long multinucleated myotubes
   typically seen in control cells were absent, instead only a few short,
   unaligned myotubes were observed following SRSF2 knockdown. In
   addition, siSRSF2 significantly decreased intensity of the
   proliferation marker Ki67 and increased p21 signals, compared to
   control siRNA (Figure [70]1E). Western blot (WB) analysis further
   demonstrated that SRSF2 knockdown led to activation of MyoD‐dependent
   genes such that levels of p21, p57, and upregulation of
   hypophosphorylated pRb. On the other hand, E2F1 itself and
   E2F1‐dependent cell cycle genes such as Ccnb1 were switched off, as
   well as deceased levels of phosphorylated pRb (p‐pRb) (Figure [71]1F).
   In addition, the effects of SRSF2 overexpression were relatively modest
   on cell proliferation, probably because of SRSF2's relatively high
   abundance in cells (Figure [72]S2, Supporting Information). Taken
   together, these data provided evidence that SRSF2 plays a key role in
   regulating proliferation and differentiation of myoblasts in vitro.

2.2. Loss of SRSF2 Caused Severe Skeletal Muscle Defects

   To further investigate the requirement for SRSF2 in myogenesis, we
   inactivated Srsf2 gene in myoblasts by crossing SRSF2^f/f mice (WT)
   with Myf5‐Cre mice to generate SRSF2^f/f/Myf5‐Cre mice (MKO)
   (Figure [73]S3A, Supporting Information). Newborn MKO pups died shortly
   after birth, and they displayed apparent skeletal muscle and rib
   defects, as well as significant weight loss (Figure [74]S3B, Supporting
   Information). Protein levels of SRSF2 were significantly decreased in
   the hindlimb muscles of MKO mice, compared to control (Figure [75]S3C,
   Supporting Information). Histological examination of skeletal muscles
   at the embryonic day 19 (E19) revealed the presence of severe
   differentiation defects (Figure [76] 2A). Hindlimb muscles exhibited a
   dramatic reduction in mass with a complete absence of mature fibers
   compared with littermate controls. Paralleled myofibers typically seen
   in WT controls were absent and replaced by short irregular fibers in
   MKO mice. Those muscle defects were also observed in the forelimb,
   intercostals and tongues, but the diaphragm was just thinning that
   seemed less affected in the MKO mice (Figure [77]S3D, Supporting
   Information). Moreover, TUNEL analysis showed high levels of apoptosis
   in the hindlimb and intercostals of mutant mice at E19 (Figure [78]2B).

Figure 2.

   Figure 2
   [79]Open in a new tab

   SRSF2 is essential for skeletal muscle development. A) Representative
   hematoxylin and eosin (HE) staining of hindlimbs and intercostal
   muscles at E19 (n = 5). Scale bars, 100 µm. Note that black dotted
   boxes are shown enlarged in the middle panels. B) Representative
   confocal images of fluorometric TUNEL staining with sections at E19.
   Scale bars, 100 µm. The percentage of TUNEL‐positive cells was shown on
   the right bar graphs (n = 3). C) Representative confocal images of Myf5
   (green) and Pax7 (red) immunostaining with forelimb sections at E15.
   Scale bars, 100 µm. The percentage of Myf5^+ and Pax7^+ cells was shown
   on the right bar graphs (n = 3). D) Representative images of
   colorimetric TUNEL staining of sections at E15. Scale bars, 100 µm. The
   percentage of TUNEL‐positive cells was shown on the right bar graphs (n
   = 3). All data are shown as the mean ± SD. 2‐tailed Student's t‐test.
   *** p < 0.001.

   At E15, mutant embryos showed severe edema and a strong reduction in
   the body wall muscle (Figure [80]S4, Supporting Information).
   Immunostaining with forelimb sections at E15 revealed that there are
   large amount of myogenic precursor cells/myoblasts that expressed Pax7
   and Myf5 in the control muscles, which were significantly decreased in
   the MKO embryos (Figure [81]2C). Consistently, a large number of
   apoptotic cells were present in the mutant forelimb (Figure [82]2D).
   These results strongly indicated that mutant muscle cells were
   undergoing abnormal apoptosis during embryonic myogenesis.

2.3. Abnormal Distribution and Differentiation of Myf5‐Derived Cells in the
Mutant Mice

   To order to track the myogenic fate of Myf5‐derived cells, we crossed
   heterozygous mice (Het:SRSF2f/W/Myf5‐cre) with a mouse line harboring a
   Rosa26‐lSL‐tdTomato allele to specifically label Myf5‐expressed cells
   by tdTomato (tdT) (Figure [83]S5, Supporting Information). We first
   prepared transverse sections of limbs from control and MKO/tdT mice at
   E12.5, E13.5, and E14.5, respectively. tdT staining revealed that
   majority of tdT+ cells (Myf5‐derived cells) were located in the
   prospective muscle regions, arranged in homogeneous coherent groups
   sharply demarcated from adjacent non‐muscle cells in control mice at
   the three different stages (Figure [84] 3A). This is called a community
   effect during muscle development, which is believed to activate muscle
   gene expression in uniform cell populations.^[ [85]^18 ^] However, such
   effect was completely disrupted in E13.5 and E14.5 mutant mice,
   although relatively normal pattern was observed at E12.5 mice. Enlarged
   images further confirmed that tdT+ cells were wildly scattered
   throughout the limb region, and they even migrated into the livers and
   bones of mutant mice (Figure [86]3A).

Figure 3.

   Figure 3
   [87]Open in a new tab

   Loss of SRSF2 resulted in aberrant clustering and myogenic
   differentiation of Myf5‐derived cells. A) Representative confocal
   images of tdT (red) immunostaining with sections from Het/tdT and
   MKO/tdT embryos at E12.5, E13.5, and E14.5 (n = 3). Scale bars, 100 µm.
   B) Representative confocal images of tdT (red) and MHC (white)
   immunostaining with sections at E14.5. Scale bars, 100 µm. Note that
   white dotted boxes are shown enlarged in the right panels.
   Qualification was shown on the right bar graph (n = 3). Data were
   presented as the mean ± SD. 2‐tailed Student's t‐test. *** p < 0.001.

