Intravital microscopy of satellite cell dynamics and their interaction with myeloid cells during skeletal muscle regeneration

Abstract

Skeletal muscle regeneration requires the highly coordinated cooperation of muscle satellite cells (MuSCs) with other cellular components. Upon injury, myeloid cells populate the wound site, concomitant with MuSC activation. However, detailed analysis of MuSC–myeloid cell interaction is hindered by the lack of suitable live animal imaging technology. Here, we developed a dual-laser multimodal nonlinear optical microscope platform to study the dynamics of MuSCs and their interaction with nonmyogenic cells during muscle regeneration. Using three-dimensional time-lapse imaging on live reporter mice and taking advantages of the autofluorescence of reduced nicotinamide adenine dinucleotide (NADH), we studied the spatiotemporal interaction between nonmyogenic cells and muscle stem/progenitor cells during MuSC activation and proliferation. We discovered that their cell-cell contact was transient in nature. Moreover, MuSCs could activate with notably reduced infiltration of neutrophils and macrophages, and their proliferation, although dependent on macrophages, did not require constant contact with them. These findings provide a fresh perspective on myeloid cells’ role during muscle regeneration.

Live animal imaging platform reveals stage-dependent role of immune cells during muscle repair.

INTRODUCTION

Skeletal muscles, a type of striated muscle, are attached to bones and responsible for voluntary body movement. They have remarkable capacity for regeneration in response to injury (1). The Paired Box 7 (Pax7)–expressing muscle stem cells (MuSCs; also known as satellite cells), residing between the sarcolemma and basal membrane of their host myofibers in quiescence (2, 3), are crucial for muscle homeostasis and regeneration. Upon injury or disease, satellite cells are activated with rapid expression of myogenic differentiation 1 (MyoD) protein and a gradual increase in cell size. Most activated MuSCs start to proliferate, differentiate into myocytes, and finally fuse into existing injured fibers or fuse with each other to form new muscle fibers (3–5). While the regulation of MuSC quiescence exit by several critical intracellular signaling pathways has been studied (e.g., phosphatidylinositol 3-kinase and Notch) (6–8), little is known about the role of niche, particularly inflammation, in MuSC activation.

The myeloid lineage dominates the inflammatory response at the early stage of muscle injury (9–12). Neutrophils are the earliest immune responders to infiltrate the damaged muscle within 1 to 3 hours after injury and are responsible for the establishment of necrotic microenvironment by releasing proinflammatory cytokines to attract other types of immune cells (12). In one study, the depletion of neutrophils delayed muscle regeneration after injection of myotoxic substances in mice (13). However, another study in mice showed that neutrophil depletion attenuated muscle injury after exhaustive exercise (14). Besides, in vitro studies demonstrated that neutrophil-mediated factors promoted proliferation but impeded the differentiation of the myogenic progenitor cells, reflecting a complex role for neutrophils in regulating myogenesis (15). MuSC activation coincides with neutrophil abundance reaching its peak, suggesting that the neutrophils may be involved in MuSC activation (16).

Following neutrophils, macrophages are the predominant myeloid lineage cells observed during skeletal muscle regeneration (17). Disrupting the CCL2-CCR2 axis by using CCR2- or CCL2-deficient mouse strains notably reduced the recruitment of MOs/MΦs and impeded muscle regeneration (18–20). Similarly, partial or total depletion of circulating monocytes leads to impaired muscle regeneration, as indicated by the smaller cross-sectional area of regenerating myofibers, intramuscular fat accumulation, fibrosis, or delayed angiogenesis in injured sites (18, 21–25). Numerous studies have shown that macrophages or neutrophils regulate the dynamics of MuSCs via paracrine factor release (15, 26–31). However, only a few studies have reported that cell-cell contacts play an essential role in regulating the fate of MuSCs (32–36), and these studies of the role of macrophages or other myeloid cells during muscle regeneration were mainly based on in vitro or ex vivo analysis and did not involve the early activation stage of MuSCs.

A recent breakthrough was made by using real-time imaging in transgenic zebrafish to systematically capture the interactions between MuSCs and the innate immune system. Ratnayake et al. (37) reported that all MuSCs at wound sites must establish prolonged, direct interaction with a specific resident macrophage before proper cell division and the contact ceased upon cell division after injury in larval zebrafish. By contrast, the origins of mammals’ skeletal muscle resident macrophages (SMRMs) are heterogeneous (38, 39), and it is extremely challenging to find a transgenic model that labels all SMRMs, infiltrating macrophages, and neutrophils simultaneously. Therefore, the spatiotemporal interaction between myeloid cells and MuSCs in mammals has not been investigated in vivo in real time due to technical challenges, and it is unknown whether constant macrophage or other myeloid cell–MuSC contact is also present and, if so, necessary for MuSC activation and proliferation in response to muscle injury in mammals.

In this study, we attempted to address the following unsolved questions of muscle regeneration in the mammals: (i) whether infiltrating myeloid cells are necessary for MuSC early activation and (ii) whether MuSCs establish prolonged contact with nonmyogenic cells during the proliferation stage of muscle regeneration. To achieve this, we developed a multimodal nonlinear optical microscope system integrating second harmonic generation (SHG) and two-photon excited fluorescence (TPEF) imaging to examine the dynamics of satellite cells, nonmyogenic cells, and their complex interactions during muscle regeneration in vivo. Combining signals from fluorescent proteins of mouse models and the reduced nicotinamide adenine dinucleotide (NADH), a metabolic coenzyme located in the mitochondria with autofluorescence, in principle, we could trace all the cell types that interacted with MuSCs/activated satellite cells (ASCs) in the injured sites. As a result, we first monitored the change in mitochondrial mass and volume of MuSCs in vivo and found that they increased synchronously after injury. Further, we demonstrated that cell-cell contacts between CCR2⁺ monocytes/macrophages and MuSCs were not essential for MuSC early activation. Notably, we found that MuSC proliferation in mice did not require continuous contact with a specific macrophage or other nonmyogenic cells as in zebrafish. However, depletion of macrophages substaintially impaired the proliferation and differentiation of MuSCs and led to intramuscular fibrosis. Besides, by reducing the infiltration of neutrophils in the CCR2 knockout (KO) mice, we demonstrated that the infiltrating neutrophils and macrophages had no influence on the initial cell volume increase in MuSCs. Together, these findings made on our novel imaging platform revealed the stage-dependent role of MuSC–myeloid cell cross-talk during muscle regeneration and enabled future studies on the underlying molecular mechanism.

RESULTS

In vivo dynamics of MuSCs could be visualized during skeletal muscle regeneration

To study the dynamics of MuSCs during muscle regeneration in vivo, we used a multimodal two-photon microscope system to visualize genetically labeled Pax7⁺ MuSCs and their descendant myoblasts in live mice (fig. S1). Pax7⁺ SCs were efficiently labeled by cytoplasmic enhanced yellow fluorescent protein (eYFP) after tamoxifen injection (6, 40). Additionally, NADH autofluorescence provided a one-of-a-kind opportunity to visualize all types of cells. This metabolic enzyme is particularly abundant in mitochondria and can be used as a label-free “mitotracker” probe for imaging cells and evaluating the distribution of mitochondria in each cell (41–43). The SHG signals from muscle sarcomeres and collagen fibers provided important information on the regeneration process (44–46).

We performed multimodal in vivo imaging on mouse hindlimb tibialis anterior (TA) muscle at different time points after needle track injury, which typically causes localized and acute damage to muscles fibers (fig. S2). Before injury, quiescent MuSCs resided on the surface of individual muscle fibers, indicated by the orderly arranged SHG pattern from sarcomeres (Fig. 1A). Twelve hours post–needle track injury (hpi), the muscle fiber degenerated, and the SHG signal from sarcomeres disappeared. However, we observed that the collagen fibers parallel to the adjacent intact myofiber, known as “ghost fibers,” which determined the direction of migration and division of MuSCs after injury (40), were still present in the injured site, as shown by arrows in Fig. 1A. Previously, based on freshly isolated MuSCs or myofibers grown in culture, activated MuSCs were found to be larger in volume than their quiescent counterparts with expanded cytoplasm and more organelles (6, 47–51). Here, we also used the cell volume change in live mice as a biomarker to investigate the kinetics of MuSC activation and growth in response to acute injury in vivo. At 12 hpi, a slight increase in cell volume compared to uninjured cells marked one of the earliest visible signs of MuSC activation (the mean value increased by about 29%). Then, the cell volume increased continuously, reaching its maximum at 2 days post-injury (dpi) when MuSCs started to proliferate, and subsequently decreased, returning to preinjury levels at 1 month post-injury (mpi) (Fig. 1B). By analyzing the autofluorescence of NADH in individual YFP⁺ cells after injury, we also observed that the increased cell volume was accompanied by increased mitochondrial mass (Fig. 1, C and D), consistent with previous in vitro analyses that proliferating myoblasts and differentiating myocytes underwent enhanced mitochondrial biogenesis (6, 52). Finally, the distribution of nuclei in myocytes and newly formed myotubes could also be visualized (Fig. 1D, indicated by the asterisks).

Fig. 1. The dynamic of MuSCs and their microenvironment could be visualized during muscle regeneration.

