We first asked whether the QSC pool is composed of a continuum of cell states during tissue homeostasis. We stained freshly isolated single muscle fibers from adult mice, with the SC identify marker, Pax7 (Sambasivan et al., 2011; von Maltzahn et al., 2013; Lepper et al., 2011) and Ddx6 (p54/RCK), an RNA helicase found enriched in P-bodies and stress granules (Buchan and Parker, 2009). Pax7 and Ddx6 are expressed in QSCs and decrease during activation and commitment (Crist et al., 2012; Zammit et al., 2006). A density map of Pax7 and Ddx6 expression shows a broad range of expression levels across the QSC pool (Figure 1A–C). A bivariate plot shows a positive correlation between Pax7 and Ddx6 (Figure 1D). Therefore, based on the expression profile of two different markers, the QSC pool is composed of a continuum of molecular states at tissue homeostasis. This supports models inferred from scRNA-seq analysis on QSCs (De Micheli et al., 2020; Dell’Orso et al., 2019, Kimmel et al., 2020).

To determine the fate bias of cells along the continuum, we isolated Pax7^high-, Pax7^medium-, and Pax7^low -expressing populations of SCs from Pax7-nGFP reporter mice (Rocheteau et al., 2012; Figure 1—figure supplement 1A–C), and analyzed cell cycle entry and differentiation. As expected, the Pax7^low population entered cell cycle and differentiated faster than the Pax7^high expressers, suggesting that Pax7^high SCs are in a more dormant state and the Pax7^low SCs exist in a primed state (Rocheteau et al., 2012). We find that the Pax7^medium population is in an intermediate molecular and phenotypic state between the Pax7^high and Pax7^low QSCs (Figure 1—figure supplement 1D, E). Therefore, the QSC pool exists in a continuum of cell states that gives rise to a continuum of fates.

The multinucleated muscle fiber functions as a niche cell for the SC, regulating the depth of quiescence and rate of activation in response to injury (Eliazer et al., 2019). The Delta-Notch signaling pathway is an evolutionarily conserved intercellular signaling pathway for cell fate specification (Kopan and Ilagan, 2009). Adult muscle QSCs display active notch signaling (Bjornson et al., 2012; Mourikis et al., 2012; Fukada et al., 2011; Low et al., 2018; Verma et al., 2018). Overexpression of NICD1 increases Pax7 expression (Wen et al., 2012) and Rbpj, a transcriptional coactivator of the notch signaling pathway, maintains the SC pool by repressing differentiation (Bjornson et al., 2012; Mourikis et al., 2012). A transgenic notch reporter mouse line where green fluorescent protein (GFP) is under the control of C promoter-binding factor 1 (CBF1) (Duncan et al., 2005) exhibits variable levels of notch activity in QSCs (Bjornson et al., 2012). Using the same reporter, we find that notch activity levels exist on a continuum across the SC pool and positively correlates with Pax7 expression levels (Figure 1E, F), raising the possibility that Notch ligands maintain the continuum of diverse cell states across the QSC pool.

To identify notch ligands specifically expressed in adult myofibers, we first examined Notch ligands on isolated single muscle fibers from postnatal day 3 (p3), postnatal day 7 (p7), and adult through microarray analysis. Dll4 transcripts in the fiber showed the most enrichment as the SC transitioned from proliferating myogenic progenitors (postnatal) to becoming quiescent (adult) (Figure 2—figure supplement 1A). RT-qPCR(Real Time-quantiative PCR) on isolated adult single muscle fibers show the expression of Dll4 transcripts, which was confirmed by RNAscope and immunofluorescence (Figure 2—figure supplement 1B, C and Figure 2A; Kann and Krauss, 2019). Dll4 transcripts are expressed in less than 1% of QSCs (Kimmel et al., 2020; De Micheli et al., 2020), suggesting that the muscle fiber is a source of SC-bound Dll4. Since the QSC pool is distributed as a continuum of cell states, we hypothesized that Dll4 expression across the multinucleated niche cell is heterogeneous. Single muscle fibers stained with anti-Dll4 show that Dll4 protein formed clusters along the muscle fiber and were enriched around the QSCs. Further analysis revealed that heterogeneous spatial expression of Dll4 protein along the muscle fiber correlated with the amount of Dll4 captured by the SCs (Figure 2A, Figure 2—figure supplement 2). We observed a positive correlation between the amount of Dll4 foci on the fibers and the intensity of Dll4 present on the adjacent QSC (Figure 2B).