   Myf5‐derived cells in control mice displayed normal myogenic commitment
   and differentiation to functional myofibers, as addressed by
   immunostaining for MHC (Figure [88]3B; Figures [89]S6A, [90]7A,
   and [91]8A, Supporting Information). Importantly, we also observed that
   they remained closely associated (merged) with MHC‐positive fibers. In
   contrast, majority of tdT+ cells in the MKO mice couldn't undergo
   normal developmental program to acquire a myogenic fate at both E13.5
   and E14.5, as a few MHC‐positive fibers were observed and ≈40–50%
   myofibers were not merged with tdT+ cells, indicating aberrant
   differentiation of mutant tdT+ cells (Figure [92]3B; Figure [93]S7A,
   Supporting Information). One day earlier at E12.5, there was not much
   differences of MHC staining between control and MKO mice
   (Figure [94]S6A, Supporting Information), indicating that the first
   wave of myogenic differentiation was relatively normal in the mutant
   mice.

Figure 8.

   Figure 8
   [95]Open in a new tab

   Loss of SRSF2 resulted in abnormal proliferation, exhaustion and
   precocious differentiation of myoblasts. A) Representative confocal
   images of Pax7 (red) and γH2AX (green) immunostaining of intercostal
   sections at E15. Scale bars, 100 µm. Quantification of Pax7 and γH2AX
   cells per area (0.1 mm^2) are shown on the right bar graph (n = 3). B)
   Representative confocal images of Pax7 (red) and p21 (green)
   immunostaining of hindlimb sections at E15. Scale bars, 100 µm.
   Quantification of p21 cells per area (0.1 mm^2) was shown on the right
   bar graph (n = 3). C) Representative confocal images of Myf5 (red) and
   Ki67 (green) immunostaining with intercostal at E15. Scale bars,
   100 µm. Quantitative analysis of Myf5 and Ki67 cells was shown on the
   right bar graph (n = 3). D) Representative confocal images of Myog
   (red), MHC (green) and TUNEL (red) staining of primary myoblasts. Scale
   bars, 100 µm. After 48 h of transient transfection, primary myoblasts
   were kept in growth medium (GM) for another 24 h or transformed into
   differentiation medium (DM) for 24 h. Cells in GM were stained for Myog
   (red), MHC (green), or TUNEL (red), respectively, while cells in DM
   were stained with MHC (green) and TUNEL (red). E) Quantitative analysis
   of Myog (GM 72 h), MHC (GM 72 h) and TUNEL (GM 72 h and DM 24 h) was
   shown in (n = 3). F) Representative WB analysis in primary myoblasts
   after RNA interference as indicated for 72 h (n = 3). All data are
   shown as the mean ± SD. 2‐tailed Student's t‐test. *p < 0.05.**p
   < 0.01.***p < 0.001.

2.4. Progressive Loss of MPCs Were Observed in the MKO Mice

   Next we wanted to examine the proliferation rates of tdT+ cells by
   immunostaining for Ki67. At E14.5, we observed up to a 5‐fold decrease
   in the number of Ki67/tdT double‐positive cells in control limbs versus
   the mutant (Figure [96] 4A; Figure [97]S8A, Supporting Information).
   Meanwhile, we detected a large number of tdT+ apoptotic cells in the
   mutant mice, but only a few apoptotic cells were present in control
   mice (Figure [98]4B). More importantly, a massive reduction of Pax7+
   cells were observed in the mutant mice at E13.5 and E14.5, which
   occurred concomitantly with the significantly decreased MHC‐positive
   fibers (Figure [99]4C; Figures [100]S7B and [101]8B, Supporting
   Information). The remaining Pax7+ cells have lost their regular contact
   with tdT+ cells, 20% of double‐positive cells were observed in the MKO
   mice, compared to ≈40% in the control mice at E13.5 (Figure [102]S7B,
   Supporting Information). With the increasing age, progressively fewer
   double‐positive cells were detected at E14.5 limbs (Figure [103]4C;
   Figure [104]S8B, Supporting Information). Apparently, the pool of MPCs
   underwent an accelerated depletion, which could be due to dispersed
   distribution and increased apoptosis of tdT+ cells.

Figure 4.

   Figure 4
   [105]Open in a new tab

   Decreased proliferation of Myf5‐derived cells and decreased expression
   of Pax7 in the mutant forelimbs. A) Representative confocal images of
   tdT (red) and Ki67 (green) immunostaining of forelimb sections at E14.
   Scale bars, 100 µm. White dotted boxes are shown enlarged in the right
   panels. Quantitative analysis of ki67^+ and tdT^+ cells was shown on
   the top bar graph (n = 3). B) Representative confocal images of tdT and
   TUNEL staining with forelimb sections at E15.5. Scale bars, 100 µm.
   Quantitative analysis of TUNEL and tdT double positive cells was shown
   on the bottom bar graph to the left of pictures (n = 3). C)
   Representative confocal images of tdT (red) and Pax7 (green)
   immunostaining with forelimb sections at E14.5. Scale bars, 100 µm.
   Note that white dotted boxes are shown enlarged in the right panels.
   Quantitative analysis of Pax7 and tdT cells was shown on the right bar
   graph (n = 3). All data are shown as the mean ± SD. 2‐tailed Student's
   t‐test. *** p < 0.001.

