Muscle Satellite Cell Cross-Talk with a Vascular Niche maintains quiescence via VEGF and Notch Signaling

Summary

Skeletal muscle is a complex tissue containing tissue resident muscle stem cells (satellite cells, MuSCs) important for postnatal muscle growth and regeneration. Quantitative analysis, biological function, and the molecular pathways responsible for a potential juxtavascular niche for MuSCs is currently lacking. We utilized fluorescent reporter mice and muscle tissue clearing to investigate the proximity of MuSCs to capillaries in 3-dimensions. We show that MuSCs express abundant VEGFA, which recruits endothelial cells (ECs) in vitro, whereas both blocking VEGFA by a VEGF inhibitor and MuSC-specific VEGFA gene deletion reduce the proximity of MuSCs to capillaries. Importantly, this proximity to the blood vessels was associated with MuSC self-renewal in which EC-derived Notch ligand Dll4 induces quiescence in MuSCs. We hypothesize that MuSCs recruit capillary ECs via VEGFA, and in return ECs maintain MuSC quiescence though Dll4.

eTOC Blurb

Verma et.al performed skeletal muscle tissue clearing and unbiased fluorescent image analysis to show that muscle stem cells (satellite cells) pattern the microvasculature to be in close proximity to them via VEGFA. In turn, this juxavascular niche keeps the satellite cells in a more quiescent state suggesting a beneficial crosstalk

Introduction

Skeletal muscle is the most abundant tissue in the body and has remarkable regenerative capacity. It is composed of several different types of cells including multinucleated muscle fibers, muscle stem cells (satellite cells; MuSCs), endothelial cells (ECs), pericytes, mesenchymal progenitors/fibro-adipogenic progenitors (FAPs), Side-Population (SP) cells, immune cells, and undefined fibroblastic cells (Asakura et al., 2002; Bentzinger et al., 2013a; Chapman et al., 2017). These diverse cell types have a complicated interplay and anatomy that changes from the basal state during acute injury and chronic disease. Understanding the functional relationships between the different cell types within the tissue is important for modeling the dynamic changes during muscle development and regeneration. MuSCs are normally quiescent stem cells located underneath the basal lamina. During injury, these cells undergo activation, proliferate, differentiate, and either form new muscle fibers or fuse with an existing muscle fiber to repair the damaged muscle tissue. A small proportion of the activated MuSCs self-renew or escape activation which is essential in order to replenish the MuSC pool (Chakkalakal et al., 2012, 2014; Schultz, 1996; Shea et al., 2010). Defects in the self-renewal ability lead to a decreased MuSC number and reduced muscle regenerative capacity; this is exacerbated in aged and diseased muscle (Chakkalakal et al., 2012; Dumont et al., 2015; Rhoads et al., 2009; Urciuolo et al., 2013). Recent work has elucidated molecular mechanisms regulating MuSC self-renewal, which include Notch, Wnt, FGF, and extracellular matrix signaling pathways (Bentzinger et al., 2013b; Bjornson et al., 2012; Krauss et al., 2017). However, while it is clear that juxtacrine interactions and/or secreted factors are involved, it remains unclear which cell types in the MuSC niche regulate self-renewal and maintenance of MuSCs. Several studies have implicated the role of the ECs and pericytes in MuSC function (Abou-Khalil et al., 2009; Christov et al., 2007; Kostallari et al., 2015). We and others have shown that increasing vascular density can cause an increase in MuSC numbers in mdx mice (Matsakas et al., 2013; Verma et al., 2010). However, the exact relationship between MuSCs and ECs in vivo has yet to be examined.

In mammalian cells, Notch signaling which includes the transmembrane Notch ligands (Jag1, Jag2, Dll1, Dll3 and Dll4) and the Notch receptors (Notch1-4) (Chillakuri et al., 2012) has been implicated in stem cell maintenance in homeostasis and disease. Notch signaling has been functionally linked to myogenic precursor cells in embryonic, fetal, and adult skeletal muscle, and is a major regulator of MuSC function. Importantly, Notch signaling has been implicated in the maintenance of quiescence in MuSC during development and ageing, (Seandel et al., 2008; Rios et al., 2011; Bjornson et al., 2012; Philippos et al., 2012) as well as implicated in having a role in formation of the MuSC niche (Bröhl et al., 2012). While the role of Notch receptors has been examined in MuSCs, the demonstration that Notch ligands are required for niche formation is lacking. ECs have been demonstrated to express Notch ligands which in turn were shown to regulate hematopoietic stem cell (HSC) and neural stem cell fates (Butler et al., 2010; Hadland et al., 2015; Kobayashi et al., 2010; Morrison and Spradling, 2008; Ottone et al., 2014; Rafii et al., 2013).

In this report, we assess the potential interactions between ECs and MuSCs and showed a molecular mechanism that mediates the interaction between MuSCs and ECs, as well as demonstrate the functional consequences of this signaling. Understanding the interaction between MuSCs and ECs would expand our knowledge about interaction between stem/progenitor cells and the vasculature.

Results

Imaging reporters of MuSCs and ECs

Previous works demonstrated the importance of the interaction between MuSCs and vascular cells for muscle regeneration and maintenance (Christov et al., 2007; Kostallari et al., 2015; Verma et al., 2010). However, these studies analyzed this information in 2-dimensions (2-D) which may greatly underrepresent the interactions between the two cell types (Christov et al., 2007; Kostallari et al., 2015). To address this, we performed muscle tissue clearing and 3-dimensional (3-D) imaging (Verma et al., 2016) to look at the interactions between MuSCs and ECs. We first established a fluorescent genetic labeling strategy that was compatible with our tissue clearing protocol. We crossed the Flk1^+/GFP (Flk1^GFP) mice to label the ECs of the vasculature (Ema et al., 2006) and the Pax7^CreERT2:ROSO26^{Loxp-stop-Loxp-tdTomato} (Pax7^CreERT2:R26R^tdT or Pax7^tdT) mice to mark the MuSCs (Murphy et al., 2011), and validated that both labels were specifically expressed in the cells of interest after tamoxifen (TMX) injection (Figures 1A, S1A and S1B). These transgenic mice were useful for both FACS analyses (Figure 1A) and imaging (Figure 1B). As the number, size and shape of the cells are especially important for evaluating the distance between cell types in tissues, we first evaluated the morphological properties and MuSC number from volumetric imaging of MuSCs. MuSC number per volume was higher in soleus muscle and extraocular muscle (EOM) than tibialis anterior (TA) muscle (Figure S1C), and comparable to those previously obtained from FACS analysis (Verma et al., 2017).

Figure 1. Tissue clearing and genetic reporters for analysis of MuSCs and ECs in muscle.

(A) Genetic reporter mice for MuSCs (Pax7^tdT) crossed with genetic reporter for capillary ECs (Flk1^GFP) allow for simultaneous detection of both cells by FACS.

(B) Maximum intensity projection (MIP) of 50 μm thickness of skeletal muscle (Pax7^tdT:Flk1^GFP) shows various shapes of MuSCs (red) along with GFP⁺ blood vessels. Scale bar indicates 50 μm.

(C) Comparison of volume of MuSCs obtained using different methods show shrinking of cell volume in single muscle fiber preparation after collagenase digestion and fixed fiber bundles compared to cleared samples. Upper and middle panels show 2-D and 3-D views of MuSCs after different treatments, respectively. Histogram shows average cell volume of each MuSC population. Scale bars indicate 20 μm (upper panels) and 5 pμ (middle panels). Values are mean ± SEM. **** denotes p < 0.0001.

We compared the size/shape of Pax7^tdT mouse-derived MuSCs after isolated collagenase digested single muscle fibers, fixed physically isolated fiber bundles, and immediately fixed muscle followed by tissue clearing. Strikingly, the Pax7-tdTomato (Pax7tdT)⁺ signal from fixed muscle bundles as well as immediately fixed and cleared tissue contained the larger cytoplasm (1285 +/− 333.9 μm³ and 1497 +/− 405.2 μm³, respectively), including long projections (Figures 1B and 1C). By contrast, the Pax7tdT⁺ signal from MuSCs obtained from collagenase digestion resulted in a significantly smaller cell size (630.4 +/− 116.4 μm³) (Figure 1C), and the label largely encompassed the shape of the nuclei (Figure S1D). We found both mono- and multi-polar projections extending from the cell bodies (Figure 1B) such as those that have been previously described in electron microscopy studies performed in rat and shark MuSCs (Kryvi, 1975; Schmalbruch, 1978). Due to the presence of these large projections, we elected to perform imaging to assess potential cell-cell interactions using the Pax7^tdT reporter rather than a MuSC nuclear reporter (Kostallari et al., 2015).