To directly link Dll4 localization with SC diversity, we isolated and stained SCs for Dll4 and Pax7. A bivariate plot of Pax7 and Dll4 expression within SCs reveals a positive correlation between the Dll4 captured by the SC and Pax7 intensity (Figure 2C). Comparison between Dll4 protein and Pax7 levels using a Pax7-nGFP reporter, confirmed the positive correlation between Pax7 and Dll4 protein expression (Figure 2D, E). In contrast, the localization of Dll4 transcript on the muscle fiber did not correlate with Pax7 expression in the adjacent SC (Figure 2—figure supplement 1C and D).

We next asked about the distribution of Dll4 protein along muscle fibers. Quantification of Dll4 expression levels on single Extensor Digitorum Longus (EDL) muscle fibers showed that Dll4 expression is variable along the fiber, and not linked to the neuromuscular junction based on α-bungarotoxin expression or the myotendinous junction based on location (Figure 2—figure supplement 2). We find no consistent pattern of Dll4 protein along the fibers. Therefore, Dll4 spatial distribution does not map to these known anatomically defined regions of freshly isolated single muscle fibers. In addition, fewer than 10% of fibers are exclusively Dll4^High or Dll4^low, suggesting SC diversity is coordinated in a fiber-autonomous manner. Dll4 expression was also variable in regions devoid of SCs, suggesting the presence of an SC does not dictate Dll4 levels along muscle fibers. Instead, the intensity of Dll4 captured by SCs and the Pax7 expression correlated with the level of Dll4 expression along the muscle fiber (Figure 2—figure supplement 2). Therefore, spatial heterogeneity and levels of niche-derived Dll4 are coupled with a continuum of diverse molecular SC states that are biased to different phenotypic fates.

To determine if Dll4 from the muscle fiber controlled SC diversity, we deleted Dll4 (Hozumi et al., 2008) specifically in the adult muscle fibers using a tamoxifen (tmx)-inducible human ACTA-CreMer mouse line (McCarthy et al., 2012) (transgenic mice are herein called as MF-Dll4^fl/fl) (Figure 3A). The abundance of Dll4 in the muscle fiber and captured by SCs decreased in MF-Dll4^fl/fl compared to control, suggesting that the muscle fiber is the major source of SC-bound Dll4 in uninjured muscle (Figure 3B–D). It is possible that cell sources other than the muscle fiber could provide Dll4 to SCs in other contexts (Verma et al., 2018). We observed that Pax7 and Ddx6 expression in SCs from MF-Dll4^fl/fl fibers decreased compared to controls (Figure 3E, G). Density maps reveal a shift in the distribution of expression levels of Pax7 and Ddx6. We calculated the variance of the expression levels to gauge the spread of the distribution. The decrease in variance in MF-Dll4^fl/fl compared to controls, indicates reduced diversity across the QSC pool, effectively reducing the range of states across the continuum (Figure 3F, H). We observed a 50% decrease in the total number of Pax7⁺ SCs in MF-Dll4^fl/fl fibers compared to controls (Figure 3—figure supplement 1A, B), suggesting a SC loss phenotype.

Deletion of Rbpj from SCs reduced SC population by over 95% (Bjornson et al., 2012; Mourikis et al., 2012). We wondered whether the partial SC ablation after Dll4 deletion was due to a more modest effect on notch signaling. To determine the contribution of niche-derived Dll4 to notch activity in SCs, we crossed mice harboring the notch reporter CBF1-GFP (Duncan et al., 2005) with MF-Dll4^fl/fl. We observed a 95% decrease in notch reporter activity after Dll4 deletion from the niche (Figure 3I), suggesting that Dll4 from the fibers nonautonomously regulates Notch activity in the QSCs and that Dll4 from the muscle fiber is the dominant source of notch ligand. This argues against differential notch activity to explain the differences in SC loss phenotypes.