(A) Typical YFP (yellow) and SHG (gray) images of muscle satellite cells/myogenic progenitor cells/myotubes and muscle environment at indicated time points after injury. Collagen parallel to the intact muscle fiber was indicated by arrows. Myoblast that fused together is indicated by brown arrowheads. (B) Time course of cell volume of individual YFP⁺ at indicated time points; N ≥ 46 cells from 3 to 5 mice per time point. (C) Time course of NADH area (mitochondrial mass) of individual YFP⁺ muscle satellite cells/myogenic progenitor cells at indicated time points. N ≥ 30 cells from 3 to 5 mice per time point. Kruskal-Wallis test. (D) NADH signal (cyan) of representative YFP⁺ muscle satellite cells/myogenic progenitor cells/myotubes (yellow). Asterisks show the nuclei distribution in the myocytes/myotubes at 4 dpi. Kruskal-Wallis test. (E) Time course of YFP⁺ cell density (left y axis, black) and YFP⁺ area density (right y axis, brown) at different time points (n ≥ 3 mice per time point). Unpaired two-tailed t test. The FOV was 300 μm × 300 μm × 60 μm. (F) Relative collagen area of the injured sites at indicated time points (n ≥ 3 mice per time point). One-way ANOVA with Tukey’s multiple comparisons test. Data were reported as mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Scale bars, 50 μm (A) and 30 μm (D).

To evaluate proliferation and differentiation of MuSCs, we measured the cell density of individual YFP⁺ cells before 2 dpi and calculated the percentage of YFP⁺ area of the injured sites after 2 dpi (Fig. 1E). Cell density at 32 hpi was similar to that of MuSCs in noninjured muscle and increased slightly at 2 dpi. At 3 dpi, YFP⁺ cell density increased significantly, and some cells started to fuse (Fig. 1A). After 4 dpi, newly formed myotubes were surrounded by collagen fibers parallel to the intact muscle cells (Fig. 1A and fig. S3A). By analyzing the SHG signal generated by type I and III collagen (53) (fig. S3B), we found that the relative collagen area in the injured sites increased significantly at 3 dpi, remained high until 5 dpi, and decreased at 8 dpi (Fig. 1F). The increase in collagen distribution may promote myoblast migration and myogenic differentiation (54). At 1 mpi, YFP⁺ muscle fibers filled the injured sites, and the periodic SHG signal indicated that the sarcomeres were regenerated (Fig. 1A). The nonmyofiber YFP⁺ cell (i.e., self-renewed MuSCs) density was slightly higher at 1 mpi than that at the noninjury stage (Fig. 1E). In addition, based on the polarization dependence of SHG intensity (55), we enhanced the SHG intensity of collagen fiber and weakened the SHG intensity of the muscle fibers by changing the polarization of the excitation laser. Then, we could quantify the relative collagen area before injury and 1 mpi and found no significant difference between them (fig. S3, C and D). The study demonstrated that multimodal nonlinear optical microscope is an excellent tool for studying the dynamics of MuSCs and the environment of the muscle system during muscle regeneration in vivo.

Cell-cell contact between RFP⁺ monocytes/macrophages and MuSCs was not required for MuSC activation

It is of great interest to investigate whether MuSC’s exit from quiescence, the first stage of myogenic program, involves the early immune response. Particularly, MuSC activation (i.e., quiescence exit) is considered distinct from previously studied MuSC proliferation (4, 8). After an acute injury, the neutrophils invade the injured muscle first and almost disappear before 2 dpi (9–11). On the other hand, monocytes/macrophages (MOs/MΦs) that express proinflammatory markers infiltrate injured muscle fibers via the CCL2-CCR2 axis (17, 21, 56). Gradually, these proinflammatory macrophages were outnumbered by the anti-inflammatory macrophages by 7 dpi. It was reported that macrophages could release ADAMTS1 that targets Notch1 protein of the MuSCs to affect the activation of MuSCs (28). In addition, MOs/MΦs can also establish cell-cell connections with MuSCs through membrane ligands and receptors, which may regulate the lineage progression of MuSCs after injury (35, 57). However, whether direct cell-cell contact between MuSCs and monocytes/macrophages is required for the activation of MuSCs is uncertain. To explore the role of monocytes/macrophages and the interaction between them and MuSCs, we performed in vivo imaging in a mouse model (Pax7^CreERT2/+;Rosa-stop-YFP;CCR2^RFP/+), in which CCR2⁺ MOs/MΦs were labeled by RFP (58) and Pax7⁺ MuSCs were marked by YFP after tamoxifen injection.

As mentioned previously, the cell volume of MuSCs increased significantly at 12 hpi, but not at 8 hpi. Therefore, we performed time-lapse imaging in the TA muscle from 4.5 to 12 hpi to monitor the dynamics of RFP⁺ MOs/MΦs and the spatiotemporal interaction between MuSCs and RFP⁺ MOs/MΦs and to trace the volume of individual MuSCs by using our well temperature-controlled stage (fig. S4). As the results showed, there were few RFP⁺ MOs/MΦs resident in the intramuscular connective tissue before injury (fig. S5A). The density of RFP⁺ MOs/MΦs slightly increased at 4.5 hpi compared to the noninjury state (fig. S5B). From 4.5 to 12 hpi, the MuSCs were motionless (fig. S5C), and RFP⁺ MOs/MΦs were gradually recruited into the injured sites in Injured-Ctrl groups (Fig. 2, A and B). Most RFP⁺ MOs/MΦs were highly dynamic (fig. S5D and movie S1). Moreover, a small population of RFP⁺ MOs/MΦs remained in the injured site during imaging, and the displacement was less than 10 μm (fig. S5E). No division of RFP⁺ MOs/MΦs was observed during imaging, consistent with the ex vivo results that there were no Ki67⁺ monocytes/macrophages before 15 hpi (21). We recorded the cell volume of individual MuSCs at 4.5, 8, and 12 hpi and analyzed the minimum distance between MuSCs and RFP⁺ MOs/MΦs every 5 min from 4.5 to 12 hpi (Fig. 2, C and D). When MuSCs were in contact with RFP⁺ MOs/MΦs, the minimum distance was zero. However, it should be emphasized that direct contact between MuSCs and macrophages cannot be confirmed definitively due to optical microscopy’s limited resolution. By analyzing the change in distance between them, we found that all MuSCs established short-lived rather than continuous or repeated direct interaction with RFP⁺ MOs/MΦs before cell volume increased at 12 hpi (Fig. 2, C and D, and movie S2). In addition, MuSCs interacted directly with different RFP⁺ MOs/MΦs but only one or two at a time immediately after injury (fig. S5F). We found that the increase in cell volume of different MuSCs was asynchronous, which may be caused by the unsynchronized muscle degeneration and intrinsic heterogeneity of MuSCs (50, 51). To exclude the possibility that long-term in vivo imaging affected MuSC activation and altered its interaction with macrophages, we conducted time-lapse imaging on noninjured muscles (fig. S5G and movie S1). In intact muscle fibers, the cell volume of MuSCs did not increase significantly compared to the Injured-Ctrl group after 7.5-hour imaging (Fig. 2C), and a portion (5 of 12 cells in four mice) of MuSCs was observed to contact different RFP⁺ MOs/MΦs during the imaging session (Fig. 2D). In addition, the duration of contact between MuSCs and RFP⁺ cells was similar to that in injured muscle (fig. S5, H and I). Together, MuSCs did establish short-lived direct contacts with RFP⁺ cells during activation immediately after injury.

Fig. 2. Both the decrease in RFP⁺ MOs/MΦs density and the absence of direct cell-cell contact between MuSCs and RFP⁺ MOs/MΦs did not impede the activation of MuSCs.

(A) Maximum z intensity projections of TPEF image stacks of YFP⁺ MuSCs (yellow), RFP⁺ MOs/MΦs (magenta), and SHG generated from collagen and muscle fiber (gray) at 4.5, 8, and 12 hpi in Injured-Ctrl, Injured-GC (glucocorticoid), and Injured-KO (CCR2 knockout) groups. The image started at 4.5 hpi. (B) Time course of RFP⁺ MOs/MΦs density at the image sites during time-lapse imaging in the Non-injury, Injured-Ctrl, Injured-GC, and Injured-KO groups. n = 4 to 5 mice per group. The FOV was 300 μm × 300 μm × 60 μm. Each line represents a traced FOV. Only the P values of cell density between different groups at 12 hpi were shown. (C) Change in normalized cell volume of individual MuSCs during time-lapse imaging in four groups. Each line represents one traced YFP⁺ MuSC (n ≥ 12 cells from 4 to 5 mice per group). (D) Time course of minimal distance from representative individual MuSCs to RFP⁺ MOs/MΦs during time-lapse imaging in Non-injury, Injured-Ctrl, Injured-GC, and Injured-KO groups. Each row represents one traced YFP⁺ MuSC. The minimal distances with a value of 0 μm were highlighted with white squares. Other values were coded with color maps. The volume of the 3D distance matrix equals the FOV. Individual data points were shown, and the bars indicated the mean value. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, two-way ANOVA with Tukey’s multiple comparisons test. Scale bars, 50 μm (A).