Proximity between MuSCs and ECs

To measure the distance between MuSCs and blood vessels we performed tissue clearing and imaged muscles from the Flk1^GFP:Pax7^tdT mouse (Figures 1B, 2A, 2B and Supplemental Video 1). Following pre-processing and image segmentation, we performed unbiased distance measurements between ECs and MuSCs (Figures S2A-S2D). We found that approximately 40% of the MuSCs from EOM and TA muscle were directly in contact with the capillaries (Figures 2C and 2D). In addition, MuSCs from TA muscle showed a larger number of cells further away from the capillary. Interestingly, approximately 80% of the MuSCs from the soleus muscle were located directly in contact with capillaries. This distinction between the distribution did not correlate with the shape (ellipticity), size (volume) or density of MuSCs in the muscle groups (Figures S1E and S2F).

Figure 2. MuSCs are found in close proximity to EC microvasculatures.

(A) Snapshot of segmented MuSCs (red) and ECs (green) in the soleus muscle of Pax7^tdT:Flk1^GFP mice (See Supplemental Video 1).

(B) Statistically coded image of MuSCs based on distance away from the closest capillary. Subset images show a magnified view. The orange cells indicate MuSCs further away from a capillary, while the dark blue cells show MuSCs in direct contact with a capillary. Color from dark blue to red indicates short to long distance. Scale bars indicate 20 μm

(C) Distribution of distance away from capillary for MuSCs in the EOM, soleus and TA muscle. N > 4 for each tissue show in the distribution.

(D) Histogram shows the mean distance away from capillary for MuSCs in EOM, soleus and TA. The distribution is scaled to the number of cells obtained in the volume. Values are mean ± SEM. **** denotes p < 0.0001, and ns denotes no significance.

In silico candidate search for interaction between MuSCs and ECs reveals MuSC-derived VEGFA

Previously, we demonstrated that the interaction between MuSCs and ECs is important for muscle maintenance, in which increased vascular density induces increased MuSC number (Verma et al., 2010). Therefore, we utilized a system based approach (Butler et al., 2016) to identify specific gene expression patterns for the correlation between the MuSC numbers and EC numbers in 11 different muscle groups (Terry et al., 2017). The correlation between the EC-identifying gene (Pecam1) and MuSC-identifying genes (Pax7 and Myf5) was 0.68 and 0.60, respectively (Figure S3A). By contrast, macrophages and FAPs identifying genes showed no significant correlation with Pecam1. Taken together, these data clearly indicate possible biological interaction between MuSCs and ECs in muscle.

To screen for the possible interactions between MuSCs and ECs, we performed transcriptome analysis on freshly isolated MuSCs, ECs, and single muscle fibers by RNA-seq data (Figure S3B and Table S1). We performed large scale analysis of a curated dataset of 2,557 cell-cell interactions based on expression of 709 ligands and 691 respective receptors for a directional interactome analysis (Table S2; Ramilowski et al., 2015; Skelly et al., 2018; Zepp et al., 2017; Zhou et al., 2017). ECs and MuSCs shared 149 common ligands and receptors, and had 109 and 75 unique ligands and receptors, respectively (Figure S3C, Table S2). Interestingly, on a global scale, ECs and MuSCs exerted a higher degree, and distinctly different type of cell-cell signaling pathways compared to those from the single muscle fibers (Figures 3A and S3D). We performed Gene Ontology (GO) analysis on the top 100 interactions in the comparisons between MuSCs and ECs. Interestingly, the VEGF signaling pathway was the most enriched signaling pathway for this cluster (Figure 3B). Closer examination showed that the ligand (VEGFA) was highly expressed by MuSCs and single muscle fibers, and the canonical VEGF receptor genes (Kdr, Flt1, Nrp1, and Nrp2) were expressed on the ECs (Figures 3C, S3E and S3F, Tables S1 and S2). We also noticed expression of several putative VEGF interacting-protein genes, such as Egfr, Itgb1, Gpc2 and ItgaV in ECs (Table S1). Interestingly, neural cell-derived VEGF along with other factors have been implicated in the patterning and differentiation of blood vessels in mouse skin during development (Li et al., 2013; Mukouyama et al., 2002).

Figure 3. MuSC to EC signaling is influenced by MuSC-derived VEGFA expression.

(A) Total connectome diagram of strength of total ligand and receptor interactions in ECs, MuSCs and single muscle fibers indicates that ECs and MuSCs show large amounts of cross-talk compared to those with the single muscle fibers.

(B) GO term analysis of the top 100 sum specificity interactions from MuSCs to ECs enriched from the total number of ligand/receptor encoding genes. VEGF signaling pathway is highlighted as the top signaling pathway.

(C) Connectome diagram shows ligand expression from MuSCs and receptor expression from ECs. Map of 100 highest specificity sum interactions. Connectome diagram of MuSC- and single muscle fiber (SMF)-derived VEGFA interacts with EC-derived KDR,FLT1, NRP1 and NRP2. Size of the bars from the cell types to the genes indicate Transcripts Per Million (TPM). Lines between the receptors and ligands indicate known ligand-receptor interactions.

(D) Immunostaining for single muscle fibers shows high expression of VEGFA in the MuSC compared to the muscle fibers in both the quiescent MuSCs (QSCs) and the day 1 activated MuSCs (ASCs). Dotted lines indicate the muscle fiber outlines. Scale bar denotes 20 μm.

(E) Immunostaining in cultured ECs and myoblasts during proliferation and differentiation shows that VEGFA expression is low in ECs but high in desmin⁺ myoblasts in growth medium (GM) and myotubes in differentiation medium (DM). Scale bar denotes 20 μm.

(F) Example of intracellular VEGFA FACS in MuSCs, CD45⁺ cells, and ECs from extensor digitorum longus (EDL) and soleus muscles.

(G) Quantification of intracellular VEGFA using FACS analysis shows that MuSCs express high VEGFA levels compared to ECs and CD45⁺ cells. Specifically, MuSCs from soleus muscle express much higher levels of VEGFA than TA muscle, EDL muscle, diaphragms or EOM. Expression was shown as geometric mean fluorescent intensity (MFI). Values are mean ± SEM. ** denotes p < 0.01.p

(H) Dynamic changes in VEGF ligands and receptors in MuSCs and ECs following cardiotoxin (CTX)-injected muscle injury were measured by Microarrays, and showed that VEGFA is transiently increased in MuSCs following injury. Heat map scale is normalized to each gene.

For these reasons, we further explored the role of VEGFs in MuSCs. We confirmed the expression of VEGF family member proteins in MuSCs and ECs in quiescent, activated, and differentiated MuSCs by RNA-seq, qPCR, ELISA, and FACS. Using RNA-seq analysis, we found that VEGFA had the highest expression levels of the VEGF family ligands (Figure S3E). We were also able to confirm this using recently available data from in vivo MuSC nascent RNA and fixed MuSCs to exclude the possibility of VEGFA expression being a consequence of the isolation (Machado et al., 2017; van Velthoven et al., 2017). Additionally, we confirmed that VEGFA was actively transcribed rather than having stabilized the transcript, as its expression was reduced following global transcriptional suppression using α-Amanitin (Figure S3G). Furthermore, we confirmed the presence of VEGFA protein in MuSCs (Figures 3D and S4A) and MuSC-derived myoblast culture (Figure 3E). Quantitative ELISA showed that myoblasts secreted 266.2 ± 13.17 pg VEGFA per 10⁵ cells over 24 hours (Figure S4B). As VEGFA was found to be the predominant VEGF family member expressed in MuSCs identified by RNA-seq analysis, as well as showing the greatest qualitative difference in immunostaining between MuSCs and muscle fibers, we focused our investigation on VEGFA. We performed intracellular FACS on freshly isolated muscle cells, and found that MuSCs showed considerably higher VEGFA levels compared to ECs and CD45⁺ hematopoietic cells (Figures 3F and 3G). We assessed VEGFA in MuSCs, ECs, and CD45⁺ cells in various muscle tissues, and found that the MuSCs expressed much higher VEGFA levels than the other cell types, and MuSCs from the soleus had distinctly higher amounts of VEGFA compared to MuSCs from other muscles (Figures 3F and 3G). In addition, MuSC-derived VEGFA was maintained during the regeneration process (Figures 3H, S4C, S4D, Tables S1 and S6). We screened for expression of VEGFA in other muscle interstitial cells using publicly available datasets. VEGFA was found in comparable levels to MuSCs in freshly isolated FAPs but not in fibroblasts or Twist2⁺ myogenic progenitors (Figures S4E and S4F). Taken together, MuSCs express high VEGFA levels in skeletal muscle, which may have a profound effect on interactions between the position of MuSCs and ECs within muscles.