2.5. scRNA‐seq Revealed Diverse Cell Populations of tdT^+ Cells and
Identified Exhausting Skeletal Muscle Cell Population in the Mutant Mice

   To characterize cell fates and dynamics of Myf5‐derived cells with or
   without SRSF2 in skeletal muscle development, we isolated mononuclear
   cells from E14 Het/tdT and MKO/tdT embryos that were decapitated and
   eviscerated, then separated tdT^+ cells by flow cytometry and applied
   them to scRNA‐seq (Figure [106] 5A). At this stage we found a
   significant decrease in the number of tdT^+ cells in the mutant embryos
   compared to control (7.43% vs 28.6%). After quality control
   (Figures [107]S9 and [108]S10, Supporting Information), we retained a
   total of 21 265 transcriptomes, including 9439 control and 118 226 MKO
   cells for the downstream analysis (Figure [109]5A).

Figure 5.

   Figure 5
   [110]Open in a new tab

   scRNA‐seq identifies distinct cell lineages of Myf5‐derived cells. A)
   Scheme of single cell preparation from control and MKO/tdT truncated
   embryos at E14, FACS isolation for tdT^+ cell, and scRNA‐seq. B) t‐SNE
   plot of combined dataset with cells colored by cell types. C)
   Expression of representative marker genes in distinct cell clusters. D)
   t‐SNE plot of overlapping control and MKO/tdT datasets (top), and the
   proportion of different cell types in control and MKO/tdT samples
   (bottom).

   The t‐distributed stochastic neighbor embedding (t‐SNE) plots revealed
   19 cell clusters, which displayed distinct transcriptional programs
   (Figure [111]5B; Figure [112]S11A, Supporting Information). Based on
   established lineage‐specific marker genes (Figure [113]5C) and
   cluster‐cluster distance analysis (Figure [114]S11B, Supporting
   Information), we assigned these 19 clusters into 10 different cell
   types (Figure [115]5B, right panel). Both cluster 4 (C4) and C8
   expressed skeletal muscle cell (SKM) markers, such as Pax7, Myf5,
   Myod1, Myog, and Myh3, which constitute the myogenic lineage. On the
   other hand, the remaining 9 cell types contributed to non‐myogenic
   lineages. Mesenchymal cells (Mesen) composing of C1, C3, C5, and C11,
   were enriched for Pdgfra and Dcn; chondrocytes (Chon:C2) highly
   expressed Chad5 and Sox5; Osteoblasts (Osteo: C10) expressed Sp7;
   neural cells (Neu: C6, C15 and C18) were enriched for Dcx and Tubb3;
   Schwann cells (Sch: C12) expressed Cdh19; endothelial cells (Endo: C7
   and C9) highly expressed Pecam1 and Kdr; immune cells (Immu: C14, C17,
   and C19) expressed Srgn and epithelial cells (Epi:C16) specifically
   expressed Krt5. While the house‐keeping gene Atcb was expressed in all
   the cell populations, Mki67 was mainly enriched in SKM and Mesen groups
   (Figure [116]5C). These results revealed high heterogeneity of
   Myf5‐derived cells during the development. But the current expression
   of Myf5 was predominantly observed in the SKM population.

   We next analyzed the proportions of these distinct cell types in
   control and mutant samples (Figure [117]5D). Approximately 21.44% of
   tdT^+ cells became myogenic in the control, however only a small
   percentage of cells (≈1.43%) belong to the SKM in the mutant mice. In
   contrast, the proportions of Neu, Endo, and Immu cells were
   significantly increased in the KO mice, perhaps resulting from the
   excessive loss of SKM cells. Besides, the Mesen group occupied ≈49% of
   total Myf5‐derived cells in either control or mutant mice, but they
   were not totally overlapped between the two samples (Figure [118]5D,
   top panel). Taken together, loss of SRSF2 significantly changed cell
   proportions among Myf5‐derived cells.

2.6. Trajectory Analysis of SKM Populations Revealed Aberrant Proliferation
and Differentiation of Myoblasts in the Absence of SRSF2

   We next performed unsupervised clustering on the SKM population, which
   identified a total of six unique subclusters that we labeled as sC1 to
   sC6 (Figure [119] 6A). Specifically, sC1 cells expressed high levels of
   Pax7, Myf5, and Myod1, also enriched for Mki67, thus we defined sC1 as
   activated myoblasts. sC2 cells were less activated than sC1 cells, as
   they expressed relatively low levels of these markers. sC3, sC4, and
   sC6 were identified as differentiated populations based on expression
   of differentiation and mature fiber markers such as p21, Myog, and
   Myh3. And sC5 cells, enriched for both activation and differentiation
   makers, might represent transitional states (Figure [120]6B,C). In
   addition, we noticed an increased proportion of differentiated cells
   and a decreased proportion of activated cells in the mutant compared to
   the control (Figure [121]6D).

Figure 6.

   Figure 6
   [122]Open in a new tab

   Subclustering of SKM cells and pseudotime analysis. A) t‐SNE
   projections of subclustered SKM cells, labeled in different colors. B)
   Violin plot indicating gene expression signatures in SKM cell
   subclusters. C) tSNE plot showing the expression of marker genes of
   Pax7, Myf5, Myod1, and Myog. D) the proportion of subclusters in
   control and MKO/tdT samples. E) Pseudotime‐ordered analysis of SKM
   cells, constructed by Monocle2. Right panel was colored by the
   subclusters. Left panel was colored by the pseudotime order. F)
   Pseudotime‐ordered analysis of SKM cells, from the mutant (left panel)
   and control (right panel) samples. G) Heatmap showing the dynamic
   expression of the selected genes along the trajectories b1 and b2.

   We then used Monocle2^[ [123]^19 ^] to analyze dynamic cell transitions
   during the myogenesis, which arranged the six subclusters along one
   major trajectory and two smaller trajectories (Figure [124]6E).
   Majority of sC1 and sC2 plotted tightly together at the start of
   pseudotime (pre‐b), and only small fraction of cells ended at the
   trajectory b2. Differentiated cells located toward or at the end of the
   major branch (b1) and small branch (b3), whereas sC5 cells did not
   exhibit a spatial bias. Even though there were only small number of
   mutant cells, they were more inclined to be distributed towards b1 than
   the control cells (Figure [125]6F). We next investigated the dynamic
   changes of indicated gene expression along the branches b1 and b2
   (Figure [126]6G). High expression of Pax7, Myf5 and Myod1was observed
   at the start of pseudotime, whereas p21 was restricted to the end of
   b1, and the quiescent marker Gpx3 was enriched at the end of b2,
   together with the total loss of Myod1. This strongly suggested that
   activated myoblasts either go to differentiation, or enter into the
   quiescence.