To further explore whether the direct cell-cell contacts between RFP⁺ MOs/MΦs and MuSCs were required to activate MuSCs, we tried to partially reduce macrophage infiltration by glucocorticoid (GC) treatment (Injured-GC group) (24, 59) or using CCR2ko (Pax7^CreERT2/+;Rosa-stop-YFP;Ccr2^RFP/RFP) mice (Injured-KO group) (19, 58). We repeated time-lapse imaging in the Injured-GC and Injured-KO mice. The infiltration of RFP⁺ MOs/MΦs during imaging was significantly impaired in the Injured-GC and Injured-KO groups as only a few RFP⁺ cells were recruited to the injured areas at 12 hpi (Fig. 2, A and B, and movie S1). Further, because of the low density of RFP⁺ MOs/MΦs, most MuSCs could not contact RFP⁺ MOs/MΦs directly before 12 hpi (Fig. 2D and movie S2). However, there was no difference in the increase in cell volume between MuSCs of the Injured-Ctrl group that had been in direct contact with RFP⁺ MOs/MΦs and MuSCs of the Injured-GC and Injured-KO group that had not directly contacted RFP⁺ MOs/MΦs before 12 hpi (Fig. 2C). (There were about ~²/₁₅ and ~¹/₁₃ MuSCs that contacted RFP⁺ MOs/MΦs before 12 hpi in the Injured-GC or Injured-KO groups, respectively, and these MuSCs were excluded.) To investigate whether the reduction in infiltrated macrophages impaired the subsequent increase of cell volume, we performed snapshot imaging at 0 hpi, 8 hpi, 12 hpi, 2 dpi, and 3 dpi and analyzed MuSC volume. The RFP⁺ cell density was significantly reduced in the Injured-GC and Injured-KO groups before 12 hpi (fig. S5J). Except that the volume of MuSC in the Injured-GC group was slightly less than that of the Injured-Ctrl group at 2 dpi (mean value decreased by ~12%), MuSC growth was similar in all these groups at all time points examined (fig. S5K). Collectively, our data indicated that both the decrease in RFP⁺ MOs/MΦs density and the loss of direct cell-cell contact between MuSCs and RFP⁺ MOs/MΦs did not impede the activation of MuSCs. Besides, the reduction in recruited macrophages did not hinder the cell volume increase of MuSCs.

Reduced neutrophil density and cell-cell contact between RFP⁻NADH⁺ nonmyogenic cells and MuSCs had no influence on MuSC activation

During time-lapse imaging sessions, along with RFP⁺ MOs/MΦs, a significant number of RFP⁻NADH⁺ YFP⁻ cells were observed to infiltrate the injured sites in the Injured-Ctrl, Injured-GC, and Injured-KO groups (Fig. 3A). Their density was higher than RFP⁺ MOs/MΦs at 4.5 hpi. To confirm that the NADH signal could be used to visualize most cells in the injured muscles, MitoTracker Deep Red (MTDR) that could stain mitochondria in live cells was applied topically to stain cells in situ. The results showed that more than 90% of cells were NADH⁺MTDR⁺ (fig. S6, A and B). Therefore, most cells could be visualized based on NADH autofluorescence without further labeling. Since the NADH signal was mainly generated by mitochondria inside cells, it could not delineate the entire cell boundary. By analyzing the RFP/YFP-labeled whole cells and their NADH signal, we found that the difference between the boundary of NADH signal and RFP/YFP signal was probably less than 2.5 μm laterally and 6 μm axially (upper 95% confidential intervals of mean value were 2.165 and 3.101 μm, respectively) (Fig. 1B and fig. S6, C and D). Therefore, if the distance between NADH⁺ cells and MuSCs was greater than 2.5 μm laterally and 6 μm axially, there was no direct contact between cells.

Fig. 3. Reduced neutrophil density and cell-cell contact between RFP⁻NADH⁺ nonmyogenic cells and MuSCs had no influence on MuSC activation.

(A) Maximum z intensity projections of TPEF image stacks of YFP⁺ MuSCs (yellow), RFP⁺ MOs/MΦs (magenta), and NADH⁺ cells (cyan) at 4.5, 8, and 12 hpi in Injured-Ctrl, Injured-GC, and Injured-KO groups. (B) Typical images for the interaction between RFP⁻NADH⁺ cells and MuSCs within 10 min in the three groups. Different color asterisks corresponded to different nonmyogenic cells at different time points. Cyan, NADH; yellow, YFP; magenta, RFP. (C) Total contact duration between RFP⁻NADH⁺ cells and MuSCs in the Injured-Ctrl, Injured-GC, Injured-KO, and Injured-KO-CP [CCR2ko mice treated with cyclophosphamide (CP)] groups. Kruskal-Wallis test. (D) Change in normalized cell volume of individual MuSCs during time-lapse imaging in the Non-injury, Injured-Ctrl, Injured-GC, Injured-KO, and Injured-KO-CP groups. Data from Non-injury, Injured-Ctrl, Injured-GC, and Injured-KO groups were the same as those presented in Fig. 2C (n ≥ 12 cells from 4 to 5 mice per group). Each line represents one traced YFP⁺ MuSC, and the bar indicated the mean value. Two-way ANOVA with Tukey’s multiple comparisons test. No significant difference was observed between the Injured-Ctrl, Injured-GC, Injured-KO, and Injured-KO-CP groups. (E) Maximum z intensity projections of TPEF image stacks of YFP⁺ MuSCs (yellow), RFP⁺ macrophages (magenta), and NADH⁺ cells (cyan) at indicated time post-injury in the Injured-KO-CP group. (F) TPEF images for the interaction between RFP⁻NADH⁺ cells and MuSCs within 10 min in the Injured-KO-CP group. Asterisks indicated an RFP⁻NADH⁺ cell that contacted the MuSCs within 10 min. Data in (C) were reported as mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Scale bars, 50 μm [(A) and (E)] and 30 μm [(B) and (F)].

As demonstrated in previous studies, neutrophils were the earliest immune cells to infiltrate skeletal muscle in response to injury (9). Therefore, we immunostained the injured TA muscle using the markers Gr-1 and Ly6G. More than 80% (mean value, ~83%) of these cells were found to be Ly6G⁺ or Gr-1⁺ neutrophils (fig. S6, E and F). Because of the high density and dynamic nature of RFP⁻NADH⁺ YFP⁻ cells, they started to interact with MuSCs frequently at 4.5 hpi, and occasionally, MuSCs were observed to contact multiple cells simultaneously (Fig. 3B and movie S2). Further, there was no significant difference in the total duration of direct contacts between MuSCs and RFP⁻ NADH⁺ cells among Injured-Ctrl, Injured-GC, and Injured-KO groups (Fig. 3C).

To further examine whether the interaction between MuSCs and RFP⁻NADH⁺ cells was essential for MuSC activation, we tried to deplete neutrophils in the CCR2ko mice. A standard in vivo method of neutrophil depletion is anti-Ly6G antibody–mediated depletion (15, 60). We succeeded in reducing the RFP⁻NADH⁺ cell density by using clone 1A8 in the CCR2-RFP mice but failed in the CCR2ko mice (fig. S6, G and H), since macrophages are required for effective anti-Ly6G–mediated depletion of neutrophils (61). Therefore, we tried another drug, cyclophosphamide (CP), which can trigger the formation of DNA crosslinks and lesions that induce cell death. Apart from neutrophils, injection of CP could also reduce the number of circulating monocytes and B and T cells by limiting the proliferation of dividing cells (62, 63). However, the density of B and T cells was much lower than that of neutrophils and macrophages during the early stage of injury-induced muscle regeneration as B and T cells are recruited to the injured site after 3 dpi (10–12, 64). Therefore, the side effects of CP on B and T cells may not be of concern before 12 hpi. Treatment of CP effectively reduced the number of RFP⁻NADH⁺ cells (mostly neutrophils) at 4.5 hpi and later time points in the Injured-KO group (fig. S6, I and J). We next repeated the time-lapse imaging in the CP-treated CCR2ko (Injured-KO-CP) mice and found that the cell volume of MuSCs in the Injured-KO-CP group was not significantly different from that in the Injured-Ctrl, Injured-GC, and Injured-KO groups (Fig. 3D), although both the density of RFP⁻NADH⁺ cells and the frequency of contact were reduced by comparison with the other three groups (Fig. 3, C, E, and F, and movies S1 and S2). Therefore, the reduction of neutrophil density and contact duration between MuSCs and RFP⁻NADH⁺ cells had no influence on the initial growth of MuSCs after injury.

Nonmyogenic cells generally interacted briefly with YFP⁺ ASCs before ASC division

The reduction of infiltrated monocytes/macrophages and the interaction between them with MuSCs did not impair the initial volume increase and the further growth of MuSC. Therefore, we studied whether proliferation of MuSCs would be impeded. Infiltration of macrophages preceded the proliferation of ASCs. At 1 dpi, many RFP⁺ monocytes/macrophages were recruited to the injured site, and granules with strong autofluorescence in the YFP channel were located inside some RFP⁺ MOs/MΦs (Fig. 4A and fig. S7A). The autofluorescence may be generated from lipofuscin inside the macrophages’ cytoplasm after the phagocytosis of muscle fiber debris (65–67). More RFP⁺ MOs/MΦs infiltrated the injured areas and surrounded the YFP⁺ ASCs at 1.5 and 2 dpi (Fig. 4A). Besides, most NADH⁺ areas were RFP⁺ after 1 dpi in the Injured-Ctrl group, suggesting that monocytes/macrophages were the predominant cell type (fig. S7B). ASC density increased significantly at 2 to 3 dpi, indicating that cell divisions were relatively frequent starting at 2 dpi. Thus, we first conducted time-lapse imaging to monitor the spatiotemporal interaction between macrophages and ASCs at 2 dpi every 5 min for 7.5 hours. Figure S7C and movie S3 showed that a single ASC migrated parallel to intact neighboring muscle fibers and was continuously surrounded by macrophages before division. Unfortunately, due to the high density of RFP⁺ MOs/MΦs, it was hard to identify individual macrophages to study their interaction with ASCs.