VEGFA regulated proximity of blood vessel to MuSCs

As VEGFA levels were distinctly higher in MuSCs from the soleus and the proximity of MuSCs to blood vessels was markedly closer in the soleus compared to other muscles, we hypothesized that high VEGFA levels in MuSCs were in part responsible for recruiting the blood vessel ECs to the location of MuSCs, resulting in the close proximity of MuSC to EC in the soleus. We performed a transwell assay with ECs on the top and myoblasts on the bottom, and showed that a myoblast-derived factor could increase the transmigration of ECs (Figure 4A). Importantly, sequestration of VEGFA using the soluble secreted portion of Flt1 (Flt1-FC) partially abrogated this transmigration. We also investigated the effects of the MuSC-specific VEGFA deletion mediated by Pax7^CreERT2 using the VEGFA^LoxP/LoxP:Pax7^tdT. We verified efficient VEGFA deletion (VEGFA^Δ/Δ) in myoblasts following Cre activation using qPCR and ELISA, compared to control myoblasts (VEGFA^+/+) (Figures S5A-S5C). We found that myoblasts lacking VEGFA displayed a deficiency in their ability to promote EC migration (Figure 4A), indicating that myoblast-derived VEGFA can induce the migration of ECs in vitro. We performed an in vivo gain of function experiment where we isolated myoblasts from VEGFA^Hyper donor mice which express VEGFA levels two-fold higher than that of wild-type (WT) control myoblasts throughout all cells and at all times (Figure S5D). We injected myoblasts isolated from VEGFA^+/+:Pax7^tdT or VEGFA^Hyper:Pax7^tdT(VEGFA^Hyper) mice into injured Flk1^GFP recipient TA muscle (Figure 4B). We performed TMX induction 13 days after the transplantation to encourage labeling of the MuSCs but not the muscle fibers, allowing us to analyze the effects of increased VEGFA expression from muscle fibers and MuSCs on the proximity of MuSC to EC (Figure S5E). We found that the average distance of VEGFA^Hyper donor myoblasts from capillaries was decreased compared in to the control WT donor myoblasts (Figure 4B).

Figure 4. VEGFA is highly expressed in MuSCs and mediates the proximity of ECs to MuSCs.

(A) Transwell assay shows myoblast-derived VEGFA induces chemotaxis of ECs. ECs are plated on the top and allowed to transmigrate to the bottom to measure chemotaxis. EC transmigration is assayed by following conditions to provide chemotaxis; no cells (control), myoblasts, myoblasts with Flt1-FC, and recombinant VEGFA (positive control). In addition, no cells (control), VEGFA^+/+ and VEGFA^Δ/Δ myoblasts showed that myoblast lacking VEGFA were deficient at transmigrating ECs.

(B) The proximity of MuSCs to ECs is increased in mice with MuSC-specific deletion of VEGFA (MuSC-VEGFA^a) compared to the control (MuSC-VEGFA^+/+ (Top) Experimental scheme. (Bottom) Distribution of distance between MuSCs and ECs in MuSC-VEGFA^+/+ control vs. MuSC-VEGFA^Δ/Δ muscle. Inset compares the average distance from nearest capillary between the two samples.

(C) VEGFA^Hyper donor myoblasts show lower average distance from capillaries than WT donor myoblasts after cell transplantation into BaCl₂-induced injured TA muscle. (Top) Experimental scheme. (Bottom) Distribution of distance from capillary for myoblast transplanted from VEGFA^Hyper:Pax7^tdT or VEGFA⁺⁺:Pax7^tdT mice into Flk1^GFP mice. Inset compares the average distance from nearest capillary between the two samples.

(D) MuSC numbers quantified in single muscle fibers at basal state and 21 days after during BaCl₂ induced muscle injury in MuSC-VEGFA^+/+ control vs. MuSC-VEGFA^Δ/Δ mice show no difference in the number of MuSC at baseline, but reduced MuSC number in the MuSC-VEGFA^Δ/Δ mice following muscle regeneration. For all figures, values are mean ± SEM. * denotes p<0.05, ** denotes p < 0.01, *** denotes p <0.001.

Next, we performed loss-of-function experiments and sequestered VEGFA using Flt1-FC during muscle regeneration, and assessed cellular proximity of the MuSCs and ECs following the unbiased distance measurements as described above. Blocking VEGFA led to an increase in the distribution of MuSCs away from blood vessels (Figure S5F). This was independent of any changes in the volume of total vasculature (Figure S5G). In addition, Flt1-FC treatment in the absence of the induction of regeneration did not result in a significant difference in the distance between the MuSCs and the ECs (data not shown), indicating that the microvasculature may be stabilized by the ECM or the structure of the muscle fiber under homeostatic conditions (Glancy et al., 2014). We deleted VEGFA in MuSC at 8 weeks of age in vivo and analyzed the muscle 21 days following VEGFA deletion. MuSC-specific deletion of VEGFA (MuSC-VEGFA^Δ/Δ) showed a stark difference in the average distance from the closest capillary, compared to the control (MuSC-VEGFA^+/+) (Figure 4C). Importantly, MuSC-VEGFA^Δ/Δ mice showed reduced MuSC number following one round of muscle injury, indicating that MuSC-derived VEGFA is essential for proper MuSC self-renewal in vivo (Figures 4D and S5H). Taken together, these data strongly indicate that MuSC-derived VEGFA is important for the patterning of the vasculature proximity of the MuSCs.

Proximity of MuSCs to ECs Regulated MuSC Quiescence

To investigate the biological result of MuSC proximity to blood vessels, we investigated whether MuSC quiescence was affected when these proximity distances were altered. Several lines of inquiry have shown that stemness and quiescence of HSCs and neural stem cells (NSCs) are mediated by their proximity to the vasculature (Kunisaki et al., 2013; Ottone et al., 2014). Truly quiescent stem cell populations can be marked because of their ability to retain the nucleotide analog EdU or genetic reporter H2B after a brief pulse during development, and are termed label retaining cells (LRCs) as these stem cells are EdU⁺ or H2B (Chakkalakal et al., 2014; Rocheteau et al., 2012). To detect the kinetics of EdU labeling in MuSCs, we noticed that one-time administration of EdU (100 ng/g body weight (BW)) into postnatal day 5 (p5) Pax7^tdT mice with a short pulse of 4-OHT (the active form of TMX) labeled 25.7% of the MuSCs as LRCs (Figures S6A-S6C). However, since the MuSCs were asynchronously cycling at this point, and because EdU has a shorter in vivo half-life (<1hr), we pulsed the mice 4 times over a 24 hrs period from p5-p6. Excessive EdU (100 ng/g BW 4-times/24 hrs) administration had a profound impact on the growth of the mice as indicated by low weight and hair loss (data not shown). However, (a dose of 50 ng/g BW 4-times/24 hrs) was tolerable and labeled 83.6% of the MuSCs at p7. There was a decline of EdU⁺ MuSC LRC over time as the cells proliferated and diluted the EdU to 1.5% by p60 (Figures 5A and 5B). The EdU⁺ MuSC LRCs trended forward an increase, but the difference was not statistically significant with imaging compared to FACS (Figure S6D). Thus, we assayed for the LRCs in our proximity in the 3-D analysis. There was a dramatic shift in the distribution of EdU⁺ LRCs, which were preferentially found close to the capillaries compared to the EdU⁻ non-LRCs (Figures 5C-5E), suggesting that the proximity of MuSCs to ECs is important for maintenance of MuSC quiescence.

Figure 5. Juxtavascular MuSCs are enriched for label-retaining stem cells (LRCs).

(A) FACS profile of the time course of EdU⁺ MuSCs.

(B) Quantification of the time course of EdU⁺ MuSCs show 83.7% of MuSCs are labeled initially, and are diluted by continued cell division to 1.56% by p60.p

(C) Experimental scheme for assaying for LRC MuSCs.

(D) Representative images of MuSCs (red) and ECs (green) with EdU labeled nuclei (purple). Big arrowhead denotes EdU⁺ MuSC, and small arrowhead denotes EdU⁻ MuSCs. Scale bar denotes 50 μm.

(E) Distribution of distance between MuSCs and ECs shows a strong bias for EdU⁺ MuSCs to be close to ECs compared with EdU⁻ MuSCs. Inset compares the average distance from nearest capillary between the two samples. Values are mean ± SEM. ** denotes p < 0.01.