   Quiescent cells account for satellite cells, and they were mainly
   responsible for postnatal skeletal muscle development.^[ [127]^20 ^] We
   next focused our attention on the differentiation trajectory. We
   excluded portion of the quiescent cells and performed the dynamic
   changes of gene expression along the differentiation process
   (Figure [128] 7A; Figure [129]S12, Supporting Information). And
   representative GO terms in biological process was illustrated in
   Figure [130]7B. Down‐regulated genes were enriched in MPC markers,
   nucleosome assembly, cell cycle, and DNA replication et al such as
   Pax7, Myf5, E2f1, Kif11, Tfam, Mki6, Esco2. Up‐regulated genes are
   mainly related with skeletal muscle contraction and fiber development,
   for example Myog, Ckm, and Myh3. We next analyzed the expression levels
   of genes along the pseudotime in the mutant and control separately.
   Interestingly, we observed that expression of Myf5 was significantly
   upregulated at the earlier pseudotime accompanied by increased levels
   of Myog and Cdkn1a in the mutant, compared to the control, whereas
   levels of Pax7 first decreased and then rapidly increased in the mutant
   (Figure [131]7C). Moreover, proliferation‐related genes Tfam, Rad21,
   Ncapd2, and E2F1 were all dysregulated, while glycolysis‐related genes
   Eno3, Pgam2, Aldoa, and Ldha were all upregulated in the mutant
   (Figure [132]7D,E). Consistently, cytoskeleton‐related genes Col1a1,
   Col3a1, Postn, and Dag1 were downregulated and expression of stress
   genes ATF3, Fos, Hspa1a, and Hspa1b were significantly increased along
   the pseudotime in the mutant compared with the control
   (Figure [133]7F,G). These data strongly indicated that myoblasts
   display abnormal proliferation and differentiation in the absence of
   SRSF2.

Figure 7.

   Figure 7
   [134]Open in a new tab

   Analysis of SKM cell transition states in MKO/tdT and Het/tdT samples.
   A) Heatmap showing the dynamic changes in gene expression along the
   differentiation pseudotime. Representative genes were listed on the
   left. B) The enriched GO terms (biological processes) changes with the
   degree of enrichment along the pseudotime. The horizontal axis is the
   pseudo‐time and the vertical axis is an enrichment item. C) Plots
   showing the expression scores for indicated myogenesis markers during
   the pseudotime order. D) Plots showing the expression scores for genes
   involved in karyokinesis and proliferation during the pseudotime order.
   E) Plots showing the expression scores for glycolysis‐related genes in
   pseudotime order. F) Plots showing the expression scores for genes
   related with cytoskeleton and stress‐response during the pseudotime
   order.

2.7. Precocious Differentiation and Depletion of the Progenitor Pool Were
Observed in the Mutant Mice and Myoblasts

   In order to validate scRNA‐seq data, we first performed immunostaining
   for Pax7 and the DNA‐damage marker γH2AX using hindlimb and intercostal
   sections prepared from E15 MKO and WT embryos. Although number of
   Pax7^+ cells in the mutant muscles were greatly dropped, more than 10
   times Pax7^+/γH2A^+ cells were observed in the mutant than in the
   control (Figure [135] 8A), suggesting that mutant Pax7^+cells tend to
   die easily, compared to control cells. Moreover, levels of p21 were
   significantly increased in the mutant muscles (Figure [136]8B). And
   Myf5^+ cells were undergoing expansion more extensively in the KO mice
   than in the control, as addressed by Myf5 and Ki67 double staining
   (Figure [137]8C). Consistently, severely impaired myogenesis occurred,
   as diminished expression of Myog, MHC and embryonic myosin chain (MYH3)
   was observed in the mutant mice at E15 (Figure [138]S13, Supporting
   Information).

   Next we transfected primary myoblasts with siNC or siSRSF‐#1, #2 in GM
   for 72 h, then stained for Myog and MHC. A lot of Myog^+ myoblasts and
   MHC^+ myotubes were observed in SRSF2‐knockdown cells, indicating that
   SRSF2 knockdown caused precocious myogenic differentiation
   (Figure [139]8D,E, GM 72 h). However, upon differentiation,
   SRSF2‐knockdown myoblasts displayed extremely decreased differentiative
   capacity (Figure [140]8D, DM 24 h). TUNEL analysis further confirmed
   high rates of apoptosis within differentiated myotubes (yellow arrows)
   as well as within undifferentiated myoblasts following SRSF2 depletion
   (Figure [141]8D,E, DM 24 h and GM 72 h). Consistent with phenotypes, WB
   analyis demonstrated that siSRSF2 induced the expression of Myog and
   MHC, upregulation of p21 and the cell‐death marker cleaved caspase 3,
   while decreased ppRb levels, compared to control siRNA
   (Figure [142]8F). Taken together, SRSF2 knockdown caused precocious
   differentiation and high rates of apoptosis in primary myoblasts.