Fig. 4. Nonmyogenic cells generally interacted briefly with YFP⁺ ASCs at the proliferation stage in the Injured-Ctrl group.

(A) Representative images indicated the infiltration of macrophages at 1, 1.5, and 2 dpi. RFP, magenta; YFP, yellow; SHG, gray; autofluorescence within macrophages, white. (B) Representative time sequence images of the dynamic contact between nonmyogenic cells and ASCs at 1 dpi. RFP, magenta; YFP, yellow; NADH, cyan. Asterisks indicated an RFP⁺ MO/MΦ that contacted ASC briefly. (C) Image sequence of interaction between a traced ASC and nonmyogenic cells in the Injured-Ctrl group at 1.5 dpi. Asterisks and white arrowheads indicated RFP⁺ MOs/MΦs or RFP⁻NADH⁺ cells that contacted the ASC at the indicated time points, respectively. Daughter cells separated from each other at time 0. (D and E) Representative compilation of the spatiotemporal interaction between different RFP⁺ MOs/MΦs (D), RFP⁻NADH⁺ cells (E), and the traced ASC before ASC division. The different color dotted lines correspond to the different time points shown in (C). Each row represented an RFP⁺ MO/MΦ (D) or an RFP⁻NADH⁺ cell (E). Each box represents one frame, and the time interval is 5 min. Green or cyan box means that RFP⁺ MOs/MΦs or RFP⁻NADH⁺ cells contact the ASC, and red box means that the contact is stopped before this frame. The number of the green or cyan boxes that dotted lines pass through is equal to the number of RFP⁺ MOs/MΦs or RFP⁻NADH⁺ cells that contacted ASC at that time. For example, at t = −200 min, there were six RFP⁺ MOs/MΦs and one RFP⁻NADH⁺ cell contacting ASC. Gray box means that the RFP⁺ MOs/MΦs temporarily dissociated from ASCs at that time but would contact ASCs again. White area means that the nonmyogenic cells are not traced. Scale bars, 50 μm [(A) and (B)] and 30 μm (C).

The density of YFP⁺ ASCs at 2 dpi was almost twice of that in noninjured mice, indicating that the first cell division already occurred in most MuSCs by 2 dpi (Fig. 1E). Therefore, we tried to perform the time-lapse imaging at 1 to 1.5 dpi to better examine the spatiotemporal interaction between RFP⁺ MOs/MΦs and ASCs. However, during 7.5-hour time-lapse imaging at 1 dpi, we did not observe the division of ASCs, which were mobile and contacted with different RFP⁺ MOs/MΦs transiently (Fig. 4B and movie S4). After 1.5 dpi, we could observe cell division, but ASCs were surrounded by RFP⁺ MOs/MΦs later as they were at 2 dpi (fig. S7D). Therefore, we reduced the number of the punctured sites from 5 to 3 on the TA muscles and avoided bleeding to reduce inflammation. Finally, in some cases, we could observe the ASC divisions at 1.5 dpi without being surrounded by RFP⁺ MOs/MΦs.

Notably, although the monocyte-derived macrophages recruited upon injury were CCR2⁺ and RFP⁺, most resident macrophages were CCR2⁻ under homeostatic conditions (38, 68), and it remains unclear whether these macrophages could express CCR2 after injury. In addition, previous studies have suggested that CCR2⁻ macrophages are involved in tissue repair (58, 69). Our data also revealed that, despite more than 90% NADH⁺ cells being RFP⁺ after 1 dpi (fig. S7B), there was a small population of RFP⁻NADH⁺ cells in the injured sites. To investigate whether constant contact with macrophages is required for ASC division, as observed in zebrafish, we need to monitor the dynamics of all macrophages. As we could not label RFP⁻ macrophages directly, we instead analyzed the interaction between all RFP⁺/RFP⁻NADH⁺ nonmyogenic cells [including RFP⁺/RFP⁻ macrophages, neutrophils, fibro-adipogenic progenitors (FAPs), and others] and ASCs. Notably, we observed that more RFP⁻NADH⁺ cells migrated to the injured sites during live imaging sessions of time window from 3.5 to 7.5 hours (fig. S7E), which was not observed in snapshot experiments conducted at 1, 1.5, and 2 dpi, with a time window less than 1 hour. These cells were likely neutrophils, which responded to the secondary inflammation caused by prolonged exposure of the muscles. However, it should be noted that if RFP⁻ nonmyogenic cells had long protrusion similar to ASCs (as shown in t = −200 min in Fig. 4C), and the protrusions could not be entirely observed by NADH signal, the contact would be missed.

The compilations of the behavior of 59 RFP⁺ MOs/MΦs and 32 RFP⁻NADH⁺ cells that contacted the representative traced ASC before ASC division are shown in Fig. 4, D and E, respectively. The frame interval was 5 min. Because of the high density and dynamics of nonmyogenic cells, a traced ASC contacted different RFP⁺ MOs/MΦs and RFP⁻NADH⁺ cells simultaneously at different time points (Fig. 4, C to E, and movie S5). For example, at t = −200 min (200 min before cell division, daughter cells separated from each other at t = 0 min), the ASC was in contact with six RFP⁺ MOs/MΦs and one RFP⁻NADH⁺ cell. Additionally, the contact duration between RFP⁻NADH⁺ cells and ASCs was significantly shorter than that between RFP⁺ MOs/MΦs and ASCs (fig. S7F). In accordance with observations in zebrafish (37), we define that constant contact between nonmyogenic cells and MuSCs must be longer than 3 hours and persist until MuSC cell division. As a result, rare RFP⁺ MOs/MΦs or RFP⁻NADH⁺ cells established constant contact with the ASCs before cell division (Fig. 4, D and E, and fig. S7F). Therefore, our data strongly suggest that the constant contact between ASCs and nonmyogenic cells is not a prerequisite for the ASC division.

To better address the role of macrophages in ASC division, we reduced macrophage infiltration by GC treatment or by using CCR2-ko mice. RFP⁺ MOs/MΦs were significantly reduced in either Injured-GC or Injured-KO mice at 2 dpi (Fig. 5, A and B). The density of ASCs in macrophage-depleted mice was slightly lower at 2 dpi compared with Injured-Ctrl mice but still higher than that in noninjured mice (Fig. 5C). Therefore, it was appropriate to investigate the interaction between ASCs and nonmyogenic cells at 1.5 dpi in Injured-GC and Injured-KO mice. Unlike RFP⁺ cells, more RFP⁻NADH⁺ cells were observed in the injured sites at 1.5 and 2 dpi in the Injured-KO and Injured-GC groups than in the Injured-Ctrl group (Fig. 5, A, D, and E). These RFP⁻NADH⁺ cells probably included neutrophils, RFP⁻ macrophages, and FAPs. While CCR2 deficiency did not impair the proliferation of FAPs (70), the density of FAPs increased at a slower rate than that of infiltrating neutrophils (9, 11, 15, 31, 71–73). Therefore, it was reasonable to assume that most RFP⁻NADH⁺ cells observed were neutrophils. Next, we performed immunostaining in the whole-mount TA muscle using fluorophore-conjugated antibodies specific to neutrophils (Gr-1 and Ly6G) and FAPs (PDGFR-a), respectively. The results indicated that more than 80% (mean value, ~85%) of cells were Gr-1⁺ or Ly6G⁺ (fig. S7, G and H), while approximately 10% of cells were PDGFR-a⁺ in both Injured-GC and Injured-KO mice at 1.5 dpi (fig. S7, I and J).

Fig. 5. Nonmyogenic cells generally interacted briefly with YFP⁺ ASCs at the proliferation stage in the Injured-GC and Injured-KO groups.

(A) Representative image for the TA muscle at 2 dpi of Injured-GC and Injured-KO groups. SHG, gray; YFP, yellow; NADH, cyan; RFP, magenta. (B and C) RFP⁺ MOs/MΦs (B) and YFP⁺ ASCs (C) density at 2 dpi of three groups (n ≥ 3). (D to F) Representative time sequence images of the dynamic contact between nonmyogenic cells and ASCs before division at 1.5 dpi in Injured-KO (D) and Injured-GC groups with low (E) and high (F) RFP⁺ MΦ density. Daughter cells separated at time 0. RFP, magenta; YFP, yellow; NADH, cyan. Asterisks and white arrowheads indicated the RFP⁺ MOs/MΦs and RFP⁻NADH⁺ cells that contacted ASC at the indicated time points, respectively. (G to L) Representative compilation of the spatiotemporal interaction between nonmyogenic cells and ASCs before cell division in the Injured-KO [(G) and (H)] and Injured-GC groups with low (I and J) and high [(K) and (L)] RFP⁺ MΦ density at 1.5 dpi. The different dotted lines in (G) to (L) correspond to the different time points in (D) to (F). Each row represented an RFP⁺ MO/MΦ [(G), (I), and (K)] or an RFP⁻NADH⁺ cell [(H), (J), and (L)]. Green or cyan box means that RFP⁺ MO/MΦs or RFP⁻NADH⁺ cells contact ASC, and red box means that the contact is stopped before this frame. The number of the green or cyan boxes that the dotted lines pass through is equal to the number of RFP⁺ MOs/MΦs or RFP⁻NADH⁺ cells that contacted ASCs at the same time. Gray box means that the nonmyogenic cells do not contact ASCs at that time but contact ASCs again. White area means that the cells are not traced. One-way (B) and two-way ANOVA (C) with Tukey’s multiple comparisons test. Data were reported as mean ± SD. *P < 0.05; ****P < 0.0001. Scale bars, 50 μm (A) and 30 μm [(D) to (F)].