A decrease in SC diversity, number and Pax7 levels can be explained by a loss of Pax7^high SCs (due to apoptosis or fusion), or a shift in the continuum toward a more committed fusion-competent state, followed by fusion of SCs expressing the lowest levels of Pax7. qRT-PCR on QSCs from control and MF-Dll4^fl/fl revealed a decrease in Pax7 (SC marker) and an increase in Myod (activation marker) and Myog (differentiation marker) transcripts (Figure 5B). The simultaneous loss of Pax7 and induction of Myog a gene not normally expressed in QSCs suggests a shift in the QSC continuum. While a loss of Pax7^high cells alone would not increase Myog expression, we cannot exclude a loss of Pax7^high SCs in addition to the shift in continuum.

To resolve these possibilities, rather than deleting, we reduced Dll4 levels in the muscle fiber by injecting control and MF-Dll4^fl/fl mice submaximal doses of tmx (75 mg/kg/day for seven consecutive days). Reducing Dll4 in fibers redistributed the SC population along the continuum toward lower Pax7 levels without decreasing their numbers (Figure 4). This suggests that heterogeneous levels of Dll4 in muscle fibers maintain a distribution of metastable states across the QSC pool during tissue homeostasis.

Deletion of Dll4 in Pax7⁺ SCs using a Pax7-CreER transgenic mouse line (Nishijo et al., 2009) (referred to as SC-Dll4^fl/fl) did not change Pax7⁺ SC number, compared to controls (Figure 3—figure supplement 2), suggesting that Dll4 does not have an autocrine role in QSCs. Together, these results suggest that Dll4 from the muscle fiber is required to maintain the continuum of diverse cellular states across the QSC pool.

To analyze the phenotypic fates of the QSCs that remained in a Dll4-depleted niche, we cultured isolated single muscle fibers from control and MF-Dll4^fl/fl mice and analyzed them at 0, 30, and 48 hr (Figure 5A). Quantification of SC numbers per fiber reveal that the differences observed between control and MF-Dll4^fl/fl becomes narrow over time in culture (Figure 5C, Figure 5—figure supplement 1A). This observation is consistent with the result that SCs on a Dll4-depleted niche entered cell cycle faster than controls (Figure 5D, Figure 5—figure supplement 1B). The absolute number of Myogenin⁺ cells per fiber at 48 hr was increased, suggesting that a fraction of the remaining SCs on a Dll4-depleted niche rapidly entered the differentiation program when exposed to mitogen (Figure 5E). These data are consistent with a redistribution of the QSC pool along the continuum toward lower levels of Pax7 (Figure 1—figure supplement 1). Induction of Myogenin is consistent with the antidifferentiative role of Rbpj (Bjornson et al., 2012; Mourikis et al., 2012). Based on prior work, we would not have predicted an increase in SC proliferation in mitogen after niche depletion of Dll4 (Bjornson et al., 2012; Mourikis et al., 2012). This likely reflects the relative fraction of SCs that remain after Rbpj deletion in SCs versus Dll4 deletion from the niche.

We next examined the effect of a Dll4-depleted niche on fate potential of the remaining SCs in response to muscle injury. Using the tmx-inducible MF-CreMer/+ genetic model, the niche is genetically modified, but the SCs are genetically wildtype. The wildtype SCs reform the adult muscle fibers during regeneration (Eliazer et al., 2019). Therefore, any regenerative phenotype observed is directly due to the loss of Dll4 in the niche prior to injury. Fourteen days after injury, muscle fiber size in MF-Dll4^fl/fl TA muscle was significantly smaller than controls (Figure 5F, G), consistent with the rapid entrance of some SCs into the differentiation program, as reported after deletion of Rbpj in SCs (Bjornson et al., 2012; Mourikis et al., 2012). The impaired differentiation phenotype in MF-Dll4^fl/fl TA muscle is also seen 40 days after injury (Figure 5—figure supplement 2). The inability to repair muscle fibers after a loss in SC diversity suggests a reduced pool of fusion-competent progenitors.