Dll4-Notch signaling induced quiescence in MuSC

Since the distribution of the LRCs showed that the majority of the LRCs were in the juxtavascular niche rather than in an even distribution within the muscles, we concluded that ECs mediated MuSC quiescence likely through juxtacrine signaling which is a cell-cell interaction mediated in multicellular organisms (Yaron and Sprinzak, 2012). Juxtacrine signaling recently was shown to be an important mechanism for cell communication within skeletal muscle including ephrin, fibronectin and Notch (Bentzinger et al., 2013b; Gu et al., 2016; Vaz et al., 2012). As such, we performed an in silico directional interactome analysis between MuSCs and ECs, but with the ECs acting as the ligands and MuSCs acting as the receptors. GO-term analysis of the top 100 interactions from ECs to MuSCs showed enrichment of several biological functions (Figure 6A). Interestingly, we noticed that the “Notch signaling pathway” was the most enriched pathway. Connectome analysis within the top 10 specificity sum interactions from MuSCs and ECs showed EC-derived Dll4 and Jag2 interacting with MuSC-derived Notch3 and Notch2 (Figure 6B). We verified expression of the Notch and Notch ligands in our RNA-seq experiment (Figure S7A). We also found that several other ligands expressed in ECs that have been shown to be important for maintaining MuSC quiescence such as Tgfb1 (Quarta et al., 2016; Rathbone et al., 2011) and Adm (calcitonin receptor ligand) (Yamaguchi et al., 2015)(Tables S1 and S2). Microarray analysis also confirmed that Dll4 expression was significantly higher in the ECs from both intact and injured skeletal muscle compared to ECs from other tissues (Figures 6C, S7B, Tables S1 and S6). We verified the cell types that expressed Dll4 by single cell RNA-seq from whole skeletal muscle (The Tabula Muris Consortium et al., 2017). Using Principal component analysis (PCA) and tSNE based clustering, we were able to distinguish 8 separate cell clusters, and assign cell identities to MuSCs, ECs, smooth muscle cells, monocytes, B and T cells, FAPs, fibroblasts, and even remnants of muscle fibers, based on expression of prototypic genes (Figure S7C). In this dataset, Dll4 was dominantly expressed in the EC compartment in skeletal muscle (Figure 6D). To confirm the EC expression of Dll4, we performed FACS for single cell suspensions from hind limb muscle and immunostained muscle fiber bundles. FACS analysis confirmed that Dll4 was primarily expressed in the ECs (Figure 6E). We also found high reactivity to Dll4 antibody in ECs along with low reactivity in the muscle fibers (Figure S7D) as recently described (Low et al., 2018). VEGFA has been shown to induce Dll4 expression in a dose dependent fashion (Ubezio et al., 2016). We also found that EC expression of Dll4 was upregulated 2.3-fold in the VEGFA^Hyper muscle ECs, compared to littermate control ECs (Figure 6F), indicating that VEGFA is able to induce Dll4 in muscle ECs. However, reciprocally, EC expression of Dll4 in EC from mice lacking VEGFA in MuSC was unchanged (Figure S7E), potentially because only a small subset of ECs in the muscle would be in close proximity to MuSC and any difference in expression in these cells would be masked in the bulk analysis. Importantly, expression of several downstream genes (Hes1, Hey1 and HeyL) known to be up-regulated by Notch-signaling were down-regulated in MuSCs isolated from mice lacking VEGFA (MuSC-VEGFA^Δ/Δ), compared to the control mice (MuSC-VEGFA^+/+) (Figure 6G). By contrast, no significant change in the expression of these Notch downstream genes was seen in cultured myoblasts isolated from VEGFA^LoxP/LoxP mice, which were treated with 4-OHT in cultures (MB-VEGFA^Δ/Δ) compared to the untreated myoblasts (MB-VEGFA^+/+) (Figure 6H), suggesting the non-cell-autonomous Notch down-regulation in VEGFA^Δ/Δ MuSCs in vivo. However, this experiment cannot fully exclude a cell-autonomous effect of VEGFA on Notch signaling since the medium used for myoblast culture may mask the effects of loss of VEGFA in vitro. Taken together, we propose that loss of VEGFA in MuSCs results in reduction of Notch activity in MuSCs, potentially through reduction of the interaction via the Notch ligand Dll4 which was highly expressed in the ECs adjacent to the MuSCs.

Figure 6. Muscle EC-derived Dll4 interacts with MuSC-derived Notch receptor.

(A) GO term analysis of the top 100 sum specificity interactions from ECs to MuSCs enriched from the total number of ligand/receptor encoding genes. The Notch signaling pathway is highlighted as the top signaling pathway.

(B)Connectome diagram for EC-derived ligands interacting with MuSCs for the top 10 specificity sum interactions shows that Dll4 and Jag2 on ECs interacts with Notch2 and Notch3 on MuSCs.

(C) Notch receptors and ligands change dynamically during CTX-induced regeneration in both ECs and MuSCs. The predominant EC ligand Dll4 is high in the basal state and decreases during injury.

(D) Single cell RNA-seq data from whole muscle shows that Dll4 expression is restricted to the EC cluster, but absent in other cell types. Cell populations were classified based on the expression of their prototypic genes.

(E) FACS from dissociated muscle cells from Flk1^GFP mice shows Dll4 is predominantly expressed in Flk1GFP⁺ ECs.

(F) ECs isolated from muscles of VEGFA^Hyper mice show statistically significantly higher expression of Dll4. Values are mean ± SEM. * denotes p<0.05.

(G) MuSCs freshly isolated from muscles of MuSC-VEGFA^Δ/Δ mice show reduced expression of Notch-downstream genes (Hes1, Hey1 and HeyL).

(H) Myoblast culture (MB-VEGFA^Δ/Δ) shows no changes in expression of Notch-downstream genes (Hes1, Hey1 and HeyL).

EC-derived Dll4 increased reserve cell formation in MuSCs

To model the interaction between MuSCs and ECs, we performed direct co-culture between MuSCs and either the B.end3 EC line or 10T1/2 mesenchymal cell line in order to determine what conditions resulted in reserve cell formation. Reserve cells are a Pax7⁺MyoD⁻ MuSC-like population that arises during myoblast differentiation, and are the correlate of MuSCs returning to quiescence following the resolution of muscle injury (Low et al., 2018). We plated a monolayer of ECs or 10T1/2 cells, overlaid myoblasts on top of them, and switched to a low serum media to induce differentiation and reserve cell formation. Interestingly, co-culture with the ECs induced many more Pax7⁺MyoD⁻ reserve cells compared to co-culture with 10T1/2 cells (Figures 7A and 7B). We found that ECs expressed Dll1, Dll4, and Jag1 several log orders higher than the expression seen in 10T1/2 cells (Figure 7C). However, there was no difference in the amount of Jag2 expression levels between the two cell lines. These expressions patterns were confirmed by FACS using antibodies against Dll4 and Jag1 (Figure 7D). To investigate whether EC-mediated reserve cell induction was due to Notch signaling, and specifically due to Dll4 that is highly expressed in muscle ECs (Figures 6C-E and S7A-D), we performed the co-culture assay in the presence of the pan-Notch inhibitor, DAPT (Figure 7E). Inhibiting pan-Notch signaling resulted in a dramatic decrease in reserve cell formation in the co-culture. We also performed siRNA-mediated gene knockdown of Notch ligands in ECs in the co-culture (Figure S7F and S7G), and demonstrated a partial reserve cell formation defect with Dll4 knockdown but not Dll1, Jag1, or Jag2 knockdown (Figure 7E). Taken together, these data strongly suggest that EC-derived Dll4 induces MuSC self-renewal through the activation of Notch signaling as a juxtavascular niche.

Figure 7. EC-derived Dll4 supports reserve cell formation in vitro.

(A) Myoblasts were added on top of the EC or 10T1/2 cell monolayer, allowed to adhere, and the cultures switched to differentiation medium. Reserve cells were detected with immunostaining as Pax7⁺MyoD⁻ mononuclear cells (arrows). Scale bar denotes 20 μm.

(B) Reserve cell formation is higher in myoblast co-cultured on ECs than 10T1/2 cells.

(C) qPCR shows Dll1, Dll4, and Jag1 are highly expressed in ECs compared to 10T1/2 cells. Expression levels of Dll4 showed the greatest difference between ECs and 10T1/2 cells.

(D) FACS confirmation of high levels of Dll4 expression in ECs compared to 10T1/2 cells or myoblasts. Myoblasts express Jag1 but not Dll4. Notch 1 is expressed in all three cells types.

(E) Downregulation of Notch signaling through the pan-Notch inhibitor (DAPT) and EC derived Notch ligand Dll4, but not Dll1, Jag1 or Jag2, strongly reduced the myoblast-derived Pax7⁺MyoD⁻ reserve cell population in the co-culture of ECs and myoblasts.

For all figures, values are mean ± SEM. * denotes p<0.05, ** denotes p < 0.01, *** denotes p <0.001, **** denotes p<0.0001.

Discussion

Although studies have proven the essential roles by MuSCs in muscle regeneration, it still remains to be elucidated how MuSC self-renewal is molecularly regulated in their specific niche. Blood vessels serve as conduits for nutrition and oxygen, but also perform perfusion-independent functions as integral components of stem cell niches for many adult stem cells. Here we utilized fluorescent reporters for ECs (Flk1^GFP) and MuSCs (Pax7^tdT) along with muscle tissue clearing to investigate the proximity of MuSCs to capillaries in 3-D. Using unbiased quantification, we found that 50-80% of the MuSCs including EdU⁺ LRCs were located near or in direct contact with blood vessels, while EdU⁻ MuSCs showed no location bias. This phenomenon was missed in previous studies and demonstrates the importance of analyzing cell-cell interactions in 3-D (Christov et al., 2007). We showed that quiescent MuSCs expressed VEGFA, and further induced in activated MuSCs after acute injury. In addition, MuSC-derived VEGFA recruited ECs in vitro, and blocking global VEGFA levels as well as MuSC-specific deletion of VEGFA gene reduces the proximity between MuSC and blood vessels, down-regulates the Notch signaling pathway, and consequently decreases MuSC number. Dll4 is most strongly expressed in ECs in the absence of injury. Co-culture experiments demonstrated that ECs could promote MuSC selfrenewal through the Dll4-Notch pathway. Therefore, we concluded that during muscle regeneration, MuSCs recruited capillary ECs via VEGFA to establish a juxtavascular niche that is important for MuSC maintenance during homeostasis via the Dll4-Notch pathway. In addition, since Dll4 is also expressed in ECs after injury, Dll4 may also be important in MuSC selfrenewal.