We conducted time-lapse imaging at 1.5 dpi in Injured-GC and Injured-KO mice, analyzing the dynamics of RFP⁺ MOs/MΦs, RFP⁻NADH⁺ cells, and ASCs before ASC division (Fig. 5, D to L). Before division, ASCs established short-lived contact with several RFP⁺ MOs/MΦs (or did not contact RFP⁺ MOs/MΦs) and more frequently with RFP⁻NADH⁺ cells during imaging in Injured-KO mice because of lower RFP⁺ cell density (Fig. 5, D, G, and H; fig. S7, F and K; and movie S6). Besides, due to the variable depletion efficiency by GC treatment, the RFP⁺ cell density fluctuated in Injured-GC mice but remained significantly higher than that in the Injured-KO mice. Therefore, the frequency and total duration of contact with RFP⁺ cells were higher in the Injured-GC mice compared with Injured-KO mice (fig. S7, F and K). Nevertheless, ASCs established transient contact with RFP⁺/RFP⁻NADH⁺ cells regardless of RFP⁺ cell density (Fig. 5, E, F, and I to L, and movies S7 and S8), and constant contact between ASCs and a specific cell before ASC division was rare in both Injured-KO and Injured-GC groups (fig. S7F and Fig. 5, G to L).

Although the overall proliferation rate of ASCs in both Injured-KO and Injured-GC groups was similarly and slightly impaired by the reduction of macrophages (Fig. 5C and fig. S7L), all ASCs in the three groups could successfully complete cell division at 1.5 dpi, indicating that cell division of ASC did not require continuous contact with a specific RFP⁺ MO/MΦ or other RFP⁻ nonmyogenic cells. Besides, the contact duration between ASCs that did not divide and nonmyogenic cells was similar to the duration between ASCs that divided and nonmyogenic cells during time-lapse imaging in the Injured-Ctrl, Injured-GC, and Injured-KO groups (fig. S8, A and B). Notably, we also found that ASCs predominantly migrated along the longitudinal axis (y axis) of the muscle fiber and the displacement in the y axis was significantly larger than in the x and z axes in the three groups, which is consistent with previous study (40). The displacement in the x and z axes did not exceed the average diameter of a muscle fiber (fig. S8, C to E). Besides, the instantaneous migration speed of ASCs in the Injured-KO groups was slightly smaller than that in the Injured-Ctrl groups (fig. S8F). In contrast to ASCs, the migration patterns of RFP⁺ MO/MΦs were found to be more intricate. However, like ASCs, their displacement along the x and z axes was considerably lower than that of the y axis (fig. S8, G and H). The Injured-GC and Injured-KO groups exhibited a reduction in the instantaneous speed of RFP⁺ MO/MΦs (fig. S8I). In summary, the depletion of macrophages impaired the proliferation rate but had no influence on the migration pattern of the ASCs.

Depletion of macrophages impaired proliferation and differentiation of MuSCs

To further investigate how the impaired recruitment of macrophages influences the differentiation of MuSCs, we performed in vivo imaging sessions in the TA muscles at 3, 4, and 15 dpi in the Injured-Ctrl, Injured-GC, and Injured-KO groups (Fig. 6). In the Injured-Ctrl group, the RFP⁺ MOs/MΦs filled the injured areas at 3 dpi, started to decrease at 4 dpi, and further decreased at 15 dpi (Fig. 6, A, B, and E). During ASC differentiation, some RFP⁺ cells exhibited strong autofluorescence in all three groups (Fig. 6A). The autofluorescence was detected at YFP channel and another autofluorescent channel that was excited by 740 nm and detected at 500 to 550 nm, in which RFP and YFP signals were not detectable (fig. S9A). As mentioned before, the autofluorescent granules may be related to the phagocytic activity of macrophages. Because of the significantly decreased RFP⁺ cell density, the RFP⁺YFP⁺ cell density was lower in both the Injured-GC and Injured-KO groups than in the Injured-Ctrl group at 3 dpi (fig. S9B). However, at 4 dpi, the density of both RFP⁺YFP⁺ and RFP⁺ cells in the Injured-GC group returned to the level seen in the Injured-Ctrl group (Fig. 6B and fig. S9B), presumably because monocyte depletion by drug treatment could increase the local proliferation of macrophage subsets (74). Besides, the ratio of RFP⁺YFP⁺ area to total RFP⁺ area was similar among all three groups (fig. S9C). The relative area of ASCs and newly formed myotubes marked by YFP⁺RFP⁻ (excluding YFP signals within RFP⁺ cells) of both Injured-GC and Injured-KO groups was similarly reduced compared to the Injured-Ctrl group (Fig. 6C), indicating impaired myogenic differentiation. In addition, some newly formed myotubes were bifurcated or shrunk with RFP⁺YFP⁺ MOs/MΦs located in their growth path in the Injured-GC and Injured-KO groups at 4 dpi (Fig. 6A). These RFP⁺YFP⁺ MOs/MΦs may be responsible for the clearance of apoptotic cells, such as FAPs (65, 66, 75, 76). Therefore, the bifurcation of myotubes may be caused by cellular debris that was not phagocytized in a timely manner.

Fig. 6. Depletion of macrophages impaired the proliferation and differentiation of MuSCs.

(A) Images for the RFP⁺ MOs/MΦs distribution (magenta), YFP⁺ ASCs/myotubes (green), and SHG generated from collagen and muscle fiber (gray) at 3 and 4 dpi in the Injured-Ctrl, Injured-GC, and Injured-KO groups. The bifurcated or shrunk myotubes at 4 dpi in the Injured-GC and Injured-KO groups were delineated by brown circles. (B to D) Statistics of RFP⁺ area (RFP⁺ MOs/MΦs) density (B), YFP⁺ area density (C), and relative collagen area (D) at indicated time points after injury of three groups. (E) Representative TPEF and SHG images of newly regenerated myotubes and RFP⁺ MOs/MΦs at 15 dpi of three groups. Green, YFP⁺ myogenic cells/myotubes; magenta, RFP⁺ MOs/MΦs; gray, SHG generated by muscle fiber and collagen. The arrow indicates a newly formed blood vessel. (F) Normalized signal profiles along the lines in (E). (G to I) Diameter of newly regenerated myotubes (G), RFP⁺ cell density (H), and YFP⁺ cell density (I) at 15 dpi of three groups. n = 3 to 6. Data in (B) to (D) were analyzed by two-way ANOVA with Tukey’s multiple comparisons test. The FOV in (H) and (I) was 300 μm × 300 μm × 60 μm. Data in (G) were analyzed by Kruskal-Wallis test. Data in (H) and (I) were analyzed by one-way ANOVA with Tukey’s multiple comparisons test. Error bars were SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Scale bars, 50 μm [(A) and (E)]. a.u., arbitrary units.

At 15 dpi, the injured areas were filled with regenerated myotubes parallel to the intact muscle fibers in all three groups (Fig. 6E). On the basis of SHG signals, we could also visualize some regenerated blood vessels where an RFP⁺ monocyte went through (indicated by the arrow in Fig. 6E). Additionally, the SHG signal generated from the sarcomeres of myotubes was stronger and had more apparent periods in the Injured-Ctrl and Injured-GC groups, suggesting that myotubes were less mature in the Injured-KO group (Fig. 6, E and F). The SHG signal of some sarcomeres was weaker in the middle of muscle fibers because the nuclei were centrally aligned in the muscle cells (fig. S9D). As expected, the diameter of the myotubes was smaller in the Injured-GC and Injured-KO groups (Fig. 6G). However, there was no significant difference in the density of individual RFP⁺ and YFP⁺ cells between the three groups (Fig. 6, H and I). Further, we could observe that some myotubes were bifurcated by intramuscular cells or collagen in the Injured-KO and Injured-GC groups, which never occurred in the Injured-Ctrl group at 15 dpi (fig. S9E). Also, although the relative collagen area of the Injured-KO group was modestly lower than that of the Injured-Ctrl group at 4 dpi, it remained at a high level at 15 dpi (Fig. 6D and fig. S9F), which might be caused by continued FAP accumulation in muscle from the Injured-KO group (70). Although the RFP⁺ cell density was higher at 4 dpi, the relative collagen area in the Injured-GC group was similar to that in the Injured-KO group at 15 dpi (Fig. 6, A, B, D, and E, and fig. S9F), which indicated that the delayed infiltration of RFP⁺ MOs/MΦs at 4 dpi could not reduce the fibrosis at 15 dpi. Together, these data indicated that the macrophage deficiency severely delayed the differentiation of MuSCs and induced intramuscular fibrosis.