We next asked whether the remaining population of Pax7⁺ SCs in a Dll4-depleted niche could reestablish the normal continuum of Pax7⁺ states in the injury model. In contrast to the contralateral uninjured muscle, the average Pax7 levels and the variance across the SC population was not different in a Dll4-replenished niche and control niche (Figure 5H, I). In addition, the number of Pax7⁺ cells was similar to control regenerated muscle (Figure 5J). This was unanticipated due to the decrease in Pax7 levels and a shift along the continuum toward a more committed state in the QSC pool (Rocheteau et al., 2012; Kuang et al., 2007). Therefore, SCs expressing modest levels of Pax7 have the potential to produce SCs with higher levels of Pax7, revealing an unappreciated level of plasticity and bidirectionality along the SC continuum during muscle regeneration.

Single nucleus RNA sequencing has uncovered substantial transcriptional diversity across muscle fibers (Kim et al., 2020; Petrany et al., 2020). However, Dll4 activity is regulated post-translationally. Mindbomb1 (Mib1), is an E3 ubiquitin ligase that activates all Notch ligands in the signal sending cell by adding mono-ubiquitin groups onto the ligands (Koo et al., 2005; Koo, 2007). Mib1 in the retinal pigment epithelium (RPE) activates Notch signaling in the adjacent retinal progenitor cell (RPC) by localizing active Notch ligands in the RPE–RPC contacts (Ha et al., 2017). The hypothalamus–pituitary–gonadal axis has been shown to induce Mib1 in muscle fibers during development, thus activating notch signaling in juvenile SCs (Kim et al., 2016). To determine if the heterogeneous pattern of Dll4 on the adult muscle fiber is directed by Mib1, we analyzed the spatial expression of Mib1 on adult muscle fibers. We stained isolated single muscle fibers with Mib1 antibody and found heterogeneous expression pattern of Mib1 on individual muscle fibers. Similar to Dll4 expression, we observed a positive correlation between the intensity of Mib1 expression on muscle fibers and Pax7 intensity within the adjacent SC (Figure 6A, B). Formation of Dll4 clusters is a marker of the activated form of the ligand (Le Borgne and Schweisguth, 2003). To test whether Mib1 played a role in clustering and activating Dll4 on the muscle fibers, Mib1 floxed mice (Koo, 2007) were crossed with tamoxifen-inducible human ACTA1-CreMer mice (transgenic: hereafter called as MF-Mib1^fl/fl). After a 30-day chase, the expression of Mib1 transcripts was decreased in the Mib1-depleted myofibers compared to controls (Figure 7A, B). Analysis of Dll4 expression in control and MF-Mib1^fl/fl isolated single muscle fibers, revealed a loss of Dll4 clusters and a reduction in the intensity of Dll4 expression in Mib1-depleted niche (Figure 6C–E). Therefore, Mib1 directs the activation and heterogeneous patterning of Dll4 within a multinucleated niche cell.

Based on the loss of SC-bound Dll4 after Mib1 deletion in the niche, we analyzed the levels and diversity of Pax7 across the QSC pool on a Mib1-depleted niche. Compared to controls, Pax7 levels decreased, the variance across the QSC pool was reduced and SC number was less in a Mib1-depleted niche (Figure 6C, F, G). RT-qPCR of myogenic markers shows that Pax7 and Myod transcript levels did not change. However, Myog was upregulated in the SCs remaining on a Mib1-depleted niche (Figure 7D). In response to activation cues, SCs on a Mib1-depleted niche proliferate faster and upregulate Myogenin compared to the SCs on control muscle fibers, indicating that the remaining SCs are primed for proliferation and differentiation (Figure 7E–G). In response to muscle injury, the size of regenerating muscle fibers was smaller than the control muscle fibers, although the number of Pax7⁺ SCs is the same as control in the regenerated muscle (Figure 7H–J). Therefore, the effect of Mib1 deletion from the niche mimics that of Dll4 deletion. In conclusion, niche-derived Mib1 expressed in a heterogeneous pattern along the muscle fiber maintains the normal continuum of diverse QSC states during tissue homeostasis.