2.8. SRSF2 Regulated Transcription of Cell‐Cycle/Cytoskeleton‐Related Genes
and p21 Depletion Rescued Defects Seen in SRSF2 Knockdown

   The skeletal muscle defects of SRSF2‐null mice are reminiscent of the
   Rb null phenotypes.^[ [143]^11 ^] We next wanted to examine whether
   SRSF2 mediated its effects on myogenesis through cell cycle‐related
   genes. Given that p21 was significantly upregulated in the mutant
   muscle and also in SRSF2‐knockdown myoblasts, we decided to test if
   knockdown of p21 could rescue defects seen in SRSF2 depletion. To this
   end, we co‐transfected primary myoblasts with siRNA against p21 (sip21)
   and siSRSF2, followed by immunostaining for p21, Ki67, Myog, or MHC,
   respectively. As shown, levels of p21 were efficiently reduced in
   myoblasts transfected with both siSRSF2 and sip21, compared to
   siSRSF2‐transfected cells. Downregulation of p21 resulted in increased
   Ki67 positivity in double siRNA‐transfected cells, compared to
   siSRSF2‐transfected cells, although it had no significant effects on
   proliferation of cells only transfected with sip21 (Figure [144]
   9A,[145]B). Moreover, knockdown of p21 significantly rescued precocious
   differentiation induced by downregulation of SRSF2, as decreased
   staining for Myog and MHC was observed in double knockdown cells, in
   comparison with SRSF2‐knockdown cells (Figure [146]9C,D). In summary,
   these results strongly indicated that depletion of SRSF2 appeared to
   activate p21‐dependent pathway in primary myoblasts.

Figure 9.

   Figure 9
   [147]Open in a new tab

   Knockdown of p21 could rescue defects of proliferation and
   differentiation induced by SRSF2 depletion. A) Representative confocal
   images of p21 and Ki67 immunostaining in primary myoblasts Myoblasts
   were transfected with siRNAs as indicated for 48 h, and then reseeded
   on coverslips (2 × 10^4 cells for siSRSF2‐#1 and siSRSF2‐#1 groups, 1 ×
   10^4 cells for other groups) before staining. Qualification of p21^+
   cells and Ki67^+ cells are shown in (B) (n = 3). C) Representative
   confocal images of Myog, MHC immunostaining of primary myoblasts. Scale
   bars, 100 µm. Qualification of Myog+ cells and MHC+ cells are shown in
   (D) (n = 3). All data were shown as the mean ± SD. 2‐tailed Student's
   t‐test. *p < 0.05.**p < 0.01.***p < 0.001.

   Previous reports demonstrated that SRSF2 functions as a potent
   transcription activator.^[ [148]^15 , [149]^21 ^] We then took
   advantage of public database and searched SRSF2 ChIP‐seq signals on the
   downregulated genes revealed by scRNA‐seq analysis. As expected, cell
   cycle‐involved genes E2F1, Hjurp, Aurka, and Aurkb,
   cytoskeleton‐related genes Col1a1 and Vimentin showed strong SRSF2
   binding signals on their promoter regions. And ChIP‐qPCR analysis and
   luciferase activity assays provided further evidence that SRSF2
   directly regulated the expression of genes Hjurp and E2F1
   (Figure [150]S14, Supporting Information). Hjurp has been identified as
   a key factor to promote chromosomal segregation and cell mitosis.^[
   [151]^22 ^] Interestingly, knockdown of Hjurp significantly decreased
   myoblast proliferation and also increased p21 levels, similar to the
   SRSF2 knockdown (Figure [152]S15, Supporting Information).

2.9. Mesen Cells Subcluster into Distinct Cell Populations

   Skeletal muscle contains a complex arrangement of connective tissue,
   which are important for skeletal muscle development.^[ [153]^23 ^]
   Next, we performed unsupervised clustering on the Mesen group and
   observed further heterogeneity with 11 subclusters, which display
   distinct proportions of cell populations between Het/tdT and MKO/tdT
   samples (Figure [154]S16A,B, Supporting Information). As shown, Mesen_1
   subcluster displayed high expression of Thbs4, Col3a1, Tnmd, Postn, and
   Cqgtnf3 (Figure [155]S16C,D, Supporting Information). Expression of
   these genes were mainly enriched in the endomysium, perimysium and
   epimysium cells of skeletal muscle.^[ [156]^23 ^] Thus, we identified
   the Mesen_1 subcluster as skeletal muscle fibroblasts. On the other
   hand, we designed the Mesen_3 subcluster as dermal fibroblasts because
   Foxd1 was enriched, and Foxd1‐expressing progenitors contributed to
   dermal fibroblast population.^[ [157]^24 ^] It is clear that loss of
   SRSF2 forced mesenchymal cell progenitors to change cell fate from
   skeletal muscle fibroblasts to the dermal fibroblasts. Moreover,
   differential gene expression analysis further revealed that expression
   of major components of extracellular matrix (ECM) Postn, CD44, Ndn, and
   Col4a1 were significantly decreased in the Mesen_1 subcluster,
   indicating that the functional ECM production was severely damaged in
   the mutant muscles.

3. Discussion

   Skeletal muscle formation is a highly precise and coordinated process,
   which requires the proliferation of myoblasts, permanent withdrawal
   from the cell cycle, differentiation and final fusion into
   multinucleated muscle fibers.^[ [158]^25 ^] Cell cycle arrest is a
   prerequisite for differentiation.^[ [159]^26 ^] The abnormal myogenesis
   observed in SRSF2‐null embryos could be the consequences of abnormal
   apoptosis and proliferative defects of MPCs, premature differentiation
   of myoblasts, and altered myogenic fate of Myf5‐derived cells.

   Examination of Myf5‐Cre//Rosa26lacZ mice revealed the broad presence of
   Myf5‐derived cells in tissues originating from the paraxial mesoderm
   and the neuroectoderm.^[ [160]^27 ^] Consistent with this, our combined
   lineage tracing and scRNA‐seq analysis identified many previously
   unrecognized cell types that transiently harbor Myf5 activity. These
   results could help explain some of previously reported discrepancies.
   For example, rib defects were initially reported in homozygous Myf5
   mutant mice,^[ [161]^5 , [162]^8 ^] and later in mice with ablated
   Myf5‐expessing cells,^[ [163]^27 ^] but it is still controversial
   whether Myf5 was directly involved in the generation of the rib
   phenotype. Our data showed that the transient expression of Myf5 occurs
   in the progenitors of developmental chondrocytes and osteocytes. Loss
   of temporal Myf5 activity during the early development could be
   responsible for these rib defects, which are another hallmark features
   observed in the SRSF2‐null mice. In addition, Myf5‐dervied cells also
   contributed to mesenchymal cells, endothelial cells, pericytes, neural
   cells and Schwann cells. These non‐myogenic cell lineages could
   potentially play important roles during skeletal muscle development.