DISCUSSION

Quiescent MuSCs exhibit morphological heterogeneity with varying cellular process patterns (77, 78). They adhere to host myofibers through cadherins (8), a basal lamina via integrin (79) and vascular endothelium through Notch signaling (80). Additionally, niche factors from the extracellular matrix or myofibers also contribute to maintaining MuSC quiescence (79, 81). The MuSC activation is a complex process involving early activation [or quiescence-to-activation (Q-A) transition], cell growth, and S phase reentry, which typically takes about 36 hours (82, 83). This process encompasses a series of molecular and morphological events, including epigenetic modifications (e.g., H3K4me3 and H3K27ac) (84), changes in transcriptome (e.g., quick induction of stress response genes) (84–86), and mitochondrial biogenesis and cell volume increase (6, 7, 48, 49). Genetic or physical disruptions in niche adhesion or factors can initiate the Q-A transition in MuSCs (8, 79–81). A recent study by Kann et al. (78) demonstrated that quiescent MuSCs respond to injury by up-regulation of Rho/ROCK signaling, leading to projection retraction and downstream activation events combining a modified ex vivo single myofiber preparation with in vivo tissue clearing. These very early events, involving swift cytoskeletal rearrangements, were shown to occur independently of exogenous growth factors. Thus, the initial Q-A transition (within 1 to 3 hpi) of MuSCs upon injury may rely on their intrinsic signaling rather than being dependent on signals from other nonmyogenic cells. Despite these recent research advancements, the mechanism underlying the subsequent activation events of MuSCs remains uncertain.

Using a homebuilt dual-laser nonlinear optical microscope platform, we simultaneously visualized various cell types and both intracellular and extracellular structures in the skeletal muscle, including RFP⁺ monocytes/macrophages, RFP⁻ nonmyogenic cells, YFP⁺ MuSCs/ASCs, collagen, and muscle fibers, which facilitates the examination of cellular dynamics of MuSCs and spatiotemporal interaction between nonmyogenic and myogenic cells under different conditions during muscle regeneration (summarized in Fig. 7). The mitochondrial biogenesis and the volume change of MuSCs were first observed during later activation in vivo using our two-photon microscope. Consequently, we found that the mitochondrial mass and volume of MuSCs increased synchronously before 2 dpi after injury and returned to basal level at 1 mpi (Fig. 1, C and D). This observation was consistent with the previous in vitro finding that mitochondrial biogenesis increased continuously in control MuSCs up to 40 hours in culture (6). Moreover, our findings revealed a significant influx of nonmyogenic cells, predominantly neutrophils and monocytes/macrophages, into the injured site within 2 dpi. Previous studies have provided evidence showing that the genetic removal of macrophages during the initial stages of injury leads to impaired muscle regeneration because macrophages play a crucial role by interacting with MuSCs in the regenerative process (18, 19, 21, 25). One specific study suggested that macrophages could promote MuSC activation by secreting ADAMTS1 (28). Additionally, myeloid cells can establish direct connection with MuSCs through membrane ligands and receptors (35, 57). These led to the hypothesis that the infiltrating myeloid cells (particularly neutrophils and monocytes/macrophages) might promote regeneration by directly engaging MuSCs. However, there is still a lack of detailed examination of how myeloid cell–MuSC interactions contribute to subsequent MuSC activation, particularly during the period when myeloid cells infiltrate the injured muscle fibers.

Fig. 7. Graphic model of the dynamics of satellite cells and their interaction with myeloid cells during skeletal muscle regeneration.

According to the time-lapse imaging at 4.5 to 12 hpi and 1.5 dpi, MuSCs and ASCs mostly established transient contact with nonmyogenic cells. The activation of MuSCs was not hindered, although there was a decrease in the infiltration of neutrophils and macrophages, and absence of contact between MuSCs and RFP⁺ macrophages. However, the depletion of macrophages impaired MuSC proliferation and differentiation, resulting in muscle fiber atrophy and fibrosis (created with BioRender.com).

To achieve this, along with our microscope platform, we introduced two macrophage depletion models to examine the behavior of macrophages during muscle regeneration. One was GC treatment (Injured-GC) to suppress the immune response, and the other was the CCR2ko mouse model (Injured-KO). The data of snapshot imaging sessions at different time points indicated that the depletion of macrophages did not impair the initial increase or further growth of MuSC volume in vivo (fig. S5K). In addition, time-lapse imaging suggested that individual MuSCs without contacting RFP⁺ macrophages in the Injured-KO and Injured-GC group increased in volume to the same extent (~20%) from 4.5 to 12 hpi as MuSCs in contact with macrophages in the Injured-Ctrl group. These findings demonstrated that the infiltration of macrophages and direct contact with RFP⁺ monocytes/macrophages were unnecessary for MuSC activation. In addition, depletion of neutrophils in the CCR2ko mice did not delay the initial increase in cell volume of MuSCs (Fig. 3D). In summary, our data provide further evidence of the dispensability of myeloid cell interaction for later activation of MuSCs, despite their prominent presence during this stage.

In contrast to the insignificant role of macrophages in MuSC activation, we found that proliferation and differentiation of MuSCs were impeded by the reduction of macrophages (Figs. 5C and 6, C and G, and fig. S7L), and fibrosis was more significant at 15 dpi in the Injured-KO and Injured-GC groups than in the Injured-Ctrl group (Fig. 6D and fig. S9F). Therefore, the newly recruited macrophages promote the proliferation and differentiation of MuSCs but are unnecessary for their activation in vivo. However, since MuSC differentiation program requires cell cycle exit after limited rounds of cell division, the balance between MuSC proliferation and subsequent differentiation dictates the regeneration outcome (4). It is difficult to discern whether macrophage depletion acts on MuSC differentiation directly or indirectly (e.g., by regulating MuSC proliferation) in vivo. Nevertheless, our findings were in line with previous reports that showed that an overall reduction of macrophages impedes satellite cell differentiation during muscle regeneration in vivo (18, 21, 23, 25, 87). Combining the autofluorescence of NADH from RFP⁻ nonmyogenic cells including RFP⁻ macrophages, neutrophils, and FAPs and TPEF of YFP and RFP, we first demonstrated the mobility of nonmyogenic cells and myogenic cells and their spatiotemporal interaction after an acute injury in the mouse model in vivo in real time. Immediately after injury, MuSCs were motionless, but the nonmyogenic cells (mostly neutrophils and monocytes/macrophages) were highly motile. Direct cell-cell contacts between MuSCs and nonmyogenic cells were mostly frequent and short-lived. Moreover, the total number of nonmyogenic cells that contacted MuSCs would be overestimated because it could not be verified whether the cells came back after leaving the tracing FOV. We also occasionally observed some immobile RFP⁻ nonmyogenic cells that had a similar spindle shape to MuSCs in the NADH channel during MuSC activation (fig. S6, K to M). In very rare instances, some of these immobile nonmyogenic cells established constant contact with MuSCs (fig. S6M). These cells are maybe mesenchymal stem cells, such as FAPs, which are located quiescently between muscle fibers in the noninjury state and share a similar temporal pattern of activation and proliferation with MuSCs after injury (71, 72).

During the proliferation stage, ASCs migrated parallel to the intact muscle fiber, slowing down when they entered the M phase and acquired a round shape (movies S5 to S8). At the same time, most nonmyogenic cells were also highly dynamic, with their mobility being asynchronous. Therefore, ASCs could not establish continuous contact with a specific nonmyogenic cell before division. However, we could not exclude the possibility that gene expression program in ASCs was altered during those time frames with short-lived contact. This observation differed from the previous in vivo finding in larval zebrafish (37). This study observed that MuSCs established prolonged direct contact with a specific dwelling macrophage before division. During the proliferation stage of ASCs in zebrafish, both MuSCs and macrophages were motionless, which facilitated ASCs to establish constant contact with a particular macrophage. However, their communication medium was a pairing between secreted cytokine NAMPT and the chemokine receptor CCR5 but not direct cell-cell signaling. Therefore, nonmyogenic cells may mainly regulate the proliferation of ASCs via paracrine cytokines or metabolite signaling (21, 26–30). As a result, continuous ASC–nonmyogenic cell contact is not required for ASC cell division, albeit the reduction of nonmyogenic cell presence in injury sites slows down the overall MuSC proliferation. The depletion of macrophages that reduced the duration and frequency of contacts between macrophages and myoblasts slightly affected instantaneous migration speed of myoblasts, but the myoblast still migrated along the longitudinal axis of the muscle fiber (fig. S8, D to F).

A few limitations in this study should be noted. First, SMRMs also contributed to muscle regeneration (39, 73), but we cannot investigate their exact role because the depletion models have no significant impact on the number of resident macrophages (38, 59), and most of them cannot be identified without labeling. Besides, current approaches to target neutrophils (including transgenic mouse models or drug treatment) either lack specificity and potency (62) or are not available to us, thus compromising more accurate analysis of neutrophils’ role in this study. Second, NADH signals can, in principle, mark all cells, but not their identity. Our data showed the interaction between nonmyogenic cells and YFP⁺ myogenic cells, but it was still unclear whether different cell types had distinct interaction patterns with myogenic cells. Therefore, mouse models that have specific markers for different cells are required in further study. Third, due to the scope of this study, we did not investigate molecular pathways involved in MuSC–myeloid cell interactions.

In conclusion, we demonstrated a promising method to study the dynamics of different cell types and their interactions in live animals, facilitating further understanding of the mechanism of muscle regeneration on the cellular and tissue level. With this tool in hand, for the first time, we documented MuSC dynamics as well as its interacting pattern with myeloid cells during the course of muscle regeneration in mice. In particular, we found that MuSC–myeloid cell interactions were always transient and prolonged contact was not present or required for MuSC to undergo quiescence exit or cell division, in contrast with a previously published live imaging study. However, depletion of macrophages impairs muscle regeneration due to moderately reduced MuSC proliferation rate and differentiation. Therefore, we would like to propose a stage-dependent role for myeloid cells to regulate MuSC functions during muscle regeneration. Further studies combining our methods, appropriate animal models, or chronic imaging windows (88) can reveal more about the mechanisms of muscle regeneration and promote the development of therapeutic strategies for muscle diseases such as Duchenne muscular dystrophy (DMD).