Directional interactome analysis between MuSCs and ECs identified VEGF and Notch signaling pathways

To screen for possible interaction between MuSCs and ECs, we performed a directional interactome analysis for ligand/receptor genes, since the co-expression of these genes does not necessitate an interaction as there are other ligand-producing and receptor-accepting cells in the tissue. Thereby, an effective interactome analysis would also include sequence data from cell types such as FAPs, fibroblasts, and infiltrating cells. The directional interactome analysis between MuSCs and ECs with the MuSCs acting as the ligand and ECs acting as the receptor and the ECs acting as the ligand and MuSCs acting as the receptor identified the VEGF signaling pathway and Dll4-Notch signaling pathway, respectively.

Effect of VEGFA on microvasculature patterning

Skeletal muscle is the most abundant source of VEGFA in the body (Logsdon et al., 2014; Yen et al., 2011). Skeletal muscle-specific VEGFA knockout resulted in a downregulation of the VEGFRs and rarefication of the capillaries around the muscle fibers (Tang et al., 2004). However, there had been no studies that described on the effects of VEGF specifically in the MuSC compartment in vivo. Here we showed that MuSCs expressed high levels of VEGFA which played an essential role in chemotaxis of ECs specifically toward the MuSCs that produce it in vivo. Due to the relatively smaller size of the MuSCs compared to the total muscle mass, MuSCs would be expected to have a relatively low effect on the total VEGFA in the body. However, computation and experimental models have shown that VEGF gradients were more important for VEGFR activation than the total concentration (Gabhann et al., 2006; Gianni-Barrera et al., 2013; Logsdon et al., 2014; Springer et al., 2003). For example, a 4-8% change in VEGF over 10 μm would result in the activation of a 50 μm long endothelial tip cell. The effects of VEGF diffusion through interstitial space is considered to be minor (Gabhann et al., 2006; Yen et al., 2011). Taken together, these data strongly indicate that hot-spots with high levels of VEGFA found in MuSCs have a dramatic effect on aggregate angiogenic response and blood vessel patterning in whole skeletal muscle.

Juxtavascular stem cell niche for MuSC maintenance via Notch signaling pathway

Currently, it is still not clear which adjacent cell types express Notch ligands in order to activate Notch signaling pathway in MuSCs. Recombinant Dll4-FC has been shown to induce Notch response genes, specifically HeyL in myoblast culture and greatly diminished the expression of MyoD (Sakai et al., 2017). Downregulation of MyoD has been shown to enhance MuSC selfrenewal (Asakura et al., 2007; Bröhl et al., 2012). Recent reports have highlighted the requirement of Notch receptors in MuSCs in order for these cells to return to quiescence and form the reserve cell populations (Fujimaki et al., 2018; Low et al., 2018). For the adjacent cell source for the Notch activator, Low et al. described that mature muscle fibers-derived Dll4 activates Notch3 expression in MuSCs to allow return to quiescence. In addition, Bi et al. also found that muscle fibers-derived Dll4 and Jag2 modulate Notch signaling in the adjacent MuSCs to enhance their regenerative capacity (Bi et al., 2016). However, in our RNA-seq data and immunostaining experiments, we found that Dll4 and Jag2 were much lower in muscle fibers than in ECs, indicating that ECs play an important role as Notch ligand producing cells to promote MuSC self-renewal in skeletal muscle.

We demonstrated that EC-derived Dll4 was essential for Notch-signaling-mediated quiescence in MuSC culture, indicating the importance of the proximity of MuSCs to the blood vessels for MuSC self-renewal. Therefore, we present a model where paracrine VEGFA from MuSCs induces Dll4 expression in the ECs in direct contact with the MuSCs. In turn, juxtacrine presentation of Dll4 interacting with Notch receptors maintains the MuSCs in a quiescent state (Graphical Abstract). This may also affect the number of MuSCs in a tissue in homeostasis as we observed a strong correlation between ECs and MuSCs, and between different muscle groups. Previous work has shown that indirect in vitro co-cultured MuSC with EC increased MuSC proliferation through soluble factors (Christov et al., 2007). Our results, that direct co-culture between EC and MuSC regulates cell quiescence, highlight the importance of direct cell-cell contact both in vitro and in vivo. Similar EC mediated stem cell maintenance through the Notch signaling pathway has been reported in neural stem cells (Ottone et al., 2014). However, since quiescent MuSCs are located underneath the basal lamina on the muscle fibers, direct cell-cell contact between MuSCs and ECs may not be possible. Interestingly, recent work from electron microscopy and imaging cycler microscopy has revealed that the basal lamina contains fractures through which cells separated by the basal lamina may communicate to each other (Koulish, J.Morphol., 1971; Hillert et al., 2016), supporting the idea that direct cell-cell interaction with the adjacent ECs is the mechanism for MuSC maintenance.

MuSC and myoblast transplantation therapy has been shown to be effective for mitigating disease and aging related muscle pathology in mouse relevant models (Cerletti et al., 2008; Hall et al., 2010). Freshly isolated MuSCs possessed robust transplantation efficiency which is rapidly lost in myoblasts after in vitro culture (Asakura et al., 2007). However, it is impossible to obtain enough cells from biopsies for transplantation, thereby requiring in vitro myoblast expansion. The ability to expand MuSCs and revert them to a transplantable state could have therapeutic potential. Inducing Notch signaling along with the other signaling pathways described above through the culture of MuSC with EC may allow for the expansion of quiescent MuSCs with more robust transplantation efficiency. The strategy of co-culturing stem cell with ECs has been utilized in previous studies to propagate HSCs prior to transplantation (Butler et al., 2010; Hadland et al., 2015).

We propose that the proximity between MuSCs and capillaries is an actively mediated process where MuSC-derived VEGFA recruits capillaries to establish a juxavascular niche for MuSCs, and that self-renewal and maintenance of quiescence of MuSCs are induced by the nearest capillaries via the Dll4-Notch pathway.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Atsushi Asakura (asakura@umn.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

B6.Cg-Pax7^{tm1(cre/ERT2)Gaka}/J (Pax7^CreERT2) mice were crossed with the B6.Cg-Gt(ROSA)^{26Sortm9(CAG-tdTomato)Hze}/J (Ai9) to yield the Pax7^CreERT2:R26R^tdT (Pax7^tdT) mice. Pax7^CreERT2, Ai9 and Vegfa^tm1.1Nagy/J (VEGFA^Hyper) mice were obtained from the Jackson Laboratory. VEGFA^LoxP/LoxP mice were obtained from Napoleone Ferrara (Gerber et al., 1999). Kdr^tm2.1Jrt/J (Flk1^GEP) were obtained from Masatsugu Ema (Ema et al., 2006). Genotyping to detect the transgenic and mutant alleles was performed by PCR using the primers described on the web site of Jackson Laboratory shown in Table S3. All primers were synthesized as custom DNA oligos from Integrated DNA technologies (IDT). The animals were housed in an SPF environment and were monitored by the Research Animal Resources (RAR) of the University of Minnesota. All protocols were approved by the Institutional Animal Care and Usage Committee (IACUC) of the University of Minnesota and complied with the NIH guidelines for the use of animals in research.

Cell isolation and culture

MuSC-derived primary myoblast isolation was performed as previously described (Asakura et al., 2011; Motohashi et al., 2014). Myoblasts were maintained in culture on collagen coated plates in myoblast growth medium (GM) containing 20% FBS and 20 ng/ml bFGF (ThermoFisher Scientific, PHG0263) in Ham’s-F10 medium. 4-OHT treatment (1 μM in EtOH) was used to induce VEGFA deletion (VEGFA^Δ/Δ) in myoblasts isolated from VEGFA^LoxP/LoxP:Pax7^tdT mice. The same myoblasts (VEGFA^+/+) were used but treated with vehicle (EtOH) only for the control experiments. Myoblast differentiation and reserve cell formation were performed by switching the medium to differentiation medium (DM, 5% horse serum in DMEM). Myogenic cells were defined as positive for anti-desmin (NeoMarkers, RB-9014). Reserve cells were defined as positive for anti-Pax7 (DSHB AB_528428) and negative for anti-MyoD (Santa Cruz sc-304) by 5 days after differentiation. The B.end3 EC line (ATCC, CRL-2299) and the 10T1/2 mesenchymal cell line (ATCC, CCL-226) were maintained in 10% FBS in DMEM. Cells were fixed with 2% PFA in PBS, permeabilized with 0.2% Triton X-100 in PBS, blocked with 1% BSA in PBS, and incubated with primary antibodies followed by secondary antibodies. Nuclei were counterstained with DAPI.