   High rates of apoptosis and severe skeletal muscle defects were
   observed in the mutant mice, which is reminiscent of Rb mice.^[
   [164]^11 ^] Interestingly, SRSF2 knockdown switched off E2F1‐dependent
   genes, while activated MyoD‐dependent genes, including increased pRb
   and decreased ppRb. This indicated a possibility that SRSF2 might
   control myogenesis through regulating cell‐cycle‐related genes.
   Significantly, scRNA‐seq analysis and immunostaining assay demonstrated
   that loss of SRSF2 resulted in precocious differentiation and a
   progressive depletion of the progenitor pool, as well as apoptosis.
   Moreover, p21 knockdown could rescue apoptosis and premature
   differentiation defects seen in SRSF2‐knockdown myoblasts. Fully in
   line with previous reports that SRSF2 functions as a potent
   transcription regulator,^[ [165]^15 , [166]^21 ^] strong SRSF2 binding
   signals were observed on the promoter regions of cell cycle‐involved
   genes E2F1, Hjurp, Aurka, and Aurkb, and cytoskeleton‐related genes
   Col1a1 and Vimentin. Interestingly, we found that there were up to nine
   potential E‐box sequences within the promoter region of E2F1 and at
   least one E‐box for Vimentin (Figure [167]S17, Supporting information).
   Therefore, it is possible that SRSF2 might regulate transcription by
   forming complex with MRFs during myogenesis, which need detailed
   investigation in the future.

   It is well‐recognized that Pax3/Pax7 positive cells constitute a
   reserve of myogenic progenitors present not just in the mature myotome,
   but also in all developing muscle masses. And during the limb muscle
   development, Pax3^+/Pax7^+ MPCs give rise to cells that express Myf5
   and/or MyoD. However, scRNA‐seq analysis revealed that Pax7 and Myf5
   had a similar expression pattern in the Myf5‐derived proliferative
   myoblasts, whereas MyoD was wildly expressed in differentiated and
   proliferative myoblasts, and Pax3 signals were only weakly detected
   (Figure [168]5C). Pax3 was weakly expressed because Myf5 genetically
   downstream of Pax3 in the myogenic context,^[ [169]^28 ^] or because
   there were two distinct myogenic pathways defined by Pax3 and Myf5,
   respectively.^[ [170]^6 ^] These results indicated a seemly different
   genetic hierarchy than previously described exists in vivo myogenesis.
   Further investigation is required for verification.

   Community effect was first described by Gurdon et al, which indicates
   that during skeletal muscle development, MPCs and myoblasts are
   required to be arranged as dense clusters in order to promote cell
   differentiation in union.^[ [171]^18 ^] In this study, we observed the
   community effect was totally lost in the SRSF2‐null mice. Myf5‐derived
   cells in the absence of SRSF2 appeared to move randomly in the limb,
   and some even migrated into the livers and bones. This randomly
   dispersed pattern was companied by drastically impaired myogenesis and
   exhaustion of MPCs in the mutant mice. scRNA‐seq analysis revealed that
   almost 50% of Myf5‐drived cells contributed to the Mesen population,
   and subclustering analysis of this population confirmed that
   Foxd1‐enriched dermal fibroblasts^[ [172]^24 ^] were abundant in the
   mutant mice at the expense of Thbs4/Col3a1‐enriched skeletal muscle
   fibroblasts. These findings indicated that loss of SRSF2 also severely
   impaired development of mesenchymal cells surrounding muscle cells. And
   the destruction of skeletal muscle cell microenvironment could lead to
   loss of community effect in the knockout mice. This was totally
   consistent with previous reports that some unknown factors in the limb
   mesenchyme would act in providing the community effect to promote
   proliferation and differentiation of limb muscle cells.^[ [173]^29 ^]

4. Conclusion

   In conclusion, our data demonstrated thatMyf5‐derived cells contributed
   to various cell lineages than previously described, and SRSF2 was a
   critical regulator for embryonic myogenesis. SRSF2 regulates the entry
   of Myf5 cells into the myogenic program and ensures their survival by
   preventing precocious differentiation and apoptosis.

5. Experimental Section

Generation of Mutant Animals

   SRSF2^f/f mice were generously provided by Dr. Fu. The SRSF2^f/f mice
   were crossed with transgenic C57BL/6J mice expressing Cre recombinase
   driven by the Myf5 promoter. Resulting progeny (SRSF2^f/w/Myf5‐Cre)
   were mated with SRSF2^f/f mice to generate skeletal muscle‐specific
   SRSF2 knockout mice SRSF2^f/f/Myf5‐Cre (MKO). The Rosa26‐LSL‐tdtomato
   (hereafter tdT) mice were purchased from The Jackson Laboratory and
   have been described previously.^[ [174]^14 ^] The SRSF2^f/f mice were
   crossed with dT breeders to generate SRSF2^f/f/tdT mice. Next,
   SRSF2^f/f/tdT mice were crossed with SRSF2^f/w/Myf5‐Cre mice to
   generate SRSF2^f/f/Myf5‐Cre/tdT mice (MKO/tdT).
   SRSF2^f/w/Myf5‐Cre/tdT(Het/tdT) mice served as controls. Primer
   sequences used in the genotyping are listed in Table [175]S3,
   Supporting Information. All experiments were conducted in accordance
   with the guidelines of the Institutional Animal Care and Use Committee
   of Shanghai Institute for Nutrition and Health Nutritional Sciences,
   Chinese Academy of Sciences.