MATERIALS AND METHODS

Animal preparation

Pax7^{CreERT2(Gaka)} (stock no: 017763), Ccr2^RFP/RFP (stock no: 017586), and Rosa-LSL-YFP (stock no: 006148) mice were from The Jackson Laboratory (Bar Harbor, ME, USA). Appropriate mating schemes were used to generate mice to study MuSC dynamics and interaction between RFP⁺ cells and MuSCs. In Ccr2^RFP/RFP mice, RFP sequence replaces the coding sequence of Ccr2 gene, abolishing gene function (58). To enable YFP expression in adult MuSCs, female and male mice, 2 to 5 months of age, were injected intraperitoneally with three doses of tamoxifen (Santa Cruz Biotechnology; 200 μg/g of body weight, diluted in corn oil) for 7 days. All experiments were conducted at least 1 week after the last injection. Mice were housed two to four animals per cage with a standard 12-hour light/dark cycle in a temperature-controlled environment (22° to 25°C with 40 to 60% humidity) and had ad libitum access to food and water. Before the injury, mice were anesthetized by intraperitoneal injection of a ketamine-xylazine mixture (87.5 and 12.5 mg kg⁻¹). First, the skin covering the TA muscle was shaved of all fur using depilatory cream (Veet) and disinfected with 75% EtOH. Next, the TA muscle was punctured at five sites with a 25-gauge needle (NN-2522R, Terumo) without incising the skin, and the skin was disinfected again. All the experiments were performed following the guidelines of the Laboratory Animal Facility of the Hong Kong University of Science and Technology (HKUST), and the animal protocols were approved by the Animal Ethics Committee at HKUST.

Two-photon microscope

The two-photon system is shown in fig. S1 and was modified from the previously reported multimodal nonlinear optical microscope system (65). In brief, two femtosecond Ti:sapphire lasers (Chameleon Ultra II, Coherent) with 80 MHz were used. One was tuned at 920 nm to excite TPEF of YFP in MuSCs and SHG signals of myosin filament and collagen. The other was tuned at 740 nm to excite TPEF of RFP in macrophages and NADH signals. The power intensity and polarization of the lasers were adjusted by a combination of a half-wave plate (AHWP10M-980, Thorlabs) and a polarizing beamsplitter cube (CCM1-PBS252/M; Thorlabs). The lasers were collimated and expanded by a pair of achromatic doublets to match the 5-mm Galvo XY-scan mirror (6215H, Cambridge Technology). Another polarizing beamsplitter cube was chosen to combine the two lasers and then direct them to the Galvo X scanner, which enables any combination of the two different wavelengths of laser (920 and 740 nm, 810 and 740 nm, or 920 and 810 nm). The XY scanners were mutually conjugated through a 4f relay formed by L5 and L6, both of which consist of two doublets (49-392, Edmunds). The Galvo Y and the rear pupil of the water-immersive objective (OL: XLPLN25XSVMP2, 25×, 1.05 numerical aperture, Olympus) were then conjugated by the scan lens L7 (two doublets, 49-392, Edmunds) and the tube lens L8 (49-365, Edmund) operating in the 4f relay configuration. The objective was mounted on a motorized linear actuator (LNR50SEK1, Thorlabs) for axial sectioning. The epifluorescence collected by the objective was reflected by a 705-nm long-pass dichroic mirror (D1: FF705-Di01-25x36, Semrock) and directed to the detection unit. In addition, a half-wave plate (AHWP10M-980, Thorlabs) was applied before the D1 to change the polarization of the exciting laser and selectively strengthen the SHG signal of collagen or sarcomere. The fluorescence was further separated by a second dichroic mirror (D2: FF495-Di03-25x36, Semrock) and directed into two current photomultiplier tube (PMT) modules (H11461-03 and H11461-01, Hamamatsu). Two band-pass filters F1 (FF02-447/60 or FF01-357/44, Semrock) and F2 (FF03-525/50 or FF01-593/46, Semrock) and two short-pass filters (FF01-680SP, Semrock) were placed before PMTs respectively to reject the excitation laser and select the fluorescence. The output currents of both PMTs were then converted to the voltage by two current amplifiers (SR570, Stanford Research and DLPCA-200, Femto) and subsequently fed into a multifunction data acquisition device (PCIe-6353, National Instruments). Custom-written C# software running in Visual Studio (Microsoft) was used to control all the hardware and acquire the TPEF images.

In vivo imaging

The mice were anesthetized with a ketamine/xylazine mixture before surgery. The skin covering the TA muscle was resected to expose approximately 0.5 cm² of the muscle. The fascia and epimysium were gently removed to avoid damaging the muscle fibers. Next, the hindlimb was put in the middle of a customized stage consisting of metal and heat-insulated materials, as shown in fig. S3. Both sides of the stage were wrapped in a soft heating plate. The coverslip (Ø = 18 mm) attached to a metal support holder was put over the muscle lightly but firmly to minimize the pressure applied to the muscles. The stage securing the mouse was placed on a five-axis stage beneath our prototype microscope. The five-axis stage allows three-axis translation and ±5° pitch and roll flexure motion. The surface of the coverslip was aligned perpendicular to the objective axis by adjusting the roll and pitch angles of the stage guided by the reflection of a laser diode (CPS635R, Thorlabs) on the coverslip. A laser tuned to 920 nm and another to 740 nm were switched sequentially to obtain the YFP/SHG and RFP/NADH signals, respectively. To minimize variability in the early stages of muscle degeneration across experiments, we imaged and analyzed field of views (FOVs) located within approximately 500 μm of the punctured site (fig. S2D). For snapshot imaging, the excitation laser power at the sample plane was kept below 35 mW, and the image integration time was less than 4 s. Optical sections were captured at 2-μm intervals to a total depth of 60 μm. The scanning FOV of the injured site was 300 μm × 300 μm (512 × 512 pixels). To measure the volume of individual MuSCs, the YFP signal of MuSCs was captured (100 μm × 100 μm, 256 × 256 pixels) with 1-μm intervals to cover the entire cells. The imaging duration was 30 to 60 min. The mice were sacrificed after each imaging session.

For several hours of time-lapse imaging, the laser power at 740 nm was less than 25 mW, and the image integration time for a single frame was less than 2 s to avoid photobleaching and photodamage. Serial optical sections were captured at 3-μm steps to a total depth of 60 to 90 μm every 5 min. The scanning FOV of the injured site was 300 μm × 300 μm/400 μm × 400 μm (256 × 256 pixels or 512 × 512 pixels). Besides, the laser power of 740 nm was higher in the time-lapse imaging of ASCs at the proliferation stage. To measure the volume of single MuSCs, the YFP signal of MuSCs (100 μm × 100 μm, 256 × 256 pixels or 50 μm × 50 μm, 128 × 128 pixels) was captured with 1-μm steps to cover the entire cells. The total imaging duration was about 8 hours, but the duration for tracing a specific FOV was about 7.5 hours. During imaging, the mice were anesthetized with 0.5 to 1% isoflurane. The heating pad beneath the mice was set to 36°C to maintain their body temperature. In addition, the muscle temperature was maintained at 28° to 30°C using two heating plates wrapping the stage holding the muscle and the lens warmer enveloping the objective. The lens warmer was set to 37°C; the heating plates were used to heat the stage to 35° to 37°C. Two thermal sensors were used to monitor the temperature of the whole mount. One was put under the lens warmer, and the other was put in the heating pad.

In situ staining of cells for in vivo imaging

MTDR (MitoTracker Deep Red FM, Thermo Fisher Scientific, catalog no. M22426) was dissolved in anhydrous dimethyl sulfoxide (DMSO) to a final concentration of 1 mM and then diluted to 500 nM using saline or phosphate-buffered saline (PBS). The dye was applied topically to the exposed TA muscle. Approximately 30 min after staining, saline or PBS was applied to the muscle to remove the unwanted dye, and the in vivo imaging started. One laser was tuned to 810 nm, and F2 was switched to another band-pass filter (FF01_650/60-25, Semrock) to obtain the TPEF image of MTDR.

Neutrophil depletion

Pax7^CreER/+/Rosa-YFP/CCR2^RFP/+ mice were intraperitoneally treated with 250 μg of anti-mouse Ly6G antibody (Bio X Cell, catalog no. BE0075-1, RRID:AB_1107721) diluted in 150 μl of saline 1 day before needle injury to deplete neutrophils in vivo. In addition, the Pax7^CreER/+/Rosa-YFP/CCR2^RFP/RFP mice were intraperitoneally injected with CP (6055-19-2, MedChemExpress) in two doses to induce neutropenia in the CCR2-deficient mice (63). Initially, 150 mg kg⁻¹ was administered in 200 μl of saline as the first dose on day 1, and the second dose of 100 mg kg⁻¹ was administered on day 4. The mice were injured 12 hours after the last injection.