Mice were housed and maintained in accordance with the guidelines of the Laboratory Animal Research Center (LARC) of University of California, San Francisco. C57BL/6 were obtained from Jackson Laboratory. Previously published Pax7-nGFP (Rocheteau et al., 2012), human ACTA1-CreMer (McCarthy et al., 2012), Dll4^flox/flox (Hozumi et al., 2008), Mib1^flox/flox (Koo, 2007), Pax7-CreER (Nishijo et al., 2009), and Notch reporter CBF1-GFP (Duncan et al., 2005) were used in this study. All mice used for experiments were adults, between 12 and 16 weeks of age. The control and experimental mice used are littermates in all experiments. Approximately equal numbers of male and female mice were used in all experiments. Animals were genotyped by PCR using tail DNA. Primer sequences are available upon request.

Tamoxifen (tmx, Sigma) was dissolved in corn oil at a concentration of 20 mg/ml. Both control and experimental mice were administered tamoxifen at a concentration of 150 mg/kg/day for seven continuous days by intraperitoneal injection. The mice were left to chase for 30 days before analysis. To reduce (not ablate completely) the levels of Dll4 in the muscle fibers, tmx was IP injected at a concentration of 75 mg/kg/day for seven consecutive days.

Control and experimental mice were anesthetized by isofluorane inhalation and 50 µl of 1.2% BaCl₂ was injected into and along the length of the tibialis anterior (TA) muscle. After 14 and 40 days of regeneration, mice were euthanized, the contralateral uninjured TA and injured TA muscle were fixed immediately in 4% PFA(Paraformaldehyde) and frozen in 20% sucrose/OCT medium. 8-µm cross-sections of the muscle were made and stained for anti-laminin. ×10 images were collected at three regions in the mid-belly of each muscle. Only mice that had more than 80% injury in their TA were analyzed. All the regenerating fibers in the entire TA section were analyzed for fiber size. The average cross-sectional area of the fibers was determined using ImageJ software.

Single muscle fibers were isolated from the EDL muscle of the adult mouse as described previously (Eliazer et al., 2019). The single fibers were fixed immediately in 4% PFA for 10 min or cultured in plating media (DMEM(Delbecco's Modified Eagle Medium) with 10% horse serum).

For in vitro cell cycle entry assays, single muscle fibers from control and Myofiber-Dll4^fl/fl mice were harvested and cultured in plating media containing EdU (10 µm; Carbosynth) for 30 hr. For in vitro differentiation assay, the single muscle fibers from control, Myofiber-Dll4^fl/fl and Myofiber-Mib1^fl/fl were cultured in plating media for 48 hr. The fibers were then fixed and stained for different antibodies. EdU staining was done using the Click-iT Plus EdU Alexa Fluor 594 Imaging Kit (Invitrogen) followed by staining with anti-Pax7 antibody.

Satellite cells were isolated from hindlimb and forelimb muscles as previously described (Eliazer et al., 2019). The mononuclear muscle cells were stained for PE-Cy7 anti-mouse CD31 (clone 390; BD Biosciences), PE-Cy7 anti-mouse CD45 (clone 30-F11; BD Biosciences), APC-Cy7 anti-mouse Sca1 (clone D7; BD Biosciences), PE anti-mouse CD106/VCAM-1 (Invitrogen), and APC anti-α7 integrin (clone R2F2; AbLab). Fluorescence-assisted cell sorting (FACS) was performed using FACS Aria II (BD Biosciences) by gating for CD31⁻/CD45⁻/Sca1⁻/α7 integrin⁺/VCAM1⁺ to isolate SCs. SCs from Pax7-nGFP mouse were sorted for GFP fluorescence. The GFP⁺ gate was divided into top 15% (GFP^high fraction), middle 45% (Pax7^medium fraction), and bottom 15% (Pax7^low fraction). The isolated SCs were fixed immediately at t0 or cultured in growth media (Ham’s F10 media, 20% fetal bovine serum, 5 ng/ml FGF2) containing 10 μm EdU for 60 hr (cell cycle entry assay) or cultured in plating media (DMEM with 10% horse serum) for 3 days (differentiation assay).