Primary Myoblast Isolation, Cell Culture, Differentiation, and Transfection

   Primary myoblasts were isolated from the dorsal muscle and limbs from
   E16 to E16.5 embryos. Briefly, muscles were isolated, minced, and
   dissociated in 0.2% (wt/vol) collagenase‐type XI (Sigma‐Aldrich) and
   2.4 U mL^−1 dispase II (Invitrogen) in DMEM (Gibco) at 37 ℃. After
   centrifugation, the pellet was filtered through an 80 µm cell strainer
   and resuspended in growth medium containing F‐10 (ThermoFisher)
   supplemented with 20% FBS (Gibco), 2.5 ng mL^−1 FGF, 1%
   penicillin/streptomycin. Cells were laid repeatedly on the dish for 3 h
   to remove the adherent cells, then unattached cells (myoblasts) were
   harvested and plated on the Matrigel‐coated (1% v/v Matrigel in DMEM)
   dishes. Cells were not allowed to subculture 75% confluence during the
   passage. Primary myoblasts were incubated in growth medium containing
   F‐10 supplemented with 20% FBS, 2.5 ng mL^−1 FGF, 1%
   penicillin/streptomycin. C2C12 cells were incubated in growth medium
   containing DMEM supplemented with 10% FBS and 1%
   penicillin/streptomycin. Myogenic differentiation of both primary cells
   and C2C12 cells was activated by replacing the growth medium with a
   differentiation medium (DMEM supplemented with 2% horse serum). For
   transient transfection, primary myoblasts and C2C12 cells were
   transfected with siRNA or plasmids using the Lipofectamine RNAiMAX
   (Invitrogen) or Lipofectamine 200 (Invitrogen) as described in
   manufacturer's protocol. The siRNA sequences used are listed in
   Table [176]S6, Supporting Information.

RNA Extraction, cDNA Synthesis, and qPCR

   RNA extraction, cDNA synthesis and qPCR were conducted as previously
   described.^[ [177]^30 ^] Total RNA was extracted from mouse tissue or
   cultured cells with Trizol Reagent (Invitrogen). cDNA was synthesized
   using the High‐Capacity cDNA Reverse Transcription kit (Applied
   Biosystems). Gene expression was assessed using standard qPCR
   approaches with SYBR Premix Ex Taq kit (TaKaRa). Analysis was performed
   on a QuantStudio Real‐Time PCR System (Applied Biosystems). Data was
   normalized relative to Rplp0. Primer sequences used in qPCR are listed
   in Table [178]S2, Supporting Information.

WB Analysis

   Cells were homogenized in RIPA Lysis and Extraction Buffer (Thermo
   Scientific) supplemented with Complete Protease Inhibitor Cocktail
   (Roche). Equal amounts of proteins were separated on SDS‐PAGE gels,
   then transferred to immobilon‐NC Transfer Membrane (Millipore).
   Membranes were incubated with primary antibodies overnight and then
   with HRP‐conjugated secondary antibodies for 1 h. Specific signals were
   detected with the chemiluminescence system (Tanon‐4100). The detailed
   information of antibodies used is listed in Table [179]S1, Supporting
   Information.

HE, TUNEL Assays, and Immunostaining

   HE and TUNEL assay on the paraffin sections were performed using a
   standard protocol. 6 µm‐thick sections of tissues were processed for HE
   staining. 5 µm‐ thick sections were processed for TUNEL staining
   according to the manufacturer's protocol (Promega, catalog no. G3250).
   Immunofluorescence on the paraffin tissues and frozen sections was
   conducted as previously described.^[ [180]^30 , [181]^31 ^] Briefly, 6
   µm paraffin tissues were deparaffinized in xylol and rehydrated with
   gradient ethanol followed by antigenic retrieval. Both frozen and
   paraffin sections were blocked with 5% goat serum (Santa Cruz
   Biotechnology) and then incubated with primary antibodies overnight at
   4 ℃, followed by incubation with secondary antibodies as well as DAPI
   for 1 h. For cellular immunostaining, cells were fixed in 4%
   paraformaldehyde for 15 min, permeabilized with 0.2% Triton X‐100 in
   PBS and blocked with 5% goat serum, followed by incubation with primary
   antibodies overnight at 4 ℃ and secondary antibodies as well as DAPI
   for 1 h. Immunofluorescence images were acquired on a Zeiss LSM880NLO
   Confocal Microscope. Quantifications were performed using ImageJ
   software (National Institutes of Health). The detailed information of
   used antibodies is listed in Table [182]S1, Supporting Information.

Chromatin Immunoprecipitation Assays (ChIP‐qPCR)

   Potential SRSF2 binding sites located on the Hjurp and E2F1 promoter
   regions were obtained from the Gene Expression Omnibus (GEO) under the
   accession number [183]GSM1106075 and [184]GSM1106077. Based on this,
   two pair of primers were designed for the ChIP‐qPCR analysis. Myoblasts
   were transiently transfected with SRSF2‐HA and vector‐HA plasmids for
   48 h. Then ChIP was performed with Magna ChIP kit (Millpore; catalog
   no. MAGNA0017) according to the manufacturer's instructions, using
   anti‐HA antibodies (Abcam) for IP with normal mouse IgG as negative
   control. The immunoprecipitated DNA and input DNA were quantified by
   qPCR analysis. The detailed information of antibodies used is listed in
   Table [185]S1, Supporting Information. Primer sequences used in
   ChIP‐qPCR are listed in Table [186]S4, Supporting Information.

Dual‐Luciferase Activity Assay

   For construction of luciferase reporter plasmids, the Hjurp promoter
   region (position −536 bp to 522 bp from the transcription start site,
   TSS) or E2F1 promoter region (position −680 bp to 767 bp from TSS) was
   inserted into the pGL3 vector (Promega). 293T cells were transfected
   using polyethylenimine (PEI; Sigma) according to the standard
   conditions with a 1:1 ratio of DNA (mg) and PEI (ml). For reporter
   assays, 293T cells were transfected with pRL‐TK plasmid and
   Hjurp‐luciferase plasmid or E2F1‐Luciferase plasmid, plus vector‐HA or
   SRSF2‐HA. Cells extracts were harvested post 48 h transfection.
   Dual‐luciferase assays were performed according to the manufacturer's
   instructions with a Dual‐Luciferase assay kit (Promega). The primer
   sequences used for pGL3‐based reporter plasmid construction are listed
   in Table [187]S5, Supporting information.