Co-culture

Co-culture was performed by plating a monolayer of ECs or 10T1/2 cells, overlaid myoblasts on top of them, and switched to a low serum media to induce differentiation and reserve cell formation (Figure 7). Dll1, Dll4, Jag1 and Jag2 siRNAs (Santa Cruz Biotechnology, sc-39668, sc-37201, sc-37203 and sc-39673) and control scramble siRNA-A (Santa Cruz Biotechnology, sc-37007) were transfected in ECs by Lipofectamine 2000 (ThermoFisher Scientific, 11668019) before co-culture following the manufacturer’s instruction (Figures 7 and S7). A γ-secretase inhibitor, DAPT (N-[N-(3,5-difluorophenacetyl)- L-alanyl]-S-phenylglycine t-butyl ester, 10 μM, Sigma-Aldrich, D5942), was used to block Notch signaling in the co-culture experiments.

Single muscle fiber isolation

Extensor digitorum longus (EDL) muscles were digested with 0.2% collagenase type I digestion (Sigma-Aldrich, C0130) of EDL muscles for single muscle fiber isolation as previously described (Asakura et al., 2002). Isolated single muscle fibers were transferred and cultured on 5% horse serum-coated tissue culture dishes with DMEM supplemented with 10% horse serum and 1% chicken embryo extract (MP Biomedicals, 092850145) for 24 hrs. Single muscle fibers were fixed with 2% PFA in PBS, permeabilized with 0.2% Triton X-100 in PBS, blocked with 10% BSA in PBS, and incubated with primary antibodies followed by secondary antibodies. Nuclei were counterstained with DAPI.

METHOD DETAILS

Histological characterization of mice

Due to challenging and unreliable labeling reagents such as antibodies for populations such as MuSCs in whole muscle, genetic reporter mouse lines encoding promoter driven fluorescent molecules (Bosnakovski et al., 2008; Keefe et al., 2015; Relaix et al., 2005; Sambasivan et al., 2009) have been developed, that allow for reliable and consistent assessment of cellular populations with high fidelity. However, current reporters for MuSCs have several limitations associated with them. First, they all use fluorescent proteins in the green spectrum. Due to the abundance of auto-florescence in that spectrum in muscle, it is challenging to obtain good signal to noise levels needed for isolation of these cells and can also result in false signal (Jackson et al., 2004; Verma et al., 2016). Second, many of reporters are localized to the nucleus which can underestimate the size of an individual cell. Therefore, for appropriate labeling of MuSCs, a cytoplasmic reporter in the red-spectrum is recommended.

To label the MuSCs, we utilized the Pax7^CreERT2 (Murphy et al., 2011) and crossed it with the Ai9 mouse reporter line (R26R^tdT) (Madisen et al., 2010), which is the brightest fluorescent reporter currently available, to yield Pax7^CreERT2:R26R^tdT (Pax7^tdT) mice on. We performed extensive validation of the fluorescent reporter mice (Figure S1). To mark the microvasculature ECs, we utilized the Flk1^GFP transgenic reporter, intravital as well as whole-mount anti-Pecam1 (CD31, BD Biosciences 551262), and tomato-lectin (Vector labs DL-1178) staining. While intravital Pecam1 and lectin staining are commonly used in the brain vasculatures, we found that them to be heterogeneous in their signal intensity in the muscle (Figures S1E and S1F), perhaps due to differences in active perfusion in different capillary beds in the muscle (Poole et al., 2013). In addition, large vessels were labeled much more strongly than the microvasculature which was detrimental to the downstream segmentation process needed for determination of single MuSCs and their relationship to the capillaries (Figure S1E). This was expected to be an issue with perfusion-based vasculature staining as well (Tsai et al., 2009). Whole mount staining with lectin showed significant staining in the microvasculatures and some muscle fibers in regenerating muscle (Figure S1F). By contrast, the Flk1^GFP mouse had homogenous labeling of the microvasculature, showed excellent co-expression with the EC markers Pecam1, Cdh5 and tomato-lectin as well as variable expression of Sca1 (Figure S1B). The GFP⁺ cells had no coexpression with leukocyte or pericyte markers such as CD45 (eBioscienc 13-0451-85), Gr1 (eBioscience 13-5931-85), F4/80 (Invitrogen MF48000), Cxcr4 (BD Biosciences 551968), PDGFRα (BD Biosciences 562777), and PDGFRβ (eBioscience 13-1402-82). Therefore, to allow for robust labeling of the MuSCs and microvasculature, we crossed the Flk1^GFP with the Pax7^tdT to yield the Pax7^tdT:Flk1^GFP mice which were utilized for imaging as well as FACS experiments. For tissue sections, 8 μm thick transverse cryosections were used for all histological analysis. Information of all antibodies and intravital dyes used in this paper are shown in Key Resources Table.

Injection of tamoxifen (TMX) or Flt1-FC

Cre inducible mice were pulsed by IP injection of 75 mg/kg body weight (BW) tamoxifen (TMX, Sigma-Aldrich T5648) three times over 5 days followed by a 5-day chase period. Neonatal mice were pulsed with IP injection of 80 mg/kg BW of 4-Hydroxytamoxifen (4-OHT, Sigma-Aldrich, H6278), the active form of TMX followed by a 48 hrs chase (Ganat et al., 2006). These protocols were shown to label 91.71% and 71.25% of MuSCs, respectively (Figure S6A). For MuSC-specific VEGFA gene knockout mouse experiments, TMX was injected into VEGFA^LoxP/LoxP:Pax7^tdT:Flk1^GFP mice or VEGFA^+/+:Pax7^tdT:Flk1^GFP mice to generate mutant (VEGFA^Δ/Δ) or control (VEGFA^+/+) mice, respectively. All experiments were performed on male and female mice aged 1.5-6-months old, unless specified. For the in vivo blocking of VEGFA, the skeletal muscles of both hind limbs were injected with 20 μl of 3 μM cardiotoxin (CTX; Sigma-Aldrich, 217503) to induce muscle injury followed by regeneration. The right tibialis anterior (TA) muscle was injected with 100 ng of Flt1-FC (20 μl of 5 ng/μl, R&D Systems, 321-FL-050) on alternative days for 14 days (Figures 4B and S5A). TMX was injected starting day 14, and the TA muscle was isolated at day 19.

Label retaining cells (LRCs)

To measure label retaining cells (LRCs), neonatal mice were pulsed with 50 μg/g BW of EdU 4 times over the span of 24 hrs (0 hrs, 6 hrs, 10 hrs and 18 hrs) to capture all possible dividing cells regardless of the time in S-phase, as the half-life of EdU is less than one hour (Cheraghali et al., 1995). Perinatal mice were pulsed with 4-OHT 2 days prior to sacrifice to label the MuSCs. Muscles from neonatal mice were obtained for tissue clearing or FACS (Figures 5 and S6). For tissue clearing, we found that glutaraldehyde fixation was incompatible with the Click-iT EdU assay kit and was omitted for all experiments requiring EdU labeling. The EdU was detected using click chemistry using the Click-iT EdU Alexa Fluor 647 Imaging kit (Baseclick, BCK-EdU647) with few modifications to the recommended protocol. Briefly, samples were incubated in the antibody incubation buffer for 2 hrs prior to the Click reaction. In addition, the Click reaction was supplemented with 1 mM SDS (0.29%) while keeping the concentration of the active reagents the same. The incubation was performed in a bacterial shaker at 37°C. Lastly the unbound Alexa-647 was quenched with two washes with the antibody incubation buffer with 10% BSA in 0.5% Triton in PBS for 1 hr at 37°C in a bacterial shaker. For FACS, dissociated single cells were fixed and permeabilized using the recommended protocol below for intracellular FACS. The EdU was detected using Click chemistry according to the manufacturers instruction.

FACS analysis

Conventional and intracellular FACS analysis was performed as previously described (Turaç et al., 2013). Cells were either trypsinized (cultured cells) or a single cell suspension was obtained following enzymatic digestion (whole tissue-derived cells). Cells were washed with FACS buffer (2% BSA and 1 mM EDTA in PBS) followed by live/dead staining using ZombieNIR (Biolegend, 423105). Cells were washed, then either immunostained for cell surface markers or fixed in 2% paraformaldehyde (PFA, Sigma-Aldrich, P6148) for intracellular FACS followed by permeabilized with 0.5% Tween-20. Blocking was performed with 1% BSA in PBS, and cells were incubated in primary antibodies followed by fluorescently-conjugated antibodies shown in Table S4. EdU Click-iT assay was performed as specified by the manufacturer (Baseclick, BCK-EdU647). FACS was performed on an LSRFortessa X-20 (BD Biosciences) with a 355 nm, 405 nm, 488 nm, 561 nm, and 640 nm lasers.