Fixed myofibers were permeabilized with 0.2% Triton X-100/phosphate-buffered saline (PBS) and blocked with 10% goat serum/0.2% Triton X/PBS. Primary antibodies used in this study were: mouse anti-Pax7 (DSHB), rabbit anti-Laminin (Abcam), rabbit anti-Dll4 (Thermo Fisher Scientific), rabbit anti-Mib1 (Sigma), rabbit anti-Syndecan4 (Abcam), rabbit anti-DDX6 (Bethyl Laboratories Inc), rabbit anti-Myogenin (Santa Cruz Biotechnology), and DAPI(4′,6-diamidino-2-phenylindole) (Life Technologies). Primary antibodies were visualized with fluorochrome conjugated secondary antibodies (Invitrogen). The stained fibers were mounted in Fluoromount-G mounting medium (SouthernBiotech). For most of the staining, the images were taken using a ×20 Plan Fluor objective of the Nikon Eclipse Ti microscope. Anti-Pax7 and anti-Ddx6 stained images were obtained with a ×40 Plan Fluor objective of the same microscope. Dll4 and Mib1 stained images were taken with ×40 oil objective of Leica DMi8 Confocal Microscope. 15 μm z-stacks around the Pax7⁺ SC were taken and the sum projection of the images was obtained. The filter settings, gain and exposure values were kept constant between experiments. The intensity of expression is determined by manually drawing a region of Interest (ROI) on a Pax7-positive SC. This will give the mean pixel intensity of the ROI in all the channels. The ROI is copied onto another region where there is no Pax7-positive cell to calculate the background intensity. The background intensity is subtracted and the mean intensity is plotted with GraphPad Prism 7. To determine the intensity of Dll4 captured by SC, a ROI is drawn around the Pax7⁺ SC. To determine the intensity of Dll4 in the muscle fibers adjacent to the SCs, a ROI is drawn in an area of 100 μm² surrounding the SC. Representative images for antibody staining were taken using Leica DMi8 Confocal Microscope.

RNAscope for Dll4 was performed on fixed myofibers as previously described (Kann and Krauss, 2019). Briefly the single muscle fibers are isolated, fixed with 4% PFA and dehydrated with 100% methanol. The fibers are then rehydrated with a series of decreasing concentrations of methanol and PBS with 0.1% Tween 20 followed by protease digestion, hybridization of the RNA probe, amplification of signal and conjugation with fluorophore.

Total RNA was isolated from single muscle fibers (from EDL muscle), or SCs using Trizol (Invitrogen) according to the manufacturer’s protocol. The RNA was DNAse treated using Turbo DNA free kit (Life Technologies). cDNA was synthesized from RNA using the Superscript First Strand Synthesis System (Invitrogen). Quantitative PCR (qPCR) was performed in triplicates from 5 ng of RNA per reaction using Platinum SYBR Green qPCR Super Mix-UDG w/ROX (Invitrogen) on a ViiA7 qPCR detection system (Life Technologies). All reactions for RT-qPCR were performed using the following conditions: 50°C for 2 min, 95°C for 2 min, 40 cycles of a two-step reaction of denaturation at 95°C for 15 min, and annealing at 60°C for 30 s. The mean Ct values from triplicates were used in the comparative 2^−ΔΔCt method. To analyze the expression of Notch ligands on the adult muscle fiber, 2^−ΔCt method was used. The results were normalized to GAPDH mRNA controls. The primers used in this study are listed in Supplementary file 2.

Total RNA was isolated from single muscle fibers of the EDL muscle from postnatal day 3 (p3), postnatal day 7 (p7), and adult mouse hindlimb using TRIzol reagent (Invitrogen). The processing of RNA, hybridization, and analysis is previously described (Eliazer et al., 2019). The expression values for Notch ligands are log2 transformed and listed in Supplementary file 1.

The density maps for Pax7 and Ddx6 intensity were drawn using R. The variance of Pax7 and Ddx6 intensity levels across the SC pool is calculated as the square of standard deviation. The statistical details of experiments can be found in the figure legends. No statistical methods were used to predetermine sample size. The investigators were not blinded to allocation during experiments and outcome assessment. No animal was excluded from analysis. All data are represented as mean ± standard error of the mean. Significance was calculated using the two-tailed unpaired Student’s t-tests (GraphPad Prism 7). The number of replicates (n) for each experiment is indicated in the figure legends. Differences were considered statistically different at p < 0.05.