Cell Viability Assays

   For cell viability assay, 1 × 10^3 C2C12 cells were plated in 96‐well
   plates, and transfected with vector‐HA or SRSF2‐HA. The proliferation
   rate of cells was detected using the CCK8 method based on the
   manufacturer's instructions. Cell viability was assessed at 1, 2, 3 and
   4 days after transfection.

Preparation of Single Cell Suspension and scRNA‐seq

   E14 MKO/tdT and Het/tdT embryos were collected, decapitated and
   eviscerated. To generate single cells, truncated embryos were incubated
   in 0.2% (wt/vol) collagenase‐type XI (Sigma‐Aldrich) and 2.4 U mL^−1
   dispase II (Invitrogen) in DMEM at 37 ℃ 30 min, followed by filtering
   through a 40‐µm cell strainer. Cells were then centrifuged, resuspended
   in cold 5% FBS in PBS, and immediately sorted with a Beckman moflo
   Astrios EQ FACS sorter. Endogenous tdT signal was detected through PE
   channel and gates were set up according to the positive and negative
   controls. Sorted tdT+ cells were resuspended in 20% FBS/PBS buffer and
   loaded onto a droplet‐based library prep platform Chromium (10x
   Genomics) with a Chromium Single Cell Reagent Kit (version 3) according
   to the manuscript's instructions. Sequencing was performed on an
   Illumina nova 6000 and 150‐base pair paired‐end reads. The sequencing
   and bioinformatics analysis were performed by OE Biotech Co., Ltd.
   (Shanghai, China). All statistical analyses, unless otherwise
   specified, were conducted using R.^[ [188]^32 ^]

scRNA‐seq Data Preprocessing

   Raw sequencing data was processed by Cell Ranger software suite (10x
   genomics, version 5.0.0) including demultiplexing cellular barcodes,
   mapping reads to the reference, and down‐sampling reads to aggregate
   data from samples, resulting in a gene count matrix. Next cells of
   low‐quality or likely multiplets were removed if the unique molecular
   identifier (UMI)/gene counts were two standard deviations (SD) from the
   mean of all cells. Cells with >10% of the counts from mitochondrial
   genes were further filtered out. NormalizedData function from Seurat
   (version 3.1.1) was then applied to cells to normalize and log
   transform the data with the option “LogNormalize”. To reduce the
   dimensionality of the data, first, top variable genes were selected
   using FindVariableGenes function in Seurat with mean.function =
   “FastExpMean” and dispersion.funciton = “FastLogVMR”, and then
   performed Principal component analysis (PCA) with RunPCA function.
   Graph‐based clusters were identified using the FindClusters function
   and visualized by t‐distributed stochastic neighbor embedding (t‐SNE)
   plot using RunTSNE and DimPlot functions in Seurat. Marker genes for
   each cluster were identified by the FindAllMarkers function (test.use =
   bimod) by comparing the gene expression in the cluster of interest
   versus all other clusters. Differentially expressed genes (DEGs) were
   idenfied using the FindMarkers function (test.use = MAST) with p value
   < 0.05 and |log2foldchange| > 0.58. Gene Ontology (GO) and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of
   DEGs were respectively tested assuming a hypergeometric distribution.

Pseudotime Analysis

   The Monocle2 package^[ [189]^19 ^] was used to study the developmental
   pseudotime. Raw counts in Seurat object were first converted to
   CellDataSet object by the importCDS function in Monocle2. Then genes
   which were informative in ordering cells along the pseudotime
   trajectory were identified by the differentialGeneTest (qval < 0.01).
   The dimensional reduction clustering analysis was performed with the
   reduceDimension function, followed by trajectory inference with the
   orderCells function using default parameters. The
   plot_genes_in_pseudotime function was applied to track gene expression
   and to track changes over pseudotime. To discover the order of
   gene‐expression and functional events during cell state transitions at
   a single‐cell resolution, GeneSwitches analyses were conducted.
   Specifically, gene expression data was first binarized into 1 (on) or 0
   (off) state with the binarize_exp function in the GeneSwitches package
   (fix_cutoff = TRUE, binarize_cutoff = 0.05). A gaussian noise with mean
   of 0 and s.d. of 0.1 added to the gene expression data, which ensures
   numerical stability in the model fitting. Genes were also filtered
   without a distinct bimodal “on ‐off” distribution. A logistic
   regression model was then fitted to model the binary states (on or off)
   by the find_switch_logistic_fastglm function with random downsampling
   of zero expressions (downsample = TRUE) to adjurst genes with high zero
   inflation. Finally, the top 50 best fitting genes (high McFadden's
   Pseudo R^2) were plotted along the pseudo‐timeline with “switched on”
   above the line, and “switched off” below the line.

Data Availability

   The data sets have been deposited in the GEO database under the
   accession number [190]GSE189088.

Statistical Analyses

   Statistical analysis was performed by GraphPad Prism 8.0 software. All
   of the data shown in the histograms were the results of at least three
   independent experiments and presented as the mean ± standard deviation
   (SD). The sample size (n) for each statistical analysis was indicated
   in detail in individual figure legend. Differences between groups were
   tested for statistical significance by using the two‐tailed unpaired
   Student's t‐test. Differences between values were considered
   statistically significant when * p < 0.05, ** p < 0.01, and *** p
   < 0.001.

Conflict of Interest

   The authors declare no conflict of interest.

Supporting information

   Supporting Information
   [191]Click here for additional data file.^ (3.2MB, pdf)

Acknowledgements