Cell chemotaxis assay

The transwell assay was performed using 8 μm transwell inserts (Corning, 3422) (Figure 4A). Myoblasts were plated and allowed to adhere to the culture well overnight. ECs were plated on top of the transwell insert, and cells were allowed to migrate for 4 hrs., then 2% FBS was used in both the top and bottom chambers to minimize the influence of the serum on chemotaxis. 50 ng/ml of Flt1-FC (R&D Systems, 321-FL-050) was used to sequester VEGF signaling. Recombinant VEGFA (50 ng/ml) was used as a positive control (R&D Systems, 493-MV-025) for chemotaxis. Cells were allowed to migrate for 3 hrs before fixation. The top of the transwell was cleared of cells using a cotton swab. transwell membranes were stained using 0.5% Crystal Violet solution (Sigma-Aldrich, HT90132) in 25% methanol and washed several times. The dye was solubilized from the cells using 1% SDS, and the relative absorbance was quantified using spectroscopic reading at 570 nm.

Cell transplantation

For myoblast transplantation studies, MuSCs were isolated from VEGFA^+/+:Pax7^tdT or VEGFA^Hyper:Pax7^tdT mice without TMX injection as described above. After cultures, 5 × 10⁵ expanded MuSC-derived myoblasts were injected into regenerating TA muscles of Flk1^GFP mice 2 days following 20 μl of 1% BaCl₂ (Sigma-Aldrich 342920) injection-induced injury (Figures 4D and S5F). Transplanted cells were Pax7tdT-labeled by IP injection of TMX starting on day 13 post-transplantation, and the muscles were harvested, cleared, and labeled cells were identified on day 21 post-transplantation. The Pax7tdT⁺ MuSC’s distance to the closest capillary analyzed as described in the preceding section.

Enzyme-linked immunosorbent assay (ELISA)

MB medium from each condition was sampled from triplicated MB cultures. VEGF protein levels in MB medium during myoblast differentiation or after treatment with or without 4-OHT were measured using Mouse VEGF Duoset® ELISA Kits (R&D Systems, DY493) according to manufacturer recommendations (Figures S4B and S5D).

RNA and genomic DNA isolation and qPCR

Approximately 100,000 FACS-sorted cells were pelleted and resuspended in TRIzol™ reagent (ThermoFisher Scientific, 15596026) for RNA isolation. For single muscle fibers, 100 collagenase-digested fibers were utilized. Cultured cells were washed with ice cold PBS and lysed on the place with TRIzol™. RNA was isolated using the Direct-zol™ RNA Microprep Kit (Zymo Research, R2062) with on-column DNase digestion followed by cDNA synthesis using the Transcriptor First Strand cDNA synthesis kit (Roche Molecular Diagnostics, 04379012001) using random primers. Genomic DNA was isolated from mouse tail snips and cultured cells with lysis buffer containing Protenase K (Sigma-Aldrich, P2308). qPCR was performed using GoTaq® qPCR Master Mix (Promega, A6001). Primer sequences are listed in Table S3. All primers were synthesized as custom DNA oligos from Integrated DNA technologies (IDT).

RNA sequencing (RNA-seq)

RNA-seq was performed on FACS-sorted MuSCs and ECs from the Pax7^tdT:Flk1^GFP mouse hind limb muscles as well as 100 single muscle fibers obtained from EDL muscle as described above. Cells were pooled from three mice to obtain enough RNA for library preparation. Total RNA was quantified using a fluorimetric Quant-iT RiboGreen RNA Assay Kit (ThermoFisher Scientific, R11490). Total RNA integrity was assessed using capillary electrophoresis (Agilent BioAnalyzer 2100) and RNA Integrity Number (RIN) of 7-10 were used. Library preparation was performed on total RNA SMARTer Stranded Total RNA-Seq-Pico Mammalian Kit (Clontech, 635007). RNA was reverse transcribed into cDNA using random primers. The Template Switching Oligo (TSO) was incorporated during cDNA synthesis and allowed for full length cDNA synthesis and strand specificity to be retained. Illumina sequencing adapters and barcodes were then added to the cDNA via limited PCR amplification. Next, mammalian ribosomal cDNA was enzymatically cleaved, and uncleaved fragments were PCR-enriched for 12-16 cycles. Indexed libraries were then normalized and pooled for sequencing using the HiSeq2500 (Illumina) with 50 bp paired ends run with 20 millions expected reads per sample.Sequences trimmed using trimmomatic to remove adapter contamination and low-quality reads. Trimmed sequences were mapped to mouse mm10 using Hisat2 (Pertea et al., 2016). Transcript assembly was performed using StringTie (Pertea et al., 2016). Gene level quantification, and analysis was performed using Ballgown or DeSeq2 (Love et al., 2014; Pertea et al., 2016) in RStudio (RStudio Team, 2015). Additional NGS data was obtained from GEO and subjected to the above pipeline and are listed in Table S5.

Bioinformatics for cell-cell interactions

To perform bioinformatics assessment of cell-cell interactions, RNA-seq was performed on MuSCs, ECs and single muscle fibers (Figures 3A-3C, 6A, 6B, S3 and S7A) and the expression data was listed in Table S1 (RNA-seq for ECs, MuSCs and SMFs, related to Figure 3). Genes that were highly expressed (> 5 transcript per million, TPM) were considered to be expressed with high confidence. The cutoff was estimated through literature values by assessing what genes were present and expressed in prior verified research. The distribution of expression was similar between the total transcriptome and the ligands or receptors, especially at a threshold of 5 TPM, suggesting there was no loss of specific compartments of genes in the analysis (Data now shown). Ligand-Receptor interactomes were mapped using the first draft of the human Ligand-Receptor interaction map by the FANTOM5 consortium(Zepp et al., 2017; Zhou et al., 2017). All the mouse gene names were successfully converted to the human orthologue using the ENSEMBLE mart through biomaRt (Smedley et al., 2015). Ligand-Receptor genes from the curated list were pulled out from the transcriptome, averaged and displayed in the Ligand-Receptor Connectome Visualization applet (Ramilowski et al., 2015). The code for the gene name conversion and formatting the input for the applet can be found at (https://github.com/verma014/Cell-Cell-interaction). GO term analysis was performed on the top 100 of the directional sum specificity interactions using the total list of Ligand-Receptor genes as the background using Gorilla (Eden et al., 2009). All GEOs and repositories used in this paper are shown in Key Resources Table.

Microarray analysis

Microarray analysis was performed using the Affymetrix Transcriptome Analysis Console (TAC) (Figures 3H, 6C and S7B). Gene expressions of ECs and MuSCs FACS-sorted from intact and CTX-injected regenerating muscle were robust multi-array average (RMA)-normalized, and the expression relative levels were acquired using the Affiymetrix Expression analysis console with gene level expression. Heat maps were generated in the heat map module in R.

Single cell RNA-seq analysis

Single cell RNA-seq analysis was performed on data obtained by the Tabula Muris consortium (The Tabula Muris Consortium et al., 2017). The microfluidics droplet data was obtained from DOI (10.6084/m9.figshare.5715025). Single cell analysis was performed using Seurat 2.2 (Butler et al., 2018) in the R-Studio environment. The data was filtered according for low nUMI expression and log-normalized and scaled to UMI counts. Principle component analysis (PCA) based clustering was performed and each population was classified based on expression of representative genes including those in Figures 6D and S7C. Fibroblast and FAP clusters were defined by gene signature from bulk sequencing of fibroblast and FAP (Chapman et al., 2017). The respective code can be found at (https://github.com/verma014/Single-cell-sequencing-results-/blob/master/MuscleSingleCell)

Correlation of genes to different muscles

Correlation between genes representing MuSCs, ECs, macrophages, and fibro-adipogenic progenitors (FAPs) from different groups of muscle was calculated (Figure S3A). ECs were represented by Pecaml and Cdh5. MuSCs were identified by Pax7 and Myf5. Macrophages were identified by the F4/80 encoding gene Emarl. FAPs were represented by Pdgfra. Gene expression in the form of FPKM from the average of 6 muscles per group from extraocular muscle (EOM), masseter, tongue, diaphragm, quadriceps, TA, EDL, soleus, gastrocnemius, plantaris and flexor digitorum brevis (FDB) muscles were obtained from an available dataset (Terry et al., 2018). Pearson’s correlation was calculated for each gene combination and the figures were produced using R. The code can be found at (https://github.com/verma014/Cell-Cell-interaction/blob/master/correlationbetweendifferentmuscle).