GC treatment to deplete macrophages

GC treatment was conducted as previously described (24). Briefly, Pax7^CreER/Rosa-YFP/CCR2^RFP/+ mice were treated with 0.9% saline drinking water supplemented with corticosterone (100 μg/ml) (Sigma, St. Louis, MO, 27840-500MG) starting from 3 days before needle injury until the mice were imaged at specific time points after the injury. The water was changed daily, and water consumption was monitored. There was no difference between the treated mice and the control mice (around 7 to 10 ml/day) during the experiments.

Whole-mount muscle immunostaining

The TA muscles were dissected at specific time points after needle injury, fixed in 4% paraformaldehyde (PFA) for 3 to 4 hours at 4°C on a shaker, and rinsed three times with PBS. Glycine PBS solution (0.1 M) was used for blocking free aldehydes. Tissue was then permeabilized using 0.5% Triton X-100 in PBS for 3 to 4 hours and blocked with blocking buffer (0.5% Triton X-100 in PBS containing 4% bovine serum albumin) for 3 to 4 hours. Samples were incubated with primary antibodies and, subsequently, with secondary antibodies diluted in blocking buffer overnight at 4°C. Washing was performed between steps by rinsing the samples three times with 0.3% Triton X-100 in PBS for 1 hour on a shaker. The samples were placed in Mowiol mounting medium and examined with two-photon fluorescence microscopy. The primary antibodies used are anti–type I collagen (SouthernBiotech, catalog no. 1310-01, RRID:AB_2753206), anti-CD45 (BioLegend, catalog no. 103108, RRID:AB_312973), anti–Pdgfr-a (BioLegend, catalog no. 135906, RRID:AB_1953269), anti-Ly6G (BioLegend, catalog no. 127613, RRID:AB_1877163), anti–Gr-1 (BioLegend, catalog no. 108406, RRID:AB_313371), and anti-CD11b (Thermo Fisher Scientific, catalog no. 48-0112-80, RRID:AB_1582237). Secondary antibodies are Alexa Fluor 488 anti-rat immunoglobulin G (IgG) (Thermo Fisher Scientific, catalog no. A-11006, RRID:AB_2534074), Alexa Fluor 594 anti-rat IgG (Thermo Fisher Scientific, catalog no. A-21209, RRID:AB_2535795), Alexa Fluor 647 anti-rat IgG (Thermo Fisher Scientific, catalog no. A48265, RRID:AB_2895299), and Alexa Fluor 488 anti-goat IgG (Thermo Fisher Scientific, catalog no. A32814TR, RRID:AB_2866497).

H&E staining

To perform H&E staining, 8-μm frozen muscle sections were fixed in 4% PFA for 10 min, washed in PBS, and then subjected to staining in hematoxylin for 30 min and eosin for 3 min. The muscle sections were subsequently dried in a series of increasing concentrations of ethanol/water solutions and 100% xylene before being mounted.

Imaging processing and analysis

The images were processed with MATLAB or ImageJ (89).

Cell volume and mitochondria mass measurement of MuSCs

After smoothing and thresholding the three-dimensional (3D) sectioning stacks, the volume was calculated using the plugin “3D Objects Counter” of ImageJ. The threshold was set to the fluorescent intensity covering the whole cell in the stacks’ maximum intensity projection (MIP) image. The NADH signal of MuSC was captured at the best focal plane of the cell. The YFP signal at that plane was regarded as a mask to depict the cell boundary. The mitochondrial mass was calculated as the area size of the NADH signal inside the mask after smoothing, thresholding, and binarizing using ImageJ “Threshold” plugin.

Cell density calculation

The FOV for cell counting was 300 μm × 300 μm × 60 μm. For RFP⁻NADH⁺ cell density, RFP⁻NADH⁺ cells were counted for the MIP of every FOV at 4.5 and 12 hpi. For myogenic cell density, YFP⁺ cells were counted for the MIP of every FOV before and at 2 dpi. For macrophage density, RFP⁺ cells were counted for every FOV before and at 12 hpi. When YFP⁺ cells started to differentiate or RFP⁺ cell density was so high that individual cells could not be identified, we calculated the YFP⁺RFP⁻/RFP⁺/RFP⁺YFP⁺ area density (%) in the injured sites as the YFP⁺/RFP⁺/RFP⁺YFP⁺ cell density. The density was calculated as the ratio of the YFP⁺RFP⁻, RFP⁺, or RFP⁺YFP⁺ area to the injured site indicated by SHG (excluding the intact muscle fiber areas based on periodic signals). We merged the YFP stacks with the RFP stacks to create RGB stacks. Then, we selected the green (YFP⁺RFP⁻) and yellow (RFP⁺YFP⁺) signals using a color threshold and calculated the YFP⁺ and RFP⁺YFP⁺ area density in Matlab, respectively. Since the YFP signal locally distributed inside the macrophages, the RFP⁺YFP⁺ macrophage density was underestimated. The cell density for each mouse was the average value of 3 to 6 FOVs.

Relative collagen area calculation

To calculate the collagen area (%), we projected the stacks of SHG channels every three slices for 15 slices (~30 μm in depth) and manually depicted the dark zone of the injured site where there was no SHG signal from muscle fibers. The contrast of the SHG of collagen in the dark zone was adjusted, and the images were median-filtered (neighborhood size was 6 × 3). The density was calculated as the ratio of the collagen area to the dark zone area after thresholding using multithresh.m in Matlab. The collagen area of one FOV was the average value of the slices, and the collagen area for one mouse was the average value of 3 to 6 FOVs.

Maximal distance of cell boundary depicted by RFP/YFP and NADH calculation

To determine the cell boundary as depicted by RFP/YFP and NADH signal, respectively, the x-y maximal projection of cells was analyzed first. The cell boundary was manually determined according to the intensity distribution by using Freehand selections in ImageJ. The maximal lateral distance of cell boundary was measured based on their boundary. To analyze the maximal axial lateral distance, the 3D sectioning stacks of different signals (z step: 2 μm) were analyzed.

Time-lapse imaging processing

Stacks of four channels from optical sections were merged, converted to hyperstacks, and aligned by reference to the SHG signal using the plugin “Correct 3D drift” in ImageJ. To calculate the minimal distance of individual MuSCs from macrophages from 4.5 to 12 hpi, RFP⁺ and YFP⁺ stacks (volume is 300 μm × 300 μm × 60 μm) for different frames were first denoised, setting the gamma value to 0.8, and smoothed and then converted to binary images using the ImageJ Threshold plugin with the “Default” auto thresholding method. The 3D distance matrix was calculated from the binary image of macrophages using bwdistsc.m (90) in the Matlab so that the further a point was from macrophages, the higher the pixel value. The area of each YFP⁺ cell was labeled using bwlabeln.m in Matlab. The minimal distance of every YFP⁺ cell from macrophages was the minimal value of the corresponding pixels in the distance matrix. If the minimum distance was 0 μm, the YFP⁺ cells were in contact with macrophages. In addition, to analyze the contact between YFP⁺ cells and RFP⁺ cells at the proliferation stage, the slices covering the whole YFP⁺ cell were stacked as a maximum z projection. The images were then denoised and smoothed. We tracked individual RFP⁺ that contacted YFP⁺ ASCs and ASCs using the ImageJ plugin “Manual Tracking.” Direct cell-cell contact was defined by at least one RFP⁺ pixel juxtaposed to at least one YFP⁺ pixel and was confirmed by the examination of individual z planes in the raw stacks. The analysis of the contact between YFP⁺ cells and RFP⁻NADH⁺ nonmyogenic cells was similar. Because the difference between the cell boundaries depicted by the NADH signal and RFP/YFP signal was less than 2.5 μm laterally and 4 μm axially, we projected two more slices in two directions for YFP⁺ cells, and if the lateral distance between YFP⁺ cells and RFP⁻NADH⁺ nonmyogenic cells was less than 2.5 μm, contact was established. Notably, only the cells with visible NADH signal (mainly located within 20 to 30 μm in depth) were analyzed. In addition, the image sequences after registration were further analyzed in Imaris (Bitplane AG, Zurich, Switzerland) to obtain the migration statistics (migration speed, migration direction, and so on). It should be noted that despite manual optimization of cell tracking analysis in Imaris, the tracking of RFP⁺ cells might not be entirely accurate owing to the high density of RFP⁺ cells in both the Injured-Ctrl and Injured-GC groups.

Statistical analysis

Statistical analysis and data visualization were performed using GraphPad Prism 7 software. All similar measurements were performed on more than two mice. Details of data replicates could be found in figure captions. Data are presented as mean ± SD, individual points, or violin plots, and α = 0.05 for all analyses. Data distributions shown as violin plots document the median (horizontal central line) and interquartile range (upper quartile and lower quartile, horizontal lines above and below central median line, respectively). For comparison between two groups or multiple groups with two to six samples, unpaired two-tailed t test or one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test was used. For comparison between two groups or multiple groups with N ≥ 7 samples (cell volume or duration of interactions and so on), data normality was checked using Shapiro-Wilk normality test first. Nonnormally distributed data were analyzed using Kruskal-Wallis test or Mann-Whitney test. Normally distributed data were analyzed using unpaired two-tailed t test or one-way ANOVA with Tukey’s multiple comparisons test. Two-way ANOVA with Tukey’s multiple comparisons test was used to analyze multiple groups’ comparison with two categorical variables. No statistical methods were used to predetermine sample sizes, but the number of animal models was similar to reported studies of skeletal muscle regeneration (27, 40, 48, 77).

Supplementary Materials

This PDF file includes:

Figs. S1 to S9

Legends for movies S1 to S8

References

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S8