Tissue clearing

Tissue clearing was performed as previously described with some modifications (Verma et al., 2016). Mice were perfused with 60 ml 4% PFA in PBS. Tissue was excised and post-fixed overnight at 4°C. 0.05% glutaraldehyde (Sigma-Aldrich, G5882) was used when no additional labeling was required. The tissue was washed extensively in PBS prior to clearing. Quenching was performed in 4% (w/v) glycine (Sigma-Aldrich, G7126) and 4% acetamide (w/v) (Sigma-Aldrich, A0500) in PBS at 37°C overnight. Samples were incubated in 4% acrylamide solution (Sigma-Aldrich, A9099) with the thermal initiator VA-044 at 0.25% (Wako Chemical, NC0632395) for 4 hrs at 4°C. Fresh and degassed acrylamide solution was replaced and polymerized for 3 hrs at 37°C. The polymerized samples were then incubated for 6 hrs in 200 mM SDS solution (Sigma-Aldrich, L6026) buffered with 20 mM boric acid (pH 8.4, Sigma-Aldrich, B6768). Next, the samples were incubated for 24 hrs at 37°C in a solution containing 5% N,N,N’,N’-tetrakis ethylenadiamine (Sigma-Aldrich, H2383), 10 mM SDS, 2% Triton X-100 (Sigma-Aldrich, T8787), and 20 mM boric acid buffer (pH 8.4). The cleared tissue was extensively washed using PBS containing 0.5% Triton X-100. The processed tissue refractory index was matched using 88% Histodenz solution (Sigma-Aldrich, D2158) or PROTOS solution (23.5% (w/v) n-methyl-d-glucamine (Sigma-Aldrich, M2004) + 29.4% (w/v) diatrizoic acid (Sigma-Aldrich, D9268) + 32.4% (w/v) iodixanol (Sigma-Aldrich, D2158, same as Histodenz or Omnipaque) + 0.01% NaN3 (Sigma-Aldrich, S2002) in H₂0). The PROTOS solution penetrated through a tissue faster than Refractive index matching solution (RIMS), and therefore requires shorter incubation time, but was prone to evaporation and was thus harder to handle, with the added concern that changes may occur in the tissues’ refractive index.

Mounting

To facilitate mounting of the muscle samples, we designed 3-dimensional (3-D) printed sample holders of different sizes, indicated by the number of grooves on the side (1 mm per groove) that could both house the samples and be stored in a standard slide holder. To use the sample holder, the bottom of the sample holder was mounted with a 22 × 22 mm glass coverslip and sealed using clear nail polish or VALAP, a mixture of vaseline, lanolin and paraffin wax. The samples were mounted in 0.5%-1.0% low melting agarose (ThermFisher Scientific, 16520050) in PROTOS (A-PROTOS) to stabilize the samples while imaging, as samples could drift in the solution, especially during long image acquisitions. A-PROTOS was made by putting agarose powder in a sealed glass container and sonicated with frequent vortexing at 55°C. Samples were then exchanged twice with the A-PROTOS solution at 37°C twice before being poured into the sample holder. We then slowly applied a 24 × 60 coverslip over the top, while avoiding air bubbles. The coverslip on both end allowed for imaging through both slides as well as second harmonics generation in the transmitted direction. If the A-PROTOS gelatinized before use, it could be rewarmed at 55°C. However, a yellow-brown hue could develop with excess or prolonged heat.

For certain experiments, 1 μm fluorescent beads were added to the agarose at a concentration of 1:20,000 (of the 1×10¹⁰ beds/ml provided by the manufacturer, ThermoFisher Scientific, R0100B). The beads could assist with troubleshooting, deconvolution, tiling reconstruction, or Multiview reconstruction. Alternatively, the tissue could be embedded in agarose made in PBS and then the gelled samples can be trimmed and then taken through 3 exchanges using PROTOS. The A-PROTOS did not show any noticeable scattering over a 2 mm-depth (data not shown), the working distance of the lens used.

Imaging

Imaging was performed on multiple types of microscopes based on their requirements at the University of Minnesota Imaging Center (UIC). Single muscle fibers, tissue sections and cultured cells were imaged using a Nikon NiE C2 upright confocal imaging system using a 20x or 40x lens (Nikon). Tissue imaging was obtained in a Nikon FN1 upright stand equipped with an A1R laser scanning head (Nikon). A Plan Apo LWD 25x water-immersion NA 1.1 (WD 2mm) objective 10x Glycerol-immersion NA 0.5 (WD 5mm) with a correction collar were used and a motorized Prior stage and piezo Z drive were used to control sample positioning and focus. The system is equipped with 405, 488, 457, 514, 561 and 640 nm laser lines. Emission bands of 425-475 nm (blue), 500-550 nm (green), 575-625 nm (orange) and 650-720 nm (red) were used in confocal mode.

A Mai Tai Dep See pulsed laser tuned to 870 nm was used for two-photon excitation for simultaneous excitation of GFP and tdTomato and fluorescence emission was collected with GaAsP non-descanned detectors. The emission bands acquired were 400-450 nm (blue), 470-550 nm (green) and 570-640 nm (orange-red). The microscope was controlled with NIS Elements 4.6 software. Channel subtraction was performed when the two-photon mode was used. Far-red fluorophore was detected in confocal mode using a 637 nm laser. Emission was detected by non-descanned detectors with a galvonomic scanner and enhanced hybrid PMTs. A linear increase in laser intensity with used for deeper imaging. A Plan Apo 25 x water-immersion, NA 1.1 (WD 2mm) or 10x Glycerol-immersion 0.5 NA (WD 5mm) with a correction collar were utilized for deep tissue imaging.

Image processing

Image processing was performed in Nikon elements (Nikon) and FIJI (NIH) (Schindelin et al., 2012). Intensity equalization was performed to normalize intensity throughout the depth. Denoising was performed on the resulting images using Nikon Elements (Figure S2A). The images were pixel-classified using machine learning through Ilastik (Sommer et al., 2011) as it showed considerably better performance for segmenting the processes in MuSCs compared to surface detection using Imaris (Figure S2B). Files were converted to Ilastik h5 format using the Fiji Ilastik plugin. A subset of files was cropped to 50 μm × 50 μm × 50 μm and used as training input. Training was performed to classify the GFP⁺ ECs and the Pax7tdT⁺ MuSCs and neither as the background using the classification uncertainty as the training guide. All features in Color/Intensity, Edge and Texture from 0.3-10 pixels were selected for training. Pixel classification was performed and the simple segmentation was exported. The raw image along with the simple Segmentation was merged and exported to Tiff using Fiji Ilastik.

The image file was converted to Imaris file format using the Imaris file format converter. Images were opened in Imaris 9, and the simple segmentation layers were used to obtain surface objects as MuSCs and ECs. The raw image was used to verify the segmentation. Distance measurements were performed in an unbiased manner (Figures S2C and S2D). Briefly, a distance matrix was calculated from the segmented capillaries so that the further a point was from a capillary, the higher the assigned pixel value. The calculated distance transform matrix value was then measured at either the minimum point of contact (Figure 2) which is an indication of the smallest possible contact point or the mean point of contact which is an indication of cell body contact. Cells that were 0 μm from the edges of the acquisition were excluded from the analysis. The representative snapshot and movie of segmented MuSCs and ECs in the muscle of Pax7^tdT:Flk1^GFP mice were shown in Figure 2A and 2B, and Supplemental Video 1 (MuSC-EC 3D Movie, related to Figure 2).

The data was exported for analysis in Prism Graphpad 7 for statistical analysis. Correlation between the mean distance from an EC, and ellipticity and volume were calculated (Figures S2E and S2F).

Measurement of cells and nuclei sizes

MuSC volume was measured using confocal microscopy as described above. Briefly, MuSC staining was imaged with a 20 x lens with a 1 μm step size to cover the entire sample. Samples were mounted in 1 mm slide chambers instead of coverslips to avoid squeezing the samples. Pax7 antibody for MuSC-nucleus staining or Pax7tdT expression for MuSC-cytoplasm was imaged in single muscle fibers or fixed muscle bundles. Images were pixel-classified as described above and imported into Imaris 9 for segmentation, object classification, visualization and volumetric analysis (Figures 1C, S1C and S1D).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis was performed using Prism 7 (Graphpad, La Jolla, CA) or RStudio (RStudio Team, 2015). For comparison between two groups, an unpaired T-test was used. For comparison between multiple groups, a one-way ANOVA was used with multiple comparisons to the control. Distributions were compared using a chi-squared test. Correlations were assessed using Pearson correlation co-efficient with two-tails. Graphing of the data was performed using Prism 7. Additional visualizations were created using ggplot2 (Wickham, 2009) and Pheatmap (Kolde, 2012) package for RStudio (RStudio Team, 2015). Vector diagrams were modified using Graphic (Autodesk). Statistics for differential gene expression from transcriptomic studies were performed at a dataset level with a False Discovery Rate (FDR) of 0.05 and represented in the graphs as such. An α level of 0.05 was used for all analyses. All values are means ± SEM unless noted otherwise. Asterisk (*), (**), (***) and (****) indicate experimental pairs where differences between the compared values were statistically significant (P < 0.05, P < 0.01, p < 0.001 and p < 0.0001, respectively).

DATA AND SOFTWARE AVAILABILITY

The accession numbers for the Microarrays and RNA-seq data reported in this paper are shown in Table S5 (GSE100505, GSE108739, GSE103684, GSE97399, GSE89633, GSE84379, GSE47067, 10.6084/m9.figshare.5715025).

Highlights.

• Skeletal muscle tissue clearing and fluorescent reporters for satellite cell niche

• Image analysis shows proximity of satellite cells to capillaries in 3-dimension

• Satellite cell-derived VEGFA mediates microvascular patterning in skeletal muscle

• Microvasculature-derived Notch ligand Dll4 promotes satellite cell self-renewal